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!
n i (r s , r d )/ O x (r s , r,)] exp . Here, x (r s , r,.) represents the excitation fluence detected at a fixed reference position, r,, in response to excitation from the source position, r s . Even though the excitation fluence, ^Trs, r,), is updated during the iteration, the Jacobian matrix can be directly calculated from Eq. (36), and
Near-Infrared Imaging
509
the change in <J> x (r s , rr) is small compared to that in <J> m (r s , r r ). As the source term of the emission diffusion equation is modified after each iteration [Eq. (22)], changes in the emission fluence are greater than that of the excitation fluence. The same consequence can be inferred from the integral equation [Eq. (36)]. Moreover, the phase of the emission fluence is greater than that of the excitation fluence and the normalized fluence, mAI>x, maintains a high phase contrast. Owing to noise and the ill condition of the Jacobian matrix and for inverting the systems of equations, updating can be accomplished using Newton's method [107] with Marquardt-Levenburg parameters A:
(37) Using excitation referencing at a single reference point, Lee and SevickMuraca [96] reconstructed an 8 X 4 X 8 cm3 phantom containing a 1 X 1 X 1 cm3 target with 100-fold greater ICG concentration, by using 8 excitation sources, twenty four detection fibers for collecting excitation light, and two reference detection fibers (one on either side of the reflectance and transillumination measurements) for collecting excitation light. Figure 28a is the original map containing two-dimensional slices of the three-dimentional geometry that demark the heterogeneity placement, while Fig. 28b is the three-dimensional reconstructed image. The results in Fig. 28 represent reconstructions based on emission FDPM measurements relative to excitation FDPM measurements at a fixed reference position: Ntziachristos and Weissleder [95] successfully reconstructed two fluorescent targets in a 2.5-cm diameter, 2.5-cm-long cylindrical vessel containing ICG and Cy5.5, and using CW emission measurements referenced to excitation measurements at each of the 36 detector fibers as a result of point excitation at 24 source fibers. The high density of measurements for reconstruction of the small simulated tissue volume is troublesome for validity of the diffusion equation used in the forward solver, but is similar to that demonstrated by Yang and coworkers [49] who reconstructed ICG and DTTCI in similarly sized phantoms and mice, presumably from absolute FDPM measurements at the emission wavelength alone. It is noteworthy that the studies of the reconstruction presented above assumed that the absorption and scattering properties were known a priori. However, using differential approaches coupled with Bayesian reconstruction approaches (see below), Eppstein and coworkers [108] demonstrated the insensitivity of reconstructions to changes in endogenous optical properties. Using a synthetic 256-cm3 volume containing 0.125-cm3 targets with 10:1 contrast in absorption owing to fluorophore and surrounded on four
510
Sevick-Muraca et al.
FIGURE 28 The reconstruction of piaxf using the excitation wave as a reference using the integral approach and Marquardt-Levenburg reconstruction. The image was required after 27 iterations with regularization parameters for AC ratio (ACR), AAC = 1.0, and for relative phase shift (RPS), A,, = 0.02 (a) optical property maps of true juaxf distribution (b) and reconstructed /uaxf distribution. Peak values of /uaxf reached 0.1205 cm 1 (c) Iteration vs. SSE.
Near-Infrared Imaging
511
sides with 68 sources and 408 detection fibers, Eppstein showed that when the absorption cross-section at the excitation wavelength, /zaxi, varied as much as 90% and was unmodeled while the scattering coefficient, ;ttsxl, varied 10% or less and was also unmodeled, the impact on the reconstruction was minimal or negligible. Similar results have recently been shown by Roy and Sevick-Muraca [ 109] who show unmodeled variations in all endogenous optical properties by as much as 50%, which did not impact reconstructions when emission FDPM measurements were individually self-referenced to excitation FDPM measurements, as was done with the CW measurements of Ntziachristos and Weissleder [95]. While it appears promising that fluorescence-enhanced optical tomography can be accomplished without much a priori information regarding the endogenous optical properties, these results are nonetheless on synthetic studies and need to be conducted on actual tissues of substantive and clinically relevant volumes for validation. 5.3
Differential Formulation of the Inverse Problem
A second approach of the full inverse imaging problem may be the differential formulation, but in reality, this time it is rewritten for measurement Z(r d > r s ), whether absolute, or relative to a reference measurement at the emission or excitation wavelength, or self-referenced relative to the excitation wavelength at each detector position, rd. We term this approach the differential formulation because a small change in the predicted measurements is directly expressed in terms of a small change in the optical properties, AX, using a Jacobian matrix, J, d(AZj)/3Xj. Considering the number of detectors to be M; then the error function is defined as the sum of square of errors between the measured and calculated values at detector i = 1..M:
F(X) = We refer to each f; as a residual and the gradients of the error function with respect to the property, X: VF(X) = 2JTf(X) |-
(39) M
V2F(X) = 2 JTJ + 2 L
i= l
-j
fi(X)V 2 f,(X)
(40) J
Consider the Taylor's expansion of function F around a small perturbation of optical properties, AX: F(X + AX) = F(X) + VF(X)-AX + - AX r - V 2 F(X)- A(X)
(41)
512
Sevick-Muraca et al.
which can be expressed as: F(X + AX) = F(X) + 2JTf(X) • AX JTJ +
+ 2 • AX
f,(X)V 2 f,(X) ' AX
(42)
and the function to be minimized, <J>(AX), can be explicitly written: 4>(AX) = F(X + AX) - F(X) = 2JTf(X) • AX + 2- AX T J T J +
L
t=i
ft(X)V2f}(X)
J
-AX
For first order Newton's methods, the term 2-AXT[E!l1 f i (X)V 2 f i (X)j • AX is neglected and the Gauss-Newton Method becomes one of minimizing, VO(AX) => 0 = JTJ • AX + J T f(X)
(44)
JTJ AX = -J T f(X)
(45)
The Levenberg-Marquardt method of optimization becomes [JTJ + A I ] - A X = -J T f(X)
(46)
The gradient based truncated Newton's method is based on retaining the second-order terms such that Eq. (43) becomes [89]: VO>(AX) => 0 = J T f(X) +
JTJ + L
fi(X)V 2 f,(X) • AX i= l
J
or, alternatively, VO(AX) => 0 = VF(X) + V2F(X) • AX
(48)
Typically, the first order Newton's methods are employed with the exception of the work by Roy [110]. In Newton's methods, it is assumed that A = J • AX and the solution is found using one of the several optimization approaches. The Jacobian can be computed either directly from the stiffness matrices of the finite element formulation or, simply but more computationally time consuming, from backward, forward, or central differencing approaches that compute the differences in the values of Z(r d , r s ) with small differences in the parameter to be updated, X(r s ). The Gauss-Newton and the Levenberg-Marquardt algorithms performed poorly in a large residual problem. Since the inverse is highly nonlinear and ill conditioned due to the error in measurement data, the residual at the solution will be large. It seems reasonable, therefore, to consider the truncated Newton method.
Near-Infrared Imaging
513
For the truncated Newton's method, the additional computational cost of computing the Hessian [associated with V2F(X)] is assisted by reverse automatic differentiation [89,110]. Using synthetic data, Roy has shown the feasibility for using the technique for three-dimensional reconstruction of lifetime, T, and absorption coefficient ^iaxf changes in frustrum and slab geometries from synthetic data containing noise that mimics experimental data [93].
5.4
Regularization and Other Approaches for Parameter Updating
In both the integral and differential formulations of the inverse problem, the tissue to be imaged must be mathematically discretized into a series of nodes or volume elements (voxels) in order to solve these inverse problems. The unknowns of the inverse problems then comprise the optical properties at each node or voxel. The final image resolution is naturally related to the density nodes or voxels. However, the dimensionality of the imaging problem is directly related to the number of nodes and can easily exceed 10,000 unknowns for a three-dimensional image. In a problem of this scale, the calculation of Jacobian matrices and matrix inversions involved in updating the optical property map are computationally intensive and contribute to the long computing times required to reconstruct the image. The instability arises because the measurement noise in the data or errors associated with the validity of the diffusion approximation can result in large errors in the reconstructed image. One of the greatest challenges associated with fluorescence-enhanced tomography is the propagation of error. In comparison with absorption imaging based on measurements of excitation light, fluorescence measurements have a reduced signal level and SNR. Lee and Sevick-Muraca [111] measured the SNR for single-pixel excitation and emission frequency domain measurements at 100 MHz to be 55 and 35 dB, respectively. In addition to the reduced signal, the noise floor of emission measurements can be expected to be elevated when excitation light leakage constitutes an increased proportion of the detected signal. Consequently, for emission tomography measurements, excitation light leakage is crucial and interference filters that attenuate excitation light four orders of magnitude (i.e., filters of optical density 4) may be clearly insufficient. Excitation light leakage will be a significant problem when emission measurements are conducted in tissue regions in which the target is absent and fluorescent contrast agents are not activated. Unfortunately, this type of error is not present in synthetic studies and is undoubtedly underestimated in the vast proportion of tomographic investigations to date.
514
5.4.1
Sevick-Muraca et al.
Regularization
Regularization is a mathematical tool used to stabilize the solution of the inverse problem and to make it more tolerant to measurement error. Regularization approaches will play an important role in the development of suitable algorithms for actual clinical screening. For example, when discretized the differential and integral general formulations result in a set of linear Newton's equations generally denoted by AY = Z, where Y are the unknown optical properties and Z are the measurements. This system is commonly solved in the least-squares sense where the object function Q = ||AY — Z||2 + A Y 2 is minimized, where A is called the regular!zation parameter. Minimization of this function results in Y = (A1 A + AI) 'ATZ. The regularization parameter is generally chosen either arbitrarily or by a Levenberg-Marquardt algorithm so that the object function is minimized [112]. Thus, the choice of regularization parameter is through a priori information and adds another degree of freedom to the inverse problem solution. While this section is not meant to be a mathematical treatise of inverse algorithm and regularization approaches, we nonetheless point out that in a recent work by Pogue and coworkers [100], a physically based rationale for empirically choosing a spatially varying regularization parameter is presented to improve image reconstruction. Bayesian Regularization. Eppstein and coworkers [87,88,91,108] use actual measurement error statistics to govern the choice of varying regularization parameters in their Kalman filter implementation to optical tomography. In their work, they developed novel Bayesian reconstruction technique, called APPRIZE (Automatic Progressive Parameter-Reducing Inverse Zonation and Estimation), specifically for groundwater problems and adapted them to fluorescence-enhanced optical tomography [104-106]. Unique components of the APPRIZE method are an approximate extended Kalman filter (AEKF), which employs measurement error and parameter uncertainty to regularize the inversion and compensate for spatial variability in SNR, and a unique approach to stabilize and accelerate convergence called data-driven zonation (DDZ). Using the notation [AX, f(X)] as described in Sec. 5.3, here the Newton's solution is formulated as [91]: AX = [[JT(Q + R) 'J + P;X'] ' • JT(Q + R ) - ' l - f ( X )
(49)
where Q is the system noise covariance which describes the inherent model mismatch between the forward model (the diffusion equation) and the actual physics of the problem; R is the covariance of the measurement error that is actually acquired in the measurement set; and Pxx is the recursively updated error covariance of the parameters, X, being estimated from the measurement error, f(X). The use of this spatially and dynamically variant co-
Near-Infrared Imaging
515
variance matrix results in the minimization of the variance of the estimated parameters taking into account the measurement and system error. The novel Bayesian minimal variance reconstruction algorithm compensates for the spatial variability in signal to noise ratio that must be expected to occur in actual NIR contrast-enhanced diagnostic medical imaging. Figure 29 illustrates the image reconstruction of 256-cm3 tissue-mimicking phantoms containing none (case 3), one (case 1), or two (case 2)
FIGURE 29 Image reconstruction with APPRIZE, a) The initial homogeneous estimate discretized onto the 9 x 1 7 x 1 7 grid used for the initial inversion iteration, and shown with the true locations of the 3 heterogeneities and the 50 detectors (small dots), b) Case 1: The reconstructed absorption due to the middle fluorescing heterogeneity, interpolated onto the 17 x 33 x 33 grid used for prediction, and shown with the locations of the 4 sources used (open circles), c) Case 2: The reconstructed absorption due to the top and bottom fluorescing heterogeneities shown with the locations of the 8 sources used (open circles), d) Case 3: The reconstructed absorption of a homogeneous phantom shown with the locations of the 4 sources used (open circles). Although the phantoms and reconstructions were actually 8 cm in the vertical dimension, only the center 4 vertical cm is shown here. (From Ref. 91.)
