ADVANCES IN PROTEIN CHEMISTRY VOLUME VII
This Page Intentionall Left Blank
ADVANCES IN PROTEIN CHEMISTRY EDITEDBY
...
22 downloads
1150 Views
23MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
ADVANCES IN PROTEIN CHEMISTRY VOLUME VII
This Page Intentionall Left Blank
ADVANCES IN PROTEIN CHEMISTRY EDITEDBY
M. L. ANSON
KENNETH BAILEY
Research Division; Lever Brothers Co. Edgewater, New Jersey
University of Cambridge Cambridge, England
JOHN T. EDSALL Harvard Medical School Boston, Massachusetts
VOLUME VII
1952 ACADEMIC PRESS INC., PUBLISHERS NEW YORK, N. Y.
Copyright 1952, by ACADEMIC PRESS INC. 125 EAST 2 NEW YORK
3 STREET, ~ ~ 10, N . Y.
All Rights Reserved No part of this book may be reproduced in any form, by photostat, microfilm, or any other means, without written permission from the publishers. Library of Congress Catalog Card Number: 44-8853
PRiNTED I N THE UNITED STATES OF AMERICA
CONTRIBUTORS TO VOLUME VII
RICHARD S . BEAR,Department of Biology, Massachusetts Institute of Technology, Cambridge, Massachusetts
G. H. BEAVEN, Medical Research Council Spectrographic Research Unit, London Hospital, London, England E. R. HOLIDAY, Medical Research Council Spectrographic Research Unit, London Hospital, London, England HILDEGARD PORTZEHL, Physiological Institute, University of Tubingen, Germany K. M. RUDALL, Department of Biomolecular Structure, The University, Leeds, England F. SANGER, Sir William D u n n Institute of Biochemistry, University of Cambridge, Cambridge, England G. B. B. M. SUTHERLAND, Physics Department, University of Michigan, Ann Arbor, Michigan HANSH. WEBER, Physiological Institute, University of Tubingen, Germany
V
This Page Intentionall Left Blank
CONTENTS CONTRIBUTORS TO VOLUME VII . . . . . . . . . . . . . . . . . . . . . .
v
The Arrangement of Amino Acids in Proteins
BY F. SANGER.Sir W i l l i a m D u n n Institute of Biochemistry. University of Cambridge. Cambridge. England I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 I1. Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 5 111. Determination of the Position of Individual Residues in Proteins . . . . IV. Methods for the Degradation of Proteins . . . . . . . . . . . . . . . 11 V. Fractionation of Peptides . . . . . . . . . . . . . . . . . . . . . . 29 VI . Determination of Peptide Structure . . . . . . . . . . . . . . . . . 41 VII . Results of Investigations on Various Proteins . . . . . . . . . . . . . 44 i7111. General Conclusions . . . . . . . . . . . . . . . . . . . . . . . . 59 The Structure of Collagen Fibrils
BY RICHARDS. BEAR.Department of Biology, Massachusetts Institute of Technology, 1. I1. 111. IV. V. VI .
Cambridge, Massachusetts Introduction . . . . . . . . . . . . . . . . . Identification and Distribution of Collagens . . . Colloidal Structure of Collagen Fibrils . . . . . The Collagen Protofibril . . . . . . . . . . . . The Collagen Molecule . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. 69 . . 76 . . 97 . 112 . . . . . . . . . . 135 . . . . . . . . . . 150
Muscle Contraction and Fibrous Muscle Proteins
BY HANSH . WEBERA N D HILDEGARD PORTZEHL, Physiological Institute, University of Tubingen, Germany
I . Introduction . : . . . . . . . . . . . . . . . . . . . . . . . . . I1. The Contractile Models . . . . . . . . . . . . . . . . . . . . . . I11. The Proteins of the Myofibril and Their Reactions . . . . . . . . . . . IV . The Proteins of the Myofibril and the Fine Structure of Skeletal Muscle . V. General Conclusion . . . . . . . . . . . . . . . . . . . . . . . .
162 163 193 237 246
The Proteins of the Mammalian Epidermis
BY K . M . RUDALL, Department 0.f Biomolecular Structure, Leeds, England I. Introduction . . . . . . . . . . . . . . . . . . . . I1. Properties of the Epidermis as a Whole . . . . . . . 111. The Extraction of Epidermal Proteins . . . . . . . . I V. Some Physical Properties of Epidermal Proteins . . . . V. Infrared Absorption Studies . . . . . . . . . . . . . VI . Summary . . . . . . . . . . . . . . . . . . . . . vii
The University,
. . . . . . . . . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
253
. . 254 . . 264 . . 270 . 279 . 286
...
Vlll
CONTENTS
Infrared Analysis of the Structure of Amino Acids. Polypeptides and Proteins
BY G . B . U . M . SUTHERLANI). I’hysics Department. University of Michigan. Anri Arbor. Michigan I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . I1. Amino Acids . . . . . . . . . . . . . . . . . . . . . . . . . . I11. Amides . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV. Polypeptides . . . . . . . . . . . . . . . . . . . . . . . . . . . V. Proteins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VI . Some General Observations . . . . . . . . . . . . . . . . . . . . .
2!)1 297 299 304 306
312
Ultraviolet Absorption Spectra of Proteins and Amino Acids BY G . H . BEAVENAND E. R . HOLIDAY, Medical Research Council Spectroyraphic Research Unit. London Hospital. London. England I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 320 I1. Methods and Experimental Aspects . . . . . . . . . . . . . . . . . 321 I11 Absorption Constants of the Aromatic and the SuIfnr-(:ontaimirig Amino Acids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323 I V . The Vibrational Fine Structure of thc Ahsorption Spectra of the Aromatic Amino Acids and of Proteins . . . . . . . . . . . . . . . . . . . 329 V . Low Temperature Spectra of Amino Acids and Proteins . . . . . . . . 331 VI . Fine-Structure Shifts in Protein Spectra-Structural Implications . . . . 336 VII The Absorption Spectra of the Aromatic Amino Acids and Proteins in Strongly Alkaline Solution (pH 12-13) . . . . . . . . . . . . . . 343 VIII . The Ultraviolet Absorption Spectrum of the Peptide Bond and of the Polypeptide Fabric . . . . . . . . . . . . . . . . . . . . . . . . . 352
.
.
AUTHORINDEX . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
387
SUBJECT INDEX . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
400
The Arrangement of Amino Acids in Proteins BY F. SANGER Sir W i k m D u n n Institute of Biochemistry. [Tniversity of Cambridge. Cambridge. England
CONTENTS I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I1. Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1. Polypeptide Chains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a. Open Polypeptide Chain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
P'age
............................. . . . . . . . c. Branched Chain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2 . Terminal Residues and Peptides. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I11. Determination of t h e Position of Individual Residues s. . . . . . . . 1. The Dinitrophenyl (DNP) Method . . . . . . . . . . . . . . . a . Fractionation of D N P Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . h N-Terminal Residues of Proteins . . . . . . . . . . . . . . . . . . . . . . . . . . . . c . N-Terminal Peptides . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 . The Phenylthiocarbamyl Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3. The Thiocarbamate Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4. The Use of S-Methylisothiourea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 . C-Terminal Residues ........................................ a. Carboxypeptidase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . h Reduction to B-Amino Alcohols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV . Methods for the Degradation of Proteins . . . . . . . . . . . . . . . . . . 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a. Non-Specific Degradation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . b . Specific Degradation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 . General Aspects of Partial Hydrolysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a . Yield of Peptides., . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . b . Complexity of Partial Hydrolyzates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . c. The Question of Rearrangement of Peptide Sequences during Hydrolysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 . Hydrolysis in Concentrated Acid., . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a . Electrostatic Effects . . . . . . . . . .
.
.
.
.
4. 5. 6.
7. 8.
2 3 3
3 3 4 4
5 5 5 6 0
7 8 9 9 9 9
10 11 11 11 11 12 12 13
15 18 18 19 ............................................ c . Bonds Involving the Amino Groups of the Hydroxy Amino Acids. . 21 Hydrolysis in Dilute Acid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 Hydrolysis in Alkali . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 Hydrolysis with Proteolytic Enzymes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 Rates of Hydrolysis of Proteins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 27 Non-Hydrolytic Methods of Degradation . . . . . . . . . . . . . . . . . . . . . . . . . 1
2
F. SANOER
Page 27 29 29 29 30 31 34 35
a. Splitting the Disulfide Bridges.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . b. Splitting by Radiation, ......................... .. ...... c. Other Methods,. , . , . , , ................................... V. Fractionation of Peptides, . . , ........................ 1. Ionophoretic Methods, , , , ....... ... .... . ...... ... .... . ... ... 2. Ion Exchange Methods, , , ... ....... . ...... . .... 3. Adsorption Chromatography. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4. Partition Chromatography. . . . . . . . . . . . . . . . , . . . . . . . . . . . . . . . . . . . . . a. Paper Chromatography. . . , , . . . , . . . . . , . . . . . . . , , , . . . , . . . . , . . b. Starch Chromatography. . , , , . . . . . . . . . . . . . . . . . . . . . . . . . . . , . . . c. Other Partition Chromatography, . . , , , , . . . . . . . . . . . . , , . . . . . . . 5. Detection of Peptides from Columns. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 . Countercurrent Distribution. , , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . , 7. Lysine Peptides.. . . . . . . . . . . , . . . . , . . . . , . . . . . . . . . . . . . 8. Cystine Peptides.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ,
9. Other Methods of Fractionation ............................... 10. Conclusions. , , . . . , . . . . . , . . . . . VI. Determination of Peptide Structure.. . , . . . . . . . . . . . . . . . . . . . . , . . . . . . . , . 1. Identification and Estimation of Amino Acids.. . , , , . . . , , , , , . , , . . , . 2. Amino Acid Sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..._ 3. Estimation of Peptides. . , . . . . . . . . . . . . . . . . . . . . . . . . . . . . , . . . , . . . . . ........................ . The Pipsyl Method. . . . . . . . . . . . . . . . s . .. . . . . . . . . . . . . . . . . . . . . . . VII. Results of Investigations on Various Pro 1. Silk Fibroin (Bombyz M o r i ) . . . . . . . . . . . . . . . . , . . . . . . . . . . , . . . . . . . . , 2. Protamines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3. Wool Keratin , . . . . . . . . . , . . , . . . . . , . . . . . . . . . . . . . . . . , . . . , , . . , , . . . , . 4. Insulin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a. N-Terminal Peptides. , . . , . , . b. Amino Acid Sequence in the Phenylalanyl Chains.. . . . . . . . , . . . . . . ....................... 5. Ovalbumin.. . . . . . . . . . . . . . . . . . . 6. 7-Globulin.. . . . . . . . . . . . . . . . . . , 7. Hemoglobin., , . , , . , . . . . . . . . . . . 8. Gelatin.. . . . , , . . . . . . . . . . . . . . . VIII. General Conclusions, , , . . . . . . . . . . . References ,
38 38 39 40 40 40 41
41 42 43 43 44 44 47 48 50 54
57
62
I. INTRODUCTION A comprehensive review of the earlier literature on the partial hydrolysis products of proteins was given by Synge in 1943. Up to that time only a few simple peptides had been clearly identified from proteins by the classical and rather laborious methods of organic chemistry and Synge concluded that ‘ I the main obstacle to progress in the study of protein structure by the methods of organic chemistry is inadequacy of technique!” Probably the greatest advance that has been made recently in this field was the development by Martin and Synge (1941) of the entirely new technique of partition chromatography. The great problem in peptide chemistry has always been to find methods of frac-
THE ARRANGEMENT OF AMINO ACIDS I N PROTEINS
3
tionating the extremely complex mixtures produced by the partial degradation of a protein. Older methods of fractional crystallization and precipitation with various reagents were as a rule inadequate to deal with these mixtures, and countercurrent methods of high resolving power, which could fractionate non-volatile, water-soluble substances, were needed. Partition chromatography, especially in the form of paper chromatography (Consden et al., 1944), is such a method, so that it has already been possible to identify as breakdown products of proteins more peptides using this technique than had previously been identified by the classical methods of organic chemistry. During the last few years, work in this field has centered largely on the development of methods, so that this review will be more a consideration of techniques and their uses than a discussion of results, which are still rather few. As an initial working hypothesis it will be assumed that the peptide theory is valid, in other words, that a protein molecule is built up only of chains of a-amino (and a-imino) acids bound together by peptide bonds between their a-amino and a-carboxyl groups. While this peptide theory is almost certainly valid (see Vickery and Osborne, 1928; Pauling and Niemann, 1939; Synge, 1943), it should be remembered that it is still a hypothesis and has not been definitely proved. Probably the best evidence in support of it is that since its enunciation in 1902 no facts have been found to contradict it. It is to be expected that investigations of the types described in the present article will throw further light on the accuracy of the peptide theory and on the possible existence of nonpeptide bonds in proteins.
11.
NOMENCLATURE
Some of the terms to be used are new and require definition. 1. Polypeptide Chains
Three types of polypeptide chains are possible, open, cyclic and branched. a. Open Polypeptide Chain. A chain of amino acids joined together by peptide bonds between the a-amino and a-carboxyl groups with a free amino group at one end of the chain and a free carboxyl group at the other end is the most usually considered type of polypeptide chain. The number of these chains in a protein may be estimated by the number of a-amino or a-carboxyl groups. b. Cyclic Chain. A cyclic chain may be derived from an open polypeptide chain by peptide bond formation between the two terminal residues, and contains no free a-amino or a-carboxyl groups. The anti-
4
F. MANGER
biotics “gramicidin S ” (Sanger, 1946) and tyrocidine (Christensen, 1945) have been shown to possess cyclic structures and the absence of any free a-amino groups in ovalbumin and certain muscle proteins suggests that they too are built up from cyclic chains. c. Branched Chain. The presence of two carboxyl groups in glutamic and aspartic acids and of two amino groups in lysine suggests the possibility that branched chains may occur in proteins. For instance, such a branched system could be formed by the formation of a peptide bond between the free a-amino group of one open chain and a y-carboxyl group of a glutarnic acid residue in another chain. For all proteins that have been studied by the dinitrophenyl (DNP) method (see below) it was found that all the c-amino groups of the lysine residues were free, indicating that these groups are not involved in any bond formation. Thus it is unlikely that there is very much branching of chains from the lysine residues. If the branching points are very few, however, it is just possible that some of the c-amino groups thus masked have escaped detection. No evidence is available as t o whether hranahing can occur from the w-carboxyl groups of aspartic or glutamic acids. The presence of the y-peptide linkage of glutamic acid in glutathione, glutamine and the capsular substances from B. Anthracis (Hanby and Rydon, 1946) suggests that it may also occur in proteins. Some proteins are built up of two or more polypeptide chains held together by a stable bond other than the peptide link. Such a bond will be referred to as a “cross linkage.” The only one that is definitely bridge of cystine. Here two cysteine known to exist is the -S-Sresidues are joined together through their side chains.
2. Terminal Residues and Peptides There are two types of terminal residues, those with a free amino group and those with a free carboxyl group. In previous publications (Sanger, 1945, 194913) the expression “terminal residue’’ has been used to denote only that residue which carries a free amino group. However, it seems that a distinction should be made. Fox (1945) has suggested that those residues containing a free amino group and a bound carboxyl group be referred to as terminal amino acids and those with a free carboxyl group as terminal amino acids. Although clear on paper, this distinction is rather difficult to make in conversation. Following a suggestion by Dr. K. Bailey, it is proposed to use the term N-terminal residue for the residue having a free amino group and C-terminal residue for that having a free carboxyl group. The same nomenclature will apply to the terminal peptides.
5
T H E ARRANGEMENT O F AMINO ACIDS I N P R O T E I N S
3. Abbreviations Throughout this review the abbreviations for the amino acid residues suggested by Brand and Edsall (1947) are used. Cysteic acid is abbreviated CySOSH. Their method of writing empirical formulae of proteins and peptides is also adopted, e.g., Gly2vAlallValzl . . . etc., the subscripts representing the number of residues of the amino acid per protein molecule. I n this type of formula the order in which the amino acids are given has no significance. Where the order of residues is known, as in the description of a peptide, the symbols are joined by a period. Thus, whereas Gly,Ala signifies a dipeptide containing glycine and alanine, Gly.Ala represents glycylalanine. Gly. (Ala,Leu) indicates a peptide containing glycine, alanine and leucine with glycine as the N-terminal residue, and the order of the alanine and leucine is unknown. As is customary, the first residue is the N-terminal residue and the last the C-terminal.
111. DETERMINATION OF THE POSITION OF INDIVIDUAL RESIDUES IN PROTEINS The terminal residues of proteins differ from other residues in the chain, since they contain free amino or free carboxyl groups, and this fact may be used to identify them. Fox (1945) has reviewed the earlier literature on the study of terminal residues. At that time the position of only one amino acid in one protein was known. This was the presence of phenylalanine as an N-terminal residue in insulin. It was identified by Jensen and Evans (1935) who isolated the phenylhydantoin of phenylalanine from a hydrolyzate of insulin that had been treated with phenylisocyanate. More recently a general method has been worked out for the study of N-terminal residues, and preliminary investigations have been carried out on three methods of stepwise degradation which promise t o be of great use in the future. 1. The Dinitrophenyl ( D N P ) Method
No20
The principle of this general method (Sanger, 1945; Porter and Sanger, 1948) for the identification and estimation of the N-terminal residues of proteins may be summarized by the following formulae: +NH2iHCO-prot. +JN /h,(
NO2 1 :2 :4-Fluorodinitrobenaene (FDNB)
Protein
k
NHCH.CO-prot. +NOJz/,
k
NH.CH.COOH
IICl
N? 2. 2: 4-Dinitrophenyl(DNP)-protein
NOz DNP-Amino acids
+
Amino acids
6
F. SANQER
The FDNB reacts with the free amino groups of the protein under mild (slightly alkaline) conditions where the peptide bond is quite stable. On hydrolysis of the protein the N-terminal residues are liberated in the form of DNP amino acids. These are bright yellow compounds that can be extracted with an organic solvent, fractionated chromatographically, and estimated colorimetrically. The accuracy varies somewhat with the particular amino acids involved, due to differences in the stability of the DNP derivatives. I n most cases the N-terminal residues of proteins and peptides may be estimated to within 10-15%. a. Fractionation of D N P Derivatives. Several methods of fractionation have been suggested, all depending essentially on partition chromatography. Originally (Sanger, 1945; Porter and Sanger, 1948; Porter, 1950c) a scheme was worked out for separating the D N P derivatives of all the known amino acids using silica gel saturated with water as the stationary phase and various organic solvents as the mobile phase. This method was found to give satisfactory and reproducible separations in the author’s laboratory; other workers (Consden et al., 194710; Blackburn, 1949), however, have found difficulty in obtaining suitable gels. These may be prepared from any type of sodium silicate by appropriate modification of the method of preparation (Desnuelle et al., 1950), and should be rather strongly adsorbent in order to hold the relatively insoluble D N P derivatives in the stationary phase. Other methods of fractionation have been suggested in which the stationary phase is a buffer, adsorbed on kieselguhr (Bell et al., 1949) or silica (Blackburn, 1949; Middlebrook, 1949). The DNP-derivatives are then partially ionized and are thus rendered more soluble in the aqueous phase. Such systems may prove of more general use, though R values are still not reproducible on different batches of silica (K. Bailey, unpublished observation). The use of paper chromatograms using buffer solutions has also been suggested (Blackburn and Lowther, 1950; Monnier and Penasse, 1950) though details have not been reported. Recently Partridge and Swain (1950) have obtained excellent separations using butanol adsorbed on rubber as the stationary phase and buffer solution as the mobile phase. Such a system with an aqueous moving phase would be expected to be more satisfactory for compounds which are more soluble in organic solvents than in water. b. N-Terminal Residues of Proteins. I n Table I arelisted the N-terminal residues of a number of proteins as determined by the DNP-technique. Besides reacting with the a-amino groups of proteins, FDNB also reacts with the eamino groups of the lysine residues, and an estimation of the e-DNP-lysine in the hydrolysate indicates how many of these amino groups are free in the intact protein. For all proteins studied reasonable
7
THE ARRANGEMENT OF AMINO ACIDS I N PROTEINS
agreement is found between the e-DNP-lysine and the total lysine content of the protein, if the protein is first denatured. It thus seems unlikely that any branching of peptide chains occurs through the €-amino TABLEI N-Terminal Residues of Proteins Protein
Assumed mol. wt.
Insulin" (Ox,pig, sheep)
12,000
Hemoglobinb (horse, donkey) Hemoglobinb (ox, sheep, goat)
66 ,000 66,000
Hemoglobinb (human adult) Hemoglobinb (human fetal) Myoglobinb (horse) Myoglobinc (whale) B-Lactoglobulind OvalbumineJ ?-Globulin' (rabbit) Edestino
66,000 66,000 17,000 17,000 40,000
Salmineh ClupeinA Myosini Tropomyosini Lysozyme j Pancreatic Trypsin Inhibitork Serum albumin' (human, horse, ox) ~-
-
160,000 300,000 -
-
14,000 9 ,000 69,000
N-Terminal residue
Number per molecule
Phenylalanine Glycine Valine Valine Methionine Valine Valine Glycine Valine Leucine None Alanine Glycine Leucine Proline Proline None None Lysine Arginine Aspartic acid
2 2 6 2 2 5 2-3 1 1 3 1 6 1 ? ? -
-
< 1 1 1
~~~
Sanger (1945). b Porter and Sanger (1948). c Schmid (1949). d Porter (1948). 0 Deanuelle and Cmal (1948). f Porter (1950a). 0 Sanger (1949d). * Felix et al. (1950). Bailey (1951). 1 Green and Schroeder (1951). k Green and Work (1951). 1 van Vunakis and Brand (1951). 0
groups of the lysine residu s. In ertain native proteins (e.g., p-lactoglobulin) some of the e-amino groups do not react, indicating that they are in some way masked in the native molecule (Porter, 1948). c. N-Terminal Peptides. The DNP technique may also be used to identify and estimate N-terminal peptides (Sanger, 194933). Thus if a DNP-protein is only partially hydrolyzed, one obtains DNP-peptides ~~
8
F. SANGER
which may be extracted into an organic solvent and fractionated on suitable chromatograms. Complete hydrolysis of these purified DNPpeptides then reveals the nature of the amino acids and the N-terminal residue present, and amino acid arrangement may be determined by a second partial hydrolysis. In this way it is possible to identify the residues that occupy positions in the polypeptide chains near to the N-terminal residues. So far it has been possible to identify peptides up to about four residues long. Longer DNP-peptides are more difficult to separate from the other unsubstituted peptides and to fractionate, though the possibilities have not yet been fully explored. For an illustration of the use of these methods the reader is referred t o the section on insulin (p. 50). 2. The Phenylthiocarbamyl Method
Recently Edman (1950) described a method for determining the sequence of amino acids in a polypeptide chain, by splitting off one residue at a time starting from the N-terminal residue. The principle of the method is illustrated by the following equations:
0’ NCS
NHzCHCO-pep.
--+o
NHCSNHCHCONH-pep.
Pyridine
---o
Phenylisothiocyanate Peptide
Anhydrous
IICl
-N---cS
CO i
‘CL R Phenylthiohydantoin
I
R
Phenylthiocarbarnyl (PTC)-peptide
-I- NH2-pep. NHI. alkali
R
I
NH2-CH-COOH N-Terminal amino acid
The formation of the phenylthiohydantoin does not require the presence of water, as does the hydrolysis of peptide bonds. Thus by heating the PTC-peptide with anhydrous HC1 in nitromethane i t is possible to break off the N-terminal residue as a phenylthiohydantoin without splitting other bonds in the peptide. The hydantoin dissolves in the nitromethane and is then separated and hydrolyzed to the amino acid which may be identified by paper chromatography. The rest of the peptide with the N-terminal residue removed is insoluble in nitromethane and the process may be repeated. The second residue is thus split off and identified. This method has given excellent results with synthetic peptides and it will be interesting to see how far it may be applied to a
THE ARRANGEMENT OF AMINO ACIDS I N P R O T E I N S
9
protein. Theoretically, it should be possible to determine the complete structure of a pure single chain polypeptide or protein. At least the method should be extremely valuable for working out the structure of smaller peptides.
3. T h e Thiocarbamate Method Another method for the stepwise degradation of a polypeptide chain has been suggested by Levy (1950). Here the N-terminal residue is split off as a 2-thiothiaaolid-5-one derivative as follows: R
R
AHCONH-pep. ""f, AHCONH-pep.
I
Alkaii
NH2
1
NHCSS-
R pH 3-4 --+
I
CH-CO NI H S1
+ NHt-pep.
'CL Peptide
Thiocarbamate
2-Thiothiazolid-5-one
Here again the process may be repeated on the peptide chain containing one residue less. This method has not yet been worked out on a small scale, but was found to give satisfactory results with synthetic peptides. The reaction of C S z with amino groups has also been used by LBonis (1948) as the basis for a titration method for estimating the total a-amino groups as well as c-amino, imino, and thiol groups. CS2 reacts about ten times as rapidly with a-amino groups as with e-amino groups, so that the two may be distinguished. 4. T h e Use of S-Methylisothiourea Christensen (1945) has used S-methylisothiourea to study the free amino groups of tyrocidine. The amino groups are converted to guanidine groups which may be estimated by the Sakaguchi reaction. This method is especially useful for detecting the free amino group of ornithine, which is converted to arginine and thus may be estimated using arginase. In the case of tyrocidine the Sakaguchi reaction was negative after the action of arginase, indicating that the only free amino groups were the b-amino groups of ornithine, and that tyrocidine was thus a cyclopeptide. It is doubtful if this method could be used to identify the N-terminal residue of more complex proteins. 5. C-Terminal Residues
a. Carboxypeptidase. Specificity studies (reviewed in Neurath and Schwert, 1950) indicate that carboxypeptidase attacks only those peptide bonds that are adjacent to a free a-carboxyl group. It appears that all
10
F. SANGER
such bonds with the possible exception of those involving glycine are attacked to some extent although the rate of hydrolysis varies greatly with the nature of the residue involved. Thus if a protein or peptide is treated with carboxypeptidase the first amino acid to be liberated in the free form is the C-terminal residue. Lens (1949) has applied the method to insulin. Samples of the digest were removed at various intervals, ultrafiltered and the ultrafiltrate analyzed by paper chromatography. Free alanine was clearly liberated first and was therefore present in insulin as a C-terminal residue. With a single pure polypeptide chain it should theoretically be possible to determine the complete sequence of residues by following the rate of liberation of different amino acids under the action of carboxypeptidase. No such experiments have been described, but it would be interesting to know how far this method could be applied in practice. Clearly it would be impossible to draw any conclusions beyond the C-terminal residues if more than one peptide chain were present, as in the case of insulin. b. Reduction to /3-Amino AZcohoZs. Fromageot et al. (1950) and Chibnall and Rees (1951) have independently worked out techniques for the identification of C-terminal residues by reduction to 8-amino alcohols. I n the former method lithium aluminum hydride is used to reduce the protein. After hydrolysis, the p-amino alcohols are extracted into ether, identified by paper chromatography and estimated by reaction with periodate. Some reduction of peptide bonds was observed, and the yields were rather low. In the method of Chibnall and Rees, the carboxyl groups are first esterified with diazomethane and then reduced with the less violent reagent, lithium borohydride. The amino alcohols are separated from the amino acids in the hydrolyzate by electrodialysis, and treated with periodate, which decomposes them according to the equation :
A determination of the extra formaldehyde and ammonia produced in this reaction gives an estimation of the number of free a-carboxyl groups, and the residue on which they are located is identified from the nature of the amino alcohol and of the aldehyde RCHO. Chibnall and Rees have also used this technique to determine the distribution of the protein amide groups between asparagine and glutamine. Residues of aspartic or glutamic acid which have a free w-carboxyl group are destroyed by the above treatment, while those in amide form remain intact and can be estimated after hydrolysis of the protein.
THE ARRANGEMENT OF AMINO ACIDS I N PROTEINS
IV. METHODSFOR
THE
11
DEGRADATION OF PROTEINS
1. Introduction
At present the only residues to which definite positions in the protein chain can be assigned, are those at or near the terminal positions, and these residues constitute only a small part of the molecule. The position of other residues in the chain can only be determined in the first place relative t o one another by the identification of products of partial breakdown. The absolute location of any one residue can be decided only when the relative positions of all the residues are known and when the complete structure of the protein is thus worked out. Thus in determining the amino acid sequence of a protein or polypeptide the main task would seem t o be to identify as many degradation products as possible. There are essentially two ways of approaching this problem, the first requiring non-specific methods of degradation and the second specific methods. a. Non-Specific Degradation. One method of determining amino acid sequence is to degrade the polypeptide or protein directly to a mixture of small peptides whose structure can be determined and to work out a unique sequence by fitting together the degradation products. It is possible to determine directly the structure of di- and possibly tripeptides. To obtain a unique solution for the amino acid sequence of a polypeptide it is necessary to obtain as many different small peptides as possible. This may best be achieved by using unspecific methods of degradation to give a mixture of maximum complexity and using reagents having different specificities. If we consider a hypothetical polypeptide whose structure may be written as A.B.C.D.E.F.G.H. where the letters represent different amino acid residues, degradation to the dipeptides AB,BC,CD,DE,EF,FG,GH would give a unique solution. If only AB,DE,FG and GH are obtained, various structures are possible, but if a different type of hydrolysis yields EF and the tripeptide BCD then the structure is determined. This method will generally be applicable to simpler peptides. I n the case of larger peptides, the complexity of the mixture will render fractionation difficult and a proportion of larger peptides will be required to work out an unambiguous result. b. Specific Degradation. Most proteins contain a t least 100 residues. It seems unlikely that a complete solution of their amino acid sequence will ever be obtained solely by the non-specific method and it will probably be necessary t o approach the problem by the second method, involving a gradual specific breakdown into a small number of large fragments which can be purified and again broken down to smaller
12
F. SANGER
products. Thus if the peptide A.B.C.D.E.F.G.H. could be split into two products A.B.C.D. arid E.F.G.H. which could be separated, their structure could be determined much more readily than that of the original peptide. In the choice of a suitable method of degradation three main factors are to be considered: 1. The reagent should cause a minimum of side reactions such as destruction of the constituent amino acids, or if such reactions do occur they should occur quantitatively to produce known products. 2. The reagent should exhibit the desired specificity. Specific methods are required for the initial breakdown and several relatively non-specific methods are needed for the final elucidation of sequence. 3. There should be a minimum of synthesis or rearrangement of peptide bonds under the influence of the reagent. The first two points will be considered in connection with the different methods of hydrolysis, and the last will be discussed separately on page 15. 2. General Aspects of Partial Hydrolysis By far the most important method of breaking down proteins is by hydrolysis of the peptide bonds. All the peptide bonds in a protein are susceptible to hydrolysis but there are great differences in the stability of the different bonds depending on the nature of the residues involved. a. Yield of Peptides. If we assume that 1 mole of a polypeptide A.B.C.D.E.F is hydrolyzed to such an extent that the mole fractions of the bonds A-B, B-C, C-D etc. that are split are b,c,d etc. respectively, the yield of the amino acid C will be cd moles, since only those molecules of the polypeptide in which the bonds B-C and C-D are broken will give rise to C. T o obtain the dipeptide CD the bonds BC,DE must be broken and CD unbroken, so that the yield will be c(l - d)e. In general the yield F of a peptide A1.A2.A.3 . . . A,. is given by:
where a1,az . . . a, are the mole fractions of the bonds involving the amino groups of A1, Az, . . . A,,, respectively that are split, and F is expressed in moles of peptide. This treatment assumes that the rate of hydrolysis remains constant as the reaction proceeds, which is not entirely true in all cases. Equation 1 indicates that the yield of a peptide depends not only on the lability of the bonds involved in its terminal residues but also on the stability of the bonds within the molecules. It is also evident that in
T H E ARRANGEMENT OF AMINO ACIDS IN PROTEINS
13
general the yield will be greater if the peptide is a terminal one since a1or a,+l is already a maximum ( = 1). Various workers (Kuhn, 1930; Montroll and Simha, 1940; Warner, 194213; Myrback, 1949) have considered the mathematical treatment of the breakdown of high molecular chains assuming that all bonds were broken a t the same rate, i.e., that a1 = a2 = a, etc. Equation 1 then becomes : F = a2(1 - a)"-' (2) Assuming the chain t o be of infinite length or cyclic, the fraction of the original chain appearing as peptides containing n residues is given by:
F,
=
na2(1 - c ~ ) ~ - - l
(3)
It is difficult to know how far this type of treatment can really be applied t o proteins where there is such great variation in the susceptibility of various bonds. Presumably in a large molecule where the intrinsic rates of hydrolysis are fairly evenly distributed about the mean, equation (3) would apply. It is interesting to note that for any value of a in equation (2), F is a maximum if n = 1 and decreases as n increases, in other words the molar yield of smaller peptides is always greater than the yield of larger peptides. On the average this is also true for equation (l), which expresses more closely the situation present in a protein, since F is never greater than alan+l;however, as each a is different it will not apply to every case and the yield of certain higher peptides may be greater than the yield of certain smaller ones. Though the values for a1,az, etc. in equation (1) can of course be expressed in terms of the hydrolysis constants for the separate bonds, the treatment becomes extremely complicated as a different constant will be required for each polypeptide in which the particular bond occurs. In other words, the constant will vary as the reaction proceeds. Thus Kuhn et al. (1932) found that a three line formula was required t o express the rate of hydrolysis of tetraglycine. Thus it does not seem that any rigid mathematical approach can be given a t present, and we shall have to be content with a few generalizations derived largely from experimental observations. b. Complexity of Partial Hydrolyzates. When a protein is partially hydrolyzed, a very complex mixture of peptides is produced, the exact complexity of which is difficult to assess. If we consider an open polypeptide chain consisting of N residues, complete hydrolysis will give rise t o N amino acids, many of which may
14
F. SANGER
be identical. If it is partially hydrolyzed, the number of possible dipeptides is N - 1, of tripeptides N - 2 and of n-peptides (ie., pep1, very few of which are likely t o be tides with n residues) N - n identical. The total number of possible peptides is N ( N 1)/2. Clearly the shorter the time of hydrolysis the greater will be thenumber of higher peptides present in significant amounts. From equation (1) (p. 12) it was inferred that the yield of the higher peptides is always less than that of the smaller ones, so that the shorter the time of hydrolysis, the greater will be the complexity of the mixture. When a protein is hydrolyzed, there is only one molecular species t o start with. The number of species present in significant amounts then increases rapidly t o a maximum and gradually falls off. The initial rise in complexity will depend on what one considers to be a significant amount. Some splitting of each bond starts immediately with the instantaneous production of small amounts of all the N ( N 1)/2 different peptides. If one defines the “significant amount,” then for a while the only species will be the original protein, and others will gradually be added to it. It is clear however that there is really no phase during the hydrolysis when the complexity of the mixture is increasing, and that it is impossible to make use of this apparently simple composition during the initial phase of hydrolysis to obtain a mixture in which only a few polypeptides are present (‘in significant amount,” as there will be too many others present in insignificant amounts. As an example we may consider a protein such as ovalbumin which is probably a cyclic polypeptide of 400 residues. If the hydrolysis were such that all bonds were split a t the same rate, then during the initial phase of hydrolysis there will be 160,000 different chemical species produced, and this number will gradually decrease to the 20 free amino acids. I n the presence of any agent, such as acid or alkali, which splits all bonds t o some extent, all the 160,000 will be produced, but some in negligible amounts depending on the specificity of the reagent. Bull and Hahn (1948) have suggested that when ovalbumin is hydrolyzed by strong acid about 50 bonds are readily broken and that the rest are broken more slowly, say a t one tenth of the rate. Consider first only the initial phase in which the 50 labile bonds are broken. When they are all broken they should give rise to 50 peptides of average length 8 residues. At the beginning of hydrolysis 502 = 2500 different combinations of these peptides will-be produced the number gradually falling to the 50 octapeptides. At the same time, however, all the other bonds within these octapeptides have been subject t o hydrolysis a t one tenth of the rate, so that 3i0 of these octapeptides are broken down and each will 1) = 36 different degradation products. The give rise t o Y$(S
+
+
+
+
T H E ARRANGEMENT O F AMINO ACIDS IN P R O T E I N S
15
hydrolysis will thus have the following composition. One-fifth will be in the form of 50 octapeptides and the remaining four-fifths in the form of 50 X 36 = 1800 smaller peptides. I n reality the situation is even more complex since there are not two types of bonds but bonds with every type of stability. However, the above example does make it clear that at no stage in the hydrolysis, except near the end, will the composition of the hydrolyzate be sufficiently simple to justify investigation. In the case of enzymic hydrolysis, a large proportion of bonds are unattacked and here again the simplest mixture and the most profitable one to investigate is the complete hydrolyzate. At earlier stages in the hydrolysis the mixture will be increasingly complex. The only other means of obtaining a relatively simple hydrolyzate would arise if there was a very sharp break in the hydrolysis curve, that is to say, if there are two types of bonds with very different labilities. For instance, if the 50 labile bonds in ovalbumin were hydrolyzed 100 times as rapidly as the other bonds, one would at a certain stage obtain a mixture in which the 50 octapeptides are present to the extent of 90%. However, it is unlikely that any one of the commonly-used enzymes could bring about such a sharp differentiation. Since methods of fractionation a t present available may be capable of separating 100 short peptides but not 1000 long peptides, it would seem advisable to confine our attention t o the later stages of hydrolysis by any agent. Special emphasis is laid on this point as it appears to be quite a common practice for protein chemists to attempt to determine the nature of proteins by splitting them to a few large peptides by partial hydrolysis, on the assumption that by starting with one compound and ending with 20, a t some stage there must be only two or three compounds. This is true only when the first molecule in the solution is split. The chance that the second molecule will split in the same place is rather remote. c. The Question of Rearrangement of Peptide Sequences during Hydrolysis. It is clear that if any synthesis or rearrangement of peptide bonds takes place during the course of hydrolysis of a protein, the amino acid sequences identified in products of partial hydrolysis may not be the actual sequences that were present in the original protein, so that this approach to the problem would be useless. The hydrolysis of the peptide bond is reversible, so that the theoretical possibility exists that any peptide bond may be synthesized. The free energy of formation of a peptide bond is probably about 3000-4000 calories per mole for a dipeptide (Huffman, 1942) and 2000 calories per mole for a peptide bond within a protein (Haugaard and Roberts, 1942) so that the equilibrium will be very much on the side of hydrolysis for most peptide bonds, and a direct reversal of hydrolysis would seem rather unlikely.
16
F. SANGER
Synthetic reactions are more likely to take place if conditions are such that the product of the reaction is rapidly removed. Bergmann and his colleagues have shown that in the presence of certain proteolytic enzymes such conditions may be obtained either when the synthetic product is insoluble and is removed from the solution by crystallization (Bergmann and Fraenkel-Conrat, 1937, 1938; Bergmann and Behrens, 1938; Bergmann and Fruton, 1938) or else when it is removed by rapid hydrolysis to other products (Behrens and Bergmann, 1939). The former possibility may be avoided by keeping the reaction mixture always in solution. However the latter possibility, is rather more difficult to eliminate. Such a reaction may be formulated as follows: A.B
+ C.D % A.B.C.D + A.B.C + D
+
If the reaction A.B.C.D -+A.B.C D is very rapid it will shift the equilibrium of the other reaction to the right and thus bring about the Such a reaction will only occur to an synthesis of the bond B-C. appreciable extent if the hydrolytic reagent has a much greater affinity for the bond C--D in the peptide A.B.C.D than in C.D, that is to say if its specificity is determined not only by the residues involved in the susceptible bond, but by residues further removed from it. This would seem to be possible in the case of hydrolysis by proteolytic enzymes but less likely when acid or alkali are used (Sanger, 1949~). Fruton (1950) has recently pointed out the possibility that “transpeptidation )’reactions may occur in the presence of proteolytic enzymes. Such reactions may be represented as follows: R’ RCO-NH
A
HCO-NHR”
R’
+ NH,R”’ % RCO-NH&HCO--NHR”’ + NHJ-l”
It was originally shown by Bergmann and Fraenkel-Conrat (1937) that benzoylglycylanilide may be synthesized more rapidly by papain from benzoylglycinamide than from benzoylglycine, thus indicating that a direct transformation of the amide to the anilide occurs without intermediate formation of the acid. Following up this work Johnston et al. (1950) showed that if papain is allowed to act on benzoylglycinamide in the presence of NH, containing N16, a small amount of the isotope is introduced into the amide. Clearly, if this type of reaction which involves very little change in free energy occurs to any great extent during a partial hydrolysis of a protein it may lead to rearrangement of peptide sequences. Another type of rearrangement that might occur is through the amino acid anhydrides or diketopiperazines. The formation of these stable six-membered rings presumably involves a smaller free energy of forma-
THE ARRANGEMENT O F AMINO ACIDS IN PROTEINS
17
tion than the synthesis of most other peptide bonds and takes place readily under various conditions. If such conditions are present the following equilibrium will exist in the hydrolyzing solution
A where A.B represents a dipeptide and [ ] the corresponding anhydride. B The relative rates of the different reactions may be such as to cause the partial or complete inversion of the order of the residues in the dipeptide, even if the anhydride intermediate never appears in the solution. Anhydrides are very much more labile both to alkali (Levene et al., 1930, 1932; Kuhn et al., 1932) and to acid (Abderhalden and Mahn, 1927,1928), than the corresponding dipeptides, presumably since they contain no free charged group. This property tends to accelerate any possible inversion of a dipeptide. The formation of diketopiperazines seems to be associated with high temperatures. Brig1 (1923) and Abderhalden and Komm (1924a, b) ware able to convert dipeptides to anhydrides in quite high yield by heating in water or dilute acids a t temperatures of 150-250’. No conversion could be demonstrated when the strength of acid was greater than 1 M . Small amounts of anhydrides mere formed when the dipeptides were refluxed in water for several days. No reaction could be demonstrated under the action of 70% HzS04 or concentrated HC1 a t room temperature. This does not necessarily prove that no anhydrides were formed, as they may have been broken down as fast as they were formed. However these results do at least indicate the possibility of anhydride formation and inversion under certain conditions. Clearly the question of whether rearrangements do actually take place in partial hydrolysis experiments can only be solved by experience. If any such reactions occur it would be impossible to interpret results in terms of a unique sequence of amino acids unless the syntheses are completely specific and quantitative, which is most unlikely. I n fact, using concentrated acid at low temperatures it has been possible to work out a unique sequence for “gramicidin S ” (Consden et al., 194713 from the peptides identified, and there was no evidence of any peptide that did not fit this sequence. Similarly a unique structure could be determined for the phenylalanyl chains of insulin (p. 54). It thus seems unlikely that any rearrangement occurs under the action of this type of reagent. On the other hand, syntheses have been definitely shown to occur in the presence of proteolytic enzymes. The formation of plastein by the action of pepsin or trypsin on concentrated peptide mixtures has clearly been shown in certain cases t o be accompanied by a decrease in amino
18
F. SANGER
nitrogen, which can only be interpreted as a net synthesis of peptide bonds (reviewed by Wasteneys and Borsook, 1930; Virtanen et al., 1950) and the results with synthetic subst,rates are even more clear-cut, although in no case have naturally-occurring peptides been used as substrates. I n conclusion it may be said th at while the theoretical possibility exists that synthetic reactions may occur during the partial hydrolysis of proteins, it is the opinion of the reviewer tha t future research will show tha t such syntheses are insignificant and will not interfere with the interpretation of data derived from partial hydrolysis experiments in terms of protein structure. 3. Hydrolysis in Concentrated Acid Acid hydrolysis is the most generally used method of degrading proteins, and i t is almost universally employed when amino acids are to be isolated or estimated, since it leads t o complete hydrolysis with a minimum of destruction. The only amino acid th a t is extensively destroyed is tryptophan, the destruction of which may be largely due to the presence of traces of heavy metals during hydrolysis and may be reduced by using very pure HC1 and quartz vessels (Jacobsen, 1949; Monnier and Jutisz, 1950). Slight destruction of serine and threonine (Rees, 1946) also takes place, but in partial hydrolyzates this would be almost negligible. No synthetic reactions or rearrangements have been shown to take place under the action of strong acids. The relative rate of hydrolysis in acid of any peptide bond and hence the yield of a given peptide is determined mainly by the number of hydrogen ions that can approach the bond. While the rate probably depends on a number of different factors, we may consider two which probably play a major role, namely, electrostatic effects and steric effects. a. Electrostatic Efects. The presence of any charged groups in the neighborhood of a peptide bond will clearly affect the approach of hydrogen ions. In strong acid all the carboxyl groups on the proteins will be uncharged but all the basic groups (amino, imidazole, and guanidyl) will be fully charged and will oppose the approach of the similarly charged hydrogen ions. From their studies of the course of hydrolysis of various proteins b y concentrated acid Gordon et al. (1941) were able to calculate the ratio of free basic amino acids in the hydrolyzate to total basic residues (free and in peptide combination). This ratio gives a n average estimate of the stability of basic peptides. In general, the values found were slightly lower than the corresponding figures for the neutral residues, indicating
THE ARRANGEMENT O F AMINO ACIDS I N PROTEINS
19
that the basic peptides were slightly, though not very much more stable, than the average peptide. Thus the ratio obtained for wool a t a certain stage of hydrolysis was 0.33 and the corresponding figure for the neutral residues 0.39. Probably the charged groups that are most effective in stabilizing peptide bonds are the a-amino groups, which are closer to a peptide bond than the e-amino, imidazole and guanidyl groups. A bond involving an N-terminal residue should thus be relatively stable to acid. An important result of this is the stability of dipeptides which leads to their accumulation a t a certain stage of hydrolysis (Gordon et al., 1941). This stability is evident from the results of Stein et al. (1944), who followed the course of hydrolysis of silk fibroin in concentrated HCl a t 40" using the van Slyke nitrous acid method for estimating the rate of liberation of free amino groups and the ninhydrin method (van Slyke et al., 1941) for free amino acids. From these results it is possible to calculate the average length of the peptides excluding in the average the free amino acids. After 43 hours hydrolysis 60% of the peptide bonds were split, and the hydrolyzate contained 25% of its nitrogen in the form of free amino acids and 75% in the form of peptides whose average length was 2.05 residues; in other words almost the whole of the N in peptide linkage was assignable to dipeptides. From equation (3) (p. 13) it may be calculated that if a completely random splitting had occurred the yield of amino acids would have been 36%) of dipeptides 29%, of tripeptides 17%) etc. Similar conclusions may be drawn from the rates of hydrolysis of gramicidin (Synge, 1945; Christensen and Hegsted, 1945) the effect being more marked a t 37" than a t the boiling temperature. On the contrary, however, it was found that the yield of free amino acids during the hydrolysis of ovalbumin with 1 N HC1 was almost exactly theoretical (Warner, 1942b). b. Steric E$ects. Apart from the effect of positively charged groups, probably the most important factor influencing the rate of hydrolysis of a peptide bond is the effective size of the amino acid side chains on either side the bond, preventing the approach of hydrogen ions by steric means. The effect will be expected to depend on the actual size of the side chain and its position relative to the bond in question. Synge (1945) has made a kinetic study of the hydrolysis of a number of simple peptides by a mixture of equal volumes of 10 N HC1 and glacial acetic acid a t 37". His results are shown in Table 11. The most stable peptides appear to be those containing valine. The bulky CH3CHCH3group is close t o the main peptide chain and effectively prevents the approach of hydrogen ions from a fairly wide angle.
20
F. SANGER
I n leucine the CH,CHCH3 is slightly further from the peptide hond, so that leucyl peptides are less stable than the corresponding valyl peptides, and peptides containing alanine and glycine are still more labile. ,4 side chain seems to be less effective in stabilizing the peptide bond if it, is on the residue whose amino group forms part of the bond (Levene et al., 1932). Val.Gly for instance, is more stable than Gly.Va1, though it should be noted that these generalizations are derived only from a study of peptides of glycine. TABLErr Hydrolysis of Dipeptides in Strong Acid (Sy nge , 1945)
Peptide Gly.Gly Gly.Ala Ala.Gly Gly.Leu Gly.Try Gly.Va1 Leu. Gly Leu.Leu IAeu.Try Val.Gly
Relative velocity of hydrolysis (G1y.Gly = 1)
1 0.62 0.62 0.40 0.35 0.31 0.23 0.048 0.041 0.015
The marked stability of valine peptides has frequently been noted. Thus Christensen (1943) was able to isolate Val.Va1 in 1.5% yield from gramicidin after boiling for 24 hours with 16% HC1, and in 5-6%) yield after 6 hours hydrolysis. Synge (1944) could find no free valine in partial hydrolyzates of gramicidin that had been treated with 5 N HCI for 10 days. The stability of valyl peptides was also apparent from the work on the partial hydrolysis of insulin (Sanger and Tuppy, 1951a). Thus tripeptides containing valine as the central residue were present in higher concentrations than other tripeptides. A lability of bonds involving the carboxyl groups of glycine was also noted. Peptides containing proline also appear to be unusually stable presumably due to steric factors (Consden et al., 1947b). In this connection the unusual stability of cyclic structures should be mentioned. In the case of carbohydrates this was clearly demonstrated by Swanson and Cori (1948) and by Myrbiick (1949), who showed that the cyclic Schardinger dextrins are considerably more stable than corresponding open chain polysaccharides. Similarly Consden et al. (1947b) found the cyclopeptide “gramicidin S” to be unusually
21
T H E ARRANGEMENT O F AMINO ACIDS I N P R O T E I N S
resistant t o acid hydrolysis. It has been suggested by Synge (personal communication) that immobilization of the peptide bond from rotation may be an important factor in inhibiting hydrolysis. Such a n effect might also partly account for the relative stability of native proteins t o enzymic hydrolysis (p. 26) and possibly for the stability of proline peptides. c. Bonds Involving the Amino Groups of the Hydroxy Amino Acids. While the electrostatic and steric effects mentioned above probably play an important part in determining the stability of a given peptide bond, they are certainly not the only factors concerned and the course of hydrolysis of a protein cannot wholly be explained in this simple manner. In fact i t has been observed th at the first bonds in a protein t o be split by strong acid are those involving the amino groups of the serine and threonine residues. Abderhalden and Bahn (1935) made use of this lability t o prepare seryl peptides from the corresponding anhydrides. Under suitable conditions only one of the bonds in the anhydrides split. Gordon et al. (1941) hydrolyzed a number of proteins with strong acid and followed the rate of liberation of the free amino groups of serine and threonine residues by means of the periodate reaction. Only those residues with a free hydroxyl and amino group react with periodate t o give ammonia. The serine and threonine residues could be differentiated by estimating the acetaldehyde produced by periodate from threonine. It was found that the free amino groups of these hydroxy amino acids were liberated much more rapidly than the average amino group (estimated by the van Slyke method). Similar results were obtained by Christensen and Hegsted (1945). More recently Desnuelle and Casal (1948) followed the relative rates of liberation of the amino groups of the different amino acids using the D N P method. During the initial stages of hydrolysis with 10 N HCl a t 30" there appeared to he a very specific breakdown of the bonds involving the amino groups of serine and threonine. Thus after one hour about 20-30% of these bonds were split whereas very few other amino groups were liberated. There was no difference between serine and threonine. I n order t o account for this effect Desnuelle and Casal have suggested that under the action of concentrated HCl an intermediate oxazoline ring is formed which breaks a t the amino group: R I -NHcHcoNHcHcoHdH2
R
R
I
-NHcHc=N-cHco-
d-CH,
I
NH2
--h'HLHCo
I
&co-
O--CH2
I
I n this way the peptide chain is considered to migrate from the amino t o the hydroxy group. The ester bond then formed will be fairly
22
F. SANGER
rapidly broken down. Such reactions have been shown to take place with simple compounds such as benzoyl serine (Bergmann and Miekeley, 1924) under the action of chlorinating agents, e.g., thionyl chloride. I n support of the above mechanism it was found that the rate of liberation of the free amino groups of the hydroxy amino acids was greater when estimated by the DNP method than when estimated by the periodate method, thus suggesting that the hydroxy group was being simultaneously masked. To sum up, when a protein is hydrolyzed with strong acid we may expect to find an initial rather specific hydrolysis liberating the free amino groups of the hydroxy amino acids followed by a more random breakdown a t different bonds, the relative rates depending largely on the nature of the residues involved, and there is likely to be a slowing up of hydrolysis and accumulation of dipeptides towards the end of the reaction. In general, the specificity of hydrolysis will be greater a t lower temperatures, since the activation energies for hydrolysis of the various bonds will be expected to show greater differences relative to the mean thermal energy of the molecules. This is evident in the results of Christensen and Hegsted (1945), who found a more random splitting at higher temperature. Desnuelle and Casal (1948) also found that the liberation of hydroxy amino residues was much more specific a t lower temperatures. 4. Hydrolysis in D i l u t e Acid While concentrated acid is usually preferred as a hydrolytic agent, it may be advantageous in certain cases to use dilute acid, which seems to exercise a rather different kind of specificity. Thus for instance very poor yields of the N-terminal glycyl peptides of insulin were obtained after hydrolysis in concentrated HC1, whereas much higher yields were obtained by boiling in 0.1 M HCl (Sanger, 194913). This difference in specificity may to some extent be ascribed to differences in the charged groups. In sufficiently dilute acid the acidic groups may become negatively charged, so that they will attract hydrogen ions and labilize any nearby peptide bonds. An attempt t o utilize this effect t o obtain a specific hydrolysis near the cysteic acid residues of oxidized insulin gave no clear-cut results (Sanger, 1949b). A very specific hydrolysis has, however, been demonstrated by Partridge and Davis (1950). They found that when a protein was boiled with acetic or oxalic acid and the hydrolyzate investigated by paper chromatography, free aspartic acid was liberated much more rapidly than any other amino acid. For most proteins during the first 8 hours of hydrolysis the only ninhydrin positive spot was that due t o aspartic acid. Free glutamic acid was the next
THE ARRANGEMENT O F AMINO ACIDS IN PROTEINS
23
amino acid to appear, but a t a very much slower rate. While it is possible that other bonds besides those next to the aspartic acid residues are also split, this type of hydrolysis seems to exhibit a marked specificity and may prove a useful method for obtaining larger polypeptides from proteins. Partridge and Davis suggest further that the presence of a free carboxyl group should labilize the C-terminal residues of a peptide chain and it may thus be possible to effect a stepwise liberation of residues from the C-terminal group. The main disadvantage of using dilute acid for hydrolysis is the possibility of anhydride formation, already referred to (p. 16). Another type of hydrolysis that may exhibit a different specificity is that catalyzed by long-chain anionic detergents. Steinhardt and Fugitt (1942) showed that the rate of hydrolysis of proteins by acid was dependent, not only on the concentration of hydrogen ion, but also on the nature of the anion present. For a homogeneous series of long chain sulfonic acids the rate of hydrolysis was proportional to the chain length of the anion. Thus the hydrolysis proceeded about a hundred times as rapidly in dodecylsulfonic acid as in HC1. It seems that the detergent is adsorbed on the protein and the presence of the -SO3- group labilizes the peptide bonds. At low concentrations of detergent the amide groups are especially labile to this type of hydrolysis. Being slightly stronger bases than the peptide bonds, they preferentially bind the anions (Steinhardt, 1941). Thus it was possible to remove amide groups with very little destruction of peptide bonds. This suggests a different type of specificity from other methods of hydrolysis. The lability of amide groups, however, seems to be a characteristic of low temperature hydrolysis in dilute acids, since Virtanen and Hamberg (1947) hydrolyzed zein at pH 1.5-1.8 and 37" for 4 months and found that 50% of the amide groups were split off while very little amino N was liberated. An obvious advantage of using this method of catalyzed hydrolysis is that there will be a minimum of destruction of the amino acids or of any bonds other than the peptide bonds. A disadvantage of the method is the practical difficulty of working with solutions of these strongly surface-active reagents. 5. Hydro&& in Alkali The main disadvantage of alkaline hydrolysis is that much more destruction of amino acids occurs than with acid. The chief amino acids to be affected are cysteine, serine, threonine and arginine. Wieland and Wirth (1949) have used paper chromatography to study the effect of strong alkali on certain amino acids. They found that serine broke down to give appreciable quantities of glycine and alanine, threonine
24
F. SANGER
gave glycine, alanine and a-amino butyric acid and cysteirie gave alariine. While relatively violent conditions were required for these reactions, they were shown to take place more readily if the amino acids were in peptide form. Hellerman and Stock (1938) and Warner (1942a) showed that, arginine is broken down by strong alkali to give ornithine and citrulline. Sanger and Tuppy (1951a) studied a n alkaline hydrolyzate of a n oxidized insulin fraction. Although the conditions used would not have been expected t o cause any destruction of free serine or threonine, it was evident that most of these residues in the protein had been broken down. Thus, for instance, Gly.Pro was found in the alkaline hydrolyzate whereas Thr.Pro was present in the acid hydrolyzate. The arginine residues were also converted to ornithine (or citrulline). Extensive racemization of the amino acids also occurs in the presence of alkali, which may complicate the results of a partial hydrolysis experiment (Levene and Bass, 1928, 1929). I n spite of these obvious disadvantages i t may, in certain cases, be worth while to employ alkaline hydrolysis, which may exhibit somewhat different specificities from those found for acid hydrolysis. Synge (1945) compared the relative rates of hydrolysis in acid and in alkali of a number of simple dipeptides of the monoamino acids, and found them to be similar in both reagents. I n these experiments the relative stabilities of the peptide bonds were probably determined largely by the steric effects produced by the side chains of the residues, which inhibit the approach of hydroxyl ions, and these effects will be similar for both acid and alkali hydrolysis. However, where the effects of charged groups are concerned one may expect to find a somewhat different specificity. I n strongly alkaline solution the basic groups are uncharged, whereas the carboxyl groups will be negatively charged and will stabilize bonds in their neighborhood. Similarly, dipeptides would be stable in alkaline solution due to the presence of the charged a-carboxyl group. It is also probable that the peptide bonds involving the amino groups of the hydroxyamino acids are more stable in alkali than in acid. Abderhalden and Bahn (1935) showed that whereas anhydrides containing serine were split by acid at the bond involving the amino group of the serine residue, they were split at the other bond by alkali. It was also found by the D N P method (Sanger, unpublished) that the relative rate of liberation of the amino group of serine from fraction A of oxidized insulin was slower in alkali than in strong acid. Alkaline hydrolysis offers a possible advantage for the investigation of tryptophan peptides since tryptophan itself is more stable in alkali than in acid (Lugg, 1938; Brand and Kassell, 1939). The possibility that long chain bases, such as hexyl trimethyl ammo-
T H E ARRANGEMENT O F AMINO ACIDS I N P R O T E I N S
25
nium, will show catalytic effects similar to those found for long chain anions is suggested by the work of Steinhardt and Zaiser (1950), who showed that they are also bourd t o protein and cause anomalous titration effects. It may be that the mild conditions necessary for such a n hydrolysis may make it possible t o avoid the undesirable side reactions that occur during alkaline hydrolysis.
6. Hydrolysis with Proteolytic Enzymes The action of proteolytic enzymes on proteins and synthetic pcptides has been extensively studied and reviewed (see Bergmann and Fruton, 1941; Bergmann, 1942; Neurath and Schwert, 1950; Linderstrgm-Lang, 1949) so t ha t the subject will not be dealt with in great detail here. One obvious advantage of using proteolytic enzymes is th a t they are very unlikely t o cause any destruction of amino acid residues, since they act under very mild conditions. Another important advantage is their specificity. Not only do they exhibit an entirely different specificity from t ha t shown b y acid and alkali but a specificity which is much more limited and exacting. Enzymes should therefore be useful for the initial specific splitting of proteins into large peptides. The finding of synthetic substrates of known structure that are hydrolyzed by the endopeptidases has made it possible to predict what bonds will be susceptible to particular enzymes. This subject has been reviewed in detail by Bergmann and Fruton (1941) and by Neurath and Schwert (1950). I n brief it may be said that trypsin splits those bonds in which the carboxyl groups of arginine or lysine are involved, chymotrypsin those involving the carboxyl groups of tyrosine, phenylalanine, tryptophan or methionine, and pepsin those involving the amino groups of the aromatic amino acids. It seems t ha t the rate at which synthetic substrates are split is considerably slower than the rate a t which peptide bonds in proteins are split (Northrop et al., 1948). The question therefore arises as to whether bonds of the type present in the synthetic peptides are the most labile in a protein or whether there are other types of bonds that are broken down. I n the case of pepsin, there is evidence that other types of bonds are also split. Thus Harington and Pitt-Rivers (1944) found that pepsin would hydrolyze the peptide Tyr.CySH, in which the bond involves the carboxyl group and not the amino group of tyrosine. Recently Desnuelle et al. (1950) have studied the action of pepsin on ovalbumin and on horse globin using the D N P technique. If the above specificity were the only one, the only free amino groups liberated should be those of the aromatic residues. I n fact this was not the case. For ovalbumin no specificity could be detected, and free amino groups of almost all amino acids were liberated simultaneously. With globin the bonds split first were those
26
F. SANGER
involving the amino groups of alanine, phenylalanine, leucine, and serine ; during a second slower phase of hydrolysis no specificity was apparent. I n a study of the action of proteolytic,enzymes on a n oxidation product (fraction B, p. 54) of insulin, Sanger and Tuppy (1951b) found that other bonds besides those adjacent to aromatic residues were split by pepsin, including those of Leu-Val, Ala-Leu, and Glu(-NH&His. It thus seems that in the case of pepsin at least there is much to be learnt about its specificity when proteins act as substrate. Trypsin and chymotrypsin were found t o split oxidized insulin with the same specificity as was found for synthetic peptides, and it seems probable th a t this specificity may be shown in their action on other proteins. Clearly a knowledge of the exact mode of action of these enzymes would greatly help in the elucidation of protein structure just a s advances in our knowledge of protein structure must throw light on the behavior of the endopeptidases. The possibility of rearrangement of sequences of amino acids under the action of proteolytic enzymes has already been mentioned (p. 15) but this danger would not seem sufficiently great to offset the advantages to be gained by using them. Nevertheless, the results must be interpreted with caution. It has already been emphasized (p. 14) that the best stage of hydrolysis a t which t o attempt the fractionation of peptides is at the point where enzyme action will proceed no farther or when there is a very sharp break in the hydrolysis curve. The disadvantage of a long incubation period is of course that the danger of rearrangements increases (see Linderstrpim-Lang and Ottesen, 1949). It may be that proteolytic enzymes act more specifically on native than on denatured proteins. The work of Linderstrgm-Lang and coworkers (reviewed by Linderstrom-Lang, 1949) has indicated th a t a t least in the case of the action of trypsin on 0-lactoglobulin, the initial step in proteolysis is some type of denaturation. Nevertheless, certain hydrolyses proceed without denaturation of the substrate. Examples are the formation of plakalbumin from ovalbumin (p. 57) a n d the splitting of globulin molecules by papain (Petermann and Pappenheimer, 1941; Petermann, 1942), which appear to be rather specific re:tctions. Presumably only a few susceptible sites are available to the enzyme on the native protein, and if these can be split without denaturing the protein and exposing the other susceptible sites, a specific type of hydrolysis may take place. 7 . Rates of Hydrolysis of Proteins I n experimental work it is very often desirable t o know how far a protein will be hydrolyzed under a given set of conditions. The course
THE ARRANGEMENT O F AMINO ACIDS I N PROTEINS
27
of hydrolysis of a number of proteins in various concentrations of acid and alkali have been recorded in the literature, though it is often difficult to find the particular and relevant data one is interested in. I n Table I11 are listed some references where such data may be found. Since protein hydrolysis follows no known kinetic laws it is impossible to define the rate in terms of any constant, so th at it is best t o refer to the original work. I n the last column are listed the approximate half-lives of the proteins in the various reagents. This is defined as the time at which 50% of the peptide bonds in the protein are broken. 8. Non-Hydrolytic Methods of Degradation
Critics of the peptide theory have claimed th a t too much emphasis has been laid on studies in which hydrolysis is used for degrading proteins, and that other methods of degradation such as oxidation and reduction should be employed. In the earlier years of protein chemistry many attempts were made to study proteins in this way but very few recognizable products could be obtained, and with those th a t could be identified it was difficult to know what was their relationship t o the original protein. This was undoubtedly due to the great complexity of the problem, the different side chains and bonds in the proteins each reacting t o give a variety of products of unknown origin. Clearly, in order to obtain any recognizable breakdown products, one must use reagents that attack only one or a very few types of bond or residue. Hydrolytic reagents, which attack almost exclusively the peptide bond, are the obvious first choice. I n the search for other reagents for degrading proteins i t would seem more profitable to consider the nature of proteins first and t o choose a reagent for a particular purpose rather than to take any reagent off the shelf and see what i t does to the protein. There is obviously a great use for reagents th at will attack a protein in a specific manner. If i t were possible to destroy one type of residue exclusively and split i t out of a protein chain, it would be a valuable step in the degradation of proteins. a. Splitting the Disulfide Bridges. Probably the only covalent linkage t ha t occurs in proteins other than the peptide bond between the amino acid residues, is the disulfide bridge of cystine; this may be split by oxidation t o sulfonic acid groups. Toennies and Homiller (1942) studied the action of performic acid on a large number of amino acids and found t ha t the only acids attacked were cystine, methionine, and tryptophan. Cystine was quantitatively converted to cysteic acid, and methionine t o the corresponding sulfone. I n the case of tryptophan the products were not identified. Sanger (1949a) used performic acid to split the disulfide bridges of insulin, which contains no tryptophan or methionine (see p. 51). This method could probably be used as a n
28
F. SANGER
TABLE I11 Rates of Hydrolysis of Proteinse
Ileferencc
Proteins studied
Reagents used
Temp. (“C.)
Ahderhaltien and Mahn. 1928 Acher et al., 1950 Bull and Hahn, 1948
Gelatin 15,38, 50,70 1 N HCI 37 Lysozyme 10 N HC1 30 Ovalbumin 7.95 N HCI 7.95 N HC1 45 7.95 N HCl 60 2.5 N HC1 Boiling Desnuelle and CasaI, 1948 Casein 37 10 N HCl 10 N HCI 30 Silk fihroin 10 N HCl 30 3.6 N HzSOa Casein Dunn, 1925 Boiling Gordon et al., 1‘341 Wool 10 N HCI 37 Edestin 10 N HC1 37 Gelatin 10 N HCI 37 Levene and Bass, 1928 Casein 125 5 N HC1 0.5, 1.0, 5 N NaOH 25 Casein Pittom, 1934 5.7 N HCI Boiling Boiling Ovalbumin 5.7 N HCI Silk fibroin 12 N HCl 40 Stein et al., 1944h Vickery, 1922 Gliadin 0.1, 0.2 N HCI 93-94 0.5 N HCI 93-94 1.0 N HC1 94-95 94-95 2.0 N HCl 4.0 N HCI 98-1 04 20% HCI 102-1 10 0.2 N HZSOa 93-94 4.0 N HzS04 96-98 0.2 N NaOH 93--94 1.0 N NaOH 93-94 0.2 N Ba(OH)2 93-94 Warner, l942b Ovalburnin I N HCl Boiling I00 20% H&O, 1.4, 4.3 N NaOH 100, 68, 35 0.43 N NaOH 100 0.2 N NaOH 68 2.3, 3.7 N Ba(OH)z 100 ‘I
b
For other earlier references see Vickery, 1922. In these experiments 1 g . protein was hydrolyzed with only 2 rnl. IICl.
Approx. half-life of protein (hours)
96 195 42 9 2 . :3 1 I5 192 144 5 95 120 35
17
35 17 7 I .8 < 1 0
16 36 4.5
THE ARRANGEMENT OF AMINO ACIDS IN PROTEINS
29
initial specific method of degrading other proteins that contain disulfide bridges, though the presence of tryptophan might lead to side reactions. Disulfide bridges can also be split specifically by mild reduction, but the thiol groups formed tend to reoxidize and cause polymerization of the products (Miller and Andersson, 1942). b . Splitting by Radiation. If a protein is irradiated with ultraviolet light of a suitable wavelength, the only residues that absorb energy are those of the aromatic amino acids phenylalanine, tyrosine, and tryptophan, and it has been suggested that the photochemical energy may be sufficient to split the peptide bonds adjoining these residues (Carpenter, 1940; McLaren, 1949; Mandl et al., 1950). Rideal and Mitchell (1937) showed that substances such as stearic anilide could be split. Here the aromatic residue is directly adjoining the peptide bond. However it was also shown that cleavage could take place when the peptide bond and aromatic residue were separated by a number of -CH2groups, since substances such as stearyl benzylamine were also split (Carpenter, 1940). Propionylphenylalanine and phenylpropionylalanine, which resemble more closely a natural peptide were both split with equal efficiency (Mandl et al., 1950). In each case the main nitrogenous end product was ammonia rather than the amino acid, indicating that deamination had also occurred. No liberation of free amino acids from Tyr.Leu or Leu.Tyr could be detected but this was probably due to deamination. However it has been shown (Carpenter, 1941; Kaplan et al., 1950) that when insulin is irradiated free tyrosine is liberated into the solution. Clearly the effects of light on a protein may be expected to be rather complex (McLaren, 1949) but it would be extremely valuable if conditions could be found for a specific photolysis near the aromatic residues. c. Other Methods. Fodor and coworkers (see Fodor, Fodor and Kuk-meiri, 1947) used anhydrous glycerol a t 130-140” to break down proteins. The products formed, which were termed “acropeptides,” appeared to be large cyclic polypeptides. Using a similar method Uchino (1934) and Tazawa (1949) reported the formation of large amounts of diketopiperazines from ovalbumin. Troensgaard (1947) has used reduction with sodium and amyl alcohol to split proteins and has isolated a number of piperazines and pyrrole derivatives. The relationship of these products to structures i n the original protein is not clear a t present.