516
Sevick-Muraca et al.
1-cm3 heterogeneities with 50- to 100-fold greater concentration of ICG dye over background levels. The spatial parameter estimate of absorption owing to the dye was reconstructed from only 160 to 296 surface reference measurements of emission light at 830 nm (as described in Sec. 5.2.2) in response to incident 785-nm excitation light modulated at 100 MHz. Measurement error of acquired fluence at fluorescent emission wavelengths is shown to be highly variable. Another important feature of the Bayesian APPRIZE algorithm is the use of DZZ. With DDZ, spatially adjacent voxels with similarly updated estimates are identified through cluster analysis and merged into larger stochastic parameter "zones" via random field union [113]. Thus, as the iterative process proceeds, the number of unknown parameters, X, decreases dramatically, and the size, shape, value, and covariance of the different "parameter zones" are simultaneously determined in a data-driven fashion. Other approaches to reduce the dimensionality of the problems involve concurrent NIR optical imaging with MRI [10,114,115] and ultrasonography [116] to compartmentalize tissue volumes and to reduce the number of parameters to be recovered in the optical image reconstruction. 5.4.2
Simply Bounded Constrained Optimization
Imposing restrictions on the ill-posed problem can transform it to a wellposed problem as discussed above. Regularization is one method to reduce the ill posedness of the problem [117]. In the optical tomography problem, its solution, i.e., the optical properties of tissue, must satisfy certain constraints, and imposing these conditions in itself can regularize or stabilize the problem. Imposing these constraints explicitly restricts the solution sets and can restore uniqueness. Provencher and Vogel [117] have suggested two techniques: (1) prior
FIGURE 30 Three-dimensional reconstruction from simply bound truncated Newton's method. Actual distribution of fluorophore absorption coefficient of background tissue variability of endogenous (50%) and exogenous (500%) properties, (b) Reconstructed fluorophore absorption coefficient of background tissue variability of endogenous (50%) and exogenous (500%) properties using relative measurement of the emission fluence with respect to the excitation fluence at the same detector point, s = 0.0001. (c) Reconstructed fluorophore absorption coefficient of background tissue variability of endogenous (50%) and exogenous (500%) properties using relative measurement of the emission fluence with respect to the excitation fluence at the same detector point. (From Ref. 109.)
Near-Infrared Imaging
517
518
Sevick-Muraca et al.
knowledge and (2) parsimony for well posedness of the problem. The first condition requires that all prior physical knowledge about the solution be included in the model. The second condition protects against the introduction of nonphysical phenomena. Tikhonov and Arsenin [118] also suggested that, to obtain a unique and stable solution from the data, supplementary information should be used so that the inverse problem becomes well posed. The basic principle of using a priori knowledge of the properties of the inverse problem is to restrict the space of possible solutions so that the data uniquely determine a stable solution. In his work, Roy showed that the constrained optimization technique, which places simple bounds on a physical parameter to be estimated, may be more appropriate for solving the fluorescence-enhanced optical tomography problem [90]. Specifically, a range of fluorescent optical properties is physically defined for the problem and the recovered parameter, X, must always be positive. Specifically, he demonstrated use of the bounding parameter, £, not only as a means to regularize and accelerate convergence but as a means to set the level of optical property contrast that is to be reconstructed using referenced emission measurements [109]. Here the possible values of parameter estimates are stated to lie between upper and lower bounds. In the first pass of the iterative solution, the optical property map is recovered and parameter estimates that lie within the upper and lower bounds plus and minus a small bounding parameter, s, are recovered and held constant for the next iteration. In this manner, the number of unknowns decrease with iteration. Indeed, the value of the bounding parameter can be used to set the resolution and the performance of the tomographic image. For example, if the bounding parameter is large, then the tomographic image will "filter" out artifacts not associated with the target, whereas if the bounding parameter is small, then the tomographic image may sensitively capture artifacts and heterogeneity that is not necessarily associated with the target. Figure 30 illustrates the reconstruction using the simply bounded truncated Newton's method, which shows that as the bounding parameter is increased, the recovered image becomes less sensitive to the background "noise." 6.
SUMMARY: THE CHALLENGES FOR NIR FLUORESCENCE-ENHANCED IMAGING AND TOMOGRAPHY
In the earlier sections, an overview of the status of fluorescence-enhanced optical imaging was presented. The opportunity to develop an emissionbased tomographic imaging modality similar to that provided by nuclear imaging but without the use of radionucleotides is offered by NIR fluorescent agents. Yet the added challenge for NIR fluorescence-enhanced im-
Near-Infrared Imaging
519
aging over nuclear imaging is that, unlike nuclear techniques, an activating or excitation signal must first be delivered to the contrast agent before there is registration of the emission signal from the tissue. Preliminary data from animals (Table 1) and phantoms (not presented herein) suggest that penetration depth and sensitivity may very well be comparable to those of nuclear techniques. A side-by-side comparison of NIR fluorescence-enhanced imaging with nuclear imaging is needed before the comparative performance can be ascertained. Another opportunity for optical imaging is the potential for tomographic reconstruction and additional diagnostic information based on the fluorescence decay kinetics of smartly designed probes. Tomography of large tissue-simulating volume has been demonstrated from experimental data as well as synthetic data as reviewed herein (Table 2), albeit with the somewhat inconvenient point source and point detector geometries. The single point source and detector geometry is a throw-back to NIR optical tomography from endogenous contrast studies and may not be the appropriate geometry for fluorescence-enhanced optical imaging, especially when transillumination through large tissues is required. Nonetheless, the tomographic algorithms as reviewed in Sec. 5, are already established for these systems. The challenge for the future is to develop tomographic algorithms for illumination and detection that are clinically feasible and adaptable for hybrid, nuclear imaging. Although the area of NIR fluorescence-enhanced optical imaging is less than a decade old, the developments are apt to continue for the coming decade, hopefully resulting in an adjuvant tomographic imaging modality for nuclear imaging. ACKNOWLEDGMENTS The review is supported in part by the National Institutes of Health grants R01 CA67176 and R01CA88082 (P. I. Eppstein, University of Vermont) and the Texas Advanced Technology Research Program. REFERENCES 1.
2.
3.
DJ Hawrysz, EM Sevick-Muraca. Developments toward diagnostic breast cancer imaging using near-infrared optical measurements and fluorescent contrast agents. Neoplasia 2(5):388-417, 2002. BJ Tromberg, N Shah, R Lanning, A Cerussi, J Espinoza, T Phain, L Svaasand, J Butler. Non-invasive in vivo characterization of breast tumors using photon migration spectroscopy. Neoplasia 2:26-40, 2000. E Reddi, A Segalla, G Jori, PK Kerrigan, PA Lidell, AL Moore, TA Moore,
520
4. 5.
6.
7.
8.
9.
10.
11. 12.
13.
14. 15.
16.
17.
18.