V. FRACTIONATION OF PEPTIDES
It has already been emphasized that a partial hydrolyzate of a protein is a complex mixture of closely related compounds, so that very sensitive
30
F. SANGER
methods of fractionation are required to separate pure peptides. I n recent years countercurrent methods, especially those depending on partition effects have been found to be most useful for this purpose. Before applying such methods it is often desirable to carry out preliminary group separations of the hydrolyxate into fractions containing fewer peptides than the original mixture. For instance, not more than about twenty peptides can be satisfactorily fractionated by paper chromatography. Thus if a partial hydrolyzate of a protein were applied directly t o such a chromatogram there would probably be considerable overlapping of the spots and interpretation of the results would be difficult. Separation into a number of relatively simple groups would make fractionation much easier. We may thus distinguish between two classes of methods of fractionation, those for preliminary group separations and general methods involving the countercurrent principle. The distinction is somewhat arbitrary as most methods may be used to some extent for both purposes. The most useful methods for group separation are those which give the most clear-cut fractionations with a minimum of overlapping, the ideal being th at each peptide should be present in only one fraction. Such an ideal is rarely achieved but it is sometimes possible, as in the case of the cystine peptides, to make use of a unique property of one amino acid, t o separate out all peptides in which it is involved. It is clearly an advantage to use for group separations methods which depend on different properties of the peptides than those on which the final general fractionation depends. Thus, whereas the methods of general fractionation usually depend on differences in the partition coefficients of the peptides, methods of group separation more often depend on differences in ionophoretic mobility or in the adsorption coefficients. 1. Ionophoretic Methods Since amino acids and peptides contain several differently charged groups, they can be fractionated by methods which make use of differences in isoelectric point or electrophoretic mobility. The various techniques available for such separations have been comprehensively reviewed by Svensson (1948). A simple compartment type of apparatus may be used to separate a peptide mixture into basic, neutral and acidic fractions (Gordon et al., 1941, 1943; Sanger and Tuppy, 1951a). Since the simplification of the mixture is usually more important than the yield of peptides, it is often advisable to repeat the ionophoresis on each fraction, and in this way clear cut fractionations may be obtained with very little “overlapping.” This method is especially useful for separating the basic peptides. By
THE ARRANGEMENT O F AMINO ACIDS IN PROTEINS
31
choosing suitable conditions of p H etc. it is also possible to carry out more extensive group separations using this type of apparatus; thus, by working a t a n alkaline p H for instance it is possible to obtain a fraction containing only arginine peptides (Sanger and Tuppy, 1951a). For more extensive separations it is best to use methods th a t depend on differences in mobility. The method that has been used most is th a t of Consden, Gordon and Martin (1946) for ionophoresis in silica jelly. It was successfully used to separate the acidic peptides from a partial hydrolyzate of wool into fractions of different mobilities (Consden et al., 1949; Consden and Gordon, 1950). This type of apparatus is less suitable for peptides containing aromatic or basic residues, due t o adsorption on the silica, which causes ((tailing” and difficulties in eluting the peptides. The use of agar-agar instead of silica may prove more suitable for such peptides (Gordon et al., 1939). Svensson and Brattsten (1949) and Grassmann (1950) have described a method of ionophoresis in which the separations are carried out in a box of glass powder. The solution is allowed to flow down the box while a voltage is applied horizontally, so that the direction of flow of each solute depends on its mobility; fractions are collected from outlet tubes a t the bottom of the box. This apparatus should prove especially useful for longer peptides and proteins, since adsorption on the stabilizer (glass powder) is unlikely. A similar continuous method using filter paper was described by Grassmann and Hannig (1950). A number of other workers have devised methods for ionophoresis on filter paper. Haugaard and Kroner (1948) have combined ionophoresis with paper chromatography by threading electrodes down the side of a paper chromatogram, which was first soaked in buffer solution. A potential was applied during the development of the chromatogram with phenol. Wieland and Fischer (1948) have used an apparatus in which a potential is applied to the ends of a strip of filter paper which is soaked in buffer and held in a small glass chamber. The ends of the strip are outside the chamber and dip into the electrode vessels. Good separations with amino acids and proteins (Turba and Enenkel, 1950) were reported. A similar principle is employed in the methods reported by Biserte (1950), Durrum (1950) and Cremer and Tiselius (1950). The method of the latter authors was designed specially for the fractionation of proteins. The whole paper is immersed in a bath of chlorobenzene, which prevents heating and loss of solvent by evaporation. 2. Ion Exchange Methods
Another method of fractionating peptides according to their charge These have frequently been
is by the use of ion exchange materials.
32
F. SANGER
used for the group separation of amino acids (reviewed by Turha, 1945; Block, 1949; Martin and Synge, 1945). I n most work the amino acids have merely been separated into two groups in each experiment, those which are adsorbed, and those th at are not adsorbed, the latter being eluted with a different solvent. I n this way it is possible to obtain quite sharp separations by choosing suitable adsorbents and conditions of pH etc. Thus, for instance, basic amino acids may be adsorbed on basic A l z 0 3 ” (Wieland, 1942), silica gel (Schramm and Primosigh, 1944), permutite (Felix and Lang, 1929) or suitable synthetic resins (Block, 1942; Wieland, 1944). Acidic amino acids may be adsorbed on “acid Alz03” (Wieland, 1942; Jutisz and Lederer, 1947; Turba and Richter, 1942) and synthetic anion exchange resins (Cannan, 1944; Tiselius et al., 1947). The synthetic resins are probably to be generally preferred because of their greater capacity and because their properties are more reproducible from batch to batch. There are also a large number of different ones now commercially available, which ex tends their use. It is probable that simple group separations of this type may also be carried out with peptide mixtures (qf. Waldschmidt-Leitz and Turba, 1941) though in such mixtures of substances with a much more scattered range of isoelectric points, considerable overlapping may be expected. Also large peptides and those containing aromatic residues will probably be held on the exchangers by ordinary adsorption which may cause difficulties. Jutisz and Lederer (1947) and Lederer and Tchen (1947) have devised a method for the group separation of neutral amino acids and peptides by making use of differences in the apparent pK of the amino groups in the presence of formaldehyde. Neutral amino acids and peptides are not adsorbed on “acid AlzOa” from water but in the presence of formaldehyde they acquire acidic properties and those with the lower pKa‘ values are adsorbed. Table IV shows the approximate pKa’ values for the three groups in 10% formaldehyde, and their behavior on columns of “acid A1203”in the presence of 10% and 1%formaldehyde, respectively. In this way three clearly separated groups may be obtained. Presumably scryl, threonyl, and cysteinyl peptides will behave as the glycyl peptides. Final elution from the columns is effected with hot water which dissociates the formaldehyde complexes. This method, which gives rather a different group separation from other techniques may prove iiseful for the initial simplification of peptide mixtures. Under suitable conditions it is possible to obtain separations on an ion exchange column by elution analysis. The p H of the developing solution should be such th at the solutes are distributed between the resin and solvent so that they move down the column as definite bands.
33
THE ARRANGEMENT O F AMINO ACIDS I N PROTEINS
I n this way Consden et al. (1948) and Drake (1947) separated glutamic and aspartic acid on columns of Amberlite IK-4 (polyamine anion exchanger) maintained a t p H 2.5, and this type of column has been used to separate acidic peptides from a wool hydrolyzate (Consden et al., 1949). More recently Stein and Moore (1949) have obtained excellent fractionations of amino acids on columns of Dowex-50 (cation exchanger, sulfonic acid groups) developed with 1.5-4 N HCl. The separations obtained are almost certainly due to differences in adsorption affinity for the resin, as well as to charge effects. At the p H used the only partially TABLEIV Group Separation of Amino Acids and Peplides according to Jutisz and Lederer (1 347) Group Monoamino acids other than those in group 2 Glycine Serine Threonine Cysteine Simple dipeptides other than those in group 3 Glycyl peptides
Approx. pKa’ Adsorption on “acid A1,03” in 10% H.CHO in 10% H,CHO in 1% H.CHO
7.0
-
-
6.8
+
-
4.2
+
+
ionized groups will be the sulfonic acid groups of the resin, the amino acids having either one or two positive charges. This technique appears to be one of the most efficient methods of fractionation and it will be interesting t o see how far it may be used for the separation of peptides. In contrast t o most other methods depending on adsorption (see next section) the bands exhibit no tailing, but give elution curves showing sharp well-defined peaks. Separations may also be carried out on ion exchangers by the principle of displacement chromatography. If the ionizing groups of a resin are completely saturated with an amino acid, and a second amino acid, having a greater affinity for the resin is introduced onto the column, the first amino acid will be displaced and will move down the column. I n this way columns may be obtained from which the amino acids are displaced in the order of their isoelectric points. Partridge (1949a, h, c ; Partridge and Westall, 1949) has developed this principle using the sulfonic acid resin “Zeo-Karb 215,” for the separation of amino acids. Elution is carried out with ammonia which has a greater affinity for the resin than most of the amino acids. This method, which is best carried
j4
F. SANGER
out on a relatively large scale, may prove useful for the group separation of peptides, but no such results have yet been reported.
3. Adsorption Chrornatography The possibilities of fractionating amino acids, peptides and proteins by adsorption chromatography have been fairly extensively explored, especially by Tiselius and coworkers (reviewed in Tiselius, 1947; Turba, 1948; Martin and Synge, 1945). Active carbon is probably the most effective adsorbent for amino acids, being one of the few on which they are retained. Aromatic amino acids are strongly adsorbed and may be separated from the other amino acids (Schramm and Primosigh 1943; Jutisz and Lederer, 1947; Partridge, 1949b) by adsorption from acetic acid solution. They can be eluted then with phenol or ethyl acetate. A similar type of group separation may also be used to separate aromatic peptides from a partial hydrolyzate (Synge and Tiselius, 1949; Sanger and Tuppy, 1951a) though larger peptides and peptides containing basic amino acids, are rather strongly adsorbed, and may appear in the “aromatic” fraction. Considerable losses are often involved in this type of fractionation, as it is difficult to elute the substances completely. Attempts t o use adsorption on charcoal for a more detailed chromatographic fractionation of peptides have not met with great success. Adsorption isotherms are usually of the Langmuir type and the only method that is generally applicable is the technique of frontal analysis. This may be used as an analytical method (Moring-Claesson, 1948; Synge and Tiselius, 1947, 1949), but does not give appreciable fractionation. Usually the bands tail too much to make separations possible by elution analysis. The most satisfactory method for separating compounds having this type of adsorption isotherm is the displacement method, though unfortunately it is not a general method and has only been used in a few special cases. One of the disadvantages of displacement analysis is that there is no intermediate zone between two consecutive bands of adsorbed substances, so th at a clear separation is almost impossible. This difficulty has been very neatly overcome by Tiselius and Hagdahl (1950) by the addition of a volatile substance having an adsorption affinity intermediate between those of the two substances to be separated. For instance when a mixture of methionine, Leu.Gly.Gly and n-butanol was subjected to displacement chromatography, the bands were eluted in the following order: methionine, n-butanol, Leu.Gly.Gly. By cutting in the middle of the n-butanol fraction it was possible t o obtain complete separation of the methionine and Leu.Gly.Gly. It is likely that adsorption methods may be useful for fractionating larger peptides, which cannot readily be separated b y other methods.
THE ARRANGEMENT OF AMINO ACIDS I N PROTEINS
35
Thus Synge and Tiselius (1947) were able to fractionate the components of tyrocidin, both by elution and displacement methods. I n a study of the partial hydrolysis of ovalbumin Moring-Claesson (1948) was able to separate by adsorption the unchanged protein from the breakdown products. The former was adsorbed much more strongly on alumina and less strongly on carbon than the smaller peptides and amino acids. On most of the more commonly used adsorbents, peptides and proteins are only weakly adsorbed from aqueous solution. Tiselius (1948; Shepard and Tiselius, 1949) has shown that in the presence of a high concentration of salt, substances are much more strongly adsorbed, thus making i t possible to use adsorbents such as paper or silica for the chromatography of proteins (see also Mitchell et al., 1949). This technique which is known as “salting out adsorption” may prove useful for the fractionation of larger peptides. Hamoir (1945) was able to fractionate amino acids into four groups by adsorption on silver sulfide. I n general those amino acids th a t form the Ieast soluble silver salts were the most strongly adsorbed. The order in which the amino acids are adsorbed on the column, is rather different from the order on other columns, so th a t a different type of group separation may be obtained. 4. Partition Chromatography Partition chromatography was originally introduced as a method for the fractionation of acetamido acids (Martin and Synge, 1941) and was early applied t o the separation of acetamido peptides from a partial hydrolyzate of gelatin (Gordon, Martin and Synge, 1943). I n this method the fractionation was carried out on columns of silica gel and the bands were located by incorporating an indicator in the aqueous phase of the chromatogram. The mixture studied was rather too complex to give really satisfactory resolution of the peptides and since much better results are obtained by direct fractionation of peptides on paper or starch chromatograms, the technique has not been used further. a. Paper Chromatography. Undoubtedly the most satisfactory procedure for fractionating amino acids and smaller peptides is paper chromatography (Consden et al., 1944) which is now familiar to most workers. The method has been the subject of a number of reviews (e.g. Consden, 1948; Martin, 1950; Jones, 1949) and its application to peptides has been described by Consden et al. (1947a). I n this method samples containing about 1 mg. of the peptide mixtures to be analyzed are fractionated on two-dimensional chromatograms. The position of the peptide spots is determined either by spraying with ninhydrin, which causes only a small amount of destruction of the peptide (see Woiwod, 1949) or from the
36
F. SANGER
fluorescence which is produced if the paper is heated (Phillips, 1948). The exact nature of this latter reaction is still obscure but it seems t o depend both on the peptide and on the paper (Patton et al., 1949). It is extremely valuable for locating spots with a minimum of concomitant destruction. The peptides may then be cut out and eluted by running a small amount of water through the ((cut.” Microtechniques for identifying the constituent amino acids and the N-terminal residues of these peptides have been described by Consden et al. (1947a) (see p. 48). The choice of suitable solvents for running the chromatograms will depend on the nature of the particular mixture to be analyzed. Probably the most generally useful ones are phenol, collidine and butanol-acetic acid mixtures (Partridge, 1948). The latter are especially useful for the larger peptides that “tail” badly on other solvents (Jones, 1948, 1949). Thus Phillips (1949a) found that a mixture of peptides from insulin which could not be satisfactorily fractionated on phenol or collidine could be separated into well-defined spots using butanol-acetic acid. By varying the acetic acid content, mixtures with different properties can be prepared. It is not possible a t present to predict the exact position of a peptide spot on a chromatogram from a knowledge of its composition, but a n approximate determination of its RF values may be obtained as follows. Martin (1949a) has shown theoretically th at the partition coefficient of a dipeptide divided by the product of the partition coefficients of the constituent amino acids is a constant for any given phase pair, ie., that for a peptide A.B ~ A . B ~- constant
O A ~ B
where aA,as and a A . B are the respective partition coefficients. Consden et al. (1944) showed th at the RF values are related to the partition coefficients by the equation
AL -
As
is a constant for any solvent system so that
RFAB
=
K(i
-
RF~RF~ RFA)(l- RFB) R F ~ R F ~
+
THE ARRANGEMENT O F AMINO ACIDS I N PROTEINS
37
Thus the Rr value of a dipeptide may be calculated in terms of the RF values of the constituent amino acids and of a constant K , which is best determined experimentally for each solvent system. The relationship is not absolutely accurate, since, for instance, peptides containing the same amino acids in different order may frequently be separated. It has, however, been found to apply satisfactorily in most cases and to be a useful check on the identity of a peptide. b. Starch Chromatography. Starch chromatography, originally introduced by Elsden and Synge (1944; Synge, 1944) has been developed into a very accurate method of amino acid analysis by Moore and Stein (1948, 1949; Stein and Moore, 1948). Amino acids and simple peptides move as sharp well defined bands and excellent resolutions may be obtained. Synge (1944, 1949) fractionated a partial hydrolyzate of gramicidin on starch and was able to identify a number of peptides. Amino acids and peptides on starch columns behave much as they do on paper chromatograms ; but the use of starch makes possible separations on a larger scale. The efficiency of separations is probably rather less on starch and it is not possible of course to use the two-dimensional technique. Starch chromatograms run extremely slowly and it is necessary to run each column for several days to obtain satisfactory fractionations. c. Other Partition Chromatography. A number of other types of partition chromatograms have been suggested and may be useful in the separation of peptides. Paper chromatography is only applicable on a micro scale and several attempts have been made to extend it to a larger scale. Mitchell and Haskins (1949) have described a " chromatopile " for such fractionations. This consists of a pile of filterpapers which are used as the column. Jones (1949) suggests the use of thick paper which may best be run by ascending chromatography. Recently there has become available preparations of powdered cellulose which are suitable for use in columns. Diatomaceous earth (Kieselguhr) may be used as an inert support for the aqueous phase of a partition chromatogram (Martin, 1949a; Bell et aE., 1949). Being only a very weak adsorbent, it is likely to be useful for larger peptides which may tail badly on other chromatograms due to excessive adsorption. Certain hydrophobic peptides are difficult to fractionate as they all tend to run fast on the usual chromatograms in which water is the stationary phase (Synge, 1949). Recently systems have been described using adsorbents such as rubber, in which an organic solvent is held as the stationary phase and the column is developed with water or a buffer solution (Boldingh, 1948; Howard and Martin, 1950).
38
F. SANGER
5 . Detection of Peptides from Columns I n any of the chromatographic methods described above it is necessary t o have some method for locating and if possible, for estimating the colorless peptides on the column or as they are eluted from it. I n paper chromatography they are located on the paper but where a column arrangement is used it is usually simpler to identify the bands as they are eluted. This is most usually done by collecting the effluent in a large number of small fractions. A number of automatic fraction collectors have been described that may be used for this purpose (Stein and Moore, 1948; Randall and Martin, 1949; Phillips, 194913) and some are commercially available. A suitable test may then be applied to each fraction. For this purpose the ninhydrin reaction is most generally used (Moore and Stein, 1948). Drake (1947) has described an automatic arrangement for spotting aliquot drops of the effluent on to filter paper, which can be developed by a suitable reagent. This could probably be used in conjunction with an automatic fraction collector. Several methods have been described for continuously recording the concentration of solute in the effluent from a column. Tiselius and Claesson (1942) have observed changes in the refractive index of the solution using a special interferometer. This method has proved very useful where adsorption chromatography has been used, but is not suitable for use with partition chromatograms as any slight changes in the composition of the solvent will cause changes of refractive index and will interfere with the recording. A small conductivity cell attached to the bottom of a column may also be used t o locate the bands of amino acids and peptides, which will cause changes in the conductivity of the solution (Randall and Martin, 1949). Recently Drake (1950) has described a polarographic method for following the chromatography of proteins. Only substances containing cystine or cysteine will be detected but it may be useful where large peptides are being studied. The method is unaffected by changes in the solvent or salt concentration.
6. Countercurrent Distribution The development of the method of countercurrent distribution is largely due t o the work of Craig and his associates (see Craig et al., 1949). The method is applicable to the separation of any substances tha t can be reversibly distributed between two immiscible solvents, and these, of course, include peptides. Several different types of apparatus
THE ARRANGEMENT O F AMINO ACIDS IN PROTEINS
39
have been devised for this purpose (Craig, 1944; Kies and Davis, 1950; Rometsch, 1950), the most efficient probably being the all-glass apparatus recently designed by Craig and Post (1949). Compared with chromatographic methods, this countercurrent distribution method is probably less efficient, the apparatus required is very much more complicated, and the labor involved is considerably greater; nevertheless, it possesses several definite advantages over chromatographic methods. The chief of these is the absence of any solid phase, which may act as an adsorbent for the solutes. In partition chromatographic methods this adsorption may often lead to distortion of the bands and render fractionation very inefficient. The behavior of a solute on a countercurrent distribution depends only on its partition coefficient, which in most cases is constant, so that it is possible to calculate the exact theoretical distribution curve and this may be used as a very sensitive test for purity. The method is especially useful for the fractionation of larger polypeptides, which cannot be fractionated easily by other techniques. Thus, for instance, Gregory and Craig (1948) studied crystalline gramicidin by this method, and found it to consist of at least three components, although it had previously been thought to be pure. Other naturally occurring polypeptides have similarly been purified (Craig et al., 1949; Barry et al., 1948; Livermore and du Vigneaud, 1949). 7. Lysine Peptides
The DNP method has been used for the separation and identification of peptides containing lysine (Sanger, 1949b). When a DNP-protein is partially hydrolyzed, the only colored products present are the DNP derivatives of the N-terminal peptides and of those peptides which contain lysine residues. The DNP-terminal peptides are extracted by an organic solvent as already described (p. 7). This procedure can be regarded as a very specific type of group separation in which only a few special peptides are separated out. The peptides containing c-DNP-lysine, which remain in the aqueous hydrolyzate solution mixed with other unsubstituted peptides, can be separated by adsorption on talc from acid solution, since the D N P group is held strongly on this adsorbent. Elution can be effected with acid ethanol and the peptides subsequently fractionated on suitable partition chromatograms. FDNB also reacts with tyrosine and histidine residues to give colorless products, which are likely to be retained by the talc. I n the case of insulin no such interference was observed, but it is doubtful if the method could be applied successfully to other more complicated proteins.
40
F. SANGER
8. Cystine Peptides Consden and Gordon (1950) have described an elegant method for investigating peptides involving cystine residues. After removal of acidic peptides from a partial hydrolyzate using an ion exchange column (Amberlite IR-4), the remaining peptides are oxidized with bromine, which converts the cystine residues to cysteic acid residues. The acidic peptides so formed can be separated from the remaining neutral peptides on another ion exchange column. Only peptides which originally contained cystine or cysteine linked with neutral amino acids appear in this fraction as cysteic acid peptides. This simplified mixture may then be fractionated by ionophoresis and paper chromatography.
9. Other Methods of Fractionation So far we have considered only the newer countercurrent methods for fractionating peptides. These will probably play a predominant role in the future, though the classical methods of fractional crystallization and precipitation should not be forgotten. They are still the most effective methods of fractionating proteins and probably larger polypeptides, such as the oxidation products of insulin (Sanger, 1949a). The aromatic sulfonic acids, which were developed by Bergmann for the specific precipitation of amino acids, have also been used for the separation of peptides from a partial hydrolyzate of silk (Stein et al., 1944). Synge and Tiselius (1950) have recently described a method for fractionating substances according to their molecular weight by electrokinetic ultrafiltration which may be applicable to the group separation of peptides. A number of ingenious methods of fractionation have recently been suggested by Martin (1949b).
10. Conclusions Many of the methods considered above have not been extensively used for the fractionation of peptides, so that it is impossible to know how far they may be applied and which are the most effective methods. Also it is to be expected that considerable improvements in these techniques will take place and that other new methods will be devised in the near future. The properties of amino acids and small peptides render them suitable to fractionation by methods employing partition chromatography. Paper chromatography is especially to be preferred for the final fractionation of a simplified peptide mixture because of the good resolutions obtained, the ease and rapidity of technique, and the possibility of using
THE ARRANGEMENT O F AMINO ACIDS IN PROTEINS
41
the two-dimensional method. The chief disadvantage is that only a very small amount of each peptide can be obtained from a chromatogram, though it is usually possible to obtain sufficient material to determine the structure of di- and tripeptides. When larger amounts of peptides are required, several methods are available; chromatography on starch, cellulose powder or ion exchange resins, or countercurrent distribution. Probably none of these gives such good resolution as paper chromatography and the work involved is greater. For the preliminary separation of a complex protein hydrolyzate into simpler peptide mixtures, ionophoretic methods are probably the most generally useful. Aromatic peptides may be separated by adsorption on charcoal and cystine peptides by oxidation. One of the main requirements at present is for methods that will efficiently fractionate large polypeptides with molecular weights between say 1,000 and 10,000. At present the best method for dealing with them is by countercurrent distribution and possibly by certain ionophoretic methods. While techniques for separating this class of substances would be extremely useful, it seems that the chief difficulty is to find suitable degradation procedures for producing a simple mixture of large polypeptides from a protein.
VI. DETERMINATION OF PEPTIDE STRUCTURE 1. IdentiJcation and Estimation of Amino Acids The first stage in the determination of the structure of a peptide is to identify the amino acid residues present in it, and the obvious technique for this is paper chromatography. By no other method is it possible to identify completely the amino acids present in a mixture so simply and so rapidly. Unfortunately no solvent has yet been found on which it is possible to separate all the naturally occurring amino acids and it is necessary to run each hydrolyzate on a two-dimensional chromatogram or else to, run aliquots on two separate one-dimensional chromatograms. Where a large number of peptides is to be analyzed, the latter technique is usually preferable (Consden et al., 1949). It is often desirable to carry out an approximate amino acid analysis to determine whether there are one or two residues of a particular amino acid in a peptide. This may be done with sufficient accuracy by carrying out the paper chromatography in a semi-quantitative manner (Polson, 1948; Consden et al., 1949). Several modifications of greater accuracy have been described (Martin and Mittelmann, 1948; Wieland and Fischer, 1948; Woiwod, 1949; Fowden and Penney, 1950; Boissonnas, 1950). The question as to how accurately it is possible to determine an amino
42
F. SANGER
acid on a paper chromatogram has been frequently discussed (Jones, 1949; Gordon, 1949; Martin, 1950) and is probably still unsettled. If careful control experiments are carried out the above methods should give results to within 5-10 %. Other methods (starch chromatography, microbiological methods, or the pipsyl method) may be used to obtain an accurate analysis but are considerably more laborious. An elegant method for determining the optical configuration of amino acids on paper chromatograms using the enzyme n-amino acid oxidase has been described by Synge (1949).
2 . Amino Acid Sequence The determination of the amino acid sequence in peptides is essentially a question of identifying terminal residues. Thus if the N-terminal residue of a dipeptide is known its structure is determined. The sequence in a tripeptide would be determined by identification of the N- and C-terminal residues. The structure of larger peptides may be worked out from their terminal residues and by degrading them to di and tripeptides. Thus, for instance, if a tetrapeptide has residue A as an N-terminal residue and on hydrolysis gives the dipeptides A.B and C.D, its structure must be A.B.C.D. An alternate method of determining the amino acid sequence is by a method of step-wise degradation from the terminal residues (p. 8). The method of Edman (1950), which has already been applied to synthetic peptides would seem especially suitable for this purpose. Consden et al. (1947a) have developed a rapid micro method for the deamination of the N-terminal residues of peptides eluted from paper chromatograms. A sample of the peptide is completely hydrolyzed and the amino acids present are identified. Another sample is then deaminated with nitrosyl chloride which destroys the N-terminal residue. The amino acids remaining are then identified after hydrolysis of the deaminated peptide. Although the reaction rarely goes completely smoothly and quantitatively, and a certain amount of destruction of the non-terminal residues usually takes place, it is nevertheless pQssible in most cases to identify the N-terminal residue. The DNP method (p. 5) may also be used for the identification of the N-terminal residues in peptides. It is rather more laborious than the above deamination procedure, but the results are usually more clear-cut. Two methods are available. Either the terminal DNP amino acid is identified by chromatography or the remaining unsubstituted amino acids are identified by paper chromatography. The former method has the advantage that it may be carried out quantitatively but requires rather more material than is usually available from the elution of one paper chromatogram.
THE ARRANGEMENT O F AMINO ACIDS IN PROTEINS
43
Bowman (1950) has shown that amino acids and peptides may be quantitatively converted to their dimethyl derivatives by reductive alkylation in the presence of formaldehyde, and has used the method to detect the N-terminal residues of peptides. A micro-modification was described by Ingram (1950) who identified the amino acid which is absent from a hydrolyzate of the dimethyl peptide.
3. Estimation of Peptides
It is clear that the estimation of the amount of a particular peptide in a partial hydrolyzate may in certain cases yield considerably more information concerning the structure of the protein than its mere identification. Unfortunately it is impossible to determine directly the total amount of a particular peptide sequence in a protein molecule. Only a minimal value can be obtained from the composition of a partial hydrolyzate. I n the case of the N-terminal peptides of insulin it was possible to estimate certain sequences in the protein from the yields of the peptides by allowing for the rate of breakdown of the bonds involved (p. 5 2 ) . Such an approach may clearly be useful in the future where the yield of a peptide can be determined. Usually only small amounts of peptide are available after chromatographic fractionation of a partial hydrolyzate, so that micro methods of estimation are required. Consden et al. (1949) obtained an approximate estimate of peptides from a paper chromatogram by the color intensities of the amino acid spots produced on hydrolysis. Clearly careful control analyses with synthetic substances should accompany any estimaitons by this type of technique to allow for losses during fractionation. The use of isotopic methods for the estimation of peptides appears to offer considerable advantages of accuracy and general applicability. I n the original “isotope dilution” technique of Rittenberg and Foster (1940) it was necessary to isolate a pure sample of the compound to be estimated, from a mixture of comparable amounts of other similar substances. In the case of peptides this would clearly be a formidable task, and has never been attempted. The Pipsyl Method. Using radioactive isotopes Keston, Udenfriend and Cannan (1946, 1949; Keston and Udenfriend, 1949) have developed an accurate micro method, which theoretically could be applied to the estimation of any amino acid or peptide in a mixture. Two different techniques, the carrier technique and the indicator technique, have been worked out. In the former the peptide mixture to be analyzed is treated with p-iodophenylsulfonyl chloride (pipsyl chloride) containing radioactive 1 1 3 1 , which reacts quantitatively with the amino groups givingpipsyl derivatives. To this is added a great excess of the non-isotopic pipsyl
44
F. SANGER
derivative of the peptide to be estimated, so diluting the corresponding isotopic derivative in the hydrolyzate. A pure sample of this pipsyl peptide is then isolated from the mixture, and from its isotope content, the amount of peptide in the original mixture may be calculated. The chief difficulty lies in the isolation of the pipsyl derivative in an absolutely pure state free from any other radioactive pipsyl derivatives of closely related compounds which tend to form mixed crystals. In many cases a large number of recrystallizations are necessary before a constant isotope concentration is obtained. To overcome the above difficulties the indicator technique was developed for the estimation of amino acids (Keston et al., 1947, 1950). The amino acid mixture is again treated with pipsyl chloride containing I1al. This mixture can then be fractionated by countercurrent distribution and paper chromatography and if all the derivatives are completely separated, the amount of amino acid can be determined from the isotope content of the spot. By adding an indicator to the mixture before fractionation it is possible to obtain an estimate even if the spots are not completely resolved. For instance, if alanine is to be estimated the indicator added would be pipsyl alanine containing S36,which may be estimated independently of the The ratio of 113' to S36in a part of the pipsyl-alanine spot may be used to estimate the original alanine content. This is only true for those parts of the alanine spot which contain no other pipsylderivatives (containing Il3'). I n such parts the to S36is constant, and this is used as the test of purity. This ratio of method has not yet been applied to peptides, but there seems no reason why it should not be, though rather complex fractionation procedure might be necessary to obtain a part of the peptide spot in a pure form. OF INVESTIGATIONS ON VARIOUSPROTEINS VII. RESULTS
In this section we shall consider the results obtained since 1943, which provide information concerning the arrangement of amino acids in proteins. Similar studies on the naturally occurring polypeptides will not be dealt with here. 1. Silk Fibroin (Bombyx Mori) Assuming a molecular weight of 30,000 for silk fibroin (Coleman and Howitt, 1946) its composition is given by the following formula: (~ly16gAlalo,Ser40Tyr lgAsp6Glu4Leu6Va18PronPhe,Thr4Arg,LyslHis1 (Levy and Slobodiansky, 1949; Tristram, 1949)
It may be noticed that this formula tends approximately but not exactly to the formula (Gly3Ala2X2),where X represents any residues
THE ARRANGEMENT O F AMINO ACIDS I N PROTIGINS
45
other than glycine or alanine, and this unusually simple composition suggests that the amino acid arrangement may also be simple. By the classical methods of fractionation the peptides Gly.Ala (Fischer and Abderhalden, 1907) Ala.Gly (Abderhalden, 1909a) and Gly.Tyr (Abderhalden , 1909b) were isolated in considerable yield from the silk fibroin of Bombyx Mori and unambiguously characterized. Several longer peptides were also isolated in small yield and fairly convincing evidence for their structure was presented, though they were not finally identified by synthesis. Thus Abderhalden and Bahn (1932) benzoylated a fraction from a hydrolyzate of fibroin which had been obtained by the action of 1 N NaOH a t 37", and obtained several benzoyl peptides. One of these contained the amino acids glycine, serine, tyrosine, and proline in the ratio 1:1:1:2 and on partial hydrolysis with 10% HzS04 gave hippuric acid, Ser.Pro and Tyr.Pro. In an experiment with synthetic peptides it was found that benzoyl Gly.Ser was completely broken down under the conditions of hydrolysis in 10% H2S04 whereas benzoyl Gly.Tyr was rather stable. From this it was concluded that the only possible structure for the pentapeptide was G1y.Ser.Pro.Tyr.Pro. A second fraction was similarly identified as Ser.Pro.Tyr.Pro. A third peptide (Abderhalden and Bahn, 1933) containing the amino acids tyrosine, serine, and proline in the ratio 2 :1: 1 was isolated from a similar hydrolyzate. On treatment with trypsin (presumably a crude preparation) this gave tyrosine, Ser.Pro and a tripeptide containing one molecule of the three amino acids. This tripeptide was then treated with benzylamine and phenylisocyanate, according to the method of Abderhalden and Brockman (1930) and from the hydrolyzate the phenylhydantoin of tyrosine and Ser.Pro benzylamide were isolated. This established the structure of the tripeptide as Tyr.Ser.Pro, and it was concltxded that the tetrapeptide was Tyr.Ser.Pro.Tyr. The presence of tyrosine as a N-terminal amino acid was assumed since the tetrapeptide was precipitable with mercuric sulfate, as were tyrosine and other dipeptides containing a free tyrosine carboxyl group, whereas the tripeptide was not precipitable. While the evidence for the structure of these peptides is fairly good, it is possible that they may have been mixtures of peptides having slightly different structures. At least they do show that there are tetrapeptide sequences in fibroin that contain no glycine or alanine. Stein et al. (1944) have made use of their method of specific precipitation with aryl sulfonic acids to isolate peptides from an acid hydrolyzate of silk fibroin. Gly.Ala was isolated in 5.597, yield using 2:5-dibromobenzene sulfonic acid and Ala.Gly in 6.0% yield using 2,6-diiodophenol4-sulfonic acid. A more precise estimation of these two peptides present at various
46
F. SANGER
stages of hydrolysis was obtained by Levy and Slobodiansky (1949) with the pipsyl carrier technique. Samples of fibroin that had been hydrolyzed with 12 N I-ICl a t 39” for 16, 24, and 48 hours respectively were analysed for glycine, alanine, Ala.Gly, Gly.Ala and Gly.Gly. The results of one such experiment are shown in Table V. It was calculated that if the arrangement were completely random, the maximum possible yield of Gly.Gly would be 18.2% and of Ala.Gly and Gly.Ala 12.2%. The above figures, especially those for Ala.Gly, which are over twice the theoretical value show th a t this is not the case. Th e relatively small yield of Gly.Gly suggests that this sequence does not occur t o any great extent whereas the yield of Ala.Gly accounts for about half of the alanine present in the protein. Levy and Slobodiansky have pointed out that these results would be expected if the minimum TABLEV Analysis of Partial Hydrolyzate of Silk Fibroin (Levy and Slobodiansky, 1949) (Yields expressed in terms of N as % of total N) Time of hydrolysis in 12 N HC1 at 39” Glycine Alanine Ala.Gly Gly.Ala Gly.Gly
16 hr.
24 hr.
48 hr.
4.4
7.1 6.3 23.3 9.0
12.9 10.5 27.0 8.3 1.8
3.1 16.9 5.4 0.1
-
repeating peptide sequence were of the type: X.Ala.Gly.Ala.Gly.X.Gly. While this is probably the simplest explanation of the results, there are certainly many other more complicated ones. It does not account for the tri- and tetra-peptides isolated by Abderhalden and Bahn which contain no glycine or alanine or for the results of Drucker and Smith (see below). The results of Goldschmidt et al. (1933), and of Grant and Lewis (1935) referred to by Synge (1943) also suggest an uneven distribution of residues throughout the molecule. At least, it may be said that the results of Levy and Slobodiansky make it likely that the sequence X.Ala.Gly.Ala.Gly.X occurs frequently in silk fibroin. Silk fibroin may be dissolved in a solution of cupri-ethylenediamine (Coleman and Howitt, 1947). On dialysis part of the material remains in solution and is said t o be “renatured.” Drucker and Smith (1950) treated this material for a short period with trypsin and obtained a precipitate, which had a n average molecular weight of about 7000 and contained only the amino acids glycine, alanine, and serine. The remaining amino acids were all left in solution. It, was suggested on the basis of this evidence
THE ARRANGEMENT OF AMINO ACIDS I N PROTEINS
47
that the silk fibroin molecule is built up of five parts, three of which have molecular weight 7000 and contain only the above three amino acids, the other two parts having molecular weight 5000 and a more complex composition. These results are in agreement with the suggestion of Meyer et al. (1940), t ha t silk fibroin is composed of two parts, an amorphous part of complex amino acid composition and a crystalline part built u p simply of Gly.Ala units linked together. This theory was based on the observation t ha t the X-ray data could be fitted by a unit cell containing four parallel Gly.Ala residues. It was considered th a t some of the alanine residues in the crystalline part could be replaced by serine since its molecular dimensions are similar. Clearly the idea of a long chain containing glycine and alanine alternately is untenable in view of the results of Levy and Slobodiansky since in such a case the yields of Ala.Gly and Gly.Ala should be equal. However such a sequence could be broken u p to a certain extent by serine residues. Abderhalden (1940, 1943) has recently renewed his investigations on the isolation of diketopiperazines from silk fibroin and has isolated considerable quantities of the anhydrides of Gly.Ala, Gly.Tyr, and Ala.Ser from partial hydrolyzates obtained b y the use of strong acid or proteolytic enzymes. No evidence could be obtained of anhydrides containing only one amino acid, making it unlikely th at such sequences occur to any great extent in fibroin. Abderhalden believes that these diketopiperazines are derived from some cyclic structure in the protein. It is possible however t ha t they may have been formed from peptides during the isolation procedure since certain quite mild conditions are known to catalyze their formation (Agren, 1940; Huang and Niemann, 1950). Hot ethanol was used to extract them, and this might have brought about ring formation. Control experiments were carried out to show that dipeptides did not undergo ring closure under the conditions of hydrolysis, but the other obvious control, boiling a dipeptide with ethanol, was not reported.
2. Protamines Little progress has been made concerning the chemistry of the protamines since Synge reviewed the subject in 1943. However it would seem that much could be learnt from the application of the newer methods of peptide chemistry to these proteins, which from their amino acid composition and molecular weight appear to have a rather simple structure (see Tristram, 1949). Porter and Sanger (1948) showed the presence of proline as a N-terminal residue in salmine, but were unable to estimate it due to the instability
48
F. SANQER
of DNP-proline. Felix and Mager (1937) had already suggested that proline was the N-terminal residue in clupein on the basis of titration data. This was confirmed using the DNP method, and it was suggested there was also a small amount of N-terminal serine (Felix et al., 1950). The presence of arginine as a C-terminal residue was indicated by Dirr and Felix (1932). If dibenzoyl arginine is treated with acetic anhydride, a n acetyl derivative is formed which on hydrolysis gives benzoyl acetyl urea and p-benzoylamino-a-piperidone. This is readily split with acid to give ornithine. It was found that ornithine was produced after benzoylation, acetylation, and hydrolysis of clupein. The above series of reactions could only take place if the a-carboxyl group was not involved in peptide linkage so that it must have originated from a C-terminal arginine residue. Since two-thirds of the residues in clupein are arginine it follows that the sequence Arg.Arg must occur frequently in the molecule, and the dipeptide has often been isolated. Felix and Schuberth (1942) have described a method for the preparation of Arg.Arg and obtained 12.5 g. of the pure diflavianate from 50 g. clupein. The fractionation of partial hydrolyzates of clupein by adsorption chromatography on filtrol-neutrol (Waldschmidt-Leitz and Turba, 1940, 1941) arid on the cation exchange resin Wofatite C (Rauen and Felix, 1948) have been studied. Considerable fractionations could be obtained, but no definite peptides were identified. 3. Wool Keratin Considerable doubt exists as to whether wool can be regarded as a single protein or whether it is a mixture of different insoluble proteins, so that it is perhaps not one of the most suitable proteins to study by partial hydrolysis methods. However, because of its practical importance the Leeds workers have applied their newer methods to it, and a number of peptides have been identified. Consden et al. (1949) studied a n acidic fraction of a partial hydrolyzate of wool. Preliminary group separation was carried out on an ion exchange column (Amberlite IR-4) to obtain an acidic fraction, which was separated into ten fractions by ionophoresis in silica jelly. Each fraction was then subjected to two dimensional paper chromatography using phenol and collidine. One of these ionophoretic fractions contained only one peptide which was identified as Glu.GIu. This was present in much greater amount than any other acidic peptide and accounted for about 10% of the total glutamic acid of wool (estimated from the total N of the ionophoretic fraction). I n Table VI are listed the peptides that were considered to be probably present. About an
THE ARRANGEMENT O F AMINO ACIDS I N PROTICINS
49
equal number of peptides were given as being possibly present. Interpretation of the results was rendered difficult by the extreme complexity of the mixture, so that only a few of the spots appeared to contain a singli: peptide. Where no difference in composition was found before and after deamination of a dipeptide spot, it was assumed that both possible peptides were present. It is doubtful if this is entirely justified, since some dipeptides are extremely stable to deamination. Clear evidence was obtained for the presence of Glu.Glu, Ala.Glu, Glu.Ala TABLEV I Acidic Peptides Identijied from Wool (Consden et al., 1949) Peptide
Yield *
(ilii.Asp Glu. G1u GIu.Gly Glu.Ala Glu .T y r Glu.I,eu Glu. Cy8 Asp.Glu Ser.Glu Gly.Glu Ala.Glu Tyr .Glu Val.Glu Leu. Glu CyS.Glu Asp.Va1 Asp.Leu Ser.Asp Leu.Asp
4 600
* mg. N of peptide per 100 g. N of
5 2!J
20 21 4
4 17
8 27 27 19 51 4 11 11
5
19
wool.
and Asp.Leu, and the peptides listed in Table VI are probably present also. The approximate yields of the peptides were estimated from the strength of ninhydrin color of the amino acids produced on hydrolysis. These yields are listed in Table VI. Consden and Gordon (1950) have studied the cysteic acid peptides derived from the cystine peptides of wool as described on p. 40. After a preliminary group separation of the cysteic acid peptides by ionophoresis, each fraction was fractionated on paper chromatograms with phenol and collidine. In Table VII are listed the peptides considered to be probably present and their approximate yield. The results were more clear-cut in this case than with the aspartic and glutamic acid
F. SANGER
50
peptides, largely due to the fact that it was possible t o separate by ionophoresis or chromatographically any two dipeptides containing the same amino acids in different order. This separation was made possible b y the low p K of the amino group in cysteic acid peptides. It is clear from the results that the structure of wool is extremely complex. Almost all of the monoamino acids occur linked t o both sides of glutamic and cysteic acids. The small number of aspartic acid peptides identified is probably due to the lower content of aspartic acid in wool so t ha t the peptide spots would probably be faint and would not show on the chromatograms. TABLEV I I Cysteic Acid Peptides from Wool (Consden and Gordon, 1960)
* m g . N of
Peptide
Yield*
Asp.CySOaH Glu. Cy SOaH Ser .Cy SOSH Gly .CySO3H Thr.CySOaH Ala.CySO3H Leu.CySO3H CyS0aH.Gly CySOaH.Thr CyS03H.Ala CyS03H.Val CyS03H.Leu CySOaH.Phe
2 2 60 24 20 30 8 10 20 14 20 10 4
peptide per 100 g . N of wool.
Middlebrook (1949) investigated the N-terminal residues by the DNP method and found rather a complex mixture, which would seem t o be a further indication of the heterogeneity of wool. 4. Insulin
Insulin has been studied in rather more detail than most other proteins due t o the interest arising from its physiological properties and to the fact that it is one of the few proteins that can be obtained in a reasonably pure form. It possesses a relatively simple structure, being built up of fairly short open polypeptide chains. For the purposes of the following discussion a value of 12,000 (Gutfreund, 1948) will be assumed for the molecular weight. The most recent analytical figures (see Tristram, 1949) indicate the following composition :
51
T H E ARRANGEMENT O F AMINO ACIDS I N PROTEINS
Gly7Ala~Va18LeulzIleu3Pro3Phe6( CyS-) 1zArg~HislLyszAsp~Glu6 (Asp-NHz)4(Glu-NH2)8Ser6ThrzTyrs The DNP method showed the presence of four N-terminal residues two of which were glycine and two phenylalanine (Sanger, 1945). From a study of the action of carboxypeptidase on insulin Lens (1949) found that free alanine was liberated before any other free amino acids, indicating the presence of alanine in the C-terminal position. Using their respective methods of reduction of the free carboxyl groups to alcohol TABLE VIII Properties of the Fractions of Oxidized Insulin (Sanger, 1949a)
Yield from insulin N-terminal residue Amino acids absent
Mol. wt. 1. From estimation of N-terminal residues 2. By ultracentrifugation (Gutfreund and Ogston, 1949)
A
B
30-40 % Glycine Lysine, arginine, histidine, phenylalanine, threonine, proline
25 % Phenylalanine Isoleucine
2900
3800
2900
7000
groups (p. 10) Fromageot et al. (1950) and Chibnall and Rees (1951) both identified alanine and glycine as the C-terminal amino acids. These results indicate that insulin is built up of four open polypeptide bridges of cystine. chains, which seem to be held together by -S-SThese are present to the extent of six residues per molecule. By oxidation with performic acid (p. 27) it was possible to split these bridges and thus to liberate the separate peptide chains (Sanger, 1949a). The oxidized insulin was fractionated by precipitation methods and two main fractions were obtained, which appeared to represent the whole of the oxidized insulin. The properties of these two fractions, designated A (acidic) and B (basic), are summarized in Table VIII. The most probable composition of fraction A is, GlyzAlalValzLeu Jleu 1(CyS) (AspzGlu4Ser~ T y (-NHz) r
4
and of fraction B, Gly3AlazVa13Leu4Pro 1Phe3(CyS) zArglHiszLyslAsplGlu 3Ser1Thr1Tyr2 (--NH2)2
Butler and coworkers have studied the action of various proteolytic enzymes on insulin. From a chymotryptic hydrolyzate they obtained
52
F. SANGER
two fractions, one precipitable by trichloracetic acid and one soluble (Butler et al., 1948). The latter fraction consisted of relatively small peptides with an average chain length of seven residues. These could be separated satisfactorily on paper chromatograms using butanol-aretic acid (Phillips, 1949a) but they have not yet been investigated further. The trichloracetic acid precipitate, which was referred to as the core” had a mean molecular weight of about 5000 and a high proportion of glycyl (approximately 3 equivalents) and a smaller amount of valyl TABLE IX U N P Deravatzves .from Fractaon B of Oxidized Insulan (Sanyer, 1949b)
Derivative
Bl €52
B3
B4
B5
Products of complete hydrolysis
Products of partial hydrolysis
Yield* from DNPinsulin
DNP-phenylalrtnine 1) N P-phenylalaninr Valine DNP-phenylalanine Valine Aspartic acid DNP-phcnylalanine Valine Aspartic acid Glutaniic arid
-
13 10
13
BI, B2
13
12
€31, l32, R3
30
55
B1, B2, B3, B4
20
-
Yield* from
IM
14
-
Total
* Moles of
92
04
peptide as % of the total N-terminal phenylalanyl residues.
and other N-terminal residues (Butler et al., 1950). On oxidation with performic acid the core was mostly converted to a fraction corresponding in properties to the fraction A of the oxidized insulin. a. N-Terminal Peptides. The N-terminal peptides of the two fractions from the oxidized insulin were determined by subjecting their DNI’ derivatives to partial hydrolysis (Sanger, 1949b). I n the case of fraction B four main bands (Bl-B4) were identified in an ethyl acetate extra(% of a partial hydrolyzate. Their properties are summarized in rl’ablc I X . I t is clear that all these derivatives are derived from the one N-terminal sequence Phe.Val.Asp.Glu. Other fainter bands (B5) were also present ; on partial hydrolysis these gave rise t o B3 and B4 and were therefore higher peptides from the same peptide sequence. The aspartic and glutamic acid residues are probably in the form of asparagine and glutamine residues in the intact protein, since other bands containing the
53
THE ARRBNGEMENT OF l M I N O ACIDS I N PROTEINS
same amino acids as H3, B4 and probably therefore amides, were also present when a shorter time of hydrolysis was used. The yields of the different peptides from a partial hydrolyzate of DNP-insulin are shown in column 4 of Table IX. It can be seen that virtually all the N-terminal phenylalanyl residues may be accounted for in terms of this one sequence. This is confirmed by comparing the yields of the DNP peptides from DNP-insulin with the yields from a sample of peptide B4 which had been hydrolyzed under similar conditions (column 5, Table IX). AgreeTABLE S Peptides Containing e-D,VP-lysine from DiVP-insulin (Sanger, 1949b)
*
Prptidc
Arnino acids present
N-tcrnrirral rcsidiic
Products uf partial hydrolysis
Structure
Yicld fronr Yield* DNP- from insulin L4
L1
c-DNP-Lysinc c-DNP-Lysine Alaninc Threonine Proline c-DNP-Lysine Threonine Proline c-DNP-Lysine Alanine
r,z
L3
1.4
-
-
c-DNP-Lysinc
-
c-DNP-Lys t-DNP-Lys.Ala
14 19
23
Thrconine
L1
Thr.Pro.(r-DN1')Lys
32
32
23
21
Thrconinc
L1. L2, L3 Thr.Pro.(a-DNP) Lys.
Ala
Other unidentified bands
6
Total
* Moles of peptide as
14
94
% of the total lysinc.
ment between the two sets of figures makes it clear that both the N-terminal phenylalanyl residues of insulin are present in the form of this one sequence, Phe.Val.Asp.Glu, and it was therefore concluded that the two phenylalanyl chains of insulin, which contain the same N-terminal tetrapeptides are in fact identical. It was also possible to separate and identify the peptides containing t-DNP-lysine from a partial hydrolyzate of the DKP derivative of fraction B. Here again four main colored bands were present and their properties are summarized in Table X. These all fit into the sequence Thr.Pro.Lys.Ala and the yields from DNP-insulin and from peptide L4 make it clear that both the lysyl residues of insulin are present in this single tetrapeptide sequence. When the DNP-derivative of fraction A of the oxidized insulin was partially hydrolyzed, four DNP derivatives were produced, all of which fitted into the sequence DNP-Gly.Ileu.Va1.Glu. When strong HCl was
54
F. SANGER
used for hydrolysis the peptides were produced in very small yields. Because of the great lability of the bond involving the carboxyl group of the glycyl residue most of the DNP-glycyl residues were present as DNP-glycine itself and as larger peptides which could not readily be fractionated. However when dilute acid was used for hydrolysis the yield of these peptides was raised and another band which appeared to be Gly.Ileu.Val.Glu.Glu was obtained. The yields of these peptides indicated that both the N-terminal glycyl residues are combined in the one amino acid sequence and it was concluded that the two glycyl chains were also identical. b. Amino Acid Sequence in the Phenylalanyl Chains. The results with the N-terminal peptides showed for the first time that the fractions A and B of the oxidized insulin were each an essentially homogeneous preparation of a polypeptide chain of 20 and 30 residues respectively. It was therefore considered worth while to investigate the peptides present in their partial hydrolysates by the methods of Consden et al. (1947b). Fraction B was subjected to hydrolysis in conc. HCI andthe hydrolyzate separated into a number of fractions by ionophoresis, charcoal adsorption, and adsorption on an ion exchange resin (Sanger and Tuppy, 1951a). The resulting peptide mixtures were then fractionated on two-dimensional paper chromatograms and their structures investigated. To illustrate the methods used a fraction containing peptides of cysteic acid will be considered. This was obtained by adsorption on an ion exchange resin (Amberlite IR-4B a t pH 2.6; Consden et at., 1948). A chromatogram of this fraction on phenol/butanol-acetic acid revealed eight peptide spots, whose compositions are given in Table XI. The N-terminal residues were determined by hydrolysis of the DNP derivatives. The structures of the peptides, as far as they can be deduced from the data in the table are given in the last column. Three dipeptides were identified: CySO,.H.Gly, Val.CyS03H, and Leu.CyS03H. Fraction B contains only two cysteic acid residues so that any peptide containing cysteic acid which is not the N-terminal residue must either contain the sequence Val.CyS0,H or Leu.CySO,H. Peptide 3 must therefore be Val.CySO,H.Gly and peptide 5 Leu.CySO,H.Gly. Thus both cysteic acid residues are joined through their carboxyl groups to a glycine residue. Peptide 8 which contains no glycine but has leucine as the N-terminal residue can only be Leu.Val.CySO,H and peptide 7 Leu.Val.CyS03H.Gly. These peptides establish therefore the presence of the sequences Leu.Va1.CySOaH and Leu.Val.CyS0,H.Gly in fraction B. In the upper part of Table XI1 are listed the various peptides identified in this work and the sequences which were deduced from them as
.
THE ARRANGEMENT O F AMINO ACIDS IN PROTEINS
55
being present in fraction B. Hydrolyzates obtained by the use of dilute acid and of alkali were also studied but only a few new peptides were found. These are also included in Table XU. From the results it was possible t o deduce five sequences as being present in fraction B, accounting for all the amino acids present except for one leucine, one tyrosine and TABLEXI Cysteic Acid Peptides from Fraction B of Oxidized I n s u l i n (Sanger and T u p p y , 1961a ) Strength of amino acid after hydrolysis of Peptide 1
2 3
4
5
6
7
8
Amino acids present Cysteic acid Glycine Aspartic acid Glutamic acid Cysteic acid Glycine Valine Cysteic acid Valine Cysteic acid Glycine Leucine Cysteic acid Leucine Cysteic acid Glycine Valine Leucine Cysteic acid Valine Leucine
Peptide
DNP-Peptide
xxxx xxxx
xxx
-
Structure CySOaH.Gly (ASP, Glu)
X X
xx
X
X
X
xx xxx xxx
xx
-
X
X
X
?
X
xxx xxx xx xx xx xx xxxx xxxx xxxx
Val. (CyS03H, Gly) Val.CyS03H Leu. (CyS03H, Gly)
xxx
Leu. CySO3 H
xx xx xx
Leu.(CyS03H, Gly, Val)
xxx xxx
Leu.(CyS03H, Val)
-
-
two phenylalanine residues. It was not, however, possible to fit these five sequences together into a single unique structure. This was partly due to the lability of certain bonds t o hydrolysis, especially those involving the amino groups of the serine and threonine residues. Thus no peptide was identified which contained these bonds intact. Another difficulty was experienced in separating the less polar peptides containing phenylalanine, leucine, valine, and tyrosine, since these residues are grouped together in the polypeptide chain and give rise to a large number of similar peptides. The action of the proteolytic enzymes pepsin, trypsin, and chymotrypsin on this fraction was next investigated (Sanger and Tuppy,
TABLEX I Peptides Identified in Hydrolyzates of Fraction B of Oxidized I n s u l i n (Sanger and T u p p y , 19510, b ) Dipeptides from acid Phe.Va1 Glu.His CySO3H.Gly His.Leu Glu.Ala CySOaH.Gly Arg.Glg Leu.Va1. Ala.Leu Leu.Va1. Gly.Glu Gly.Phe Val.Asp His.Leu and alkaline hySer.His Val.Glu Val.CySO3H Glu.Arg Asp.Glu Leu.CySOaH drolyeates Leu.CySOaH.Gly Val.Glu.Ala Tyr.Leu.Va1 Gly.Glu.Arg Tripeptides from acid Phe.Val.Asp and alkaline hyGlu.His.Leu Ser.His.Leu Val.CySO3H.Gly drolyeates Val.Asp.Glu Leu.Val.Glu Leu.Va1.CySOaH His.Leu.CyS08H Ala.Leu.Tyr Higher peptides from Phe.Val.Asp.Glu Ser.His.Leu.Va1 Tyr.Leu.Val.CySO3H acid and alkaline His.Leu.CySOsH.Gly Leu.Va1 .Glu.Ala Leu.Val.CySOaH.Gly hydrolyzates Phe.Val.Asp.Glu.His Ser.His.Leu.Val.Glu Glu.His.Leu.CySO3H His.Leu.Val.Glu Ser.His.Leu.Val.Glu..41s Sequences deduced Phe.Val.Asp.Glu.His.Leu.CySOtH.G1y Tyr.Leu.Val.CyS01H.Gly from above pepSer.His.Leu.Va1.Glu.Ala Gly.Glu.Brg.Gly tides Peptides identified in Phe.Val.Asp.Glu.His.Leu.CySO~H.Gly.Ser.His.Leu Leu.VaLCyS03H.Gly.Glu.Arg.Gly.Phe peptic hydrolyzate Val .Glu.Ala.Leu
Lys.Ala Thr.Pro Pro.Lys.Ala
Thr.Pro.Lys.Ala
Thr .Pro .Lys .Ala
Tyr.Thr.Pro.Lys.Ala
His.Leu.CySOaH.Gly.Ser.His.Leu Tyr.Thr.Pro.Lys.Ala PeDtides identified in Phe.Val.As~.Glu.His.Leu.C~SO~H.Glv.Ber.His.Leu.Val.Glu.Ala.Leu.Tpr chymotryptic Leu.Val.CyS01H.Gly.Glu.Arg.Gly.Phe.Phe hydrolyzate Peptides identified in G1y.Phe.Phe.Tyr.Thr.Pro.Lys. tryptic bydrolyeate Ala Phe.Va1.Asp.Glu.His.Leu.(CyS-) .Gly.Ser.His.Leu.Val.Glu.Ala.Leu.Tyr.Leu.Val.(CyS-) .Gly.Glu.Arg.Gly.Phe.Phe.Tyr.Thr.Pro.Lys. Structure of the phenylalanyl chain Ala of insulin
THE ARRANGEMENT OF AMINO ACIDS I N PROTEINS
57
1951b). The larger peptides present, which contained up to about fifteen residues, could be satisfactorily separated on paper chromatograms and with the knowledge obtained from the lower peptides it was possible to deduce the structure of many of them from their amino acid composition and N-terminal residues and to work out a unique sequence for fraction B. These results are also shown in Table XII, where the structure of those peptides which played a major part in the elucidation of the sequence are given. The structure of fraction B was worked out as the only possible sequence which would fit all the experimental results, assuming that it was a single polypeptide chain of about thirty residues. The fact that it was possible to find a unique sequence was regarded as proof that such an assumption was correct, and that there is only one type of phenylalanyl chain in insulin. 5 . Ovalbumin Using the DNP method it has been shown that ovalbumin has no N-terminal residue (Desnuelle and Casal, 1948; Porter, 1950a). Either the amino groups at the end of the chains are masked by the carbohydrate moiety or the protein contains one or more cyclopeptide units. A very specific type of hydrolysis occurs when ovalbumin is incubated with a proteolytic enzyme prepared from B. subtilis. A new crystalline protein, plakalbumin, is produced (Linderstrgm-Lang and Ottesen, 1949), which differs from ovalbumin in crystalline form and in solubility and has a somewhat lower molecular weight (Guntelberg and Linderstrgm-Lang, 1949). A t pH values below 7.0 it has a slightly different electrophoretic mobility which is concordant with the loss of two acidic groups per molecule (Perlmann, 1949). The conversion is accompanied by the liberation of six atoms of non-protein nitrogen per molecule of ovalbumin (Eeg-Larsen et al., 1948). This fraction contains two free amino groups, 3.6 free carboxyl groups and 4 atoms of peptide bond N per six nitrogen atoms; it contains no free amino acids but on hydrolysis yields alanine, glycine, valine and aspartic acid (Villee et al., 1950). The results suggest that about two small peptides are present. Since the ratio of the rate of formation of plakalbumin to the rate of formation of nonprotein N was not constant, it was evident that the reaction takes place in a t least two stages, probably the rupture of two peptide bonds. On further incubation with the enzyme, the plakalbumin is broken down slowly to a mixture of products. Clearly this is a case where it is possible to obtain characteristic degradation products at the early stages of an hydrolysis. It would seem that about two peptide bonds are split about ten times as rapidly as any others.
58
F. SANGER
6. 7-Globulin
Porter (1950a) found rabbit y-globulin to have one N-terminal alanine residue per molecule of molecular weight 160,000, which suggests that it is probably a single long chain of over a thousand residues. IIt: also determined the N-terminal peptide sequence. All the DNP-peptides found fitted into the one sequence DNP-Ala.Leu.T’al.Asp and the fifth residue was probably glutamic acid. There were no other D N P peptides present and the whole of the terminal alanine residues could be quantitatively accounted for in peptides fitting the above sequence. In contrast t o the results of physicochemical studies, these findings suggest that the protein is chemically homogeneous, since it would seem unlikely that two different proteins would have the same terminal tetrapeptide sequence. It is difficult t o be absolutely certain about this question, since nothing is known about the principles that govern the order of amino acid residues in proteins, but it may be pointed out that 011 a purely random basis there are 1g4 possible terminal tetrapeptide sequences. The fact t ha t various different residues have been found to occupy the N-terminal positions in proteins indicates that there is no generd principle that defines closely the amino acid that occupies a particular position in a protein, so that the presence of a single N-terminal peptide sequence in a preparation can best be explained on the basis of chcmiral homogeneity. Porter also compared the N-terminal peptides of normal 7-globulin and of purified antivoalbumin, which was studied in the form of a specific precipitate with ovalbumin. No difference could be found, again suggesting the chemical similarity between the antibody and the normal y-globulin, from which it is formed. An alternative explanation is th a t only a small part of the polypeptide chain, “the active center,” has a different amino acid sequence in the antibody. In an attempt to identify the “active center” of the antiovalbumin molecule, Porter (I 950b) studied the action of proteolytic enzymes on it. By the action of papain a molecule about a quarter the size of the original 7-globulin was split off, which acted as a specific inhibitor in the antibody reaction. Since it had a N-terminal alanine group it appeared to come from the N-terminal quarter of the y-globulin molecule. On further hydrolysis all activity was lost and it was not possible t o obtain an active molecule small enough for chemical investigation. These results illustrate the possibilities of applying chemical methods to biologically active proteins, most of which have been studied so far only by physicochemical techniques.
T H E ARRANGEMENT O F AMINO ACIDS I N PROTEINS
59
7. Hemoglobin Porter and Sanger (1948) determined the N-terminal residues of hemoglobins from a number of different animal species. These results are summarized in Table I. Considerable species differences were evident both in the nature of the N-terminal residues and in the number of open polypeptide chains present. I n the case of horse hemoglobin which was studied in more detail, there are six valyl N-terminal residues and therefore six open polypeptide chains. It is interesting to note th a t the cystine content is not more than three residues, so that some of the chains must be held together by another type of cross-linkage. An investigation of the N-terminal peptides by the DNP-technique showed the presence of the following N-terminal sequences : Val.Leu, Val.Glu (--NH2) .Leu Thus there are a t least two different types of polypeptide chains present in the molecule (Sanger, 1948).
8. Gelatin Gordon et al. (1943) attempted to fractionate the acetyl derivatives of peptides from gelatin on silica gel chromatograms. The hydrolyzate was separated into neutral and basic fractions by ionophoresis in a three compartment cell. The neutral peptides were then fractionated on a silica chromatogram using ethyl acetate, and each fraction was refractionated using butanol-chloroform. I n general the fractionation was not sufficient t o give clear-cut results though the presence of the following peptides was suggested : Leu.Gly, Gly.Leu, Gly.Pro, a dipeptide containing proline and alanine, and a t tripeptide containing proline, alanine and glycine. These investigations are largely of historical interest, since they mere carried out before the introduction of paper chromatography. The amino acids present in the peptides were identified and estimated as their acetyl derivatives on silica chromatograms. VIII. GENERALCONCLUSIONS It is clear from the above that considerable progress has been made in recent years in the development of methods for investigating the arrangement of amino acid residues in proteins. These methods have not yet been extensively applied and a vast amount of work is still required in this field. Such work may be expected t o be rather unrewarding a t first. The separation of a few peptides from a protein is not likely to
60
F. SANGER
make possible the formulation of any general theory of protein structure or to explain the physiological action of a protein. Only by the accumulation of a large amount of experimental evidence can such objectives be attained. It does not appear, however, a t present th a t there is any easy short cut t o the solution of the problem of protein structure and action, Probably only when more is known about the exact chemical structure of proteins will it be possible to understand the unique part played by proteins in the living organism. Every peptide identified will contribute towards this ideal, even though it may appear to have no significance in itself. By the application of the methods described in this review the structure of a pure polypeptide containing thirty residues has been determined (p. 54) and there seems no reason why it should not be possible to work out the complete amino acid sequence in proteins which are as simple as insulin. How far it will be possible t o apply these techniques t o more complex proteins is difficult to say. The larger the polypeptide chains in a protein, the greater the necessity of isolating larger peptide breakdown products. Probably the chief need in this field is for techniques for the specific breakdown of proteins into larger peptides and for the fractionation of such peptides. Most of the more commonly studied proteins contain more than 300 residues but it is possible that some of them, when studied in greater detail may be found to have a simpler structure than is a t present believed. The relative simplicity of insulin may be merely apparent as insulin has been studied in more detail than have other proteins. It has frequently been suggested that proteins may not be pure chemical entities but may consist of mixtures of closely related substances with no absolute unique structure. The chemical results so far obtained suggest that this is not the case and that a protein is really a single chemical substance, each molecule of one protein being identical with every other molecule of the same pure protein. Thus it was possible to assign a unique structure to the phenylalanyl chains of insulin. Each position in the chain was occupied by only one amino acid and there was no evidence that any of them could be occupied by a different residue. Whether this is true for other proteins is not certain but it seems probable that it is. The N-terminal residues of several pure proteins have been determined (Table I) and this position is always found t o be occupied by a single unique amino acid. These results would imply a n absolute specificity for the mechanisms responsible for protein synthesis and this should be taken into account when considering such mechanisms. It is impossible with the small amount of experimental evidence a t present available to form any general theory of protein structure or to
THE ARRANGEMENT OF AMINO ACIDS IN PROTEINS
61
formulate any principles that govern the arrangement of amino acids in proteins, though several such theories have been put forward on very much less evidence. Certainly there is no simple periodic arrangement of residues along the chains of the type suggested by Bergmann and Niemann (1936). The presence of peptides such as Glu.Glu in a hydrolyzate of wool (Consden et al., 1949) precluded this possibility and no periodicity whatsoever was evident in the structure of the phenylalanyl chains of insulin (Table XI). Although they are not immediately apparent, it still seems probable that there may be certain principles which determine amino acid sequences. The mechanisms of protein synthesis, about which almost nothing is known, would be expected to have their limitations so that one might at least expect to find certain sequences that occur more frequently than others. The results at present available do suggest that this may be the case. Thus, for instance, in the phenylalanyl chain of insulin there are three dipeptide sequences (His.Leu, Leu.Va1, CyS.Gly) that occur twice in the chain of thirty residues. One of these (CyS.Gly) also occurs in glutathione. Other dipeptide sequences in this chain were Thr.Pro, which was also detected in the antibiotic actinomycin (Dalgliesh et al., 1950) and Ala.Leu, which was found in gramicidin (Synge, 1949). The sequence Glu.Glu occurs both in wool (Consden et al., 1949) and in insulin (Sanger, 1949b) and is probably also present in gliadin, since Nakashima (1927) obtained a peptide fraction that contained tyrosine and glutamic acid in the ratio 1 :3. It may be that this sequence has a special significance since glutamic acid and glutamine residues frequently occur linked together in natural products, such as folic acid, the capsular substance from B. Anthracis (Hanby and Rydon, 1946) and in the derivative of triglutamine isolated by Dekker et al. (1949) from a marine alga. In this connection Woolley (1949) has suggested that the same amino acid sequence occurs both in insulin and in trypsinogen, since two similar fractions were obtained from tryptic digests of the two proteins. How far these results do reflect a general principle of protein structure will only be known when considerably more experimental evidence is available. The results obtained with wool (p. 48) on the contrary suggest that almost every possible dipeptide containing glutamic acid or cystine is produced on hydrolysis of this protein and that there are therefore no obvious limitations to the type of sequences that can occur. It is certain that proteins are extremely complex molecules but they are no longer completely beyond the reach of the chemist, so that we may expect to see in the near future considerable advances in our knowledge of the chemistry of these substances which are the essence of living matter.
62
F. SANGER
ACKNOWLEDGMENTS I wish to express my thanks to Dr. A. C. Chibnall, Dr. R. L. M. Synge, and Dr. H. Tuppy for valuable discussions and criticisms which have greatly helped in the preparation of the present review.
REFERENCES Abderhalden, E. (190%). 2.physiol. Chem. 62, 315. Abderhalden, E. (1909b). 2.physiol. Chem. 63,401. Abderhalden, E. (1940). 2.physiol. Chem. 266, 23. Abderhalden, E. (1943). Z. physiol. Chem. 277, 248. Abderhalden, E., and Bahn, A. (1932). 2. physiol. Chem. 210, 246. Abderhalden, E., and Bahn, A. (1933). 2. physiol. Chem. 219, 72. Abderhalden, E., and Bahn, A. (1935). 2. physiol. Chem. 234, 181. Abderhalden, E., and Brockmann, H. (1930). Biochem. 2. 226,386. Abderhalden, E., and Komm, E. (1924a). 2. physiol. Chem. 134, 121. Abderhalden, E., and Komm, E. (1924b). Z. physiol. Chem. 139, 147. Abderhalden, E., and Mahn, H. (1927). 2. physiol. Chem. 169, 196. Abderhalden, E., and Mahn, H. (1928). 2. physiol. Chcm. 174,47. Acher, R., Jutisz, M., and Fromageot, C. (1950). Biochim. et Biophys. Acta 6,493. Agren, G. (1940). Arkiu. Kerni Mineral. Geol. 14B,No. 21. Bailey, K. (1951). Biochern. J . 49, 23. Barry, G.T., Gregory, J. D., and Craig, L. C. (1948). J . Biol. Chem. 176, 485. Behrens, 0. K., and Bergmann, M. (1939). J . B i d . Chem. 129, 587. Bell, P.H., Hone, J. F., English, J. P., Fellows, C. E., Howard, K. S., Rogers, 31. M., Shephard, It. G . , and Winterbottom, R. (1949). Ann. N . Y . Acad. Sci. 61,8!17. Bergmann, M. (1942). Advances in Enzyrnol. 2, 49. Bergmann, M., and Behrens, 0. K. (1938). J . Bid. Chem. 124, 7. Bergmann, M.,and Fraenkel-Conrat, H. (1937). J . Biol. Chem. 119, TOT. Bergmann, M., and Fraenkel-Conrat, H. (1938). J . Bid. Chem. 124, 1 . Bergmann, M., and Fruton, J. S. (1938). J . Biol. Chem. 124,321. Bergmann, M., and Fruton, J . S. (1941). Advances in E’nzywiol. 1, 63. Bergmann, M., and Miekeley, A. (1924). 2. physiol. Cheni. 140, 128. Bergmann, M., and Niemann, C. (1936). J . B i d . Chem. 116,77. Biserte, G. (1950). Biochim. el Biophys. Acta 4, 416. Blackburn, S. (1949). Biochem. J . 46, 579. Blackburn, S.,and Lowther, A. G. (1950). Biochem. J . 46, sxvii. Block, R. J. (1942). Proc. SOC.Exptl. Biol. Med. 61,252, Block, R. J. (1949). I n Ion Exchange, edited by Nachod, F. C. Academic Press, New York, p. 295. Boissonnas, R. A. (1950). Helv. C h i m Acla 33, 1975. Boldingh, J. (1948). Experientia 4, 270. Bowman, R. E. (1960). J . Chem. SOC.1349. Brand, E., and Edsall, J. T. (1947). Ann. Rev. Biochem. 16, 224. Brand, E., and Kassell, B. (1939). J . Biol. Chem. 131,489. Brigl, P. (1923). Ber. 66, 1887. Bull, H. B., and Hahn, J. W. (1948). J . Am. Chern. SOC.70, 2128. Butler, J. A. V., Dodds, E. C., Phillips, D. M. P., and Stephen, J. M. L. (1948). Biochem. J . 42, 116.