Sevick-Muraca et al. D Gust. Carotenoporphyrins as selective photodiagnostic agents for tumors. Br J Cancer 69:40-45, 1994. T Takemura, S Nakajima, I Sakata. Tumor-localizing fluorescent diagnostic agents without phototoxicity. Photochem Photobiol 59:366-370, 1994. R Biolo, G Jori, JC Kennedy, P Nadeau, R Potteir, E Reddi, G Weagle. A comparison of fluorescence methods used in the pharmacokinetic studies of Zn(II) phthalocyanine in mice. Photochem Photobiol 53:113-118, 1991. RC Straight, RE Benner, RW McClane, PMNY Go, G Yoon, JA Dixon. Application of charge-coupled device technology for measurement of laser light and fluorescence distribution in tumors for photodynamic therapy. Photochem Photobiol 53:787-796, 1991. M Kohl, U Sukowski, B Ebert, J Neukammer, HH Rinneberg. Imaging of superficially growing tumors by delayed observation of laser-induced fluorescence. Proc SPIE 1881:206-221, 1993. R Cubeddu, G Canti, P Taroni, G Valentini. Time-gated fluorescence imaging for the diagnosis of tumors in a murine model. Photochem Photobiol 57:480485, 1993. R Cubeddu, G Canti, A Pifferi, P Taroni, G Valentini. Fluorescence lifetime imaging of experimental tumors in hematoporphyrin derivative-sensitized mice. Photochem Photobiol 66:229-236, 1997. V Ntziachristos, AG Yodh, M Schnall, B Chance. Concurrent MRI and diffuse optical tomography of the breast after indocyanine green enhancement. Proc Natl Acad Sci 97:2767-2772, 2000. JR Lakowicz. Principles of Fluorescence Spectroscopy. Plenum Press, New York, 1983. JP Houston, AB Thompson, M Gurfinkel, EM Sevick-Muraca. Sensitivity and penetration depth of NIR fluorescence-enhanced contrast imaging. Photochem Photobiol 2002 (submitted 07/02). AB Thompson, EM Sevick-Muraca. NIR fluorescence contrast enhanced imaging with ICCD homodyne detection: Measurement precision and accuracy. J Biomed Opt in press (2003). J Lakowicz, K Berndt. Lifetime-sensitive fluorescence imaging using an if phase-camera. Rev Sci Instrum 62:1727-1734, 1991. EM Sevick, JR Lakowicz, H Szmacinski, K Nowaczyk, M Johnson. Frequency-domain imaging of obscure absorbers: principles and applications. J Photochem Photobiol 16:169-185, 1992. JS Reynolds, TL Troy, EM Sevick-Muraca. Multipixel techniques for frequency-domain photon migration imaging. Biotechnol Progr 13(5):669-680, 1997. A Pelegrin, S Folli, F Buchegger, J-P Mach, G Wagnieres, H van den Bergh. Antibody-fluorescein conjugates for photoimmunodiagnosis of human colon carcinoma in nude mice. Cancer 67:2529-2537, 1991. S Folli, G Wagnieres, A Pelegrin. JM Calmes, D Braichotte, F Buchegger, Y Chalandon, N Hardman, DC Heusser, JC Givel, G Chapuis, A Chatelain, H van Den Bergh, JP Mach. Immunophotodiagnosis of colon carcinomas in
Near-Infrared Imaging
521
patients injected with fluoresceinated chimeric antibodies against carcinoembryonic antigen. Proc Natl Acad Sci 89:7973-7977, 1992. 19. S Folli, P Westermann, D Braichotte, A Pelegrin, G Wagnieres, H van den Bergh, JP Mach. Antibody-indocyanin conjugates for immunophotodetection of human squamous cell carcinoma in nude mice. Cancer Res 54:2643-2649, 1994. 20. MM Haglund, DW Hochman. AM Spence, MS Berger. Enhanced optical imaging of rat gliomas and tumor margins. Neurosurgery 35:930-939, 1994. 21. S Mordon, JM Devoisselle, V Maunoury. In vivo pH measurement and imaging of tumor tissue using a pH-sensitive fluorescent probe (5,6-carboxyfluorescein): instrumental and experimental studies. Photochem Photobiol 60: 274-279, 1994. 22. B Ballou, GW Fisher, AS Waggoner, DL Parkas, JM Reiland, R Jaffe, RB Mujumdar, SR Mujumdar, TR Hakala. Tumor labeling in vivo using cyanineconjugated monoclonal antibodies. Cancer Immunol Immunother 41:257263, 1995. 23. JM Devoisselle, S Soulie, SR Mordon, G Mestres, TMD Desmettre, H Maillols. Effect of indocyanine green formulation on blood clearance and in vivo fluorescence kinetic profile of skin. Proc SPIE 2627:100-108, 1995. 24. I Rokahr, S Andersson-Engels, S Svanberg, M-A D'Hallewin, L Baert, I Wang, K Svanberg. Optical detection of human urinary bladder carcinoma utilising tissue autofluorescence and protoporphyrin IX-induced fluorescence following low-dose ALA instillation. Proc SPIE 2627:2-12, 1995. 25. MM Haglund, MS Berger, DW Hochman. Enhanced optical imaging of human gliomas and tumor margins. Neurosurgery 38:308-316, 1996. 26. K Sakatani, M Kashiwasake-Jibu, Y Taka, S Wang, H Zuo, K Yamamoto, K Shimizu. Noninvasive optical imaging of the subarachnoid space and cerebrospinal fluid pathways based on near-infrared fluorescence. Journal of Neurosurgery 87:738-745, 1997. 27. D Neri. Targeting by affinity-matured recombinant antibody fragments of an angiogenisis-associated fibronectin isoform. Nat Biotechnol 15:1271-1275, 1997. 28. B Ballou, GW Fisher, J-S Deng, TR Hakala, M Srivastava, DL Farkas. Fluorochrome-labeled antibodies in vivo: assessment of tumor imaging using Cy3, Cy5, Cy5.5, and Cy7. Cancer Det Prev 22(3):251-257, 1998. 29. C Eker, S Montan, E Jaramillo, K Koizumi, C Rubio, S Andersson-Engels, K Svanberg, S Svanberg, P Slezak. Clinical spectral characterization of coIonic mucosal lesions using autofluorescence and d aminolevulinic acid sensitizations. Gut 44:511-518, 1999. 30. JS Reynolds, TL Troy, RH Mayer, AB Thompson, DJ Waters, KK Cornell, PW Snyder, EM Sevick-Muraca. Imaging of spontaneous canine mammary tumors using fluorescent contrast agents. Photochem Photobiol 70:87-94, 1999. 31. A Becker, K Licha, M Kresse, B Riefke, U Sukowski, B Ebert, H Rinneberg, W Semmler. Transferrin-mediated tumor delivery of contrast media for optical imaging and magnetic resonance imaging. Proc SPIE 3600:142-150, 1999.
522
Sevick-Muraca et al.
32.
R Weissleder, CH Tung, U Mahmood, A Bogdanov Jr. In vivo imaging of tumors with protease-activated near-infrared fluorescent probes. Nat Biotechnol 17:375-378, 1999. U Mahmood, C-H Tung. A Bogdanov, R Weissleder. Near-infrared optical imaging of protease activity for tumor detection. Radiology 21(3):866-870, 1999. A Becker, B Riefke, B Ebert, U Sukowski, H Rinneberg, W Semmler, K Licha. Macromolecular contrast agents for optical imaging of tumors: comparison of indotricarbocyanine-labeled human serum albumin and transferrin. Photochem Photobiol 72:234-241, 2000. M Gurfinkel, AB Thomson, W Ralston, TL Troy, AL Moore, TA Moore, JD Gust, D Tatman, JS Reynolds. B Muggenburg, K Nikula, R Pandey, RH Mayer, DJ Hawrysz, EM Sevick-Muraca. Pharmacokinetics of ICG and HPPH-car for the detection of normal and tumor tissue using fluorescence, near-infrared reflectance imaging: A case study. Photochem Photobiol 72:94102, 2000. K Licha, B Riefke, V Ntziachristos, A Becker, B Chance, W Semmler. Hydrophilic cyanine dyes as contrast agents for near-infrared tumor imaging: synthesis, photophysical properties and spectroscopic in vivo charactarization. Photochem Photobiol 72:392-398, 2000. M Yang, E Baranov, F Sun. X Li, L Li, S Hasegawa. M Bouvet, M AlTuwaijri, T Chishima, H Shimada, A Moossa, S Penman, R Hoffman. Wholebody optical imaging of green fluorescent protein-expressing tumors and metastases. PNAS 97:1206-1211, 2000. RM Hoffman. Visualization of GFP-expressing tumors and metastasis in vivo. Biotechniques 30:1016-1026, 2001. JE Bugaj, S Achilefu, RB Dorshow. R Rajagopalan. Novel fluorescent contrast agents for optical imaging of in vivo tumors based on a receptor-targeted dyepeptide conjugate platform. J Biomed Opt 6:122-133, 2001. S Achilefu. RB Dorshow, JE Bugaj, R Rajogopalan. Novel receptor-targeted fluorescent contrast agents for in vivo tumor imaging. Invest Radiol 35:479485, 2000. S Achilefu, JE Bugaj, RB Dorshow, HN Jimenez, R Rajagopalan. New approach to optical imaging of tumors. Proc SPIE 4259:110-114, 2001. S Achilefu, JE Bugaj, RB Dorshow, HN Jimenez, R Rajagopalan, RR Wilhelm, EG Webb, JL Erion. Site-specific tumor targeted fluorescent contrast agents. Proc SPIE 4156:69-78, 2001. A Becker, C Hessenius, K Licha, B Ebert, U Sukowski, W Semmler, B Wiedenmann C Grotzinger. Receptor-targeted optical imaging of tumors with near-infrared fluorescent ligands. Nat Biotechnol 19:327-331, 2001. C Bremen C Tung, R Weissleder. In vivo molecular target assessment of matrix metalloproteinase inhibition. Nat Med 7:743-748. 2001. B Ebert, U Sukowski, D Grosenick, H Wabnitz, KT Moesta, K Licha, A Becker, W Semmler, PM Schlag, H Rinneberg. Near-infrared fluorescent dyes for enhanced contrast in optical mammography: phantom experiments. J Biomed Opt 6(2): 134-140, 2001.
33.
34.
35.
36.
37.
38. 39.
40.
41. 42.
43.
44. 45.
Near-Infrared Imaging 46.
523
JC Finlay, DL Conover, EL Hull, TH Foster. Porphyrin bleaching and PDTinduced spectral changes are irradiance dependent in ALA-sensitized normal rat skin in vivo. Photochem Photobiol 73:54-63, 2001. 47. BW Rice, MD Cable, MB Nelson. In vivo imaging of light-emitting probes. J Biomed Opt 6(4):432-440, 2001. 48. NS Soukos, MR Hamblin, S Keel, RL Fabian, TF Deutsch, T Hasan. Epidermal growth factor receptor-targeted immunophotodiagnosis and photoimmunotherapy of oral precancer in vivo. Cancer Res 61:4490-4496, 2001. 49. Y Yang, N Iftimia, Y Xu, H Jiang. Frequency-domain fluorescent diffusion tomography of turbid media and in vivo tissues. Proc SPIE 4250:537-545, 2001. 50. CM Leevy, F Smith, J Longueville. Indocyanine green clearance as a test for hepatic function. Evaluation by dichromatic ear densitometry. JAMA 200: 236-240, 1967. 51. K Kogure, NJ David, U Yamanouchi, E Choromokos. Infrared absorption angiography of the fundus circulation. Arch Ophthalmol 83(2):209-214, 1970. 52. EM Sevick-Muraca, G Lopez, TL Troy, JS Reynolds, CL Hutchinson. Fluorescence and absorption contrast mechanisms for biomedical optical imaging using frequency-domain techniques. Photochem Photobiol 66:55-64, 1997. 53. ML Landsman, G Kwant, G Mook, WG Zijlstra. Light-absorbing properties, stability, and spectral stabilization of Indocyanine green. J Appl Physiol 40: 575-583, 1976. 54. S Mordon, JM Devoisselle, S Soulie-Begu, T Desmettre. Indocyanine green: physiochemical factors affecting its fluorescence in vivo. Microvas Res 55: 146-152, 1998. 55. KA Schafer, G Kelly, R Schrader, WC Griffith, BA Muggenburg, LA Tierney, JF Lechner, EB Janovitz, FF Hahn. A canine model of familial mammary gland neoplasia. Vet Pathol 35:168-177, 1998. 56. V Ntziachristos, X Ma, B Chance. Time-correlated single photon counting imager for simultaneous magnetic resonance and near-infrared mammography. Rev Sci Instrum 69:4221-4233, 1998. 57. KM McMaster, AE Giuliano, MI Ross, DS Reintgen, KK Hunt, VS Klimberg, PW Whitworth, LC Tafra, MJ Edwards. Sentinel lymph node biopsy for breast cancer—not yet the standard of care. N Engl J Med 339:990-995, 1998. 58. D Krag, D Weaver, T Ashikaga, F Moffat, S Klimberg, C Shriver, S Feldman, R Kusminsky, M Gadd, J Kuhn, S Harlow, P Beitsch, P Whitworth, R Foster Jr, K Dowlatshahi. The sentinel node in breast cancer—a multicenter validation study. N Engl J Med 339:941-946, 1998. 59. K Licha, A Becker, F Kratz, W Semmler. New contrast agents for optical imaging: acid-cleavable conjugates of cyanine dyes with biomolecules. Proc SPIE 3600:29-35, 1999. 60. MM Huber, AB Staubili, K Kustedjo, MHB Gray, J Shih, S Fraser, RE Jacobs, TJ Meade. Fluorescently detectable magnetic resonance imaging agents. Bioconj Chem 9:242-249, 1998. 61. X Li, B Chance, AG Yodh. Fluorescence heterogeneities in turbid media,
524
62.