THE ARRANGEMENT O F AMINO ACIDS I N PROTEINS
63
Butler, J. A. V., Phillips, D. M. P., Stephen, J. M. I,., and Creeth, J. M. (1950). Biochem. J . 46, 74. Cannan, R. K. (1944). J. Biol. Chem. 162,401. Carpenter, D.C. (1940). J . Am. Chem. SOC.62,289. Carpenter, D. C. (1941). J . Franklin I n s t . 232, 76. Chibnall, A. C., and Rees, M. W. (1951). Biochem. J . 48, xlvii. Christensen, H. S. (1943). J . Biol. Chem. 161,319. Christensen, 11. S. (1945). J. Biol. Chern. 160, 75. Christensen, I-I. N., and Hegsted, D. M. (1945). J . Biol. Chem. 168, 593. Coleman, D., and Howitt, F. 0. (1946). Fibrous Proteins. Society of Dyers and Colourists, &adford, p. 144. Coleman, D., and Howitt, F. 0. (1947). Proc. Roy. SOC.London A190, 145. Consden, R. (1948). Nature 162, 359. Consden, It., and Gordon, A. H. (1950). Biochem. J . 46, 8. Consden, R., Gordon, A. H., and Martin, A. J. P. (1944). Biochem. J . 38, 224. Consden, R., Gordon, A. H., and Martin, A. J. P. (1946). Biochem. J . 40, 33. Consden, R., Gordon, A. H., and Martin, A. J. P. (1947a). Biochem. J . 41, 590. Consden, It., Gordon, A. H., and Martin, A. J. P. (1948). Biochem. J . 42, 443. Consden, R.,Gordon, A. H., and Martin, A. J. P. (1949). Biochem. J . 44, 548. Consden, R., Gordon, A. H., Martin, A. J. P., and Synge, R. L. M. (194713). Biochem. J . 41, 596. Craig, L. C. (1944). J. Biol. Chem. 166, 519. Craig, L.C., Gregory, J. D., and Barry, G. T. (1949). Cold Spring Harbor Symposia Quant. Biol. 14,24. Craig, L. C., and Post, 0. (1949). Anal. Chem. 21, 500. Cremer, H. D., and Tiselius, A. (1950). Biochem. Z. 320, 273. Dalgliesh, C.E., Johnson, A. W., Todd, A. R., and Vining, L. C. (1960). J . Chern. SOC.2946. Dekker, C. A., Stone, D., and Fruton, J. S. (1949). J . Biol. Chem. 181, 719. Desnuelle, P., and Casal A. (1948). Biochim. et Biophys. A d a 2, 64. Desnuelle, P., Rovery, M., and Bonjonr, G. (1950). Bzochim. et Biophys. Acta 6, 116. Dim, K., and Felix, K. (1932). 2. physiol. Chem. 206, 83. Drake, B. (1947). Nature 160,602. Drake, B. (1950). Acta Chem. Scand. 4, 554. Drucker, B.,and Smith, S. G. (1950). Nature 166, 196. Dunn, M. S. (1925). J. Awl. Chem. SOC.47,2564. Durrum, E. L. (1950). J . Am. Chern. SOC.72, 2943. Edman, P. (1950). Acta Chem. Scancl. 4, 283. Eeg-Larsen, N., Linderstr@m-Lang, I<., and Ottesen, M. (1948). Arch. Biochem. 19,340. Elsden, S. R., and Synge, R. I,. hl. (1944). Biochem. J. 38, ix. Felix, K., Fischer, H., Krekels, A , , and Rauen, H. M. (1950). 2. physiol. Chem. 286, 67. Felix, K., and Lang, A. (1929). Z. physiol. Chem. 182, 125. Felix, K., and Mager, A. (1937). Z.physiol. Chem. 249, 111. Felix, K., and Schuberth, H. (1942). Z. physiol. Chem. 273,97. Fischer, E., and Abderhalden, E. (1907). Ber. 40, 3544. Fodor, A., Fodor, P. J., and Kuk-meiri, S. (1947). Enzymologia 12, 101. Fowden, L., and Penney, J. R. (1950). Nature 166, 846.
64
F. SANGER
Fox, S. W. (1945). Advances i n Protein Chem. 2, 155. Fromageot, C., Jutisz, M., Meyer, D., and Pknasse, L. (1950). Biochim. et Biophys. A d a 6, 283. Fruton, J. S. (1950). Yale J . Biol. Med. 22, 263. Goldschmidt, S., Freyss, G., and Strauss, K. (1933). Ann. 606, 262. Gordon, A. H. (1949). Discussions of the Faraday SOC.7, 128. Gordon, A. H., Keil, B., and Sebesta, K. (1949). Nature 164, 498. Gordon, A. H., Martin, A. J. P., and Synge, R. L. M. (1941). Biochem. J . 36, 1369. Gordon, A. H., Martin, A. J. P., and Synge, R. L. M. (1943). Biochem. J . 37, 92. Grant, R. L., and Lewis, H. B. (1935). J . Bid. Chem. 108, 667. Grassmann, W. (1950). Angew. Chem. 62, 170. Grassmann, W., and Hannig, K. (1950). Nat,uru,issenschajten 37, 397. Green, F. C., and Schroeder, W. A. (1951). J . Am. Chem. SOC.73, 1385. Green, N. M., and Work, E. (1951). Biochem. J . 49, xxxvii. Gregory, J. D., and Craig, L. C. (1948). J . Biol. Chem. 172, 839. Guntelberg, A, V., and LinderstrZm-Lang, K. (1949). Compt. rend. lab. trav. Carlsberg. Sdr. chim. 27, 1. Gutfreund, H. (1948). Biochem. J. 42, 544. Gutfreund, H., and Ogston, A. G. (1949). Biochem. J. 44, 163. Hamoir, G. C. M. (1945). Biochem. J . 39, 485. Hanby, W. E., and Rydon, H. N. (1946). Biochem. J . 40, 297. Harington, C. R., and Pitt-Rivers, R. (1944). Natiire 164, 301. Haugaard, G., and Kroner, T. D. (1948). J . Am. Chem. SOC.70, 2135. Haugaard, G., and Roberts, R. M. (1942). J . Am. Chem. Sor. 64, 2664. Hellerman, L., and Stock, C. C. (1938). J . Biol. Chem. 126, 771. Howard, G. A., and Martin, A. J. P. (1950). Biochem. J . 46, 532. Huang, H. T., and Niemann, C. (1950). J . Am. Chem. SOC.72, 921. Huffman, H. M. (1942). J . Phys. Chem. 46, 885. Ingram, V. M. (1950). Nature 166, 1038. Jacobsen, C. F. (1949). Compt. rend. lab. trav. Carlsberg. Shr. chini. 26, 455. Jensen, H., and Evans, E. A. (1935). J . Bid. Chem. 108, 1. Johnston, R. B., Mycek, M. J., and Fruton, J. S. (1950). J . Biol. Chem. 186, 629. Jones, T. S. G. (1948). Biochem. J . 42, xxxv. Jones, T. S. G. (1949). Discussions of the Faraday Sor. 7, 285. Jutisz, M., and Lederer, E. (1947). Nature 169, 445. Kaplan, E.H., Campbell, E. D., and McLaren, A. D (1950). Biochim. et Biophys. Acta 4, 493. Keston, A. S., and Udenfriend, S. (1949). Cold Spring Harbor Symposia Quant. Biol. 14, 92. Keston, A. S., Udenfriend, S., and Cannan, R. K. (1946). J . Am. Chem. SOC.68, 1390.
Keston, A. S., Udenfriend, S., and Cannan, It. K.
(1049). J . Anz. Chem. SOC.71,
249.
Keston, A. S., Udenfriend, S., and Levy, M. (1947). J . A m . Chenz. SOC.69, 3151. Keston, A. S., Udenfriend, S., and Levy, M. (1950). J . A m . Chem. Soc. 72, 748. Kies, M. W., and Davis, P. L. (1950). Federation Proc. 9, 189. Kuhn, W. (1930). Ber. 63, 1503. Kuhn, W.,Molster, C. C., and Freudenberg, K. (1932). Ber. 66, 1179. Lederer, E., and Tchen, P. K. (1947). Biochim. et Biophys. A d a 1, 35. Lens, J. (1949). Biochim. et Biophys. A d a 3, 367.
THE ARRANGEMENT O F AMINO ACIDS IN PROTEINS
65
LBonis, J. (1948). Compt. rend. lab. trav. Carlsberg. Sdr. chim. 26, 315. Levene, P. A., and Bass, L. W. (1928). J . Biol. Chem. 78, 145. Levene, P. A., and Bass, L. W. (1929). J . Biol. Chem. 82, 171. Levene, P. A., Rothen, A., Steiger, R. E., and Osaki, M. (1930). J . Biol. Chem. 86, 723. Levene, P. A., Steiger, R. E., and Rothen, A. (1932). J . Biol. Chem. 97, 717. Levy, A. L. (1960). J . Chem. SOC.404. Levy, M., and Slobodiansky, E. (1929). Cold Spring Harbor Symposia Quant. Biol. 14, 113. Linderstrgim-Lang, I<. (1949). Cold Spring Harbor Symposia Quant. Biol. 14, 117. Linderstrflm-Lang, K., and Ottesen, M. (1949). Compt. rend. lab. trav. Carlsberg. Sbr. chim. 26, 403. Livermore, A. H., and du Vigneaud, V. (1949). J . Biol. Chem. 180, 365. Lugg, J. W. H. (1938). Biochem. J . 32, 775. McLaren, A. D. (1949). Advances in Enzymol. 9, 75. Mandl, L., Levy, B., and McLaren, A. D. (1950). J . Am. Chem. SOC.72, 1790. Martin, A. J. P. (1949a). Biochem. SOC.Symposia 3, 4. Martin, A. J. P. (194913). Discussions of the Faraday SOC.7, 332. Martin, A. J. P. (1950). Ann. Rev. Biochem. 19, 517. Martin, A. J. P., and Mittelmann, R. (1948). Biochem. J . 43,353. Martin, A. J. P., and Synge, R. L. M. (1941). Biochem. J . 36, 1358. Martin, A. J. P., and Synge, R. L. M. (1945). Advances in Protein Chem. 2, 1. Meyer, K. H., Fuld, M., and Klemm, 0. (1940). Helv. Chim. Acta 23, 1441. Middlebrook, W. R. (1949). Nature 164, 501. Miller, G. L., and Andersson, K. J. I. (1942). J . Biol. Chem. 144, 465. Mitchell, H. K., Gordon, M., and Haskins, F. A. (1949). J . Biol. Chem. 180, 1071. Mitchell, H. K., and Haskins, F. A. (1949). Science 110, 278. Monnier, R., and Jutisz, M. (1950). Bull. SOC. chim. biol. 32, 228. Monnier, R., and Penasse, L. (1950). Compt. rend. 230, 1176. Montroll, E. W., and Simha, R. (1940). J . Chem. Phys. 8, 721. Moore, S., and Stein, W. H. (1948). J . Biol. Chem. 176, 367. Moore, S., and Stein, W. H. (1949). J . Biol. Chem. 178, 53. Moring-Claesson, I. (1948). Biochim. et Biophys. Acta 3, 389. Myrback, K. (1949). Arkiv. Kemi 1, 161. Nakashima, R. (1927). J . Biochem. (Japan). 7, 441. Neurath, H., and Schwert, G. W. (1950). Chem. Revs. 46, 69. Northrop, J. H., Kunitz, M., and Herriott, R. M. (1948). Crystalline Enzymes, 2nd ed., New York, p. 73. Partridge, S. M. (1948). Biochem. J . 43, 238. Partridge, S. M. (1949a). Discussions of the Faraday SOC.7, 296. Partridge, S. M. (194913). Biochem. J. 44, 521. Partridge, S. M. (1949~). Biochem. J. 46, 459. Partridge, S. M., and Davis, H. F. (1950). Nature 166, 62. Partridge, S. M., and Swain, T. (1950). Nature 166, 272. Partridge, S. M., and Westall, R. G. (1949). Biochem. J . 44, 418. Patton, A. R., Foreman, E. M., and Wilson, P. C. (1949). Science 110, 593. Pauling, L., and Niemann, C. (1939). J . Am. Chem. SOC.61, 1860. Perlmann, G. E. (1949). J . Am. Chem. Sac. 71, 1146. Petermann, M. L. (1942). J . Biol. Chem. 144, 607. Petermann, M. L., and Pappenheimer, A. M. (1941). J . Phys. Chem. 46, 1.
66
F. SANGER
Phillips, D. M. P. (1948). Nature 161,53. Phillips, D.M. P. (194%). Biochim. et. Biophys. Acta 3, 341. Phillips, D. M. P. (1949b). Nature 164, 545. Pittom, W.W. P. (1914). Biochem. J. 8, 157. Polson, A. (1948). Biochim. et Biophys. Acta 2, 575. Porter, R. R. (1948). Biochim. et Biophys. Acta 2, 105. Porter, R. R. (1950a). Biochem. J. 46, 473. Porter, R. R. (1950b). Biochem. J. 46, 479. Porter, R. R. (1950~). Methods in Medical Research 3, 25G. Porter, R. R., and Sanger, F. (1948). Biochem. J. 42, 287. Randall, S. S., and Martin, A. J. P. (1949). Biochem. J. 44, ii. Rauen, H. M., and Felix, K. (1948). 2. physiol. Chem. 283, 139. Rees, M. W. (1946). Biochem. J. 40, 632. Rideal, E.K., and Mitchell, J. S. (1937). Proc. Roy. SOC.London A169, 206. Rittenberg, D., and Foster, G . L. (1940). J . Biol. Chem. 133, 737. Rometsch, R. (1950). Helv. Chim. Acta 33, 184. Sanger, F. (1945). Biochem. J. 39, 507. Sanger, F. (1946). Biochem. J. 40, 261. Sanger, I?. (1948). Nature 162,491. Sanger, F. (1949a). Biochem. J. 44, 126. Sanger, F. (1949b). Biochem. J. 46, 563. Sanger, F. (1049~). Cold Spring Harbor Symposia Quant. Biol. 14, 153. Sanger, F. (1949d). Hiochem. Soc. Symposia 3, 21. Sanger, F., and Tuppy, H. (1951a). Biochem. J. 49, 463. Sanger, F.,and Tuppy, H. (1951b). Biochem. J. 49, 481. Schmid, K. (1949). Helv. Chim. Acta 32, 105. Schramm, G.,and Primosigh, J. (1943). Ber. 76, 373. Schramm, G., and Primosigh, J. (1944). Ber. 77, 417. Shepard, C. C., and Tiselius, A. (1949). Discussions of the Faraday SOC.7,275. Stein, W. H., and Moore, S. (1948). J . Biol. Chem. 176, 337. Stein, W. H., and Moore, S. (1949). Cold Spring Harbor Symposia Quant. Biol. 14, 179. Stein, W. H., Moore, S., and Bergmann, M. (1944). J . Biol. Cheni. 164, 191. Steinhardt, J. (1941). J. Biol. Chem. 141,996. Steinhardt, J., and Fugitt, C . If. (1942). J . Research Natl. Biir. Standards 29, 315. Steinhardt, J., and Zaiser, E. M. (1950). J . Biol. Chem. 183, 789. Svensson, H. (1948). Advances in Protein Chem. 4, 251. Svensson, H.,and Brattsten, I. (1949). Arkiv Kemi 1, 401. Swanson, M. A., and Cori, C. F. (1948). J . Riol. Chem. 72, 797. Synge, R. 1,. M. (1943). Cheni. Revs. 32, 135. Synge, R. L. M. (1944). Biochem. J. 38, 285. Synge, R. L. M. (1945). Biochem. J . 39, 351. Synge, R. 1,. M. (1949). Biochem. J . 44, 542. Synge, R. L. M., and Tiselius, A. (1947). Acta Chem. Scand. 1, 749. Synge, R. 1,. M., and Tiseiius, A. (1949). Acta Chem. Scand. 3, 231. Synge, It. L. M., and Tiselius, A. (1950). Biochem. J. 46,xli. Taeawa, Y. (1949). Acta Phytochim. (Japan) 16,73. Tiselius, A. (1947). Advances i n Protein Chem. 3, 67. Tiselius, A. (1948): Arkiv Kemi Mineral. Geol. 26B, No. 1. Tiselius, A , , and Claesson, S. (1942). Arkiv Kemi Mineral. Geol. 16B,No. 18.
T H E ARRANGEMENT O F AMINO ACIDS I N P R O T E I N S
67
Tiselius, A., Drake, B., and Hagdahl, L. (1947). Experientiu 3, 651. Tiselius, A,, and Hagdahl, L. (1950). Acta Chem. Scand. 4,394. Toennies, G., and Homiller, R. P. (1942). J . Am. Chem. SOC.64,3054. Tristram, G. R. (1949). Advances in Protein Chem. 6, 83. Troensegaard, N. (1947). Acta Chem. Scand. 1, 672. Turba, F. (1948). 2. Vitamin-, Hormon- u. Fermentforsch. 2, 49. Turba, F., and Enenkel, H. J. (1950). Naturwissenschuften 37, 93. Turba, F.,and Richter, M. (1942). Ber. 76,340. Uchino, T. (1934). J. Biochem. (Japan) 20, 65. van Slyke, D. D., Dillon, R. J., MacFadyen, D. A., and Hamilton, P. (1941). J. Biol. Chem. 141, 627. van Vunakis, H., and Brand, E. (1951). Abst,racts, 119th Meeting, American Chemical Society, April, 1951, p. 28c. Vickery, H. B. (1922). J . Biol. Chem. 63,495. Vickery, H.B., and Osborne, T. B. (1928). Physiol. IZevs. 8,393. Villee, C. A,, Linderstrbm-Lang, Ti., and Ottesen, hl. (1950). Federation Proc. 9, 241. Virtanen, A. I., and Hamberg, U. (1947). Acta Chem. Scand. 1, 847. Virtanen, A. I., Kerkkonen, H., Hakala, M., and Laakonen, T. (1950). Xatwwissenschaften 37, 139. Waldschmidt-Leitz, E., and Turba, F. (1940). J . prakt. Chem. 166, 55. Waldschmidt-Leitz, E., and Turba, F. (1941). J . prakt. Chem. 168, 72. Warner, It. C. (1942a). J. Biol. Chem. 142,705. Warner, R.C. (1942b). J . Biol. Chem. 142,741. Wasteneys, H.,and Borsook, H. (1930). Physiol. Revs. 10, 110. Wieland, T. (1942). 2. physiol. Chem. 273,24. Wieland, T. (1944). Ber. 77, 539. Wieland, T., and Fischer, E. (1948). Naturwissenschaften 36, 29. Wieland, T., and Wirth, 1,. (1949). Ber. 82, 468. Woiwod, A. J. (1949). Biochem. J . 46, 412. Woolley, D. W. (1949). J . Riol. Chem. 179, 593.
This Page Intentionall Left Blank
The Structure of Collagen Fibrils
BY RICHARD S. BEAR Department of Biology, Massachusetts Institute of Technology, Cambridge, Massachusetts
CONTENTS I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1. Significance of Collagen.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. Hierarchies of Structural Element., . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3. Methods.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11. Identification and Distribution of Collagens. . . . . . . . . . . . . . . . . . . . . . . . . . . . 1. General Characteristics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. Physical Criteria for Classification.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3. Types of Mesenchymal Collagen. . . . . . . . . . . . . . . . . . . ............. 4. The “Secreted” Collagens.. . . . . . . . . . . . . . . . . . . . . . . . ............. 5. The Chemical Significance of the Physical Classification.. . . . . . . 111. Colloidal Structure of Collagen Fibrils. . . . . . . . . . . . . . . . . . . . . . . . . . . 1. Optical Evidence 2. A Fibrillar Mode 3. Diffraction Theory for 4. Electron Optical Evide IV. The Collagen Protofibril. . . ......... 1 . General Characteristics 2. Pseudoperiodic Axial Structure of Small Size.. . . . . . . . . . . . . . . . . . . . . . . . 3. Structure Transverse to Fibril Axes; Hydration.. . . . . . . . . . . . . . . . . . . . . 4. Protofibril Extensibility, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5. Protofibril Contractility; Thermal Contraction. . . . . . . . . 6. The Content per Unit of Protofibrillar Length.. . . . . . . . . . . . . . . . . . . . . . . V. The Collagen Molecule.. ......................................... 1. Description of the Molecule., . . . . . . . . . . . . . . . 2. Distribution of Side Chains. . . . . . . . . . . . . . . . 3. Swelling Phenomena. . . . . . . . . . . . . . . . . . . . . . . ................
Page 69 69 71 73 76 76 78 84 90
112 115 117 119 127 135 146
........................ . . . . . . . . . . . . . . . . 150 2. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
I. INTRODUCTION 1. SigniJicance of Collagen
The collagenous tissues of animals are in some respects similar to the cellulosic structures of plants. As a major fibrous constituent of skins, 69
70
RICHARD S. BEAR
bones, tendons, fascia, and other types of loose and dense connective tissue occurring in practically all body organs (149) collagen is undoubtedly the most prevalent single organic constituent of animal organisms. Medical concern with collagen is derived from its importance as a constituent of tissues intimately involved in aspects of the aging process, in neoplastic growth, in diseases connected with vitamin deficiencies or hormonal imbalances, and, indeed, in many general systemic disorders involving the connective tissues in a diffuse or inflammatory way (124). As the principal constituent of animal skins, collagen is manipulated in the tanning of leather. Gelatin and glue are well known degradation products, both of which find wide spread industrial use, the former, for example, being indispensable to the production of photographic emulsions. The immunological inertness of collagen gives rise to its usefulness in surgical sutures. h recent bibliography of the literature of collagen (37) lists o \ w a thousand references. h’otwithstanding this volume of activity the state of fundamental concepts in the field is unsatisfactory. The present review summarizes some of the basic information supplied by X-ray diffraction, electron microscopy, and chemical studies, developing a viewpoint regarding the structure of the ultramicroscopic element of all collagenous substances, the collagen fibril. I n thus narrowing the present subject many medically and commercially important aspects will not be immediately served, for these often depend on tissue organization, which involves interrelation between collagen and other components. A perusal of modern literature dealing with the physiology and pathology of connective tissues (GO, 124) leaves the general impression that the fibrous components, collagen and elastin, are proving of less immediate importance in problems of crucial medical concern than are the mucopolysaccharide-containing “ground substance l 1 and the cellular elements. Kevertheless, the ubiquitous (204) collagen fibrils can scarcely fail to affect in fundamental ways the avenues of movement and growth open to cells, as well as to act as barriers or filters screening the metabolites passing between individual cells, the blood and lymph. Lloyd (138) has pointed out that, on the basis of the rough correlation between the water contents of tissue elements and their biological or biochemical activity, the collagen fibril is one of the more moribund of tissue elements, being chemically designed, in fact, for stability in normal physiological environments. As the present discussion advances it becomes clear that a collagen fibril is essentially a long, thin, single “crystal”; as such its formation or dispersion, synthesis or degradation, is expected t o be largely at the mercy of its surroundings. Nevertheless,
THE STRUCTURE OF COLLAGEN FIBRILS
71
it is capable of more variation than is met with in simpler crystalloids, so th a t when methods for the examination of its chemical and physical structure are refined, these variations may furnish information concerning the physiological milieu from which it was derived. I n order that these practical aims may be served it is necessary first to develop in as much detail as possible the structure of normal collagen fibrils. A wide variety of these objects are available. They provide particularly suitable material for structural studies, which should furnish important bases from which t o attempt extrapolation to protein systems of more distinct biological activities.
2. Hierarchies of Structural Element
To make clear the aspects of structure herein treated, a few definitions are of importance a t the outset. Figure 1 indicates the major structural elements found in connective tissues, as well a s the correspondence between these elements and the methods applicable for their resolution and study. The terminology adopted is similar to th a t employed by Kuntzel (131), who described as fibers the microscopically visible aggregates with diameters of 100 to 200 p in tendon, 20 to 40 u in skin. Smaller components are primitive fibers ( 2 to 10 u ) . The latter may be further subdivided by mild mechanical or chemical means into fibrils whose diameters range from figures close to the resolution of light microscopes down to hundreds of Angstroms and require the electron microscope for adequate resolution. While the possibilities for longitudinal cleavage do not stop with the fibril, it is generally a fairly coherent unit whose properties provide the ultramicroscopic basis for collagenous tissue elements. I n fact, since the grosser fibers may contain also cellular, polysaccharide, and elastic components, it is with the fibril that one first expects t o approach something like a “pure” collagen. When thinner, subfibrillar elements are discernible, these are termed generally $laments, and, as Schmitt et a l . (192) further suggest, it is useful to consider a hypothetical thinnest unit filament, the protojibrill which carries the essential chemical and configurational structure of collagen. A collagen molecule would be smaller, though perhaps not thinner, than the protofibril. It is suggested below th at the protofibril is no wider than a very few polypeptide chains (probably one, diameter ca. 12 to 17 A), so that chain termini (defined as the points where primary valence connections between atoms become lacking or weak) will mark off molecular boundaries along the protofibril. The classical term, micelle, is no longer very useful in describing for collagen the volume over which the quasi-crystalline arrangement of intrafibrillar structure extends to produce the oriented X-ray diffraction
72
RICHARD S. BEAR
FIQ.1. See opposite page for legend.
T H E STRUCTURE O F COLLAGEN FIBRILS
73
effects that are observed. The fibril turns out to have diffracting structures-within-structures, of sizes ranging from hundreds to a few Angstroms. The micelle for large features of structure would be the entire fibril, or much of it, while the micelle of the smaller details would approach classical significance, including a length much less than th a t of the molecule but a width which is that of a few clustered protofibrils. There are certain descriptive terms which apply to intrafibrillar volumes whose boundaries cut across protofibrillar or molecular limits. These terms are useful, particularly in describing the appearance of fibrils in electron micrographs. Therein one observes bands, which are disk-like enlargements of fibrillar diameter with greater than average power to deflect or absorb electrons and to take u p electron stains. T h e bands form a characteristic pattern of cross striations which periodically repeat along the fibril axis. The fibrillar levels remaining between bands may be called interbands. The discussion which follows deals with the means of recognizing fibrils of the collagen type and with the description and significance of the subfibrillar entities just enumerated.
3. Methods For present purposes electron microscopy is particularly useful because it is the only available method for directly observing individual fibrils with any degree of adequacy. Th e X-ray diffraction method offers the next most direct approach, since, although the results are obtained from relatively large fiber bundles (diameters of about 1 mm.), the diffraction effects are the summation of the effects of individual fibrils or subfibrillar components. One takes precautions t o see th a t the fibrils are as nearly parallel as possible, and if individuals are not too different in structure, the entire diffraction pattern becomes essentially t h a t of a single fibril. The diffraction method, while suffering from certain inherent limitations of interpretation, offers the greatest resolving power for the elucidation of configuration th at has been applied to collagen as yet. The X-ray information is obtained from diffraction patterns which FIG.1. Illustrating the various hierarchies of element involved in present considerations. A t the top are shown the distinctive components of connective tissue. The ground substance, which contains the mucopolysaccharides, hyaluronate and chondroitin sulfate, has been studied chemically by Karl Meyer (150, 151) and by K. H. Meyer (158),among others, and electron optically by Gross (87), b u t this component does not receive significant further attention herein. Occasional reference is made to elastin, but the text considers principally the collagenous elements, shown in steps of increasing magnification.
74
RICHARD S. BEAR
represent a peculiar sort of inverted image of structure. Since many readers will not be accustomed to analysis of these diffraction images, Fig. 2 is given in explanation of some of the unavoidable technical terms.
FIG.2. Illustrating the X-ray diffraction method as applied to collagenous fibers. Electrons, e, emitted hy an incandescent filament, F , are driven by *. pofmtiLt1 of about 40,000 volts onto the target of copper, Cu, which yields the X-radiation that emerges from the X-ray tube through a window of beryllium, Be. A thin foil of nickel, N i , transmits principally the characteristic Cu Ka radiation (wavelength A = 1.54 A . ) to the collimating system, which consists of fine pinholes cut in lead disks, Pb. Beyond the collagen fibw a t C emerge the undiffracted beam lf and various diffracted or deviated beams represented by the one marked D. The relatively intense ccntral beam is attenuated by the beam trap, Y,whose shadow is wen overlaying the center of each of the diffraction patterns of Figs. 3 through 20. The final diffraction pattern, as registered on a photographic film consists of layer lines (vertical dashed lines) and row lines (horizontal dotted lines) of spots. Prominent features of the collagen wide-angle pattern are indicated by M , the important 2.86 A arc on the meridian; E, the 10 to 12 A. equatorial spot; and H , the diffuse “half-halo” (4.6 A.) referred to in the text. The small-angle layer lines a t S generally require finer pinholes and longer collimators and specimen-to-film distances to permit their resolution. In some cases collimating apertures are horizontal slits, which diffuse the diffraction spots along layer lines but retain layer-line sharpness vertically and permit more rapid registration of patterns.
There are two general size ranges which the diffraction method can cover. One includes sizes approximating those of atoms, atom groups, amino-acid residues, and the thinner dimension of collagen protofibrils and molecules (20 A. or less) ; these quantities are investigated at wide diffraction angles, just as the large apertures of high-power objectives in
THE STRUCTURE O F COLLAGEN FIBRILS
75
ordinary microscopy are employed for the resolution of small details. The second size range (20 to 1000 A.) overlaps with that of the electron microscope in indicating large fibrillar or subfibrillar features of structure; the diffraction angles are small (up t o about 4" with Cu K a radiation of wavelength 1.54 A.), corresponding to the small apertures ordinarily used by lower-power microscope objectives. On either type of diffraction pattern the individual diffraction spots can be used t o calculate spacings between planes of equivalent structure in the specimen. The spots often occur in row or layer lines as shown in Fig. 2 . Particular importance attaches to the layer lines, which are expected t o cross the central row line (meridian) a t regular intervals and may be numbered (indexed) according to their distance from the central layer line (equator). The separation between layer lines usually permits unequivocal establishment of the size of the axial period of the fibril's structure, i.e. the distance between equivalent locations along the fibril. Row lines can be used to determine structure transverse to the fibril axis, but unless a large number of row lines are available the study of transverse structure is limited. I n collagen investigations the true fibrillar period is gained at small angles. Wide-angle patterns exhibit a smaller, apparent or pseudoperiod, indicating t ha t within the large pattern there are smaller sub-patterns. I n diffraction by such fibrous systems other unusual effects (21) are encountered because the fibrils are not ideal three-dimensional crystals, or because all fibrils of a massive specimen may not be perfectly parallel. In the latter case individual spots show arcing about a circle of constant radius, the full circle being completed when orientations are entirely random. Factors such as distortion of internal structure or limitation in external dimensions of fibrils cause the spots to spread along row or layer lines, and the manner in which this diffusion is exhibited indicates the nature of the cause. I n collagen each small-angle layer line consists of a single, meridionally centered spot which acquires appreciable length along the layer line because of these effects. At wide angles non-meridional spots are observed, but they become particularly diffuse as they depart from the meridian. Ordinary optical, organic chemical and physical chemical approaches also offer information which is indispensable for the present purposes. These commonly suffer from the difficulty th a t they may include extraneous, noncollagenous tissue elements in the analysis, but they furnish useful auxiliary information about molecular composition, size, shape and orientation. Some of these molecular characteristics are particularly difficult t o derive from the methods of electron microscopy and X-ray diffraction directly.