63.
64.
65.
66.
67. 68. 69.
70.
71.
72.
73.
74. 75.
76.
77.
Sevick-Muraca et al. limits for detection, characterization, and comparison with absorption. Appl Opt 37:6833-6843, 1998. XD Zhu, S-P Wei, XW Guo. Imaging objects in tissue-like media with optical tagging and the diffuse photon differential transmittance. J Opt Soc Am 14: 300-305, 1997. RC Haskell, LO Svassand, T-T Tsay, T-C Feng, MS McAdams, BJ Tromberg. Boundary conditions for the diffusion equation in radiative transfer. J Opt Soc Am A 11:2727-2741, 1994. MS Patterson, B Chance, B Wilson. Time resolved reflectance and transmittance for the non-invasive measurement of tissue optical properties. Appl Opt 28:2331-2336, 1989. AH Hielscher, SL Jacques, L Wang, FK Tittel. The influence of boundary conditions on the accuracy of diffusion theory in time-resolved reflectance spectroscopy of biological tissues. Phys Med Biol 40:1957-1975, 1995. TJ Farrell, MS Patterson, B Wilson. A diffusion theory model of spatially resolved, steady-state diffuse reflectance for the noninvasive determination of tissue optical properties in vivo. Med Phys 9:879-888, 1992. EM Sevick-Muraca, CL Burch. Origin of phosphorescence re-emitted from tissues. Opt Lett 19:1928-1930, 1994. R Richards-Kortum, EM Sevick-Muraca. Quantitative optical spectroscopy for tissue diagnostics. Ann Rev Phys Chem 47:555-606, 1996. E Kuwana, EM Sevick-Muraca. Fluorescence lifetime spectroscopy in multiply scattering media with dyes exhibiting multi-exponential decays kinetics. Biophys J 83:1165-1176, 2002. V Chenomordik, D Hattery, I Gannot, AH Gandjbakhche. Inverse method 3-D reconstruction of localized in vivo fluorescence—application to Sjogren syndrome. IEEE J Sel Top Quant Electron 54:930-935, 1999. D Hattery, V Chernomordik, I Gannot, M Loew, AH Gandjbakhche. Fluorescence measurement of localized, deeply embedded physiological processes. Proc SPIE 3978:377-382, 2000. MS Patterson, BW Pogue. Mathematical model for time-resolved and frequency-domain fluorescence spectroscopy in biological tissue. Appl Opt 33: 1963-1964, 1994. CL Hutchinson, JR Lakov.'icz, EM Sevick-Muraca. Fluorescence lifetimebased sensing in tissues: a computational study. Biophys J 68:1574-1582, 1995. S Fantini, E Gratton. Fluorescence photon density waves in optically diffuse media. Optic Comm 173:73-79, 2000. RH Mayer, JS Reynolds, EM Sevick-Muraca. Measurement of fluorescence lifetime in scattering media using frequency-domain photon migration. Appl Opt 38:4930-4938, 1999. AE Cerussi, JS Maier, S Fantini, MA Franceschini, WW Mantulin, E Gratton. Experimental verification of a theory for time-resolved fluorescence spectroscopy of thick tissues. Appl Opt 36:116-124, 1997. A Chen. Effects of fluorescence and phosphorescence lifetime on frequency
Near-Infrared Imaging
78.
79.
80. 81. 82.
83.
84.
85. 86.
87. 88.
89.
90. 91.
92.
93.
525
domain optical contrast for biomedical optical imaging. MS thesis, Purdue University, West Lafayette, IN, 1997. MA O'Leary, DA Boas, B Chance, AG Yodh. Reradiation and imaging of diffuse photon density waves using fluorescent inhomogeneities. J Lumin 60, 61:281-286, 1994. J Wu, Y Wang, L Perleman, 1 Itzkan, RR Dasari, MS Feld. Time-resolved multichannel imaging of fluorescent objects embedded in turbid media. Opt Lett 20:489-491, 1995. J Chang, RL Barbour, H Graber, R Aronson. Fluorescence Optical Tomography. Proc SPIE 2570:59-72, 1995. MA O'Leary, DA Boas, XD Li, B Chance, AG Yodh. Fluorescence lifetime imaging in turbid media. Opt Lett 21(2):158-160, 1996. DY Paithankar, AU Chen, BW Pogue, MS Patterson, EM Sevick-Muraca. Imaging of fluorescent yield and lifetime from multiply scattered light reemitted from random media. Appl Opt 36:2260-2272, 1997. J Wu, L Perelman, RR Dasari, MS Feld. Fluorescence tomographic imaging in turbid media using early-arriving photons and Laplace transforms. Proc Natl Acad Sci 94:8783-8788, 1997. J Chang, HL Graber, RL Barbour. Improved reconstruction algorithm for luminescence when background luminophore is present. Appl Opt 37:35473552, 1998. H Jiang. Frequency-domain fluorescent diffusion tomography: a finite-element-based algorithm and simulations. Appl Opt 37(22):5337-5343, 1998. EL Hull, MG Nichols, TH Foster. Localization of luminescent inhomogeneities in turbid media with spatially resolved measurements of cw diffuse luminescence emittance. Appl Opt 37:2755-2765, 1998. MJ Eppstein, DE Dougherty, DJ Hawrysz, EM Sevick-Muraca. Three-dimensional optical tomography. Proc SPIE 3497:97-105, 1999. MJ Eppstein, DE Dougherty, TL Troy, EM Sevick-Muraca. Biomedical optical tomography using dynamic parameterization and Bayesian conditioning on photon migration measurements. Appl Opt 38:2138-2150, 1999. R Roy, EM Sevick-Muraca. Truncated Newton's optimization scheme for absorption and fluorescence optical tomography: Part II Reconstruction from synthetic measurements. Opt Exp 4:372-382, 1999. R Roy, EM Sevick-Muraca. Active constrained truncated Newton method for simple-bound optical tomography. J Opt Soc Am A 17(9): 1627-1641, 2000. MJ Eppstein, DJ Hawrysz, A Godavarty, EM Sevick-Muraca. Three-dimensional, near-infrared fluorescence tomography with Bayesian methodologies for image reconstruction from sparse and noisy data sets. Proc Natl Acad Sci 99:9619-9624, 2002. DJ Hawrysz, MJ Eppstein, J Lee, EM Sevick-Muraca. Error consideration in contrast-enhanced three-dimensional optical tomography. Opt Lett 26(10): 704-706, 2001 R Roy, EM Sevick-Muraca. Three-dimensional unconstrained and constrained image-reconstruction techniques applied to fluorescence, frequency-domain photon migration. Appl Opt 40 (13):2206-2215, 2001.
526 94.
Sevick-Muraca et al.
R Roy, EM Sevick-Muraca. A numerical study of gradient-based nonlinear optimization methods for contrast-enhanced optical tomography. Opt Exp 9(l):49-65, 2001. 95. V Ntziachristos, R Weissleder. Experimental three-dimensional fluorescence reconstruction of diffuse media by use of a normalized Born approximation. Opt Lett 26(12):893-895, 2001. 96. J Lee, EM Sevick-Muraca. 3-D fluorescence enhanced optical tomography using references frequency-domain photon migration measurements at emission and excitation measurements. J Opt Soc Am A 19:759-771, 2002. 97. EM Sevick-Muraca, DY Paithankar. Fluorescence imaging system and measurement. U.S. patent #5.865,754, issued 2/2/99. 98. JM Schmitt, A Knuttel, JR Knutson. Interference of diffusive light waves. J Opt Soc Amer A 9:1832-1843, 1992. 99. DY Paithankar, A Chen, EM Sevick-Muraca. Fluorescence yield and lifetime imaging in tissues and other scattering media. Proc Soc Photo-Opt lustrum Eng 2679:162-175. 1996. 100. B Pogue, T McBride, J Prewitt. U Osterberg, K Paulsen. Spatially varying regulari/ation improves diffuse optical tomography. Appl Opt 38:2950-2961. 1999. 101. MJ Holboke, AG Yodh. Parallel three-dimensional diffuse optical tomography. Biomedical Topical Meetings, OS A, Miami Beach, FL, 2000, pp 177-179. 102. M Xu, M Lax, RR Alfano. Time-resolved fourier diffuse optical tomography. Biomedical Topical Meetings, OSA, Miami Beach, FL, 2000, pp 345-347. 103. SR Arridge, JC Hebden. M Schweiger, FEW Schmidt, ME Fry, EMC Hillman, H Dehghani, DT Delpy. A method for three-dimensional time-resolved optical tomography. Int J Imaging Sys Technol 11:2-11, 2000. 104. MJ Eppstein, DE Dougherty. Optimal 3-D traveltime tomography. Geophysics 63:1053-1061. 1998. 105. MJ Eppstein, DE Dougherty. Efficient three-dimensional data inversion: soil characterization and moisture monitoring from cross-well ground penetrating radar at a Vermont test site. Wat Resources Res 34:1889-1900. 1998. 106. MJ Eppstein. DE Dougherty. Three-dimensional stochastic tomography with upscaling. U.S. patent application 09/110,506, (9 July 1998). 107. TJ Yorkey, JG Webster, WJ Tompkins. Comparing reconstruction algorithms for electrical impedance tomography. IEEE Trans Biomed Eng BME-34, 843— 852, 1987. 108. MJ Eppstein, DE Dougherty, DJ Hawrysz, EM Sevick-Muraca. Three-dimensional Bayesian optical image reconstruction with domain decomposition. IEEE Trans Med Imaging 20(3): 147-163, 2000. R Roy. A Godavarty, EM Sevick-Muraca. Fluorescence-enhanced, optical tomography using referenced measurements of heterogeneous media. IEEE Trans Med Imaging (in press). R Roy. EM Sevick-Muraca. Truncated Newton's optimization scheme for absorption and fluorescence optical tomography: part I theory and formulation. Opt Exp 4:353-371, 1999. J Lee. EM Sevick-Muraca. Fluorescence-enhanced absorption imaging using
Near-Infrared Imaging
112. 113. 114.
115.
116.
117.
118.
527
frequency-domain photon migration: tolerance to measurement error. J Biomed Opt 6(l):58-67, 2000. SR Arridge. Optical tomography in medical imaging. Inverse Problems 15: R41-R93, 1999. MJ Eppstein, DE Dougherty. Simultaneous estimation of transmittivity values and zonation. Wat Resour Res 32:3321-3336, 1996. Y Pei, F-B Lin, RL Barbour. Modeling of sensitivity and resolution to an included object in homogeneous scattering media and in MRI-derived breast maps. J Biomed Opt 5:302-219, 1999. BW Pogue, KD Paulsen. High-resolution near-infrared tomographic imaging simulations of the rat cranium by use of a priori magnetic resonance imaging structural information. Opt Lett 23:1716-1718, 1998. MJ Holboke, BJ Tromberg, X Li, N Shah, J Fishkin, D Kidney, J Butler, B Chance, AG Yodh. Three-dimensional diffuse optical mammography with ultrasound localization in a human subject. J Biomed Opt 5:237-247, 2000. SW Provencher, RH Vogel. Regularization techniques for inverse problems in molecular biology. Numerical Treatment of Inverse Problems in Differential and Integral Equations. Boston: Birkhauser Press, 1983. AN Tikhonov, VY Arsenin. Solution of Ill-Posed Problems. Washington, DC: V. H. Winston & Sons, 1977.
15 Fluorescence in Photodynamic Therapy Dosimetry Brian C. Wilson Ontario Cancer Institute, University of Toronto, and Photonics Research Ontario, Toronto, Ontario, Canada Robert A. Weersink Photonics Research Ontario, Toronto, Ontario, Canada Lothar Lilge University of Toronto and Photonics Research Ontario, Toronto, Ontario, Canada
1.
INTRODUCTION
Photodynamic therapy (PDT) is a technique for treating a variety of malignant and nonmalignant conditions based on the use of light-activated drugs (photosensitizers). Typically, the photosensitizer is administered either systemically (intravenously or orally) or topically to the tissue to be treated. After allowing time for uptake of the photosensitizer to the target tissues or tissue structures, light of an appropriate wavelength to activate the drug is applied. This results in the photoproduction of one or more cytotoxic agents, leading to the intended cellular and tissue effects. For most photosensitizers used or under investigation clinically, it is likely that the main photophysical pathway is production of singlet oxygen, 'O2. Singlet oxygen is an excited form of oxygen that is highly reactive with biomolecules, leading typically 529
530
Wilson et al.
to oxidative damage. The Jablonski energy diagram for this so-called type II process is shown in Fig. 1. The absorption of a photon by the groundstate photosensitizer activates the molecule to the excited singlet state. This short-lived state (typically nanoseconds) may de-excite to the ground state either nonradiatively or by fluorescence emission, or may undergo a change of spin to the triplet state. This is relatively long lived, since decay to the ground state is a quantum mechanically forbidden transition. Energy exchange with ground-state oxygen is, however, an allowed transition, since 3 O2 is also a triplet state, and this excites the oxygen to 'O2. (Note that type I processes are those in which the reactive species are generated directly from the photosensitizer-excited singlet or triplet state and may or may not be oxygen dependent. They also lead to photosensitizer fluorescence from the singlet state.) Most of the current photosensitizers have some fluorescence emission. (The in vivo measurement of nonfluorescent drugs is discussed briefly later.) For the PDT efficiency to be as high as possible, a high triplet state quantum yield is desirable, which competes with the fluorescence quantum yield. However, using the fluorescence emission to monitor the photosensitizer does not require a high yield (e.g., a few percent), so that in practice a high triplet-state yield can be selected. Figure 2 shows the absorption and fluorescence emission spectra of some common PDT drugs. Note that the fluorescence excitation spectrum is generally very similar to the absorption spectrum, but that these may both change in biological media compared to the spectra in simple solutions, due to substrate (e.g., protein) binding. Shifts in the absorption peaks by several nanometers are common.
SINGLET STATES
"
BIOLOGICAL »-"-• ^
EFFECT
GROUND STATE
OXYGEN
FIGURE 1 Jablonski diagram for type II photodynamic reactions. The vertical axis indicates the energy for the distinct electronic states of the photosensitizer and oxygen molecules.
Photodynamic Therapy Dosimetry
531
O CD
•j=> Q) X O LU Q
400
600
500 600 Wavelength (nm)
700
800
(a)
FIGURE 2 (a) Absorption and fluorescence emission spectra of some common PDT photosensitizers. (b) Corresponding excitation-emission matrices.
Most PDT photosensitizers have spectra across the visible range, often with a strong Soret band in the UV-A/blue region. In order to obtain maximal light penetration in tissue for treating larger lesions (such as solid tumors), many newer PDT drugs have strong absorption bands in the far-red (650700 nm) and near-infrared (700-850 nm), and correspondingly their fluorescence is also at these long wavelengths. Figure 3 summarizes the classes of disease for which PDT is being investigated as a possible therapeutic method. Several of these treatments have regulatory approvals and are entering routine clinical practice. The potential applications are wide, reflecting the facts that (1) PDT can be applied to many body sites through the use of fiberoptic light delivery, as summarized schematically in Fig. 4, and (2) with different photosensitizers the mechanisms of action are varied, and include direct cell kill (by necrosis
Wilson et al.
532
Photofrin E •S c g ro
•3x ill
o in. cs
300
400 500 600 Emission (nm)
700
AIPcS4
o 10 c ,2 o
5 8
'Q x Hi
i 300
400 500 600 Emission (nm)
700
.m
(b) FIGURE 2
Continued
and/or apoptosis), vascular destruction, and immune stimulation or suppression. The photosensitizer fluorescence may be utilized in a number of ways, including: 1. 2. 3.
Lesion detection in vivo Lesion localization in vivo for therapy guidance, both PDT and surgical Quantification of the concentration of the photosensitizer in the target and other tissues in vivo for the purposes of a. Determining the drug pharmacokinetics in order to select the optimum time for light activation, or
Photodynamic Therapy Dosimetry In Vivo
photosensitizer
533
treatment light • solid tumors • dyslasias • papillomas • rheumatoid arthritis • age-related macular degeneration • cosmesis • psoriasis • endometrial ablation • localized infection • prophylaxis of arterial restenosis
Ex Vivo
• extracorporeal photophoresis • blood purging: HIV, hepatitis B, protozoa • bone marrow purging photosensitizer
I
FIGURE 3
treatment
Potential clinical applications of PDT.
b.
4.
5. 6.
Individualizing the light delivery to produce the required biological response ("dosimetry") Measurement of the photosensitizer photobleaching during PDT light activation, in order to estimate the effective PDT "dose" delivered Photosensitizer quantification in tissue samples (e.g., biopsies) ex vivo Determining the microdistribution of photosensitizer in tissues (ex vivo) or cells (in vitro) by fluorescence microscopy
Tissue autofluorescence may be used in addition or as an alternative to photosensitizer fluorescence for purposes 1, 2, and 4, or as a monitor of tissue response to treatment. Fluorescence imaging of other "contrast agents" may be of value for PDT in particular sites, such as the use of fluorescein angiography in PDT of age-related macular degeneration, a dis-
Wilson et al.
534
JLJJ
Superficial
Interstitial
Endoscopic
Whole Surface Intracavitory
FIGURE 4 Light irradiation methods for PDT, showing the wide variety of target geometries.
ease that leads to blindness due to growth of abnormal blood vessels in the choroid layer of the eye near the macula, which is the area of central acute vision. In this case, as illustrated in Fig. 5, fluorescein angiography is used both before treatment to show the area of choroidal neovasculature so as to target the treatment light and after treatment to evaluate the area and completeness of vascular closure. In this chapter, the emphasis will be on topics 2-6, since lesion detection and localization for treatment guidance are dealt with elsewhere, in the general context of fluorescence "contrast agents." It is worth noting, however, that many of the in vivo fluorescence-based methods and instruments that have been developed over the past 20 years have been strongly
Photodynamic Therapy Dosimetry
535
FIGURE 5 Fluorescein angiograms before and after PDT for age-related macular degeneration, showing closure of the area of abnormal neovasculature.
associated with the corresponding development of PDT as a treatment modality, since they provide complementary "search and destroy" capabilities. Thus, for example, autofluorescence bronchoscopy (Chapter 11) grew from the original endoscopic spectroscopy and imaging work with the photosensitizer hematoporphyrin derivative (HPD) and was originally aimed at localizing tumors in the lung for PDT treatment. (Interestingly, HPD itself was developed from hematoporphyrin in an attempt to improve its fluorescence properties and was only subsequently discovered to be an effective PDT drug.) Studies were performed with successively decreasing doses of HPD in order to reduce the associated skin photosensitization, and in the case of early bronchial cancer/dysplasia it was found that the detection accuracy was greatest with zero dose, i.e., using only the autofluorescence. However, fluorescence of HPD continues for the other objectives listed above, and many developments in fluorescence spectroscopy and imaging in vivo continue to be tightly linked to improving and monitoring PDT treatments, although they are also developing independently. In addition to PDT applications exploiting the photosensitizer fluorescence, fluorescence-based optical fiber probes are being developed for light fluence monitoring in tissue. In these, a small point-like volume of specific fluorophore is incorporated into an optical fiber and placed on or within the tissue. The PDT treatment light then excites the fluorescence, a fixed fraction of which is transmitted along the fiber and detected. Probes incorporating several different fluorophores at different positions along the fiber allow
536
Wilson et al.
multipoint measurements to be made simultaneously by spectral deconvolution. The concept can also be extended to tissue oxygen measurements using phosphorescence lifetime measurements. This technology and its applications will be discussed at the end of the chapter. 2. 2.1
PHOTOSENSITIZER QUANTIFICATION IN VIVO Relative Measurements
Figure 6 shows a simplified schematic diagram of noninvasive measurement of photosensitzer fluorescence in vivo. The excitation light from a light source (laser, or filtered lamp or light-emitting diode) at wavelength Aex is delivered to the tissue surface by an optical fiber (or fiber bundle). Then one or more collection fibers pick up some fraction of the light remitted through the tissue surface; this comprises elastically scattered light (diffuse reflectance), photosensitizer fluorescence (at wavelength A cm ), tissue autofluorescence, and inelastic (Raman) scattered light. Typically, the order of magnitude of total photons remitted for each of these components, relative to the incident light, is around 10 ', 10~ 4 , 10 \ and 10 7, respectively, a small fraction of which is collected by the detector fibers. The elastic scattering is removed by optical filtering, typically, placing a long-pass filter (>Aex) between the output end of the collection fibers and the photodetector. The Raman signal is usually negligible, whereas the photosensitizer fluorescence is observed superimposed on the tissue autofluorescence background. Figure 7 shows such fluorescence spectra, for different administered doses of a specific agent, in this case the prodrug aminolevulinic acid (ALA), m
source
f photodetector i
xpPH;!H tissue
FIGURE 6 Schematic of fiberoptic detection of fluorescence at a tissue surface.