76
RICHARD S. BEAR
11. IDENTIFICATION AND DISTRIBUTION OF COLLAGENS 1. General Characteristics
The determination of a set of properties which can be used to form a definition of collagen is as difficult as this substance is complex. Collagen fibers are often recognized histologically b y their ability to swell i n acids and alkalies; by rather non-specific affinities for some dyes, such as acid fuchsin or aniline blue, particularly when mordanted; by lack of extensibility comparable with that of elastic tissue components; hy an immunological inertness and resistance to degradation by enzymes ; by a pronounced shrinkage in length at elevated temperatures or in the presence of certain electrolyte solutions; and by a transformation into gelatin in boiling aqueous environments. These general characteristics have sufficed to permit investigators to identify collagenous fibers in practically all parts of vertebrate animals, and, indeed, in all animals except those of small size or cell number (Protozoa). It is desirable, however, t o possess more definite means of establishing relationships between these diverse substances. The types of difficulty facing the more common means of identifying collagms may be illustrated by a few examples. Histologically visible fibers of collagenous nature are complex objects (see Fig. 1). They may, on occasion, be expected t o show chemical and physical properties characteristic of any of the major connective tissue components, even though they may most often resemble collagen predominantly. Reliance on the microscopic appearance of collagen fibers has often led to difficulty, such as th at involving the relation between reticulin and collagen. The former occurs as very fine fibers which commonly stain with silver much more strongly than do the relatively massive, typical collagen fibers. A similar distinction may be made with Schiff reagent after periodic acid oxidation, a method which stains many tissue components containing polysaccharides and also colors reticular fibers strongly (137). Embryonic skins, involving reticulin and almost no typical collagen, have also been reported to contain abnormally high sulfur content, which was judged to have been responsible for variations in physical properties from those normal for collagen-containing material (100). I n view of the well known fact that embryonic skins are high i n polysaccharide content (204) it would seem possible th a t the peculiar properties and high sulfur contents of reticular tissues may be ascribed to polysaccharide which has been in part adsorbed on or occluded within the fine reticulin fibrils. Partridge (166) has recently suggested that
THE STRUCTURE OF COLLAGEN FIBRILS
77
chondroitin sulfate plays an important role in the organization of collagen in developing connective tissue. I n any event, further discussion below definitely shows that reticulin is basically a collagen, as has long been suspected (162). Factors such as fiber diameter, compactness, and the difficult solubility of collagen would seem of importance in determining such properties as resistance to hydrolytic attack by enzymes. This inertness is probably not entirely derived from the chemistry of collagen but depends also on physical factors which prevent an enzyme from gaining access to its substrate. Pepsin, acting under acid conditions on swollen fibrils, is able to cause degradation (195), and increased temperature or physical maceration facilitates even the attack by trypsin in neutral or slightly alkaline environments (194, 200). On the other hand, collagenase capable of hydrolytically attacking, rather specifically and at neutral pH, both gelatin and collagen has been reported to be produced by cultures of certain anaerobic bacteria (various members of the genus Glostridium, 29, 113). This enzyme has not been more than partially purified, which is an important point, for it will be interesting to determine whether and how a single enzyme species can attack the bulky collagen fibril, overcoming its insolubility and causing hydrolysis simultaneously. This problem does not arise to the same degree with gelatin, but even here the particular mode of hydrolytic action is not yet demonstrated. Certain of the physical properties commonly recognized as characteristic of collagen are not infallible guides. I n ordinary microscopy extensibility has been found limited for collagen fibers, but this has not proved the case for individual fibrils observed electron optically (216). Mammalian and other collagens are normally transformed into gelatin by action of hot water, but the elastoidin of elasmobranch fishes, otherwise much like typical collagens, is resistant to this transformation (69). Elastoidin also shows, in its thermal shrinkage properties, a degree of reversibility which many collagens exhibit only after formalin fixation. These are just a few of the differences which otherwise clearly related collagenous substances exhibit, and along with other variations they are considered more fully a t appropriate places below. Chemical criteria for distinguishing collagens are also not satisfactory. This is in part because present evidence is not available in sufficient quantity to determine whether significant chemical definitions can be formed. As is elaborated below, however, it seems likely that these adequate chemical criteria will be elusive for some time. I n any case, in dealing with substances as complex as collagen, one will not immediately wish to demand chemical identity of all collagens, for
78
RICHARD S . BEAR
physical configuration of molecule and fibril may be as important as constitution. Required are methods which reach the basic structure of the ultramicroscopic collagen fibrils and are sensitive enough t o show the features which are fundamental. At the same time they must be sufficiently insensitive t o overlook minor variations such as are encountered in passing from tissue to tissue in the same organism, or possibly even from animal to animal, A method which meets these requirements has been available in X-ray diffraction for some time, and somewhat more recently electron microscopy has been found similarly useful. 2. Physical Criteria for Classification The first X-ray investigations of collagen-containing tissues or derivatives involved wide-angle diffraction (99, 104, 186) and were carried out in 1920 t o 1926. Early techniques failed in some cases to yield very significant results, but there soon followed a recognizable description of the typical wide-angle diffraction of native collagen fibers (104) and a demonstration that even the degradation product, gelatin, yields, upon stretching, the same oriented diagram as tendon (82, 120). It was evident quite early that collagens from a number of mesenchymal tissues exhibit essentially identical patterns, differing only in the degree of orientation indicated. Gone11 and Kratky (84) were among the first clearly t o contrast the collagenous tissues with the keratinous ones on the basis of their diffraction, but Astbury (7) finally crystallized the views regarding the wide-angle diffraction classification of the fibrous proteins in 1938. Astbury pointed out that it was possible to recognize common structural features in a number of protein fibers which yield practically identical diffraction patterns in spite of diverse origins. Distinction could readily be made in this way between members of a collagen class and those of another broad and considerably more heterogeneous set, now termed the keratin-myosin-epidermis-fibrinogen (k-m-e-f) group, comprising as the name suggests fibers from hair, muscle, epidermal and blood sources (12). Although the significance of Astbury’s classification has been questioned (154), there is as yet no convincing reason for discarding it. The diffraction patterns originally used for this classification were of the wide-angle type. I n 1936 Clark, Parker, Schaad, and Warren (54) and Wyckoff, Corey and Biscoe (217) discovered th a t toward the center of the collagen diffraction pattern, in a region hitherto obscured by bulky traps for the undiff racted beam, occur a number of diffractions indicating structures along the fibril axes with periods of possibly several hundred
THE STRUCTURE OF COLLAGEN FIBRILS
79
A. These were resolved a few years later (18, 128) t o reveal a large fibrillar period of 640 A., which was almost simultaneously photographed directly with electron microscopes (96, 207). The lc-m-e-f group of fibers also show small-angle diffraction (19, 58, 147)) and corresponding large structures may be seen in electromicrographs (59, 68, 97). Criteria useful for distinguishing the two great groups of protein fibers in the realm of large structure have been suggested (19, 148). The structural significance of the X-ray diffraction and electron optical observations is discussed in further detail below. At the moment these physical observations may be regarded as empirical means of segregating or distinguishing members of a class. Those aspects of the physical observation which are important for identification and classification of collagens now follow. Figures 3 through 5 show a number of the variations of the wide-angle collagen diagram that are obtained from several dried mesenchymal tissues. In all of these cases an attempt has been made, by means of proper selection of t,issue for purity and natural orientation, plus application of tension during drying, to secure as near an approach as possible to the best wide-angle diagram of collagen, typified by Fig. 3. However, in examining the many diverse tissues which contain members of the class the desired purity and orientation are not always secured. I n the worst instances one has to contend with the contribution by impurities of overlying diffuse background, which is most likely to obscure the diffraction areas corresponding to spacings in the neighborhood of 5 A. Lack of orientation throws the normal spots or arcs into rings, wiping out or running together the fainter and more diffuse details. Nevertheless, even under poor conditions dry specimens yield the 10 to 12 and the 2.86 A. diffraction rings, which normally are equatorially and meridionally accentuated, respectively. The collagens also show characteristic wideangle diffraction when moistened, as shown in Figs. 7 and 8, with differences in detail (Figs. 9 and 10) which are discussed further in Section IV, 3. While all types of collagen fibril diffract in nearly identical fashion a t wide angles, this is not the case at small angles, where individual differences become apparent. Figures 11 through 20 present a number of small-angle diffraction diagrams of members of the collagen class, serving to illustrate the points: (a) that the fundamental large structures of the different members studied to date are all of essentially the same size (600-660 A. normally in the dry state), as is evidenced by the fair constancy of the separations between consecutive lines of each pattern, and (b) that individual members of the class do, however, exhibit differences
80 RICHARD 6. BEAR
T H E STRUCTURE O F COLLAGEN FIBRILS
81
in the way in which they distribute diffracted intensity among the several lines of the separate patterns. Figures 11 and 15 show that the diffraction patterns of different members become more nearly alike when the specimens are moistened, whereupon the axial period also becomes somewhat enlarged (665 to 680 A.) and there is a distinct tendency for many of the odd layer lines to be more intense than the alternating even ones. By appropriate manipulation the small-angle diagrams of a single type of collagen may be varied, with results of the kind shown in Figs. 12, 13, 14, and 16, whose significance is considered further below. The small-angle diffraction method of identifying and comparing collagens has not often been applied as yet (148), though the cases available include material of quite widely separated origins and offer illuminating indications as to what may be expected of a general survey. Added t o the difficulties of the type mentioned above for the wide-angle examination of collagenous tissues, are the generally increased technical troubles of small-angle work, which particularly requires massive, fairly pure, well oriented specimens in order to facilitate registration of the diagrams during reasonable exposure times. Also fibrous protein specimens lose their ability to exhibit diffraction effects after physical or chemical manipulation more readily for the small- than for the wide-angle range (18), which reduces the permissible variety of pretreatments of material, designed for purification or orientation. The electron optics should furnish structural criteria for collagen classification nearly equivalent to those of small-angle diffraction. Electron micrographs disclose the large periodic structure, running along fibril axes, as a sequence of bands and interbands, seen in various aspects illustrated by Figs. 21 through 27. The over-all size of the sequential pattern corresponds in magnitude to the fundamental large spacing revealed by the small-angle diffraction line separations, while the details of different band and interband patterns correspond to the variFIGS.3 THROUGH 8. Wide-anglc diffraction patterns of collagens. All are a t the same magnification, which is about three quarters of the original size as obtained at a specimen-to-film distance of 5 cm. (rf. Fig. 2). Individual patterns were obtained from the following specimens: 3. Dry kangaroo tail tendon (KTT). 4. Dry shark-fin ceratotrichium (elastoidin). 5. Thyone body wall, dried under tension. 6. Dry chrome-tanned KTT. 7. Moist KTT. 8. Moist elastoidin. FIGS.9 A N D 10. Enlarged views of the central equatorial streaks of patterns 7 and 8, respectively. See Section IV, 3.
82
RICHARD S. B E A R
FIGS.11
THROUGH
20.
See opposite page for legends.
THE STRUCTURE OF COLLAGEN FIBRILS
83
ous distributions of diffracted intensity among the small-angle lines. Although the electron microscope has not as yet been used for systematic identification and comparison of widely different collagens, this correspondence between image characteristics and small-angle diffraction suggests that electron optics will disclose relatively constant large periods but quite different band and interband structures within them. The electron optical method suffers, however, from several defects. The most important of these is that its specimens are necessarily subjected to more or less drastic pretreatments in order to separate fibrils for individual examination. The result is often a considerable variation between the individual fibrils relative to size of period and the details of appearance of the band pattern (164, 191), a variability greater than would seem indicated b y the diffraction effects observed in massive samples which have received minimal manipulation (see Section 111, 2 ) . Consequently, conclusions as to the representative condition of the fibrils of a given population must be reached after statistical treatment of many observations (90, 192). Because of various technical limitations, it is not generally possible to demand of the electron optical studies clear resolution of structures with size less than about 50 A. in organic substances of this type. Consequently, the possibility arises that a given structure, departing appreciably from what is usually observed for collagens, may be difficult t o classify (cf. the case of “neurotubules,” 190). The X-ray method offers the possibility of employing both wide- and small-angle criteria, in effect establishing that certain structural details as small as 2.86 A. are those of a collagen. For this reason X-ray examination, where applicable, furnishes more definitive classification. FIGS.11 THROUGH 20. Small-angle diffraction patterns of collagens. All are a t the same magnification, which is about 5.4 times the original size yielded a t a specimen-to-film distance of 15 cm. Patterns 11 through 16 were obtained from a camera with resolving power for small angles much better than that of the one used for patterns 17 through 20, which came from poorer diffractors. (The cameras were those described as having “ f d ” values of 800 and 400 A,, respectively, by Bolduxn and Bear, 34.) Layer lines whose indices are multiples of 3 are indicated by dots to the left of the upper halves of each pattern. Specimens are as follows: 11. Moist K T T . 12. Dry (“unfanned”) KTT. 13. Dry (“fanned”) KTT. 14. Dry K T T with ca. 1.5 % phosphotungstate stain. 15. Moist elastoidin. 16. Dry chrome-tannrd KTT. 17. Oven dried carp swim bladder (ichthyocol). 18. Thyone body wall, dried undcr tension. 19. Dry elastoidin. 20. Dry decalcified Ralticina axial stalk (cornein).
84
RICHARD S. BEAR
There are two aspects to the identification problem: (a) in the first, one is concerned with determining the characteristics of collagens which are found a s components somewhere within massive samples, possibly originating from digerent tissues; but (b) one might also desire to localize and distinguish the several components of a single tissue. The second problem, which involves employing the physical methods discussed above as adjuncts to histochemical studies, is technically more difficult than the first. With X-ray diffraction, the limitations arise largely from the weakness of the diffraction effects which can be obtained from the small amount of collagen localized in a small area of a thin tissue section. With available techniques and X-ray sources, the wide-angle and small-angle diagrams can be obtained in feasible exposure times (a few hundred hours) from reasonably pure and well oriented volumes of the order of ml., corresponding t o short segments of fibers tens of microns in diameter. Feitelberg and Kaunitz (76) have used wide-angle X-ray diffraction in this way t o study changes with age in the alignment of collagen in human chorda tendineae. I n electron microscopy individual fibrils may be photographed, but the great magnification normally employed makes difficult the localization of groups of fibrils relative to structure at the ordinary microscopic level. Nevertheless, with improvement of sectioning methods for electron microscopy, this procedure will undoubtedly prove to be the one of choice for this purpose. Punching of small, microscopically located portions of tissue sections, followed by electron optical study of the isolated bits, is an alternative method which may be used in studying the state of collagen in localized areas of tissue sections.
3. T y p e s of Mesenchymal Collagen Collagen fibrils are generally expected in mesodermal or mesenchymal tissues, or in the mesoglea when typical mesoderm is not developed, as in the Coelenterata and Porifera. Table I summarizes the varieties of collagen which have been identified in sources of this kind and b y the physical methods described immediately above. As is easily understandable, most of the available information deals with the mammalian tissues. In this class of vertebrate animal the small-angle diffraction evidence suggests that all common forms of collagen are basically similar. For example, their fundamental fibrillar periods are in the neighborhood of 640 A. for dry specimens, and they show prominent intensification of such diffraction lines as the sixth, ninth, and eleventh (18). Examples examined to date are too numerous t o mention individually in the table, and, indeed, they furnish much of the
THE STRUCTURE O F COLLAGEN FIBRILS
FIGS.21
THROUGH
24.
See page 87 for legends.
85
86
RICHARD S . BEAR
FIGS.25
THROUGH
28. See page 87 for legends.
THE: STRUCTURE O F COLLAGEN FIBRILS
87
wide-angle diffraction and electron optical evidence regarding collagen structure cited throughout the present paper. The small-angle diff raction investigations include tendons of various sources (rat, beef, kangaroo), decalcified bone (guinea pig), skin (human, calf), cornea (beef), intestinal submucosa (sheep surgical gut), and demineralized mammoth tusk. Beef spleen was examined after concentration of the collagenous elements by partial digestion with trypsin (18). This organ is usually considered to provide a good source of reticulin, and the small-angle diffraction evidence indicated the presence of typical mammalian collagen. Since histologically recognizable ordinary collagen was present, this may not have constituted conclusive proof th a t reticulin and collagen are structurally similar. Nevertheless, electron optical study places reticulin definitely in the collagen class, since it differs from normal collagen chiefly in the matter of fibril diameter a s far as appearance in electron micrographs can determine (89). Wide-angle diffraction evidence serves only to make identification definite and is able to make no distinctions except a s to general purity and orientation. Heringa and Weidinger (100) reported a n exception to the rule, namely, that reticulin from the skin of newborn rats exhibits a shortened equatorial spacing, 9.7 A., instead of the usual value of about 10 to 12 A., observed with other d ry collagens. Experiments of the present author with Dr. Jerome Gross (89), using similar material, ~-
~
FIGS.21 THROUGH 28. Electron micrographs of collagenous preparations as follows: 21. Rat tail tendon (RTT) fibril stained with phosphotungstic acid (PTA), reprinted from Hall (95). The stained bands are labelled after Schmitt and Gross (191). Dots mark bands a to e, reading up. 22. RTT fibril frayed by swelling in 0.1 % pepsin and 0.5 % acetic acid, stained with 0.1 % ’ PTA. Unpublished photograph of Schmitt, Hall, and Jakus (192). 23. Chromium-shadowed collodion replica of moist human skin fibrils, with two dried fibrils adhering. Reproduced from Gross and Schmitt (90). 24. Similar shadowed replica of parallel human skin fibrils whose periods are in lateral register. Also after Gross and Schmitt. 25. Unstained R T T fibril, greatly stretched by accidental breaking of supporting collodion film. Reprinted from Schmitt, Hall and Jakus. 26. Fibril from chromc-tanned calfskin, otherwise unstained. Reprinted from Schmitt and Gross. Dots mark bands b l , b2, c, d, e and a, which form three pairs of bands whose centers are equally spaced. 27. Part of a human skin fibril left embedded in a formvar replica and shadowed with Cr. One portion, a single periodic segment in length, has been displaced, leaving a clean gap. Unpublished photograph of Gross and Schmitt. 28. Long spacing fibrils precipitated by dialysis from a mixture of plasma mucoprotein and an acetic acid solution of ichthyocol. Stained with PTA. Prepared from a photograph of Highherger, Gross and Schmitt (110).
88
RICHARD S. B E A R
indicated t ha t the observation of an anomalous value for the reticulin spacing is in error, and th at Heringa and Weidinger were confused b y a background of diffuse diffraction supplied b y the voluminous “ground substance” present in immature skin. Again it seems clear th a t reticulin is basically a collagen, since it possesses normal wide-angle diffraction. The elastic fibrous component of connective tissue, elastin, has often been examined by use of wide-angle diffraction techniques, applied t o ligaments as sources of high elastin content. Earliest studies showed collagen-type diffraction (104) , and Kolpak (126) found that stretching TABLEI Mesenchynial and Mesogleal Collagens Identified by Physical Criteria Phylum Vertebrate.
Class Mammalia
Aves Pisces Teleostomi Teleostomi Elasmobranchii Mollusca Echinodermata
Name
Common sources
Collagen
Connective tissues, tendons, bones, skin corium
Reticulin
Embryonic skin, spleen
Collagen
Tendons
Ichthyocol
Skins, tendons, swim bladders Scales Ceratotrichia of shark fins Squid connective tissue Thvone body wall Arbacia peristonie Asterias body wall
lchthylepidin Elastoidin
Cephalopoda Holothuroida Echinoida Asteroida
I
Coelenterata
Anthozoa
Cornein
Axial stalks, acontia
I’orifera
Demospongiae
Spongin
Skeletal fibers
Identification* W.A., S.A., E.M.
W.A. (89, 100) S.A. (18) E.Al. (89) W.A. (121)
I {
W.A. (49, 98, 108) S.A. (18) W.A. (84, 104) W.A. (16,47, 84, 104) S . A . (this paper) E.AI. (192)
W.A., S.A. (148) W . A . (9,50, 148) S.A. (118) W.A. (104, 148)
* Identification by means of W.A.. wide-angle X-ray diffraction; S.A., sniall-anple diffraction; E. M., electron microscopy. References to original papers are given where not too numerous.
beef liqamentum nuchae beyond 75 % was effective in developing oriented collagen diffraction against the normal background of diffuse, unoriented halos. Astbury included elastin in the collagen family (7, 8) but recognized that the collagen diffraction of ligament is probably due to typical collagen, which is always present in small amounts (9). He suggested that elastin itself, which apparently cannot be induced t o yield sharp, oriented diffraction, may he a collagen whose thermal shrinkage temperature (see Section IV, 5) is abnormally low, below room temperature. Small-angle diffraction disclosed a faint collagen-type diffraction in beef ligament (18), such as would result from a small amount of collagen impurity included with the elastin. Elastin has also been studied eler-
THE STRUCTURE O F COLLAGEN FIBRILS
89
tron optically (78,88,208). It seems clear that elastin must be excluded from the collagen family. Of the other possible collagens of vertebrate animals, those obtained from fish tissues have received most attention, ichthyocol and elastoidin in particular. Small-angle diffraction yields for ichthyocol a normal axial period but line intensities which are similar but not identical with those of mammalian collagen (Fig. 17). Chemical and physical properties, often cited below, also suggest that ichthyocol differs appreciably from other vertebrate collagens. Unpublished diffraction studies have produced evidence that carp swim bladder tunic requires more intensive drying than does mammalian collagen to develop the strong sixth, ninth, and eleventh small-angle layer lines. Elastoidin, however, shows still more definite departure of its large structure from that of mammalian collagen, as is clearly seen in Fig. 19. The nature of these variations appears in more detail below, but suffice it at the moment that elastoidin shows diffraction phenomena typical of a fibrillar structure which is considerably distorted internally. The over-all spacing (dry) is reduced to 600 A., and the small-angle line intensities and “shapes ” are quite distinctly altered; for example, the sixth line is almost lacking and most lines show greatest strength off the meridian. It becomes difficult to observe the small-angle diffraction because the distortion makes the fibril into a poor diffractor. Wideangle diffraction shows the evidence for unusual distortion only in subtle ways, being for the most part normal. Of the collagens from invertebrate animals listed in Table I only those of the Echinodermata and Coelenterata have received more than incidental attention, Three echinoderm classes are represented, whose collagens tended to show the usual marked intensity of the sixth smallangle diffraction, but an abnormal depression of the ninth line’s intensity. Other normally (i.e., as in mammalian collagen) weak lines, such as the fifth and tenth, often showed moderate expression. Among the anthozoan axial stalks examined, Balticina yielded the best small-angle diagram, which provided a result quite different from all other collagens: only the fifth and fourteenth small-angle lines had appreciable intensity, although the sixth still showed some persistence. The specimens from these two phyla usually exhibited somewhat larger periods (650-675 A., dry) than normally encountered, but not beyond reason. I n several instances of Table I only wide-angle diffraction is used to identify the members as collagen-type fibers. Spongin, however, is the sole one of these cases for which the large spacings have actually been sought so far. No conclusive result was obtained (148). Negative evidence is difficult to assess because tissue impurity or internal structural
90
RICHARD S. B E A R
distortions can easily render the small-angle diffractions difficult to observe. Spongin is, however, phylogenetically farthest removed from the collagens of typical mesenchymal origin and may represent a n early evolutionary stage of the collagen type of fiber. Systematic coverage of all possible varieties of mesenchymal collagen has obviously not been secured as yet. Nevertheless, certain general trends seem apparent in the available evidence. Over wide phylogenetic excursions the members of the family show a remarkable degree of constancy in the magnitudes of their large and small structural spacings. The variations which occur involve chiefly those structural features th a t develop the alterations in small-angle line intensities, whose significance is considered subsequently. These variations show some relation to phylogeny: witness, for example, the fact th a t the elastoidin, obtained from the most primitive vertebrates listed in Table I, shows greatest departures from the properties of the other vertebrate cases, and th a t in moving t o other phyla, the echinoderms and coelenterates, in each instance one encounters distinctly new characteristics. Members of the collagen class of fibrous proteins form a more homogeneous group structurally than do those of the keratin-myosin class. While one must make appreciable steps across or between phyla in order to discover distinctly new collagens, with the keratin-myosin group one encounters, within the same organism, tissues (hair and muscle, for example) whose short spacings are comparable but whose large structures are very different (19). However, a given tissue type (muscle is the best known example, 189) shows fibrous components with perhaps even greater phylogenetic stability than the collagens possess, suggesting that it may be more proper t o compare separate sub-groups of constant tissue type. Although the wide-angle diffraction serves a useful function in permitting the classification of fibrous proteins according to certain features of fundamental structure, it is now clear that the larger size ranges covered by small-angle diffraction and electron microscopy are also necessary in obtaining deeper insight into the meaning of such classification. 4. The “Secreted” Collagens Not all of the fibrous proteins which have been reported to yield the collagen-type wide-angle diffraction diagram are listed in Table 1. Several others are segregated in Table I1 because it is possible th a t they may prove of special significance. Among these are two substances which Champetier and FaurB-Fremiet (48) included under the term ‘‘ secreted keratins.” Their results showed, however, that designations such as “ovo-keratin” and “bysso-keratin ” of the older literature are inappropriate according to diffraction evidence.
THE STRUCTURE O F COLLAGEN FIBRILS
91
It is widely considered that all collagens are produced extracellularly, although cells are actively engaged somehow, from the fibroblasts of mammalian tissues to the spongioblasts of Porifera. Until the mechanism of formation is better understood, it remains possible that all collagens are essentially secreted, but in the sense applied to the present special group of “secreted’’ collagens, this designation refers t o the rather more obvious secretory origin of typical members. The special group would scarcely be recognized as collagens except for the diffraction evidence. They arise not from obviously mesenchymal tissue but from epithelial cells or glands, sometimes of non-mesodermal origin. I n a sense, therefore, they may not deserve the term “collagen,” TABLE I1 “Secreted” Collagens
Phylum
Special name
Source
Reference
Vertebrata Ovokeratin Skate egg capsule (48) Mollusca Byssokeratin Bivalve byssus threads (45, 46, 48) Annelida Earthworm cuticle (11, 174, 182) Nematoda Ascaris cuticle (74, 174) Echinodermata “Ejected filaments” of sea cucumber (8, 9)
although this term has no particular biological significance except through custom. Since the wide-angle criteria indicate close similarity to the more typical collagens, these special members of the class need not be excluded. To date none of the secreted collagens have been recognized as such except through wide-angle diffraction evidence. Small-angle diffraction or electron optical evidence for large structure is lacking, either for the reason that no attempt has been made or because of failure to secure evidence when sought. The situation in this regard may be summarized as follows: The diffraction studies cited in Table I1 were confined to wide angles. In the author’s laboratory preliminary small-angle attempts were at one time unsuccessful with byssus threads. Recently more exhaustive efforts were expended on the skate egg capsule’s ovokeratin. The origin, histology and biochemical and physical properties of this material have been the subject of extensive investigation by Faur6-Fremiet and collaborators (17,70,72,73). The fibrous layers of the capsuleprovidedpreparations which seemed macroscopically well oriented but which yielded, rather reluctantly and with disappointing diffraction orientation, the typical collagen wide-angle pattern. Small-angle methods, satisfactory
92
RICHARD 5. BEAR
for the most refractory cases of Table I, failed to produce clear evidence of large spacings, though faint indications were sometimes suspected. Reed and Rudall(l82) studied the cuticle of the earthworm (AZZoZobophora longa), securing by an ingenious stripping technique electron micrographs of layers of the crossed collagen-type fibrils secreted by columnar epithelial cells. Most important for the present discussion they reported inability to observe in their material the banded appearance characteristic of the usual connective tissue fibrils. There may be other substances related to the secreted collagens. Champetier and FaurB-Fremiet (48) mention also the membrane immediately within the shell of a bird’s egg and a fibrous secretion (koilin) of the avian gizzard. Neither of these, however, clearly diffract in detail like collagen at wide angles. Fraenkel and Rudall (77) point out that the collagenous cuticles of annelids and nematodes are replaced by chitinous ones in arthropods. Insect cuticles were found to have a layered structure composed of chitin and a protein belonging to the P-subgroup of the k-m-e-f family. Chitin, a polymer of acetylglucosamine, is chemically related to the hyaluronate and chondroitin sulfate of connective tissue “ground substance.” The question as to whether the collagenous and chitinous tissues are chemically homologous or analogous provides an interesting problem. The secreted collagens, if indeed they possess significance as a group, at present pose more questions than they illuminate. Are they truly without large structure? Since, as is developed subsequently, the wideangle diffraction of collagen arises largely from a certain type of polypeptide chain configuration, is this the only common feature possessed by the secreted and mesenchymal collagens? Gelatin, a degraded collagen, is unable to form banded fibrils and does not exhibit the diffraction evidence for large structure (18, 5 5 ) . Are the secreted collagens, like gelatin, incomplete collagens, produced by feeble synthetic mechanisms or resulting from degradation of complete collagens? Do they represent separate evolutionary developments of collagen-like polypeptides, accomplished by epithelial cells?
5. The Chemical Significance of the Physical Classification
It is customary to apply the term collagen most properly t o the limited group of mammalian members of the class, although evidence is available for the presence of collagenous substances in all but the simplest animals. Nevertheless, recognizing that it may be improper to include all these chemically diverse substances under one name, investigators have invented the separate ones shown in Tables I and 11.
THE STRUCTURE O F COLLAGEN FIBRILS
93
The diffraction evidence shows clearly that there is a basic structural similarity connecting all of the collagens and provides justification for use of a term to cover the whole class. From the biochemical point of view, however, it may be feared that this action carries along with it an unwarranted assumption of chemical identity. This is not the case, for the physical methods can not be presumed to indicate even that two closely related mammalian collagens are chemically identical. The situation is analogous to the phenomenon of isomorphism encountered in simpler substances, wherein chemically similar but not identical molecular species may replace one another in almost identical crystal lattices. I n the collagens, amino-acid residues may be presumed to be the replaceable units within fibrils which retain a basic type of organization. While chemical constitution will undoubtedly be found eventually a t the basis of collagen configuration, the chemical evidence now available is disappointing. The difficulties are, of course, great, since adequate results would require securing of sufficiently massive and pure specimens, the latter being hard to obtain in the absence of definite chemical criteria of purity. Excellent summaries have been prepared concerning the results of past attempts to arrive at the amino-acid contents of collagens and the gelatins derived therefrom (31, 38). Table I11 collects together the information suitable for a comparison of the class members of predominant present interest. It is desired to consider to what extent the chemical data parallel the results of the physical comparisons of the collagens. There are, of course, many studies of the composition of gelatins derived from mammalian sources and from fish skins. I n Table 111the relatively recently available results on the whole collagens have been cited. Mammalian collagen is represented largely by the analyses summarized by Bowes and Kenten (40), performed on a preparation of ox hide. The data account for nearly 100% of the total substance and have been found in reasonable agreement with acid-base titration curves. This set of values may be used to define a standard collagen with which other substances may be compared. The general finding of the physical methods, namely, that mammalian collagens are similarly constructed, seems in agreement with the existing chemical information. Neuman (163) observed .that a number of collagens from different tissues and from several mammals, including also an avian source, were scarcely distinguishable chemically. A. Pirie et al. (175) obtained paper partition chromatograms of hydrolysates of collagens from ox vitreous humor and cornea, finding them very similar to those yielded by a gelatin hydrolyzate. Bowes and Kenten (42) used
94
RICHARD S. BEAR
TOLE I11 The Chemical Composition of Collagens and Elastin (Grams of residues or of N in 100 g. of protein)
Bovine collagen
Bovine elastin
Glycine Alanine Phenylalanine Leucines Valine Proline Hydroxyproline Glutamic acid Aspartic acid Arginine Lysine Hydroxylysine Hist dine Serine Threonine Methionine Cystine Tyrosine Tryptophan
(40) 19.9 7.6 3.7 4.8 2.9 12.7 12.1 10.0 5.5 7.9 4.0 1.1 0.7 2.7 2.0 0.7 0.0 1.3 0.0
(85, 163, 196) 22.5 15.1 4.4 10.1 12.5 13.4 1.7 2.4 0.35 0.88 0.39
-
-
Total Total N Amide N
99.6 18.6 0.65
87.2 16.9
Amino Acid
0.04 0.68 0.87 0.18 0.28 1.4 0.0
0.04
Ichthyocol (28, 163) 14.9 9.5 1.9 5.2 2.0 13.3 4.0 8.6 5.1 8.4 4.4 0.32 0.91 4.2 2.9 2.2 0.09 0.39 0.0
Herring scales Cornein Spongin (30, 33) (30, 53) 10.9 11.8 0.2 3.3 6.7 6.1
(32) 15.7 2.6 4.7 2.4 7.2
6.0 3.2
5.2 2.6
4.3 0.0 14.6 3.5 4.6 3.0
2.2 6.2 2.4 2.5 0.60 2.5 0.28
0.4
Lo.0
5.5 12.8 0.0
2.6 1.5
6.3
-
-
-
88.3 18.3 0.47
64.8 16.0
44.0 14.7 ~
0.0 54.6 14.6 0.6
~~
Original sources are indicated a t column heads and include only the most recent reports needed t o secure as complete cover'age as possible for these significantly different preparations. All data have been reduced to expression as grams of residues or of N per 100 g. of ash- and moisture-free protein. Bovine collagen and elastin are those of purified hide and ligamentum nuchae, respectively. Ichthyocol samples were derived from fish swim bladders and skins, whose analyses are sufficiently similar to be combined for present purposes. Similarly, available analyses of the corneins, gorgonin of Gorgonia flabellum and plexaurin of Plexaurella dichotoma, have been combined, as have also results derived from the spongins of Hippospongia equina and of the bath sponge. Analyses of herring scales present the most complete available information for this sort of material, which is complex, though all isolated components have presented similar collagen-like patterns of residue composition according to chromatographic studies.
T H E STRUCTURE O F COLLAGEN FIBRILS
95
paper chromatograms and titration curves to show that elastins differ from normal collagen, but that reticulins are essentially collagens. Data of Table I11 agree in some detail with the physical conclusion that elastin is not a collagen. The former is relatively low in residues with polar side chains : hydroxyproline, glutamic and aspartic acids, arginine, lysine, serine, and threonine. It is particularly rich in glycine, alanine, the leucines and valine, residues whose side chains are nonpolar. Elastin and collagen are similar only in their moderately high proline contents and in their low proportions of histidine, cystine, tyrosine, and tryptophan. Just as the small-angle diffraction suggested, one still finds ichthyocol recognizably similar to mammalian collagen according to the chemical results. Here, however, there begin to be indications of significant differences. Neuman (163) pointed out that fish-skin collagen contains appreciably greater quantities of serine, threonine and methionine than does mammalian collagen, and the earlier results of Beveridge and Lucas (28) on hake swim bladder indicated the same thing. It should also be noted that according to the latter authors the hydroxyproline content of swim bladder is relatively low, so much so that the over-all content of hydroxyl-containing residues (serine, threonine, and hydroxyproline) may be significantly lower in ichthyocol than in collagen. It is unfortunate that beyond this in the direction of the more primitive collagens the chemical data are disappointing. Figures for these cases are open to the serious criticism that they are incomplete and were often obtained from impure specimens after dubious pretreatments. Glycine, alanine, serine, and hydroxyproline have been identified qualitatively but not estimated quantitatively in elastoidin (63). I n view of similarities between actinotrichia of teleosts and the ceratotrichia of the elasmobranchs (go), plus their general physical resemblance to teleostean fish scales, the analyses for herring scales of Table I11 may possibly provide the best available indication of what may be expected for these substances. It is probable, however, that these collagens, as usually studied, are far from being pure substances. They contain possibly two proteins in appreciable proportion (80, 32), one soluble and the other a more resistant collagen-like component, and considerable amounts of aminopolysaccharide (63), according to available descriptions of individual cases. If the analyses of the corneins and spongins have any significance it would appear that certain aspects of composition which have often been considered typical of collagen, judging from mammalian examples, may be questionable. For example, cystine which is practically absent in mammalian collagens is said to be represented appreciably in the corneins
96
RICHARD S. BEAR
and spongins; the glutamic acid content of spongin is reported much greater than t ha t of normal collagen; hydroxyproline and proline, often regarded as the distinctive components of familiar collagens, are either absent or diminished in spongin; and both gorgonin and spongin are reported as rich in tyrosine. I n these instances the tyrosine and possibly other residues are halogenated considerably with iodine and bromine (1, 183). Very little is known regarding the chemical composition of the secreted collagens. Analyses of several show reasonable nitrogen values for protein and higher sulfur content than is typical of mammalian collagen (0.9 t o 1.2% in the selachian ovokeratin, 72). According to Baudouy (17) a precursor of ovokeratin (prokeratin) contains sulfhydryl groups which become oxidized in the final stabilization of the selachian egg capsule. One gains the general impression th at the relatively primitive collagens often contain more sulfur than do mammalian collagens, though whether the sources are methionine and carbohydrate sulfates, in contrast to cystine, is not always clear. Lloyd (140) pointed out that in swelling behavior normal collagen, elastoidin and ovokeratin, progressively richer in sulfur, also show increasing resemblance to keratin, whose resistance to swelling is attributed to stabilization of fibrillar structure by disulfide bridges between polypeptide chains. The chemical evidence is disappointing just at the points where it could be most illuminating. When available, it will be most definitive in suggesting the underlying causes for the structural variations of members of the collagen class. This problem is twofold, requiring (a) indications as t o the chemical features common t o all collagens and permitting the molecular chains to adopt the small-spacing arrangement typical of all, and (b) information relative to variations in chemical composition which could account for the larger structural features recognized by small-angle diffraction and electron optics. Astbury’s original definition of the collagen class anticipated only part (a) of the problem, but part (b) seems also essential if the differences between members of the class are t o be understood as well as the similarities. The chemical results to date are at least in general qualitative agreement with the facts disclosed b y small-angle diffraction. Both approaches indicate that t o find collagens which are significantly different from the more common varieties it will be necessary to study a wider range of animal phyla than has been the general practise. Until this is accomplished in more detail than a t present the definitive chemical features which link the collagens will remain elusive. The remaining sections use the best understood collagens to provide
THE STRUCTURE O F COLLAGEN FIBRILS
97
type-models for collagen fibrils. These models illuminate further the nature of the structural problems involved and provide working hypotheses in the light of which the future chemical and physical evidence may be considered in attempting to arrive eventually at complete structures.
111. COLLOIDAL STRUCTURE OF COLLAGEN FIBRILS 1. Optical Evidence
Before the application of the small-angle X-ray diffraction and the electron optical methods of study to collagen, there existed a largely unfilled gap between the histological or microscopic size range and the molecular or chemical level. This intermediate region, which may be termed colloidal, was then represented by extensions of ordinary light optics beyond true resolution, i.e. by the indirect optical methods of ultramicroscopy and polarization optics. The wide-angle X-ray diffraction entered the scene with details of structure so far below those clearly discernible by the light-optical methods, that a great deal of imagination had to be exercised in drawing together the information available a t the lowest levels with that discovered by optical means. The wide-angle diffraction method also has, a t least with collagen, a critical blindspot, in th a t its information (Section IV, 1) deals with only a portion of the structure of the collagen fibril. For these reasons it seems desirable to introduce the present discussion of fibril organization with a consideration of structure in the colloidal size range before progressing to the finer details, even though the historical development reversed this order in some respects. While ultramicroscopy and polarization optics have not offered all that might be desired, they did contribute much to early ideas. The darkfield investigations supplied knowledge of the existence of collagen fibrils, estimates of their diameters, and indications of how they swell (65). Studies of collagen double refraction were more illuminating, although even now many of these observations remain incompletely explained. Relatively recent data of Pfeiffer (172) are used in Fig. 29 to show th a t collagen fibers, like most fibrous proteins and polysaccharides, possess both form and intrinsic types of double refraction. Under all normal conditions the sign of the first kind of double refraction in collagen is positive relative to a n optic axis paralleling the fiber axis. According to optical theory of colloidal systems (79, 188) this means that the macroscopic collagen fiber contains submicroscopic particles of anisodiametric shape with long dimensions aligned with the fiber axis.
98
RICHARD S. BEAR
The identification of the particles is not immediately clear, however. It has been the fashion to speak of these as rodlet micelles, but further discussion below indicates that the only elements which might be involved here are the fibril and the protofibril. The latter are closely packed into fibrils and offer limited and peculiar possibilities for an interpenetration of the immersion fluids used to demonstrate the form double refraction. Perhaps the fibrils supply the major form effects.
I l 4
FIG.29. Double refraction of collagen fibers produced by fibroblast cultures of chick embryo explants, after Pfeiffer (172). Retardation, r, in mp is plotted, as ordinate, against the refractive index, n, of the immersion medium. Curve A is for fibers fixed in Kelly’s fluid (containing formalin), and curve B shows the result of fixation in tannic acid.