Photodynamic Therapy Dosimetry
537
0.005
0.000
?„ (nm)
0.05
0-00
tQmg/kg@3hf
'
'
02
0.0 500 —-Barrett's
"-""I
(nm)
30mg/kg @ 6 Nr
600
700 Squarnous
FIGURE 7 In vivo fluorescence spectra in Barrett's and normal human esophagus, at different times post ALA administration and for different ALA doses. Individual spectra represent measurements at different points on the tissue and are normalized to the autofluorescence background at 500 nm. (See color insert.)
which induces increased biosynthesis of heme, the penultimate step of which is the fluorescent photosenisitizer protoporphyrin IX (PpIX). These data illustrate a number of points. First, the fluorescence intensity is highly variable from patient to patient and even point to point in the same patient for a given drug level and time point. This type of variability has been reported for other photosensitizers and is, of course, a major reason why quantitative measurements are required. Second, the drug fluorescence spectrum is known and can be easily identified on the autofluorescence background, at least for higher levels of drug; however, at low levels, the added drug spectrum may not be easily seen, especially if the autofluorescence itself is highly variable (which is certainly the case for diseased tissue, and indeed is the basis of autofluorescence-based diagnostics). Third, the increase in the PpIX fluorescence with ALA dose and the dependence on time after administration can be observed, at least in the average spectra of each tissue type. In order to obtain a quantitative estimate of the PpIX concentration, a fit may be made to the autofluorescence background: in the simplest case, by a linear interpolation between points just outside the main PpIX fluores-
Wilson et al.
538
cence peak, as illustrated in Fig. 8a, or by a polynomial fit, taking the (local) shape of the background into account (Fig. 8b). The residual photosensitizer signal, PSF*, is then proportional to the (average) tissue concentration, C ps . If all factors other than C—namely, the excitation light, excitation and detection geometry, fluorescence quantum yield, and tissue optical absorption and scattering coefficients at Aex and A ox —remain constant, then the FSF* values yield the relative drug concentrations. The assumption of these factors remaining constant usually holds best in following the photosensitizer pharmacokinetics in a given location in an individual patient. Even in this idealized case, however, complications arise. For example, photosensitizer aggregation/disaggregation depends in many cases on the local microenvironment (pH, polarity, etc.) and this alters the fluorescence yield. The yield can also change when the photosensitizer binds
550
600
650
700
750
Wavelength (run)
b
> CO
O
o C
O
550
600
650
700
750
Wavelength (nm) FIGURE 8 Subtraction of the tissue autofluorescence background to measure the photosensitizer level, (a) Linear background, (b) Polynomial fit to the background.
Photodynamic Therapy Dosimetry
539
to biomolecules in the tissue. Hematoporphyrin derivative (commercially, Photofrin, Axcan, Canada), for which most clinical approvals are in place at present, is a mixure of many porphyrin molecules, with different pharmacokinetics, binding, fluorescence yield, and photoactivity, so that even the relative fluorescence signal in an individual patient can be difficult to interpret in terms of how much photodyamically active drug is present in the tissue. Nevertheless, the relative PSF* measurements can provide useful guidance, even if not exact information, on, for example, the time to maximal photosensitizer uptake or time to maximal tumor-to-normal tissue ratio, and drug clearance in individual patients. 2.2
Absolute Measurements
In order to obtain the true value of the tissue concentration, C, the effects of the other factors must be taken into account. The most important of these is the effect of the tissue optical absorption coefficients, /Aa(Acx) and /A a (A em ), and transport scattering coefficients, /u-s(A ex ) and /u^(Aem), that attenuate the excitation and emission light. The general expression for the detected fluorescence signal, F*(Aex, Aem, r), measured at some point at radial separation, r, from an incident pencil excitation beam of intensity, I0, is F*(Aex, A em ) = Tr*(A ex x,
r')'E(A e m , r -
r')} dr'
(1)
where 17 and e are, respectively, the photosensitizer fluorescence quantum yield and extinction coefficient and (f> [W.crrT2] is the light fluence rate at point r' in the tissue. S is the sensitivity (responsivity) of the fluorescence collection/detection system, including the geometric efficiency for light entering the collection fiber (essentially given by the numerical aperture for a fiber in contact with the tissue, corrected for internal reflection at the tissuefair]-fiber interface) and the detector sensitivity at the emission wavelength. The term E(A cm , r — r') is the "escape function" that describes the probability that a fluorescent photon generated at point r' will reach the tissue surface (or measurement point within the tissue) at position r. If the photosensitizer concentration, C(r), is uniform, at least within the sampled tissue volume, then this can be taken outside the integral. The issues related to 17 and e have already been mentioned, and essentially these factors depend on the tissue microenvironment and drug binding. The strong dependence on the tissue optical absorption and scattering properties arises from the terms (f> and E. There have been a number of different approaches to handle the complexity of this dependence, as
540
Wilson et al.
discussed below. We have grouped these techniques into three broad categories: 1.
2.
3.
Calibration techniques, in which the fluorescence is measured as a function of known photosensitizer concentration in different tissues (or tissue-simulating phantoms) and used as a standard for comparison to determine the unknown concentration in vivo Model-dependent techniques in which attempts are made to determine the tissue optical properties and to apply a rigorous or approximate model of light propagation in tissue, essentially to derive the excitation fluence and fluorescence escape functions and, thereby, correct F* to yield F, and Model-independent methods, in which the measured fluorescence signal is (empirically) corrected by a separate measurement on the tissue (e.g., diffuse reflectance or tissue autofluorescence) or in which the excitation and detection geometry is selected to be independent of the optical properties.
There is only a very limited literature on each of these approaches, and each will be illustrated briefly. 2.3
Calibration Techniques
The simplest approach is to measure the photosensitizer fluorescence and then compare the signal with the fluorescence signal from tissue samples with known photosensitizer concentration and similar escape function (optical properties). An example of this type of measurement was that of Panjehpour et al. [1], who determined the correlation of photosensitizer fluorescence signal to the true in vivo concentration for several tissues, using phthalocyanine as the photosensitizer. A simple surface contact probe with a single excitation fiber surrounded by six fluorescence collection fibers was used for the measurements. The collection fibers delivered the collected light to a photodiode array spectrometer. The actual tissue concentration was determined by extraction assay (see below). For many tissues, a correlation was found between the measured fluorescence signal and the C value. However, the signal strength at constant C was strongly dependent on the tissue type because of different optical attenuation, as seen in Fig. 9a. In principle, these calibration curves can be used to estimate C, but this requires calibration for each tissue type. Durkin and Richards-Kortum [2] compared several methods of photosensitizer quantification, including two-flux Kubelka-Munk (KM) and partial least-squares (PLS) analysis. For the former, model-dependent method a derived transfer function was used to relate the fluorescence in dilute
18
o • * v
16 ro 141 c CD 12
liver tumor kidney spleen
o
4
6
CO
IS' CO
0 O
r:
64-
0
2
8
Concentration by Tissue Extraction (mg/kg)
(a)
Source (lx) F (AJ
R (AX)
1
—J Tissue — PR 2.5 O)
o liver ^ skin • muscle
2.0 03
QC
g = 1.0"ro "c o 0.5c o O 0.0 0 (b)
0.5
1.0
1.5
2,0
2.5
Concentration by Tissue Extraction
FIGURE 9 Correlation between measured fluorescence signal (F*) and the true concentration of photosensitizer, C, in different tissue for three different measurement techniques. Also shown are schematic representations of the principle of each technique, (a) Simple surface probe (Panjepour). (b) Fluorescence normalized to diffuse reflectance at fixed source-detector separations (Weersink). (c) Confocal probe (Pogue).
Wilson et al.
542
500
400-
liver muscle
o CO
g> 300 c/? 0
c 200J CD O
b 100 2
(c) FIGURE 9
4
Concentration by Tissue Extraction (mg/kg) Continued
solution to that in an optically turbid medium. PLS is one of a number of multivariant statistical techniques that are used widely in spectroscopic analysis, and here is considered as a "black box" calibration technique. It requires simulated or measured training data sets (here, fluorescence spectra) for a range of known C values. This set must encompass the range of possible variations in the spectra and expected photosensitizer concentrations. These two methods were compared in phantoms, where the prediction error was —5% for PLS and — 1 8 % for the KM model. To achieve this low prediction error in PLS, the input data set consisted of the combination of fluorescence spectra collected at two excitation wavelengths. The limitation of PLS is that building of an adequate training set requires an extensive set of measurements that encompasses all possible C values across the range of tissue types. This must be repeated for each photosensitizer. Furthermore, a wide spectral measurement range must be used to provide an adequate spectral "signature" in the tissues. Applying it to layered tissues or inhomoge-
Photodynamic Therapy Dosimetry
543
neous drug distributions may also present problems. Finally, although it can be very accurate, it provides no physical insight, so that it is difficult to assess what would be the effect of altering any of the measurement conditions. 2.4
Model-Dependent Techniques
To avoid the need for tissue-dependent calibration, assumptions can be made about the fluorescence escape function and excitation fluence distribution, based on a priori knowledge of the tissue optical properties. One approach is to collapse the escape function into 1-D by employing a broad-beam excitation with point collection and assume a single exponential dependence on depth. Another simplification is to assume that the optical properties at Aex and Aem are the same. Thus, Potter and Mang [3] postulated that the fluorescence signal was proportional to C*§2, where 8 is the penetration depth of light in the tissue at Aex = 630 nm. Separate measurements of 8 using diffuse reflectance were required using a second optical fiber probe. This was tested in vivo in an animal model with Photofrin. However, correcting the fluorescence signal by 82 still resulted in errors in the prediction of C. Profio et al. [4] used similar approximations but derived a relationship between the ratio of the fluorescence signal and the quantity (1 — y)/(l + y), where y is the diffuse reflectance. At Acx = 633 nm in tissue-simulating phantoms containing Photofrin, the prediction error for C was —20%, for a range of background absorption coefficients tissue/Ji and E, as summarized in earlier chapters. This requires some method of measuring the tissue optical properties /AU and /AS at both the excitation and emission wavelengths. This then must also be done noninvasively, such as by spatially or time-resolved diffuse reflectance spectroscopy [5]. This complete approach has not been reported. A compromise has been to incorporate the total diffuse reflectance spectrum, which is a measure of the transport aldebo and depends on /XU//AS, into the analysis of the measured fluorescence signal, as, for example, the work of Feld and colleagues [6,7], Patterson and Pogue [8], and Jacques and colleagues [9,10]. In most studies using this approach, the primary goal has been to determine the true (intrinsic) fluorescence spectral shape rather than absolute quantification of the fluorophore concentration. Feld's technique, explained in more detail in Chap. 4, performs well in recovering the intrinsic fluorescence shape, in comparison with Monte Carlo simulations and tests on ex vivo tissue samples. Patterson applied their fluorescence model to frequency domain measurements of liquid phantoms and was able to accurately predict phase, modulation, and DC fluorescence signals.
544
Wilson et al.
Gardener et al. [9] derived an analytical method for extracting the intrinsic fluorescence spectral shape and C value. A 1-D pathlength factor for the penetration of excitation light into the tissue and fluorescence escape was defined for a given geometry. This is a function of the diffuse reflectance, as measured at Acx and A em , as well as the penetration depth, <5. The method was tested on a set of tissue-simulating phantoms and yielded a prediction accuracy of about ±20%. The technique was limited to point excitation/area collection, with the source and collection detectors located far from the tissue surface. For other geometries, the pathlength factor needs to be reformulated. 2.5
Model-Independent Techniques
In these methods, no attempt is made to extract the tissue optical properties. Rather, (semi)empirical normalization factors are applied to compensate for differences in the optical properties between tissues, or the measurement geometry is made independent of these. Finlay et al. [11] were primarily interested in extracting the true intrinsic fluorescence spectra of PpIX and its photoproducts during PDT. They corrected the measured fluorescence spectrum, F*(A), by the measured diffuse reflectance spectrum, R(A), and the diffuse reflectance at the excitation wavelength. It was assumed that the light paths are the same for F and R at the same wavelength, so that the simple ratio F*(A)/R(A) recovers the intrinsic fluorescence spectrum shape, while the reflectance at the excitation wavelength corrects the signal intensity to recover C. The measurement probe consists of a ring of 24 fibers, with 8 fluorescence excitation fibers, 8 reflectance source fibers, and 8 detection fibers used to collect both F and R signals. This ring of fibers encircles a central fiber bundle to collect the diffuse reflectance at A ex . The true fluorescence spectrum was given by FT(A) =
R(A)
This functional relationship is a simplification of the relationship derived by Wu et al. [6] for recovery of the intrinsic spectral shape. The value of 0.8 for the power of the excitation reflectance signal was determined empirically. Measurements on scattering emulsion phantoms demonstrated that FT <* C and is insensitive to background tissue absorption. Absolute quantification requires only a further calibration on a single sample of known C. The probe worked well for the specified geometry, but the source-detector displacement (8 mm) may be too large for some applications, such as in endoscopy. Capulot and Mourant [12] use a contact probe in which excitation light was delivered by one optical fiber and the fluorescence and diffusely re-
Photodynamic Therapy Dosimetry
545
fleeted excitation light by a second fiber at radial distance p. They assumed that at small p the number of scattering events for both fluorescence and diffuse reflectance is the same. Consequently, the effect of tissue optical properties on the fluorescence signal can be reduced by normalizing to the reflection signal. Experiments were performed on scattering solutions over a range of collection distances, p = 0.24-3 mm, with a porphyrin fluorophore (excitation at 442 nm, fluorescence detection at 650-700 nm). Only a limited number of different absorption and scattering coefficients were used, so that the measurements were primarily made to check for linearity of the ratio with fluorophore concentration, which was best for p = 0.64 mm. Between solutions with different scattering properties, there was a 25% difference in the estimated value of C. Weersink et al. extended this approach by using two separate fibers for the fluorescence and diffuse reflectance signals [13]. Again, a contact probe was used with one fiber for delivery of excitation light, a second at PP to collect the fluorescence signal, and a third at pR to collect diffusely reflected excitation light. The ratio F(pp)/R(pR) was proportional to C in a given tissue, PF and PR were chosen to minimize the effects of tissue scattering and absorption, the optimal choices being pp = 0.65 mm and pR = 1.35 mm. This technique provided a prediction error of 15% over the ranges /A, = 0.5-2.0 and /Aa = 0.004-0.1 mm '. In vivo experiments on skin, muscle, and liver tissues demonstrated the utility of this technique. The fluorescence signal alone was strongly dependent on the tissue type, as expected, but using the ratio technique, a linear relationship was found between F/R and C (as determined by ex vivo assay), independent of the optical properties (Fig. 9b). As with many methods, external calibration on a phantom of known fluorophore concentration was required to provide the absolute C value, i.e., to give the slope of this curve. The technique samples an effective tissue volume of several cubic millimeters. The effect of layered tissue or inhomogeneous drug distribution are under investigation. Tissue autofluorescence has been investigated as a surrogate for the escape function by normalizing the photosensitizer fluorescence to the autofluorescence signal, assuming that this varies only with the optical properties between different tissues. This technique has typically used UV excitation to limit the optical penetration depth and hence the number of scattering events before the fluorescence escapes through the tissue surface. Doiron et al. [14] used a system designed specifically for the photosensitizer SnET2, with excitation at 442 nm (HeCd laser), with the autofluorescence and SnET2 fluorescence measured at 570 and 670 nm, respectively. The ratio auu,F*/psF* was compared with C determined by tissue extractions on several tissue types. In most cases, there was a strong correlation over the range C = 1-60 /tg/g, with prediction errors of about ±20%. However, the sensitiv-
546
Wilson et al.
ity at low concentrations (C < 0.1 yu-g/g) was limited by autofluorescence at the same wavelength as the photosensitizer fluorescence. Sinaasappel and Sterenborg [15] alleviated this problem by measuring the autofluorescence signal at the photosensitizer fluorescence band with a second excitation source that was off the photosensitizer excitation peak. They derived a function that relates C to the fluorescence signals such that the measured signal was independent of tissue optical properties and excitation/collection geometry. Four fluorescence signals were used, and the corrected signal takes the form Fj m * F; „
1 + aC
Y = —^ —* F,.m * F ,n 1 + bC
(3)
where i and j refer, respectively, to the excitation wavelengths on and off the photosensitizer excitation peak (405 and 435 nm for HpD) and m and n refer to emission wavelengths of the photosensitizer (>660 nm for HpD) and tissue autofluorescence (577 nm). A series of phantom measurements was used to simulate the detection of HpD fluorescence on an autofluorescence background. A significant improvement was found in the correlation between Y and C compared to F i-m and C for phantom solutions simulating weak autofluorescence. However, for phantoms simulating strong autofluorescence, the fluorescence signal was already strongly correlated with the concentration, and so the improvement using Y was not as appreciable. Also, Y was not independent of the strength of the background autofluorescence, with different calibration curves used for phantoms with different simulated autofluorescence signals. The signal Y was independent of background absorption at the excitation wavelengths (AY = ± 12% over 0.001 < /xa < 0.01 mm '). These autofluorescence-based techniques have a distinct advantage for imaging applications, since the processed signal is insensitive to the collection geometry and only limited spectral information is required. Measuring F* in a tissue volume in which there is minimal scattering can reduce the effects of varying tissue properties. Pogue et al. [16,17] used a confocal geometry to achieve this. This involved limiting the excitation and detection light to a very small subsurface volume. The fluorescence signals were found to be insensitive to local pigmentation (^J and nearly linear with C (Fig. 9c). However, there was still a strong dependence on the tissue scattering. A single, small-diameter delivery and collection fiber in contact with the tissue can be used to approximate the confocal geometry. However, this gives a very weak signal. Hence, multiple independent fibers can be used, separated by a distance large enough to avoid cross-talk (~0.51.0 mm) [17]. On phantoms with varying background absorption, the prediction error using a single large probe (1 mm diameter) and measuring only
Photodynamic Therapy Dosimetry
547
the fluorescence signal was 40%, while a multifiber probe (37 fibers each of 100 ^m diameter) had only a 10% prediction error. The fluorescence signal taken with the multifiber probes was still strongly dependent on the background scattering properties. However, taking the ratio of the fluorescence signal to reflectance of the excitation light reduced the prediction error to approximately 30% over the range of /u,s = 5-40 mm '. This technique measures C only near the probe tip, so that an interstitial probe has also been developed comprising 7 fibers with a total outside diameter of 0.3 mm. All of the techniques described above provide reasonable predictions within a limited set of tissue properties and measurement geometries. It is interesting to note that for accurately predicting C, calibration and modeling techniques have largely given way to model-independent techniques. Calibration techniques can provide excellent accuracy but require separate calibrations for each tissue type and measurement geometry. Modeling techniques have been very useful in extracting the true shape of the fluorescence spectrum. They also can provide some insight into the causes of the distortions in the fluorescence spectra. However, they have had limited success in accurately predicting C, due to errors introduced by approximations in the models and in estimates of the tissue optical properties, required as model inputs. When tissue optical properties are not explicitly used in the models, a separate measurement of the diffuse reflectance is required, introducing further measurement error and complicating the measurement process. The model-independent techniques are the easiest to use and the most accurate to date. They are geometry specific, with each geometry useful for particular applications, i.e., point surface probe measurements measuring tissue volumes, confocal techniques for measuring endothelial layers. All of these techniques require an external calibration to a sample with known C and approximately similar optical properties as the tissue samples being measured. 2.6
Layered Tissues
None of the above methods has been adapted to layered tissue. Recently, Farrell and colleagues [18,19] derived a diffusion theory model of spatially resolved fluorescence, F*(p), for two-layer tissues. This is an extension of earlier work on spatially resolved diffuse reflectance in homogeneous [20,21] and layered media [22,23]. The intention is to use the model to determine the concentration of photosensitizer localized in either the upper or the lower layer, or to derive the upper layer thickness, t. The diffusion theory predictions were compared to two-layer phantom measurements and Monte Carlo simulations. For photosensitizer localized in the top layer only, the shape of F*(p) was almost independent of t, but for drug in the low layer only, F*(p) varied significantly with the top layer thickness. In general, there was
548
Wilson et al.
excellent correspondence between the measured fluorescence, Monte Carlo simulations, and the diffusion model. However, to date, the evaluations have been done only for forward calculations, i.e., for predicting F(p) knowing Cupper* Ciowen and t. The more challenging inverse problem has not been reported. 2.7
Diffuse Reflectance Spectroscopy
Diffuse reflectance measurements [20-27] have also been used to measure in vivo concentrations, with the advantage that they can be used for photosensitizers wth little or no fluorescence. Spatially resolved diffuse reflectance spectroscopy [20,21,24] explicitly estimates the scattering and absorption spectra of the tissue, and C can then be estimated from the absorption spectrum, subtracting the tissue background spectrum. Layered tissue can affect the accuracy of this prediction, but light propagation models that account for the layering can improve the prediction of C in the lower layer [22,23]. An additional advantage of what is, in effect, in vivo absorption spectroscopy of the photosensitizer, is that it allows tracking of effects such as spectral changes due to disaggregation of the drug in vivo, which is very important for optimizing the drug-light time interval and the treatment wavelength. Determining these by fluorescence is not very straightforward because of the other factors that affect the fluorescence properties in vivo. Thus, reflectance spectroscopic results may be used to complement fluorescence information. In another diffuse reflectance technique, the source/collector fiber separation, p, may be chosen so that the pathlength of the diffusely reflected photons through the tissue is constant across a wide range of scattering values [25-27]. With such a constant optical pathlength, C can be determined from changes in the absorption spectrum seen as the photosensitizer is taken up in the tissue.
3.