The intrinsic double refraction expresses the essentially anisotropic or internally crystalloidal character of the aligned particles themselves. Since formalin-treated fibers show positive intrinsic double refraction relative to the fiber axis (curve A of Fig. 29) this is probably the normal condition. Because of colloidal variability of native collagen fibers the analysis of form and intrinsic components by means of immersion methods requires prior fixation. Von Ebner (61, 187) long ago observed that certain phenols, phenolic acids and higher aldehydes exert a profound effect on collagen double refraction, reversing its sign. Kuntzel (130) demonstrated that the same effect is accomplished by treatment with several pyrogallic tanning
THE STRUCTURE O F COLLAGEN FIBRILS
99
agents (curve B of Fig. 29) and that fibers of this sort owe their reversal largely to a change of the intrinsic double refraction from positive to negative. Formalin, chrome tan, and other reagents do not accomplish the reversal, although many of them cause a decline in the positive magnitude. Kuntzel also observed anomalous immersion curves, with double refraction continually rising along with refractive index of the immersion medium, encountered in fibers treated with pine bark extract and sulfite cellulose. The normal positive sign of the intrinsic double refraction indicates that " intra-micellar " (intrafibrillar) molecular chains run parallel to the fibril axis. The variations produced by the substances described in the last paragraph have been ascribed to intramicellar penetration by the added agents (130), either with oriented absorption (188) or accompanied by chemical reaction (79). The last word has probably not been written about the complex optical properties of collagen. The situation is undoubtedly so difficult that it will be clarified only after more detailed models for fibril structure appear from other directions. At the moment one should conclude from the optical studies that in collagen fibers there are oriented long particles (fibrils), that within these are also regularly arranged thinner elements (protofibrils or molecular chains), and that the particles are penetrable in optically indeterminable ways by chemical agents whose molecules can be of some size. 2. A Fibrillar Model Derived from Small-Angle X-Ray Difraction
After 1942 the diffraction techniques were improved to the point where resolution of consecutive small-angle layer lines became possible (18, 128). Success was secured by the use of finely collimated beams, large specimen-to-film distances, and long exposure times. Economy of radiation was improved by using slits to form ribbon beams, which because of cross-fire of great aperture orthogonal to the fiber axes of specimens, contributed adequate intensity to meridionally separated points of the registering film (see Fig. 2). These slit-camera patterns were capable of exhibiting excellent angular resolution, corresponding to spacings of thousands of Angstroms, along one direction of the resultant pattern, but lacked resolution in the direction paralleling the large dimensions of the slits. Consequently, although the early results were capable of separating the layer lines, they were utterly inadequate for the demonstration of true line shapes, i e . , the manner of distribution of diffracted intensity along individual layer lines. Since collagen appeared to exhibit only a single set of diffractions,
100
RICHARD 5. BEAR
Bolduan and Bear (35) suspected that the collagen fibril was effectively a one-dimensional diff ractor at small angles. They looked for certain theoretical consequences by means of which order deficient systems (i.e., those lacking the full three-dimensional lattices of normal crystalline materials, 21) may be recognized. It became clear that the large structure of the collagen fibril was indeed one-dimensional, with periodic regularity extending solely along the fibril axis. Certain unexpected diffraction behavior was observed, however, which suggested that it would be desirable to view line shape. In the meantime experimental methods employing efficient pinhole diffraction cameras capable of accomplishing this purpose had been developed (34). Preliminary studies (35), using combinations of results obtained with pinhole- and slit-camera photographs, furnished comparisons between the diffraction effects of collagen and those expected of simple models for one-dimensional diffractors (21). After these investigations it became possible t o obtain the full, rather complex small-angle diagram of collagen and to develop a structural model for its general explanation. A brief summary of these studies is available (20). Specimens of collagen kept wet during examination show smallangle diagrams like Figs. 11 and 16. These are particularly simple, but because of overwhelming influence of water (see Section IV, 3) they are not as directly related to the intrafibrillar distribution of protein as are the patterns of dry material. Even the latter often yield patterns such as Figs. 12, 17, 18, and 20, which are not as detailed in diffraction phenomena as are the patterns of Figs. 13, 14, 16, and 19. The latter exhibit a characteristic “fanning” of the layer lines, which also show maxima and minima of intensity. The development of this fanned appearance often depends on the individual collagen source. It was readily obtained in certain kangaroo tendon and elastoidin samples, and many other dry preparations showed at least some tendencies in the same direction. Apparently, the excellence of the development of fanning is a function of variables involving the previous history of a sample, the degree of hydration permitted prior t o final drying under tension, the tension then applied, etc. (23). The dry fibril is, as demonstrated below, a statistical structure, and the fanning results from this fact. It is not surprising, therefore, that to obtain a massive sample in which the fibrillar structure is statistically constant throughout may not always be a matter that is readily controlled, At any rate, these fortunate cases are particularly valuable and furnish the diffraction patterns now described. Note that in a pattern such as Fig. 13 all lines are straight, extending normal to the pattern meridian, and they are equally spaced along
THE STRUCTURE O F COLLAGEN FIBRILS
101
the meridian. The straightness indicates that the fibrils of which the specimen was composed were all effectively rendered parallel by the tension which was applied to the massive sample during preparatory drying to accomplish this purpose. The separation between lines is reciprocal in magnitude to the large axial structure of the fibril, which is found from evidence of this sort to be usually near 640 A. in air-dried specimens. No evidence is observed for the existence of periodic structure across the fibril, i.e., no true row lines are present. One concludes that as far as structure revealed at small angles is concerned the fibril shows only one-dimensional periodic structure, along its axis. There is, however, wide-angle evidence of transverse periodic structure, as developed below, but this is of another smaller order of magnitude. The spectacular fanning feature of the layer lines is seen in their progressive lengthening as the index increases. Line lengths transverse to the meridian are normally expected to be, with one-dimensional diffractors, inversely proportional to the diameter of the fibril over which coherent diffraction to the line occurs, hence constant for all lines yielded by a well organized, cylindrical fibril (21). A continually changing line length can only mean that as smaller structural details, corresponding to lines of greater index, are brought into consideration, a smaller portion of the fibril cross section is able to contribute coherently to each line. The line spreading phenomena clearly indicate internal distortion within each fibril, and the unit of cross sectional area which can be removed by distortion from coherence with the remainder would seem small, since there are no large increments in length from line to line. The above major characteristics of the small-angle diffraction suggested the model for fibrillar structure shown in Fig. 1 (22, 23). The fibril is regarded as composed of much thinner filaments or protofibrils, each of which contains along it various specifically ordered structural details (chemical variations), which are considered alike in all protofibrils. The diffraction characteristics which establish collagen fibrils as onedimensional diffractors with respect to large structure require that the protofibrils be adjusted axially so that equivalent structural details of neighboring ones match transversely. Entire fibrillar cross sections are thus characterized by a degree of constancy of chemical composition and electron density. The forces of attraction between adjacent protofibrils are not everywhere of sufficient strength to maintain rigid matching, however. Consequently, along the fibril one has a periodic variation in degree of matching, corresponding to the periodic alteration in the chemical composition of protofibrillar levels. Wherever matching is maintained one may say
102
RICHARD S. BEAR
that a cross-section (interband) of perfect structure is present. At those levels where matching is relaxed, displacements of protofibril elements across and along the fibril cause development of bands in which distortion or imperfection is present. In attempts to calculate the diffraction expected of the model (22) one of the difficult problems involved the particular manner in which the distortions were to be introduced. A rather general type of statistical displacement of elementary protofibril sections was found satisfactory. Corresponding elements of length along adjacent protofibrils were assumed to be randomly shifted by small amounts, both along the fibril axis and normal to it. Neither the axial (longitudinal) nor transverse (radial) distortion alone accounted for the line shapes observed, a fact which provides possible evidence for longitudinal interconnection of intrafibrillar volume elements into protofibrils. It is difficult with a filamentous structure to displace a small segment normal to the axis of the filament without a t the same time causing some axial distortion, and vice versa. The diffraction effects of the perfect interbands and of the two kinds of distortion a t bands may be separately recognized: (a) perfection accounts for the short line components closely applied to line centers and seen most purely a t the fourth and seventh lines of Fig. 13; (b) longitudinal imperfection is related to the components which are also maximal at line centers but are more extensively spread along the lines, as a t the ninth line; and (c) radial imperfection is responsible for the non-central maxima observed particularly clearly on the eighth and eleventh lines. Only the diffraction components of imperfection expand in length with increase of line index. All of these effects are well accounted for by the application of diffraction theory to the model. The particular way in which the bands and interbands are distributed along the axis of the typical fibril is related to the observed manner of distribution of various types of diffraction component among the several layer lines. It is not possible, however, to derive these features of fibril structure directly from their corresponding diffraction effects because of the well known difficulty of determining the relative phases of the several diffraction lines or line components. The electron optical methods determine the band and interband locations more directly. In spite of these limitations of the diffraction method one may use the various characteristics mentioned above (line separation, length, and shape) to determine size of periodic structure, fibril diameter, and the types of perfection and imperfection present. The collagen small-angle layer lines are remarkable for their sharpness, i e . , their small breadth in the meridional direction, and their large
THE STRUCTURE O F COLLAQEN FIBRILS
103
number. Only the first dozen lines reproduce well in Fig. 13, but on the original films it may be seen that lines continue with increasing length out through those with indices of about twenty. The components of line shape due to imperfection become, however, quite faint and almost indistinguishable at these large indices, with major emphasis then placed on the central components of perfection. Beyond an index of 20, direct pinhole registration has not been accomplished, but slit-camera patterns have shown in some instances lines with indices as high as 32. These facts indicate that the collagen fibrils in the population of an intact tendon are predominantly of the same sized period. It may easily be calculated that consecutive diffraction lines of two different fibrils should fuse near the kth layer line if approximately l / k is the fraction of period by which their fundamental structures differ in size. The same fraction represents also a generously maximal deviation from average intraperiod positions permitted for the structural details of an individual fibril if every detail is to contribute coherent scatter to a kth line. Since the highest k value observed is 32, the fraction representing faithfulness of structure, both within and among fibrils, is about 3%. In absolute size (20 A.), however, this is still large compared to single amino-acid residues. Kaesberg et al. (114) determined the absolute line breadth of the first layer line of the collagen small-angle diagram to be less than 20 seconds radiation), which places a minimal value of roughly of arc (with Fe KLY thirty periodic segments on the effective lengths of the fibrils for coherent diffraction to the first line (calculated from the theory of line breadth of reference 21). Although this result refers to moist specimens, which may be structurally more perfect than dry ones (IV, 3), this estimated length seems consistent also with the number of lines which have been observed with dry specimens. If, for example, within a single fibril the statistical variations in axial position of comparable intrasegment structure reduce the diffraction coherence between adjacent periods a t a rate such that in thirty periods the coherence for diffraction to the first layer line is destroyed, then one would expect that diffraction could not be observed much beyond the thirtieth line. Electron micrographs show that collagen fibrils are much longer than the effective lengths indicated here from the diffraction observations, so that intrafibrillar distortion is clearly necessary to reconcile these facts. Examination of the model suggests other consequences which have not been given quantitative expression as yet. First of all, the fibrillar period (640 A.) may be regarded as a quantity which shows the constancy of magnitude indicated above, from one unit segment to the next, largely as a result of a statistical summation of the projections of protofibrillar
104
RICHARD S . BEAR
elements upon the axis, not because each consecutive segment’s corresponding elements are identically contorted. Increase in the general degree of distortion should shorten the over-all period, while decrease in imperfection will lengthen it. One may also expect that bands, i.e., portions of the fibril which possess distortion, will have protofibrils less densely packed in the fibrillar cross sections than would be the case if perfection existed there. The actual density at a given fibrillar level is a function also of the particular chemical content, but unless protofibrils have compensating variations in diameter at different levels, the bands will be distinguishable as locations at which fibril diameter is greater than average. Because of the poorer packing of protofibrils at band locations, these should be more readily invaded by foreign substances. Large invading molecules may actually increase the distortions of protofibrils at bands or actually “pry open” the structure to increase the spread of distortion along the fibril. On the other hand, molecules which are small compared to protofibril diameters may serve to “smooth out the kinks” and increase perfection, or perhaps to “fill in the holes” and reduce the electron-density fluctuations a t the imperfect cross sections, thus causing similar reduction in the diffraction evidence of imperfection. Chemical attack on the fibril should most easily be accomplished at band locations. The general, qualitative properties just described as expected of the fibrillar model are used to illuminate various aspects of problems which arise subsequently. 3. Digraction Theory for the Model
A complete understanding of the structural significance of the smallangle diffraction diagrams of collagens is difficult to secure without more detailed reference to the diffraction theory applicable to the model developed in the last section. This model presents a kind of structure which, though of a rather general type, as yet is not well known or often encountered. It should apply to other collagens as these are discovered, or as known ones are made to produce the fanned diagrams. Elastoidin, in particular, has already yielded (Fig. 19) an extreme version of this sort of diffraction, whose meaning would have been difficult to determine without prior experience with kangaroo tail tendon. In determining the diffraction expected of structures of the type of the model shown in Fig. 1, parameters Ub and ua were used to express the root mean square longitudinal and radial displacements, respectively, experienced between corresponding points of any pair of adjacent protofibrils (22). In terms of these parameters the fibrillar structure may be described
THE STRUCTUEE OF COLLAGEN FIBRILS
105
as follows: in any cross section a t a given axial level of the fibril the parameters remain constant, though not necessarily equal for the two kinds of distortion. They vary with axial level, being periodic functions of position along the fibril. One distinguishes bands from interbands by arbitrarily letting the former comprise levels at which the u’s are appreciable and the latter those where the a’s are zero. Without entering into details of derivation, the final useful result may be written where IO may be regarded as a proportionality constant connecting the intensity, I k t , observed at the kth layer line at the angular separation [ from its center, with the other experimentally or theoretically interesting quantities. The respective cross sections of total fibril and of individual protofibril are r R 2 and aRo2. The Fk’s are structure factors for the kth layer line, distinguished with subscript p , in the case of the factor involving only axial levels of perfection, and with b and a, respectively, for those concerned with regions of longitudinal and radial imperfection. The 8’s are functions of [ indicating the theoretical line-component shapes, while B is a factor which largely determines the general effect of the degree of distortion on the intensities of both types of imperfection component. The theoretical shape functions (8’s) turn out to be much like those observed experimentally. Expressions for them are :
8, E
[Ji(2rGR5)] Lybk
Sak [k
+ exp [ - ( ~ f i R [ ) ~ ] .
1/(1 fk2/(1
tk2)3
+
tk2)3
(%OuOa/kUOb)
l.
I n these equations i is the reciprocal of the X-radiation wave length and bo is the magnitude of the fibrillar axial period (ca. 640 A.). The c o b and uoaare mean values for the respective displacement parameters averaged over the fibrillar levels a t which distortion occurs. The symbol J 1 signifies the first-order Bessel function whose argument is included after it in parentheses. The S , shape function is independent of k, and if R is large (1000 A. for kangaroo tail tendon, see 35) it falls rapidly with increasing 5 to produce the short line components typical of diffraction by fibrillar perfection. S b k has greatest value at f or t k zero, and declines as t. increases, a t a rate determined by (i%OuOa/kuOb), which becomes less as k increases. Sob has maximum intensity a t 4 = (Icuob/t/Z fboaoa) but falls to zero at
106
RICHARD S. BEAR
6 = 0 (the line center) or a t large 4. These various shape characteristics are easily identified in the experimental observations. The structure factors are unknown a priori and provide the greatest barriers t o attainment of complete analysis of fibril structure. They contain the distributions of matter and of distortion along the fibril axis. Each has the form
where uo is the average distortion and a the actual distortion a t level v along the fibril, the latter expressed in fractions of the axial period (bo). Both of these a’s are further specified b y subscripts b or a, depending upon which type of distortion the structure factor concerns. For perfection, (uo/u)is always unity, and then variations in N ( v ) , the number of electrons per unit cross sectional area of the fibril, provide the only structural details involved. Integration is restricted to the portions of the unit fibrillar period a t which imperfection or perfection exists, whichever is involved. The F’s are complex numbers whose phases are not encountered (because of squaring) in the intensity expression. Consequently, the functions u and N can not be experimentally determined, and in this way the usual limitation of X-ray studies is encountered in the present case. Nevertheless, certain qualitative consequences of the structure factors can be realized. Note that the structure factors are functions of k and will not be the same for all layer lines, explaining why different lines have not the same relative expression of the components of perfection and imperfection. The components of imperfection will lose particularly the contributions to scatter from fibril elements whose a’s are unusually large relative t o u o . The factor B of the intensity expression is given by b o 4 / 2 ( k ~ ~ 0 band )4 is a potent one in determining the influence of the general degree of distortion upon the expression of the diffraction components of imperfection. Note that increase in (kuoa)rapidly diminishes this factor, showing that increased distortion can rapidly wipe out diffraction evidence of imperfection, or that components of imperfection should be lost a t layer lines of high index. The latter effect has been mentioned before in describing the collagen small-angle diagram. The factor B is a large one, if, as is expected, (Tab is a small fraction (0.01) of the size of bo. However, the total coefficient for the diffraction components of imperfection contains also r R o 2 . The corresponding coefficient for the component of perfection is nR2 alone. Since with kangaroo tendon R is about two hundred times Ro ( 2 2 ) , the net effect of
THE STRUCTURE O F COLLAGEN FIBRILS
107
these quantities is to favor, for layer lines above about the fourth, expression of only the components of perfection. One actually observes the components of imperfection, however, hence it must be concluded that normally in dry collagen the structure factors of perfection are small in comparison to those of imperfection, as was indeed assumed in deriving the form of intensity expression given. The theory provides for direct measurement of the ratio of uOa/u06 from the positions of the non-central intensity maxima observed on layer lines with outstanding components of radial imperfection. If trnis the position along the hth line at which the intensity maximum occurs, the equation go. = k
fiGbotm provides means for calculating the ratio of mean displacements. One cannot, however, obtain absolute values of the go’s, except as these may eventually be estimated from the general rate a t which the components of imperfection fade, relative to the components of perfection, as layerline index increases. Because the mathematical treatment did not permit allowing full freedom of range to all the parameters involved, the usefulness of the given intensity expression is limited. Implicit in the theory, but not well expressed in the formulation given, is the possibility that components of perfection may dominate the entire small-angle pattern under two extreme conditions: (a) when the preponderant portion of the fibril lacks distortion, or (b) when the imperfect regions are so heavily distorted as to lose capability of coherent scatter. One may attempt to distinguish these cases by examining the values of bo, which, as mentioned above, should be large when the fibril is relatively perfect and decrease when imperfection becomes serious. This method can not, however, be expected to correlate in all instances with diffraction evidence for degrees of perfection. It is possible that certain fibrillar regions will be so badly distorted as not to contribute much to the structure factors of imperfection while markedly decreasing bo, and other parts of the fibril may remain slightly distorted, contributing to imperfection structure factors without great change in bo. The former kind of distorted level can cause variable bo without leaving an impression on observed diffraction intensities. The theory also presumes both types of imperfection to be present together, and can not be used in the form given when one desires to pass continuously to a situation in which either or both types of imperfection vanish. Nevertheless, one is able to discuss all cases encountered thus
108
RICHARD S. BEAR
far in terms of three types of situation: (a) occasions in which chiefly the diffraction components of perfection are observed, when essentially the first term of the intensity equation applies; (b) instances which show the complete set of phenomena shown in Fig. 13, accounted for by the complete equation; and (c) the case of elastoidin (Fig. 19), whose dry specimens show largely the diffraction components of imperfection, requiring chiefly the part of the theory dealing with distortion. I n effect; because they are not immediately determinable except by observation, the squared structure factors of the intensity equation may be given any relative positive values (including zero) that may be required by a given situation. The case of elastoidin is particularly interesting because it presents a structure in which distortion is very prevalent. Figure 19 shows that, except for the first three layer lines, practically all of the diffraction components are those recognized as typical of radial imperfection alone. One can not conclude, however, that relatively little longitudinal imperfection is present, because the noncentral intensity maxima are even more widely separated from line centers than in kangaroo tendon diagrams. The latter are consistent with a ratio of uOa/uObof about 1.85, whereas the same ratio for elastoidin is 0.77. One concludes that elastoidin fibrils are so badly distorted axially that no diffraction evidence of this type of imperfection remains, except possibly a t the lowest layer lines. The excessive degree of general distortion in dry elastoidin also shows up in its unusually low bo (600 A.) and in the fact that several times the normal exposure to X-radiation is required to yield its small-angle diagram. The shape factors for components of imperfection must be regarded as approximate, though qualitatively correct and becoming more exact at small [ and large k, because of certain approximations of the theoretical treatment. On actual diagrams, such as that of Fig. 13, each line has non-meridional end locations beyond which diffracted intensity of all sorts becomes reduced to practically nothing. The envelopes of all these “cut-offs ” give the impression of linear boundaries for diffraction, extending from the pattern center at about 12” to the meridional direction in the case of kangaroo tendon. The fairly abrupt line cut-offs are not well explained by the theory, but this is a matter which is also influenced by the particular statistical distribution of the distortions assumed over fibril cross sections. Gaussian distributions were employed which failed to take complete account of the steric hindrances to distortion exerted by one protofibril on another or the interconnection of consecutive protofibrillar elements. Nevertheless, the general significance of collagen small-angle diffraction seems capable of being accounted for in fair detail by the theory in its present form.
THE STRUCTURE O F COLLAGEN FIBRILS
109
Hess and Kiessig (107) have suggested that the large axial periods of fibers of polyamides (nylons) and polyesters are related to structural patterns of well ordered regions (the present perfect interbands) alternating with regions of disorder (bands). The diffraction theory developed for collagen should, therefore, have some relation to the synthetic fibers also. Arnett et al. (3) have, indeed, reported for nylon 610 and polythene fibers small-angle diffraction resembling that of collagen, particularly elastoidin, in that non-meridional intensity maxima were observed. They concluded, however, that a second translational period transverse to the fiber axis is present in their material, since it failed to show a certain diffraction consequence (a shift of layer line along the meridion with tilt of the specimen in the X-ray beam), characteristic of one dimensional diffractors (21). Their own published pinhole photographs suggest that this may be a matter of experimental difficulty and not proof of the presence of a transverse spacing. The collagen small-angle diffraction is, however, much more highly developed in terms of the number and sharpness of layer lines than is that of the synthetic fibers. While in the latter the large spacings are a sort of crystallization artifact, with the collagens the large structures are clearly related to complex chemical structure.
4. Electron Optical Evidence Many of the features which appear in the model for the collagen fibril and are designed to account primarily for small-angle diffraction, can be photographed directly in the electron microscope. Indeed, much of the model development was facilitated by the availability of the electron optical evidence. Perhaps the chief difficulties preventing the latter method from immediately arriving at the same result as the former are derived from the fact that electron micrographs are not reliable, with objects of organic composition, for details much below 50 A. in size. The objects themselves are lacking in contrast and often undergo appreciable alteration or deterioration during preparation for examination and while under the action of the electron beam in the microscope. Consequently, the electron optical images can not furnish clear views of some of the details of fibril structure which have been mentioned. Nevertheless, the electron micrographs do yield direct images of fibrils, whose band and interband distributions can be seen in rough outline a t least. Inferences of finer structure can often be gained by noting physical changes resulting from accidents of manipulation. The following paragraphs collect together the observations made in this way and believed to bear structural significance.
110
RICHARD S. BEAR
First of all, the large periodic structure is readily observed in a number of ways. The fibrils, examined with minimal pretreatment (Fig. 25) show alternating dark and light segments designated, respectively, a s A and B by Schmitt et al. (192) and as D and H by Wolpers (207). Two of these, one of each kind, form a single period. Staining methods reveal th at the gross divisions just described contain fine structure (Figs. 21 and 26). Wolpers, using chiefly osmic acid, at first described (209) two subcomponents, the 61 and 6 2 disks, which develop from the broad D level by interposition of a lighter region, designated y . Later he observed the disks each to split into two lamellae (210) but believed the lamellar state to be abnormal, a sign of structural weakness. Wolpers also saw an occasional lightly staining area a t the H region. Schmitt et al. (193) used phosphotungstate to develop essentially the same five stained levels per period. Nutting and Borasky (164) obtained a similar number in formalin-tanned material stained with uranyl acetate, but also indicated that in other cases the number of stained levels can reach as high as six. Schmitt and Gross (191) found it convenient to specify five primary staining levels, a, b, c, d and el of which the first four comprise the original A or D division, while e is at the B or H region. Both b and e show a tendency t o split into two components, so that as many as seven staining components have been observed. Higher numbers are not excluded and the present limit may be largely a n experimental one. The states showing the maximum number of staining components, contrary t o Wolpers, are probably not abnormal developments, since similar numbers of levels characterized by greater than average diameter can be demonstrated by shadowing techniques on fibrils with minimal handling (90). Also, the shadowing methods, which demonstrate unevenness of fibrillar surface by overlay of metal (Cr, Pt, Au) atoms, directed onto the fibrils at a n angle to their supporting film, demonstrate that the levels which can stain most heavily also have greatest diameter (90, 164, 191), with or without prior introduction of stain. At this point, one may identify the levels which take up stain and possess enlarged diameter as the bands of the model. It was pointed out above that imperfection a t band levels should most readily permit entry of stain and cause the less compact organization leading to enhanced diameters. X-Ray evidence exists th at phosphotungstate, commonly used t o bring out the electron optical bands, does in fact choose the imperfections for entry. Graded sorption of this substance by kangaroo tendon specimens caused eventual fading of the diffraction components of imperfection arising from the bands as required b y the diffraction theory when distortion is increased sufficiently by addition of stain.
THE STRUCTURE O F COLLAGEN FIBRILS
111
The band regions attract attention in electron micrographs because of staining and enlarged diameter, but by default leave the intermediate regions of small or non-existent staining and depressed diameter to be interpreted as the more perfectly organized interbands. The size of period disclosed by the electron optics is in agreement with that of small-angle diffraction, except that allowance must be made for a degree of variability observed with the former method. While there seems to be a fair constancy of period along each individual fibril, those of a popu1at)ionmay show considerable variation (90, 164, 191). Schmitt, Hall and Jakus reported individual fibrils of rat tail tendon whose periods measured between 400 and 1000 A., although the average was 644 A. This spread is so wide that it seems doubtful from the sharpness of smallangle diffraction lines that the distribution could be as great in intact tissues. One presumes that it is at least partly due to the manipulations designed to separate fibrils for observation. A regrettable consequence of inadequate electron optical resolution has been the rise of a number of descriptive terms which give an erroneous impression of band and interband structure. For example, Wolpers (210) writes of “disks” and “lamellae.” Porter and Vanamee (177), failing to resolve the detailed band structure, described 270 A. “beads” along fine collagen fibrils, implying that they may be closely related to particulate collagen macromolecules. Wyckoff (215) prefers to call the band elevations “rings,” although alternatively they are termed “disks ” or “threads,” the latter being presumed to run transverse to fibril axes. These descriptions are suitable as long as it is realized that in most instances elements with bounding surfaces normal to fibril axes are only apparent and not basic. (One exception will be cited below in Section V, 1.) Diffraction evidence favors the assumption that the primary structural cleavage of collagen fibrils is longitudinal, continuing down to elements of protofibrillar thinness, resolvable only by means of wideangle diffraction. The nearest electron optical equivalent of this evidence is now summarized. In some collagen fibrils (rat tail tendon) there is a marked tendency for cleavage to develop along fibrils (Fig. 22), so that mechanical damage causes them to fray into finer filaments which show the original banded structure of the parent fibril (192). Others are more tightly integrated (human skin) and are rarely observed to split longitudinally (90). Diameters of collagen fibrils show a continuous range of sizes. They have been observed as small as the resolution of the microscope (specimens frayed by treatment with acetic acid, fine fibrils developed in tissue cultures, 177) up through diameters of several hundred Angstroms (beef spleen and cornea, 192) to quite large thicknesses, thousands of Angstroms
112
RICHARD S. BEAR
(mammalian skin and tendons, 192). Some tissues contain fibrils with a great diversity of diameters (tendons), and those of others show relatively constant widths (skin, 90). Observations of the type briefly covered in the last two paragraphs demonstrate that fibrils must be constituted of thin protofibrils and that depending on circumstances of fibril formation a continuous range of diameters may be achieved. Occasional reports have appeared in which thickenings or coagulations of protofibrillar elements have arisen to provide a punctate appearance across the transverse fibrillar bands or interbands (210) or to develop intrafibrillar columns or filaments (164, 191). These are readily explained as a local consequence of the same type of cleavage between protofibrils which permits their extensive fraying. This incipient cleavage occurs even in skin fibrils (90) in which the lateral forces between protofibrils are unusually strong. The apparently continuous cross striations seen in most cases indicate that the unresolved protofibrils are placed in the fibril in such a way as to align corresponding chemical features normal to the fibril axis, as was assumed for the small-angle diffraction model. This tendency toward transverse matching can be demonstrated in another way. Fibrils contain along their surfaces the same chemical array as do thc! protofibrils, so that even fibrils often occur in parallel bundles with a marked degree of correspondence between locations of band levels (90). This is shown in Fig. 24. The electron optical observations reviewed to this point have most direct bearing on those features of fibrillar structure which also can be derived from small-angle X-ray diffraction. Other important disclosures are more conveniently covered in relation to special problems below. IV. THE COLLAGEN PROTOFIBRIL The evidence discussed above seems consistent with the view that collagen fibrils are composed of very thin elements whose long dimensions are parallel to the fibril axis and whose corresponding periodic chemical features match transversely with varying degrees of perfection. None of the information hitherto cited has been capable of demonstrating the absolute size of the thin dimension of these elements. In anticipation of the conclusion soon reached below, namely, that the elements are as thin as individual polypeptide chains, they have been termed protofibrils, since they represent the thinnest structural prototype capable of carrying the characteristics of collagen. Other protofibrillar properties, such as the content of matter per unit length, the possibilities for extension, and ability to be subdivided into collagen molecules, are also derived in the next few sections, which
113
THE STRUCTURE O F COLLAGEN FIBRILS
make use principally of various types of physical and physical chemical evidence relating to these matters. 1. General Characteristics of Collagen Wide-Angle X-Ray Difraction
The wide-angle patterns yielded by collagens have been the subject of numerous investigations, none of which, however, have led t o proposal of reliable models of collagen structure. This is due t o the fact that organic fibrous systems yield faint and diffuse diffraction phenomena. The maintenance of the X-ray apparatus has often been tedious or expensive, leading the impatient or impecunious scientist t o sacrifice quality of pattern for faster and cheaper results. Even when adequately investigated, collagen will not yield a t wide angles the sharpness and wealth of diffraction detail which is desirable for orthodox crystallographic procedures. One must adopt special methods of approach and be satisfied with but rough outlines of structure. On the other hand, the proposition can be defended that a given diffraction pattern, if analyzed in all details, will eventually yield the structural order that is permissible for the material involved. Mammalian tendons are readily available and provide the best degree of fibrillar orientation, Patterns obtained from them (Figs. 3 and 7) permit observation of the maximum number of diffractions. Instead of tracing the historical development of the subject, this discussion is based on data obtained from the best patterns available t o the author. Table IV shows the average measurements for the positions of individual spot TABLEIV Small Spacings of Collagen (Kangaroo Tail Tendon) k
0 2 4 5 7
Meridian A.
Pseudoperiod
Prominent row line, A.
Secondary row line, A.
Other
A.
(a)
19.1
11.6 11.6
6.7
4.6
20.
12.
6. 6. 6.
4.
9.66 (5.0) 3.97 2.86
19.9
20.
A.
The bold figures represent the stronger wide-angle diffractions commonly observed. The column headed “meridian” indicates the spacing components of t h e layer lines parallel with the fiber axis and, except for the fourth line, most accurately measured by means of the corresponding meridional spots. The “pseudoperiod” column indicates the product of the layer-line index, k , and the related meridional spacing. Other columns give the spacing components of non-meridional diffractions transverse to the axis, on the appropriate layer lines. The equatorial diffractions are those on the line with index zero.
114
RICHARD S. B E A R
or arc centers, compiled from several air-dried specimens whose smallangle patterns had previously proved th at their fibrils were accurately parallel. Bold figures indicate diffractions whose existence has often been recognized because of sufficient intensity, while a few additional but rather faint and diffuse spots of subsequent usefulness are given without emphasis. I n spite of the good jibrillar orientation known t o exist in the specimens, the diffractions show considerable arcing, such as is generally attributed t o imperfect orientation of constituent elements. This indicates that intrafibrillar elements (protofibrils) causing the wide-angle diffraction are not as well oriented as the fibrils. Disregarding the arcing, one notices that the spots are sharpest on and in the direction of the pattern meridian and become increasingly diffuse as they depart from the meridian. Qualities of this general sort are evident on all diffraction patterns of the fibrous proteins, both a t wide and small angles. They are expected of fibrous structures because the individual diffracting elements (fibrils or aggregatcs of protofibrils) are larger and also probably less distorted in the direction of the axis than normal thereto. Indeed, one suspects that in collagen the intrafibrillar units may be packed into fibrillar cross sections with a degree of regular arrangement intermediate between that of molecules in a liquid and that in normal crystals. I n the early days of wide-angle study the dimensions of collagen and gelatin diffractions were used to estimate micellar dimensions (83, 106, 133) on the assumption th at the diffuseness was due entirely to a limitation of the volume domains over which regularity of crystalline structure was maintained. The subsequent experience with collagen small-angle diffraction, as outlined above, indicates widespread distortion within fibrils. One learns from this experience th a t distortion is a relative matter: that which does not greatly influence the length of the first small-angle layer line becomes more obvious a t higher ones. It follows that a distortion sufficient t o influence the large structures in evidence a t any small-angle layer line should have considerably diffused and deteriorated diffraction to wide-angles. The wide-angle diffraction by collagen, diffuse as it may be, must largely arise from the more perfect interbands of the fibril, since the bands contain excessive distortion. The diffuseness arid incomplete orientation of the wide-angle diagram shows that even in interbands appreciable imperfection is present, even though not of a degree to have prevented coherent diffraction by the interbands t o the small-angle line components of perfection. The interband structure, however, shows some of the diffraction characteristics of crystalline substances, since rudimentary layer and row lines are distinguishable in the wide-angle pattern.
THE STRUCTURE OF COLLAGEN FIBRILS
115
2. Pseudoperiodic Axial Structure of Small Size I n this section are considered the layer lines, from which one may obtain estimations of any small sized pseudoperiodicities extending along protofibrils at interbands. Historically the recognition of a sufficient number of layer lines for this purpose with collagen has been difficult. This is because most patterns yield only two clearly evident layer lines: those cited in Table IV as having meridional spacing components and 9.55 A. The strong meridional arcs at spacings of 3.97 and 2.86 A. were not clearly ascribable to layer lines of simple relationship to the others. Failing to recognize more than the two distinct layer lines, investigators have in the past, when they have considered the problem at all, usually adopted one of two procedures. Some have assigned index one to the 9.55 A. line, arriving at values for the period between 9.3 and 9.8 A. (102, 125, 133, 201). Others essentially have assigned index three to the strong 2.86 A. arc, thereby obtaining in the neighborhood of 8.58 A. for the period (generally cited as 8.4 A., 117, 152). The difficulty of correlating the distinct layer lines with the positions of the meridional arcs, led K. H. Meyer (153) t o suggest that the collagen wide-angle pattern may be in reality a dual pattern, resulting from the superposition of two independent diffraction systems, representing two quite different types of structure located a t various situations within the collagen fibrils. In the light of the modern small-angle diffraction and electron optical results, these two kinds of structure might arise from interbands and bands, or alternatively from different interbands, if it be agreed that wide-angle diffraction arises largely from interbands. It is, however, not entirely impossible to relate the distinct layer lines and the prominent meridional arcs. Heraog and Gonnell (105) long ago pointed out that a period of approximately 20 A. is capable of accounting for the collagen wide-angle layer lines. In terms of the measurements of Table IV the basis for a 20 A. pseudoperiod may be stated as follows: The two strong meridional arcs are not hyperbolic intersections of the flat film with widely extended reciprocal disks ( Z l ) , but their circular arc shape indicates clearly that their expression is caused by disorientation. That layer lines exist at these levels is indicated by the fainter diffuse lines with approximately the same axial spacing components. In addition, there is evidence for another layer line at about 5.0 A. These, along with the distinct ones, can be assigned indices 0, 2, 4, 5, and 7 to very nearly account for their relationships. Astbury on one occasion (9), although not citing measurements or attempting index assignments, published an idealized rotation diagram for collagen wide-angle diffraction. It is difficult t o conclude whether
116
RICHARD S. BEAR
all of the spots there shown are reliable, but they very nicely fall on the layer lines for a 20 A. period, within the probable accuracy. Indices of the layer lines there shown would be 0, 2, 3, 4, 5, and 7. Astbury has generally disregarded the desirability of arriving at a period and emphasized instead an interpretation of the strong meridional arc at 2.86 A. (discussed further in Section IV, 6). The 20 A. pseudoperiod has been largely ignored, and the reason may be the following: although the spacings of the more diffuse diffractions of Table IV are not measurable with sufficient accuracy to deny or affirm the index assignment conclusively, the stronger and sharper spots near or on the meridian are not exactly in agreement. It does not seem possible to account for these discrepancies by recourse to any of the conceivable sources for accidental or systematic error. The difficulty involves something like a 5% difference between the period (19.1 A.) calculated for the second layer line and that derived from the seventh (20.0 A .) . Nevertheless, the layer lines so nearly support the 20 A. value that one is inclined to accept it and suggest that these small variations are not unreasonable in the light of the complexities of fibril structure disclosed by the small-angle diffraction and electron optics. Using a variation of Meyer’s idea, one may suppose that the several interbands, or different parts of these, contribute in different degrees to the various layer lines, but that all such localities possess only slightly different pseudoperiods instead of the very different ones implied by Meyer’s original suggestion. This view recognizes that wide-angle diffraction represents only the smaller details of a very large structure. Because of this fact the wideangle layer lines need not be indexed in the exact proportions 0, 2, 4, 5, and 7, but could be given very large indices with nearly these relative values in the small-angle system. Although this cumbersome procedure could be adopted to satisfy crystallographic ideals, it is not a practical one. The wide-angle layer lines are so broad in the meridional direction that they could include something like ten or more layer lines of the large period. Instances have been observed where the large fibril period varies as much as 20 yo or more without accompanying measurable alteration of the wide-angle layer lines (18). Cameras capable of resolving the small-angle layer lines never resolve the wide-angle ones into several components. For these reasons, tendencies to assign large indices to the wide-angle layer lines (10, 127, 128) seem unnecessary, and it would appear more in line with real conditions t o consider the wide-angle system of diffraction as practically independent of the small-angle one. At the same time one must realize that the wide-angle pattern in this case may have
THE STRUCTURE O F COLLAGEN FIBRILS
117
special characteristics because it is the result of several, effectively independent portions of the large fibrillar structure. 3. Structure Transverse to Fibril Axes; Hydration An examination of the wide-angle diffraction row lines indicated in Table I V shows that, in addition t o the meridian, an outstanding one occurs a t a spacing of 11.6 A. transverse to the fibril axis. Still another, which may be a second order of the one just mentioned, is faintly and diffusely represented at 5.7 A. on the equator and a t higher layer lines. The prominent 11.6 A. row line, represented most obviously by its equatorial spot, is famous in the historical development as providing an outstanding example of the “intramicellar l 1 type of swelling (117, 133). When collagen is moistened this row line moves closer to the pattern meridian, corresponding to an increase of the spacing. Detailed studies have disclosed that the prominent row line’s position is a continuous function of humidity; the spacing changes from about 10.4 A. in very dry material t o an upper limit of 15 t o 16 A. with moist collagen (133) and about 17 A. in gelatin (83, 102, 118). The diffractions a t the meridian are not appreciably displaced by hydration, except perhaps for a small decrease in the spacing of the second layer line (to about 9.3 from 9.55 A.). Only the stronger layer lines, with indices 2, 5, and 7, remain visibly represented a t the meridian. The over-all picture presented by collagen wide-angle diffraction, and its alteration by hydration, is that of a structure composed of parallel, thin elements whose longitudinal structure is relatively constant during introduction of water, but whose separation, transverse to their common axial direction, is increased by the hydration. These elements meet the specifications of the protofibrils postulated in earlier sections. The 10 to 15 A., hydration sensitive, spacing furnishes a measure of the distance between planes passing through their centers (not necessarily between their centers directly, cf. Section IV, 6). There are other significant alterations to collagen wide-angle diff raction caused by hydration. Figures 7 and 8 show that the spots of the prominent row line occurring a t the equator and second layer line almost fuse with extensions of their meridional neighbors, the result being an approach t o the diffraction which might be expected of isolated protofibrils. The latter would provide solely meridionally centered, continuous layer lines, such as are expected of long, thin periodic structures (21). A difference between hydrated kangaroo tendon and elastoidin is illustrated in Figs. 9 and 10, which shows the central equatorial streaks extending out t o the location of the normally prominent 15 A. equatorial
118
RICHARD S. BEAR
diffraction. Elastoidin exhibits only the streak, while kangaroo tendon patterns possess the additional bulbous terminal spot corresponding to the spacing mentioned. This equatorial streaking was previously described for elastoidin by Astbury and Lomax (16), who cited i t as evidence of a certain amount of breakdown of regular structure transverse t o fibril axes. Apparently elastoidin, which also showed, a t small diffraction angles, evidence of unusual development of distortion in the dry condition (Section 111, 3), also undergoes extraordinary relaxation of attachment between protofibrils when hydrated. It is of interest to consider the concomitant changes introduced by hydration into the structure seen a t small diffraction angles. Figures 11 and 16 show t ha t a t small angles wet specimens lack diffraction evidence of imperfection and exhibit longer periods (665 t o 680 A.) than when dry (GOO t o 645 A. See also 18, 213). This behavior suggests that the protofibrillar distortions have been partially smoothed out by introduction of water. Empirical support for this view is added b y the observation that the lines with even indices below about ten, which often have the strongest components of imperfection with dry kangaroo tendon specimens, are the ones which lose most of their total intensity after hydration (23). Evidence t hat fibrils, when hydrated, lose to a certain extent the differentiation between band and interband regions is furnished also by electron optical experiments in which surface replicas of moist fibrils were compared with the same fibrils after drying, the latter remaining attached t o the replicas (90). The distinctly corrugated appearance of the dry fibrils is not as evident in the replicas of the moist fibrils (Fig. 23). The wide-angle and small-angle views of the effects of hydration a t first sight seem inconsistent. For example, the relaxation of protofibrillar attachment, which takes place during hydration, scarcely seems compatible with the increased perfcrtion suggested b y t)he small-angle observations. It is possible, however, th at the great amount of water introduced between and around protofibrils largely overbalances the contributions of protein to small-angle diffraction. This seems indicated by the fact t hat substances as different as are kangaroo tendon and elastoidin, when dry at small-angle diffraction (Figs. 13 and 19), become more alike in this diffraction region after moistening (Figs. 11 and 15). The type of perfertion introduced by hydration would seem t o be one which straightens protofibrillar kinks a t bands to provide the longer over-all period. The wide-angle pseudoperiod is not correspondingly enlarged, because it is related to the structure at interbands. At the latter, hydration largely influences protofibrillar separation. Additional non-meridional wide-angle diffraction is very diffuse,
THE STRUCTURE O F COLLAGEN FIBRILS
119
much of it intrinsically unresolvable into separate spots. Most peculiarly characteristic of collagen is the exceedingly diffuse half-halo, which extends above and below the equator at a radius corresponding to a spacing of about 4.6 A., nearly enveloping the equatorial spot (5.7 A.) on its inner margin. Hydration does not greatly displace the half-halo, but it becomes considerably less prominent. It is also sharpened and somewhat resolved into a diffuse row line of spacing about 4.5 A. Two possibilities seem available for this row-line: either (a) it represents finer details of structure within protofibrils, or (b) it indicates that protofibrils have a second transverse direction of approach which allows them to occur in planes as close together as 4.5 A. Either view takes cognizance of the fact that this spacing is essentially independent of changes in the separation of protofibrils brought about by hydration. Relatively large, equatorially oriented spacings, verging into the small-angle range, have occasionally been reported for collagen. One of the most persistently encountered of these was measured a t 23 A. by Gerngross, Herrmann and Abitz (83, l02), but Clark and Schaad (55) found this to be an artifact arising because of the continuous general radiation and the bromine absorption edge of the registering film. The spacings 19.9, 30.0, and 47.6 A. reported by Corey and Wyckoff (58) are also of doubtful validity, since spacings of this sort often arise in tissues because of lipid impurity. In any event, these anomalous cases are weak and by no means constantly represented on all tendon diagrams (18). Other remarks regarding the significance of wide-angle diffraction by collagen and relative to the phenomena of swelling and hydration appear more conveniently in Section V, 3. At the moment it may be said that wide-angle diffraction does not supply sufficient information for direct application of orthodox crystallographic procedures. Nevertheless, successful models for collagen fibril and protofibril organization must account for: (a) the approximately 20.0 A. axial pseudoperiod, and the attachment of particular intensity to the second, fifth and seventh layer lines of this period; (b) the spacing of protofibrils transverse to axes in planes separated by 10.4 to 15 A., depending on degree of hydration; and (c) the peculiar equatorial half-halo or very diffuse row line, whose spacing remains relatively constant at 4.5 to 4.6 A. Unfortunately, most of this information appears to apply to interband structure, and little can be derived regarding details of band organization. 4. Protofibril Extensibility
Massive fibers of collagen are capable of being extended only a little beyond normal relaxed length before rupture occurs. This fact has been widely used as an indication that the thinnest elements (polypeptide
120
RICHARD S. BEAR
chains) are alao nornially in a fairly inextensible condition (9). Electron optical studies have, however, made it abundantly clear that collagen fibrils are far from being as limited in this respect as are the grosser Jibers. For example, in Section 111, 4 it was mentioned that fibrillar periods commonly vary between something like 400 and 1000 A., the latter figure being more than 50% greater than the normal average value of 640 A. It is obvious that massive fibrils will break down when interfibrillar bonding collapses, which may occur before sufficient tension has been communicated t o individual fibrils to cause them to become very greatly stretched (109). The only way which has arisen for experimentally observing the result of application of great stress upon single fibrils has been that of relying upon accidents of electron optical study, wherein the chance breaking of a supporting film occasionally causes individual fibrils t o become caught between retracted parts of the film. An outstanding example is that described by Schmitt et al. (192) and reproduced in Fig. 25. I n this case more than a ninefold increase in the over-all length of the periodic band pattern was observed. However, when the period exceeded 4000 A. the boundaries between A and B regions became indistinct. The fine details of band pattern were not under examination a t this time. A fibril is still massive in diameter compared to a protofibril, and it is the extensibility of the latter which is of fundamental interest. One can expect with an elongating fibril to encounter at least two phenomena in addition t o simple protofibrillar lengthening: (a) a slippage of protofibrils past each other, and possibly (b) the breaking of protofibrils a t weak levels and their retraction therefrom (see further Section V, 1). The second effect will probably be encountered most seriously under the highest extensions, so that i t is desirable t o avoid this by considering only the lower and intermediate degrees of stretching. If protofibrils do not themselves extend but only slip past one another, then unless the slipping elements are very short (which is inconsistent with the markedly filamentous nature of collagen) the band pattern will have been obliterated. Since Schmitt and Gross (191) cite detailed band pattern measurements for fibrils whose periods ranged from 550 t o 1310 A., i t seems inescapable that the protofibrils must be capable of concerted extension. The question arises as t o whether all levels along a fibril or protofibril are capable of equal extension. It has been reported that the H (or B) regions rupture more easily and become thin or lengthen more readily than do the D (or A) regions when a fibril is stretched (209,219), but that lengthening of the period proceeds at expense of material drawn from the
THE STRUCTURE O F COLLAGEN FIBRILS
121
A (or D) regions (192). Finer details relative t o individual bands and interbands are difficult to secure. Although the several kinds of fibrillar or protofibrillar level may conceivably yield variously t o tension a t different stages of extension, it seems likely that all levels are not too different in content. The suggestion of Lelli and Marotta (134) that molecular chains are doubled back on themselves a t bands would place more material there than seems possible. The band elevations in unstretched fibrils are of the order of 15% or less of fibrillar diameter (164). The longitudinal cleavage phenomena indicate t h a t as many protofibrils course through cross section a t bands as a t interbands. Consequently, in dealing with protofibril extensions one should not count on more than about 30% greater extensibility from bands than from interbands. This rough calculation indicates that one might expect much of the normal variability of the large period to come from alterations in the protofibrillar distortion at bands, b u t it would seem that quite another degree of extensibility is required of both bands and interbands t o explain the greatly extended fibril of Schmitt, Hall and Jakus cited above. This must be several fold, since some evidence of band pattern could be seen when the period was as high as 4000 A., although by then some protofibrillar slippage and breakage had undoubtedly occurred. 5. Protofibril Contractility; Thermal Contraction Collagen fibrils are also capable of undergoing considerable shortening. Not infrequently they are observed with axial periods as low as 400 A., and down to a t least 550 A. the details of the band pattern have been discerned (191). There is, however, an even more extensive contraction possible, which has been recognized for almost 150 years as an outstanding characteristic of connective tissue (for the early history see 206). This is the so-called thermal shrinkage, which resides in the collagen component. When collagen fibers are heated to GO'C., more or less, depending on previous history and environmental conditions, they shorten to something like a third or a quarter of their initial length. Various hypotheses have been advanced to account for the thermal contraction. Some of the main lines of thought are briefly described below. An early suggestion was provided by the coagulation theory of Hermann (101), who argued from the similarity of the conditions which cause protein coagulation (denaturation) and collagen shrinkage that the two phenomena are closely alike. Subsequent thermodynamic studies pave shown why this is so: both denaturation (67, 159) and the shortening hrocess (203) require preliminary, rate-determining, activation processes
122
RICHARD S. BEAR
involving large increases in heat content (AH”) and in entropy (AS*). Because of large AH*, the temperature coefficients are also large and the reactions proceed with an apparent but not actual abruptness near a temperature I”, which is about 60°C. under minimal swelling conditions in aqueous environments. Then TAS* nearly balances AH* to yield small enough free energy change, AF*, to permit appreciable activation to occur. From the kinetic standpoint the large AH* is accounted for by assuming that the large molecuIes invohed require for activation the breaking of a number of relatively weak bonds which normally serve to hold the protein in its native configuration. In the case of the collagen fibril this would mean the breaking of links between the neighboring protofibrils, and possibly also the release of intra-protofibrillar bonds tending t o stiffen them individually, in sufficient number to permit the subsequent relaxation. The large AS* corresponds to the increased freedom of motion permitted the protofibrillar segments after this release. Weir (203) gives typical values for the thermodynamic quantities involved in the shrinkage of kangaroo tail tendon in water a t 60°C. : AH* = 141 kcal./mole, AS* = 349 cal./mole-degree, and AF* = AH* TAS* = 24.7 kcal./mole. These figures apply to a “molecule” whose weight is not known with certainty. On the assumption that the subsequent contraction process is adiabatic (AH = 0) and that the AH* of the activation process is the only heat change involved in a reaction whose over-all heat absorbed has been measured to be 12.5 cal./g., Weir calculates a “molerular weight” of 11,300 for collagen. This need not be the weight of a complete collagen molecule or protofibril, but can represent the size of segment which is required to be activated a t one time. The assumption that the contraction process is ideally adiabatic, while perhaps not entirely permissible practically, seems indicated by modern theory of the behavior of molecular chains, which pictures these as undergoing, when freed of restraints, a sort of segmental diffusion, much like the adiabatic expansion of an ideal gas into a vacuum (155). I n the case of the molecular chain, it diffuses to the most probable, randomly coiled configuration, which is much less asymmetric, hence shorter, than a n initially extended chain. Because rubber most nearly presents this ideal behavior, those fibers which develop increased tension (a measure of the tendency toward assumption of the contracted form) when held isometrically under conditions of increasing temperature (favoring the “diffusion”) are said t o be “rubber-like.” Most normal elastic solids upon stress are strained from some stable structure and relax as the temperature is raised. Protein molecular chains often show more nearly the normal elastic
THE STRUCTURE OF COLLAGEN FIBRILS
123
behavior, presumably because of a tendency to prefer particular folded configurations. This is the case with members of the k-m-e-f group of fibrous proteins a t moderate temperatures and swelling (212), but after increased swelling and above a sort of “rubbery melting point,” which depends on the conditions, the preferred or quasicrystalline configurations “melt” and rubberlike behavior ensues (62, 141, 156). The collagens have not been as extensively studied in these respects as might be desired, but some evidence is available. Wohlisch and de Rochemont (206) believed contraction t o be a reversible transformation, a t constant temperature, between two states I and 11, corresponding, respectively, to the solid or crystalline state and the melted or dissolved condition of ordinary crystalloids. Somewhat earlier it had been suggested that during contraction collagen is transformed into gelatin. Working initially with beef tendon, Wohlisch did not immediately identify collagen I1 with gelatin, though he did so after experience with the more easily gelatinized rat tail tendon (205). Cherbuliez, et al. (51) believed it possible to distinguish the reversible “melting l 1 of the polypeptide chains from the irreversible hydrolysis which produces gelatin. These authors compared pig achilles tendon, elastoidin, and ichthyocol in these respects, finding that ichthyocol most readily yields gelatin, while elastoidin does so only with difficulty (see also 69). Elastoidin furnishes particularly suitable material for study of thermal shrinkage because it provides an approach t o a reversible transformation between states I and I1 without gelatinization complications. Native elastoidin (state I) shows many characteristics of a collagen. It has a quasicrystalline structure yielding the usual wide-angle diffraction (16, 47) and a small-angle diagram which, while distinctive, as discussed in Section 111, 3, is still recognizably that of a collagen. Some of the other physical and chemical properties, such as the positive uniaxial double refraction (69), the resistance to tryptic hydrolysis (69) and a thermoelastic behavior like that of normal solids, with negative temperature coefficient (173) also indicate that elastoidin I has a n ordered structure. Transformation t o state I1 occurs in water near 62”C., but a restraining tension raises the temperature required and swelling media lower it (75). The elastoidin 11, which is produced and is stable a t temperatures above the transition point, has quite new properties. It is contractile and rubbery, shortening spontaneously to 27y0 of the original length (69) ; its thermoelastic coefficient is large and positive, attesting t o its rubberlike nature (173). In shortening it loses its ordered structure, yielding a single wide-angle diffraction halo characteristic of a n amor-
124
RICHARD S. BEAR
phous state (47). The double refraction diminishes and is said to change sign from positive to negative (69), though this behavior is inconsistent with a n amorphous condition and may result from internal strains set up in an accompanying non-collagen matrix (see further below). The resistance t o trypsin is markedly diminished (69). When cooled, elastoidin I1 shows signs of reversion t o state I (69). It regains spontaneously over half of the initial length at 20"C., the positive double refraction and the normal wide-angle diffraction return, and there is even some regain of resistance t o trypsin. Elastoidin differs from the general run of collagens in respect to these reversionary tendencies. Gerngross and Katz (81), in a n early study of tanned and untanned avian tendons, found that the untanned preparations shorten and remain shortened, losing their normal wide-angle diffraction (except for the diffuse halo characteristic of the amorphous condition) unless forcibly re-extended. Formalin-treated, or lightly chrome-tanned specimens, on the other hand, showed the behavior, originally described by Ewald (66), of spontaneous partial regain of length, and, correspondingly, the collagen wide-angle pattern also returned. A similar formalin-fixed rat tail tendon has been carried through the shortening and spontaneous lengthening cycle, with examination of the ability of the final preparation t o diffract a t small angles (18). A diagram was obtained consisting solely of the first three layer lines of a period which had been reduced to 550 A., corresponding closely t o the macroscopically observed net loss of 13% in length. Normally, without the formalin stabilization, the tendon loses all discrete small-angle diffraction, even after forcible re-extension. This result throws light on the meaning of the formalin stabilization, which must link the protofibrils of each fibril so that they do not become hopelessly disarranged during the shortening phase and can return on lengthening t o reconstitute, in part a t least, the original fibrillar structure. The native stabilization of elastoidin is probably produced in another fashion; as mentioned in Section 11, 5 , elastoidin is impure, and the fibrillar stabilization may result from confinement in a n amorphous matrix which tends to keep the protofibrils of a fibril together. The stabilized collagens show the effect of restraints; for example, Gerngross and Katz found their untanned tendons could shorten 89% at 60 t o 65°C. but the formalin-treated ones contracted only about 60% at 85 t o 90°C. In elastoidin the stabilization is not intrafibrillar, as seen by its normal shortening temperature (62"C.), but it is restrained extrafibrillarly from shortening more than 73%. After the first cycle of shrinkage and re-extension, these cases may be carried with more com-
THE STRUCTURE O F COLLAGEN FIBRILS
125
plete reversibility through subsequent cycles, though the transformation temperature is lower than initially and the condition in state I is never that of the native collagen. The essentially cyclic nature of the contraction and re-extension process has been emphasized because it illustrates particularly clearly the facts that (a) the cyclic phenomena are primarily protofibrillar properties, the fibril stabilization being only the means by which the massive sample behaves as do the individual protofibrils; and th a t (b) the normal state of protofibrils, ie. a t ordinary temperatures in the absence of agents producing excessive swelling, is not the contracted, random form, but the extended condition, possessing the specific configuration which yields the collagen wide-angle diffraction pattern when several protofibrils are arranged properly with their long axes parallel. I n other words, the amorphous state I1 is not the normally stable condition; the ordered configuration of state I is preferred. The situation is very much like that in gelatin sols and gels. Katz, Derksen and Bon (116, 119) have shown that a freshly prepared gelatin solution has amorphous structure, which on cooling transforms into a gel yielding the collagen wide-angle diffraction diagram. I n this sense, the earlier investigators were correct in comparing the thermal contraction of collagen t o a transformation into gelatin, even though the actual production of gelatin may not be demonstrable. An early rival of Hermann’s coagulation theory was the swelling theory of Engelmann (64), and, indeed, the problem of thermal contraction can not be dismissed without consideration of the relation of swelling to the phenomenon. It has long been known th a t a marked swelling of the collagen is not obviously involved in contraction, although there is a small increase in volume. Nevertheless, prior swelling of the collagen is quite generally effective in reducing the temperature required t o effect it (123, 135, 199). Perhaps the most elegant way of expressing the effects of the swelling reagents is by way ‘of the changes in the thermodynamic quantities involved. Weir (203) gives data of this sort for kangaroo tail tendon, showing t ha t acids, alkalies and neutral salts generally decrease A H * , AS* and AF*, except th at concentrated salts increase AF*. Tans, which generally oppose swelling, increase AF*, but in ways dependent on the tan: most of them reduce AS* more than AH*, but chrome increases both of these markedly. Lowered AF*, of course, means a lesser temperature required for activation leading t o contraction, and decreased AH* corresponds t o a broader temperature interval for measurable contraction rates. The interpretation of the thermodynamic quantities in terms of
126
RICHARD S. BEAR
detailed kinetic pictures is not easy, however, since the activation process involves not only collagen but the water and other reagents present. The thermodynamic quantities represent the net result of energy and entropy changes in all components of the system involved in activation. This may be made clearer by a consideration of the quantities cited above for contraction of tendon in normal aqueous environment. According to the chemical analyses of mammalian collagen (Table VI) the average residue weight is 93. If a protofibrillar segment of weight 1 1,300 is involved in the activation process, then about 121 residues participate. On the assumption that the activation energy, AH* = 141 kcal., is largely used to break weak hydrogen bonds of 5 kcal. each within the protein structure, 28 bonds are involved, or one for every four to five residues. In actual fact, however, a rupture of more than this number is probably required before the structure is sufficiently weakened. While these are being broken within the protein framework, others between protein and water are undoubtedly being formed, so that AH* is the result of heat absorbed in releasing intraprotein links and heat evolved as hydration occurs. It is well known that removal of water greatly increases the shrinkage temperature. Thermodynamic properties of the contraction by relatively anhydrous collagen would come closer t o furnishing information directly interpretable in terms of what is happening to the protein itself. At different places in a massive sample of collagen the activation necessary to initiate contraction is not always the same. Thus, shortening can arise at local regions while others remain unaffected. The local contraction probably also distorts neighboring fibrils to reduce their activation requirement, so that the contraction spreads from the initial loci. This phenomenon is illustrated particularly clearly by Weir (203). Small-angle diffraction studies disclose that as shortening progresses, during stages short of completion the sample still exhibits fading normal diffraction rather than continuous other change (18, 214). The contracted portions lose ability to diffract at small angles while the uncontracted ones retain it, and the proportion of the latter diminishes gradually. Contributions of electron optics to this problem have been disappointingly meager, possibly because of difficulties in properly arranging for observation of contracted individual fibrils. Nutting and Borasky (164) briefly indicated that heating produces fibrils with greatly enlarged diameters and probably shortened lengths, though banding is lost. The clearest indication of shortening was seen in fibrils with a spiral contra(.tion of a central undispersed core produced by shortening of the outer layers. Wolpers (211) also found a tendency for band levels to disperse
THE STRUCTURE O F COLLAGEN FIBRILS
127
transversely t o fibril axes, and when the fibrils were sufficiently shortcned all visible structure was lost. 6. The Content per Unit of ProtoJLibrillarLength I n the last two sections compelling evidence has been cited t o indicate that collagen protofibrils, while normally in a configuration characterized by the 640 A. period when dry, may be forcibly extended t o several times this length, or they can spontaneously contract to a fraction of their normal length when thermally activated. Models for the detailed protofibrillar structure must contain enough polypeptide chain per unit of length t o allow, in particular, the remarkable extensions observed. The wide-angle diffraction data do not clearly enough indicate the exact interpretation of protofibrillar dimensions normal t o the long axis to permit one to derive unequivocally the amount of protein which density considerations permit to exist per unit axial projection. Nevertheless, a consideration of density limitations in relation t o other protofibrillar properties produces some interesting suggestions, as now follow. The most widely circulated views relative to this matter are those of Astbury. His suggestions are as follows, revised to expression in terms of magnitudes employed in this paper. The argument essentially starts with the assumption th a t the 10.4 and 4.6 A. spacings of collagen, transverse t o fibril axes (Section IV, 3) are analogous, respectively, t o the “side-chain l 1 and “back-bone ” spacings of &keratin structures (13). I n this way two dimensions of a protofibril (polypeptide chain) would be established normal t o the long axis. Assuming that these dimensions may not be strictly orthogonal, but include a n unknown angle p, one readily calculates th a t a single residue of average weight 93 would satisfy a typical protein density of 1.35 if it occupied an axial length of about 2.4 sin ,8 A. (15). Thus, the length per residue is of the order of 2.86 A., the prominent meridional spacing of collagen, which has, therefore, been regarded as having this significance. Proceeding in the reverse direction, one calculates a density of 1.12/sin @ for collagen or gelatin if the average residue occupies a volume 10.4 X 4.6 x 2.86 sin ,B A.3 (see 13). While this density value is low, uncertainties regarding p and the value of the very diffuse 4.6 A. spacing could allow correction upward. According t o the present suggestions that wide-angle diffraction may be related to only a part of the total structure, it is conceivable that the average residue may not be involved, and heavier than average residues at interbands could also increase the calculated density. As subsequent considerations show, however (see
128
RICHARD S . BEAR
Section V, 2) i t seems likely that interbands contain lighter than average residues. A maximum extension of about 3.64 A. per residue is permitted (57) for polypeptide chains arranged as shown in Fig. 30a. Silk yields the highest observed value of 3.5 A. (157), and normally the /3-keratins show still lower values, 3.1 to 3.3 A. (14). While it is possible to suppose th a t silk and p-keratin have polypeptide chains which are somewhat crumpled and contracted from maximum extension, collagen would require rather special treatment for explanation of a residue extension as low as 2.86 A. /
OC\
T-!
HN\
To
"
R-CH
7.0
OC\ HN
\ a
b
C
FIG.30. Polypeptide chain configurations.
Astbury has from time t o time proposed different reasons for this low value in collagen. His most recent model (9) is shown in Fig. 30b. A partial cis configuration of the main chain causes shortening, at the same time allowing the prevalent pyrrolidine (of proline and hydroxyproline) and glycine (hydrogen) side-chains (see Section V, 2) to be on the same side of the main chain, opposite the other bulky ones, t o avoid side-chain congestion. Attempts to stretch this model are said to cause it to coil upon itself, which is believed to explain the supposed inability of collagen polypeptide chains to be extended. There are several objections which might be made to the Astbury
THE STRUCTURE O F COLLAGEN FIBRILS
129
model. The extension per residue is not invariable but only averages 2.86 A., and one would expect the distance between pyrrolidine residues (3 X 2.86 = 8.58 A.), rather than 2.86 A. itself, to be heavily represented in the wide-angle diagram. The meridional spacing nearest t o 8.6 A. is observed a t 9.5 A.; indeed, no serious attempt appears to have been made to consider what fiber pseudoperiod may be present in collagen and how the model could explain it. Huggins (112) has also made more general types of objection to the models proposed by Astbury. The main polypeptide atom chain of the Astbury model does not possess a screw axis of symmetry. The unbalanced forces on opposite sides of such a chain would tend to prevent it from readily adopting a straight course, bending it continuously in some direction. Huggins also believed it likely that corresponding atoms along the main chain should be surrounded with similar configurations of atoms (except possibly for differences in side chains), which the Astbury model does not accomplish. The high content of pyrrole ring side chains in mammalian collagen may obviate the validity of the second objection, but the first is less easily escaped. Huggins offered the coiled model for a polypeptide chain shown in Fig. 30c, believing this suitable for collagen. Such chains were placed side by side in a thick plane, joined by hydrogen bonds such as > N-H . . . 0 = C <. The distance between chain centers in the layer thus formed corresponded to 4.5 A., or essentially the spacing of the wide-angle diagram’s diffuse equatorial half-halo. The layers were regarded as stacked, with a separation of 11 A. determined by the side chains which extend above and below the layer planes. This corresponds to the spacing of the moisture-sensitive, prominent row line of the diffraction pattern. The most serious objection to the Huggins model is found in the axial pseudoperiod of 5.8 A., which represents the projected length of two residues. While half this value is close to the 2.86 A. meridional spacing, the discussion of Section IV, 2 shows that other prominent layer lines are present which bear no simple relation to 5.8 A. Ambrose and Elliott (2) have recently suggested that a slightly modified form of the Huggins model can account for the facts regarding collagen wide-angle diffraction and infrared dichroism. The objections applicable to the Huggins model may be made again here. The modifications were designed largely to satisfy the infrared data, which suggest an apparent preponderance of orientation of carbonyl and imino bonds of the main chain normal to the fiber axis. The Astbury, Huggins and Ambrose-Elliott models give essentially the same interpretation to the spacings 11, 4.6, and 2.9 A., differing only in the manner in which the main chain is contracted to present one residue
130
RICHARD S. B E A R
per 2.86 A. of axial projection, or in other details. Full extension could produce a t the most only about 28% increase beyond normal protobrillar length. The evidence presented in Section IV, 4 seems, therefore, t o exclude these models. Similar objection can be raised with regard t o the models of Zahn (218), which are essentially the result of exploration attempting to preserve the 2.86 A. spacing as the projected length of a single residue. The interpretation pf the 2.86 A. spacing as the length of a single residue in collagen has been fixed so firmly in the thinking of investigators that a number of recent, otherwise provocative trends of thought have become seriously confused. For example, Mustacchi (160) realizing the significance of the great extensibility of collagen fibrils and protofibrils (micelles, to him), was led, in part by a misunderstanding of the significance of X-ray diffraction data, to postulate unnecessarily complex deductions regarding the numbers of chain molecubs in collagen " micelles." Pauling and Corey (170), following lines of thopght initially very similar t o those developed below, were also induced by a belief in the essential inextensibility of collagen molecular chains, t o devise a model for collagen which fails to account for the actual extensibility of collagen protofibrils. It seems, therefore, desirable to attempt a n entirely new departure in arriking a t a conception of the protofibrillar configuration. The discussion of small-angle diffraction and electron optical evidence of pdqeding sections has disclosed that there is every reason to believe that fibrils are cylindrical objects, with reasonably- good integration over entire cross sections (particularly a t interbands whfch yield. the wide-angle diffraction) , and that longitudinal cleavage does-not produce laminar but essentially cylindrical filaments or protofibrils, as far as can be seen. For the new approach, suppose that the protofibril, like thedbril, is essentially cylindrical. I n the fibril these thin cylinders will achieve closest packing if they adopt a hexagonal arrangement of centers a t cross sections as shown in Fig. 31. The spacing between planes through centers is assumed t o be the 10.4 A. spacing of dry collagen. This corresponds t o an actual separation between protofibrils of 10.4/cos 30" or 12.0 A. The cross sectional area available to a single protofibril is 124.8 A.2 It is then easily determined that the average residue of weight 9 3 accounts for a density of 1.35 if the residue has an axial projection of about 0.91 A. According t o a model of this sort, a meridional spacing of 2.86 A. would correspond more nearly to the projected length of three residues. at 0.95 A. per residue, rather than t o that of a single residue. Indeed, examination of the measured layer-line spacings of Table I V discloses that a variability of residue extension between 0.91 and 0.95 A., with the lower value for the second layer line and the higher extreme for the
THE STRUCTURE OF COLLAGEN FIBRILS
131
10.44
20.0A
FIG.31. Illustrating helical models for the collagen polypeptide chhin (protofibril). Wn the left is the skeleton of the helix, with the locatioq of atoms filled in for the bottom turn. The numbefed carbon (solid black) atoms of the skeleton represent the 21 plaqes (CY carbon atoms) at which residue side chains are attached in one full period of the helix. The number of turns per period illustrated is 4, indicated in general a s n. The diameter of the helix skeleton is d. The upper right part of the drawing indicates cross sectional plans for three helices and their mode of packing in the collagen fibril. I n one helix the atoms of a single turn are shown; in another, lines directly joining the Q carbon atoms of the residues are given for a full period; and in .the third, their general circular locus of projection is indicated. Dotted lines show rough domains within which penetration of side-chains from outside each helix is largely prohibited by van der Waals repulsions of the helix skeleton. At the lower right is shown one unrolled coil, with one residue indicated in detail. Stippled circles represent nit,Eogen atoms, -small open circles are hydrogen, large open circles oxygen, and shaded circles show the general direction for emergence of R groups (side chains).
132
RICHARD 6. BEAR
seventh, indicates that 21 residues would be accommodated in the 19.1 to 20.0 A. pseudoperiod. To account for small residue projections of this sort a polypeptide chain must clearly be coiled. When this is done according to the above specifications one obtains a model which is capable of being extended maximally almost fourfold, though somewhat less will be realizable if the pyrrolidine residues prevent full extension (see 167). Coiled models for polypeptide chains have been proposed with increasing frequency in recent years. Even before Huggins, Taylor (197) had suggested a coiled model suitable for a keratin, with three residues per turn. Bragg et al. (43), in particular, have explored a wide variety of possible folded chain configurations, emphasizing the usefulness, in providing fold stabilization, of hydrogen bonds between intrachain > NH and >CO groups, separated in position along the chain but by folding or coiling brought into the proximity required for bonding. Pauli