PHOTOSENSITIZER PHOTOBLEACHING FOR PDT DOSIMETRY
The determination of the effective "dose" in PDT is complex in comparison with, say, radiation therapy because of the strong dependence on the local concentration and microdistribution of the photosensitizer, on the local light fluence and fluence rate, and on the local tissue oxygenation [28]. In addition, these factors are interdependent and may change dynamically during treatment. Thus, for example, a high initial concentration of photosensitizer increases the optical attenuation, thereby altering the light distribution. Through photobleaching, the photosensitizer absorption coefficient decreases
Photodynamic Therapy Dosimetry
549
during irradiation in a nonuniform way, with the greatest effect being closest to the tissue surface (external irradiation) or light source (interstitial). This then shifts the light distribution toward that in the absence of photosensitizer. The photosensitizer photobleaching may be oxygen and/or singlet oxygen dependent [29], and the local pO2 may change during treatment due to rapid vascular response of the tissue and/or photochemical depletion if the rate of conversion from 3O2 to 'O2 exceeds the replenishment rate through perfusion [30]. This conversion rate depends, in turn, on the local drug-light product, Much of the development of dosimetry techniques and devices for PDT has focused on making separate measurements of these factors, as in the case of measuring C discussed above. The measured or calculated values of C, <£, and pO2 are then combined through a photophysical/photobiological model. One alternative to this is to measure the 'O2 generation directly. In solution, this is commonly done by measuring its near-IR (1270 nm) luminescence. In tissue this has proved elusive because singlet oxygen is so reactive that its lifetime is greatly shortened in biological media, making the luminescence signal very weak. Recently, an extended near-IR PMT has become available, and we have achieved successful measurements of 'O2 luminescence from PDT drugs in cells and in vivo [31]. A technologically simpler and cheaper alternative at present is to monitor the photobleaching of the photosensitizer. The use of photobleaching as a dose metric is based on the idea that, as the photosensitizer goes through each photoactivation cycle (as per Fig. 1), there is a probability that it will be "destroyed," either directly or by indirect action of the singlet oxygen. The greater the level of photoactivation, the greater the photobleaching. It is then assumed that this also reflects a greater singlet oxygen production and, thereby, greater photodynamic effect. Four main issues arise. First, if the photosensitizer fluorescence is used to monitor the photobleaching, then the same problems of making quantitative noninvasive measurements are encountered as discussed above for determination of C. The photobleaching rate, (3 [cm2.J '], can be obtained from a series of relative measurements during irradiation, but to determine the absolute number of photosensitizer molecules photobleached per unit volume the absolute value of the concentration at the start of treatment, C0, is required. For firstorder kinetics, the concentration after a total fluence of O is given by
Coe-
(4)
In the case of indirect photobleaching, such as by 'O2, the kinetics are more
550
Wilson et al.
complex (second order if there is no other photobleaching mechanism), since the rate of 'O2 production depends on the instantaneous value of C. Thus: C(4>) = C () [1 + C () /3<£J '
(5)
Second, depending on the molecular changes to the photosensitizer, there may be changes in the fluorescence without an exactly corresponding change in the photodynamic efficacy of the drug. An example of this is where the photoproducts may be photosensitizing but have a different fluorescence spectrum from the parent photosensitizer, as in the case of ALAgenerated PplX where the original main single emission peak around 630 and 710 nm decreases with fluence but a new peak around 675 nm appears, indicating the appearance of a second photoactive compound [11]. Hence, simply monitoring the 700 nm fluorescence signal underestimates the amount of photoactive drug present and so underestimates the effective PDT dose. If these effects are known for a particular photosensitizer, as determined in solution or in preclinical cell or animal models, then they can be taken into account, although it may be necessary to monitor the fluorescence in multiple-wavelength bands or across the full emission spectrum to obtain adequate data on photobleaching and photoproducts. Third, and most difficult, the interpretation of the photobleaching as a unique reporter of the PDT dose may be an oversimplification. For example, with m-THPC it has been shown even in solution that not only is the photobleaching oxygen (and singlet oxygen) dependent but that this dependence changes with the tissue microenvironment and, especially, with binding of the photosensitizer to biomolecules such as proteins [29]. Thus, the photobleaching measurement depends on the microlocalization of the drug, which of course profoundly affects the photobiological damage. As a result, the photobleaching may by itself not provide sufficient information, and simultaneous measurement of the other factors, such as the local changes in pO2, may also be required. There is little information on this aspect of photobleaching for most of the PDT drugs under investigation. A final issue is that noninvasive or minimally invasive measurements of fluorescence will at best provide a measure of the local value of C averaged over some tissue volume. In practice, with the possible exception of the use of confocal microimaging in vivo, the microdistribution of drug will not be known. It has been observed that the variability in the microdistribution within a single tumor can be comparable to the variation in the volume value of C between subjects [32]. However, it has not been established, other than in specific instances, what, if any, is the relationship between the PDT response and the volume-average value of C. In some cases, the relationship seems to hold, but this may not be generally true. For example, in
Photodynamic Therapy Dosimetry
551
drugs that act primarily on the tissue vasculature, it may be the fraction in the vessels or bound to the vascular endothelial cells that determines outcome, not the total tissue content of photosensitizer. These complicating factors do not mean that photobleaching is of no value in PDT dosimetry but that much more detail work is required to understand the role of the effects in practice. For such studies to be possible, continuing development of minimally invasive techniques and instrumentation for monitoring the micro- and macrokinetics of photobleaching in vivo is essential, as are photophysical and photobiological models of the effects. Even if photobleaching proves to be of limited value as a quantitative measure of the PDT dose, it has a useful role in showing the uniformity and completeness of irradiation. For this purpose, fluorescence imaging is most appropriate, and Fig. 10 illustrates this for two different clinical situations. As discussed earlier, Farrell and colleagues [18,19] have developed a diffusion model of spatially resolved fluorescence for layered tissue. The model can be used to determine the depth of photobleaching in a nonlayered tissue with an initial uniform photosensitizer distribution. As light irradiation progresses, the photosensitizer is differentially photobleached in the top "layer" of tissue, and so the effective thickness of this layer increases. Changes in F*(p) can then be used to derive the effective thickness and, thereby, the depth of photobleaching. If the degree of photobleaching corresponds to depth of necrosis, then it provides a simple means of determining this. As an alternative dosimetry technique, it has been proposed [33] to use photobleaching of the tissue autofluorescence as a dose metric rather
FIGURE 10 Imaging of photosensitizer photobleaching during clinical PDT with Photofrin. (a) In esophageal tumor. (Courtesy of Dr. N. Marcon, Toronto, Canada.) (b) In the surgical resection in (residual) malignant brain tumor. (Courtesy of Dr. P. Muller and V. Yang, Toronto., Canada.) (See color insert.)
552
Wilson et al.
than that of the photosensitizer itself. This is based on the molecular or metabolic changes caused by photodamage. The specific endogenous fluorophore studied to date is NADH in RIF1 cells in vitro treated with BPD and in vivo on normal muscle tissue in a mouse model treated with BPD. The NADH fluorescence decreased relative to the total autofluorescence signal as treatment progressed. The NADH fluorescence lifetime did not change, indicating that the decreasing signal was due to a change in the concentration of NADH rather than a change in the nonradiative processes of the NADH. The authors speculate that the decreasing NADH concentration is a byproduct of the loss of mitochondrial function during treatment, which disrupts the production of NADH. The relationship between the change in NADH and the tissue response (e.g., necrosis) has yet to be established and is likely to be complex and tissue dependent. As in the case of measuring C, diffuse reflectance spectroscopy can be used for photobleaching measurements as an alternative or supplement to fluorescence photobleaching. However, the interpretation of the results may not be identical, since the mechanisms that cause reduced fluorescence are not necessarily identical to those that reduce the photosensitizer absorbance, depending on the particular photosensitizer. 4.
PHOTOSENSITIZER QUANTIFICATION IN TISSUE OR CELL SAMPLES
In preclinical studies in animal models and clinical treatments, it is often useful to measure the photosensitizer concentration in tissue samples (e.g., biopsies). This information may then be used for several purposes: (1) to determine the pharmacokinetics in individual cases where in vivo monitoring is not feasible for either technical reasons or clinical impracticality but where sequential tissue samples can be taken; (2) to give absolute normalization of sequential relative in vivo measurements, e.g., by making a series of measurements at different times (and locations) following administration and then taking a single tissue sample at the end, from which the absolute concentration can be determined and applied to the whole data set; and (3) to give the absolute value of C at the time of treatment as one of the independent outcome-determining factors in PDT dosimetry (and possibly during treatment to monitor photobleaching). For in vitro experiments, corresponding population-average measurements of the intracellular concentration of drug can be made from cell pellets obtained by centrifugation in order to construct dose-response curves. The methods below are essentially identical for cell pellets and tissue samples of the same weight. Several methods have been reported for such quantification, including absorption spectroscopy, (spectro)fluorimetry, high-pressure liquid chroma-
Photodynamic Therapy Dosimetry
553
tography, and radioscintigraphy. The essential factors are to ensure that the whole photosensitizer content of the tissue sample is assayed and that the assay technique indicates the relevant quantity in vivo that determines the PDT efficacy. The photosensitizer can be extracted from the tissue using multistage chemical processing with alcohols, ethers, or alkaline reagents. The challenge is to ensure that all, or at least a known and reproducible fraction, of the photosensitizer is extracted without chemical destruction of the photosensitizer. Alternatively, the tissue can be "solubilized," i.e., mechanically and enzymatically disaggregated into a colloidal suspension [34]. This ensures that there is no loss of material. The enzymatic digestion can also destroy the photosensitizer. The rate of destruction is different for each drug but predictable for most photosensitizers, so that the assay can be reliable if the timing and temperature of the solubilization are well controlled. Radiolabeling is the most sensitive assay, but it requires that a radiolabeled version of the photosensitizer is available that has the same pharmacokinetics as the unlabeled drug (usually the case if 3H or I4C labeling is done), that there is no free label, and that the labeling is stable in vivo. HPLC is particularly useful either if the photosensitizer is nonfluorescent or if it comprises multiple molecular species with potentially different kinetics and binding. Spectrofluorimetry has high sensitivity and can be used either with extracted photosensitizer or on solubilized tissue. Spectral scanning is preferable to simply measuring at a single emission peak in order to subtract the autofluorescence background. The calibration factor relating the signal to the true concentration must be determined independently for each tissue type and photosensitizer. This can be done by referencing against another calibrated technique such as radiolabeling, by generating a "standard curve" in which known concentrations of the photosensitizer are added to homogenized tissue and then extracted or solubilized, or by "spiking" the extracted or solubilized material with a known quantity of photosensitizer. The main factors to consider are (1) how best to reproduce the in vivo photosensitzer binding in the tissue, especially for drugs that have multiple components, and (2) how to account for photosensitizer aggregation that alters uptake and fluorescence yield. The lower detection limit depends on the mass of the tissue sample, the tissue type (because of the autofluorescence background and the loss of fluorescence signal due to light absorption), and the photosensitizer fluorescence yield and solubilization stability. For example, with 2-mg samples, concentrations as low as 0.01 jjug per g tissue can be reliably measured for Photofrin in brain, whereas concentrations of 2 ^tg/g are required for benzoporphyrin derivative in (highly pigmented) liver. For Photofrin, these values are about 5- to 10-fold lower if ether extraction is
Wilson et al.
554
used. These spectroflourimetric methods have excellent linearity over several orders of magnitude, as illustrated in Fig. 11. For nonfluorescent photosensitizers, absorption spectroscopy may be used as an optical assay method on the extracted or solubilized samples. In this case, the lower limit of the measurement is typically an order of magnitude higher than for fluorescence, or worse for highly pigmented tissues, due to the intrinsic tissue absorption background. The assay is most sensitive for photosensitizers with a strong absorption peak in the far-red/near-IR range, where the hemoglobin absorption is lower than at shorter wavelengths. 5.
FLUORESCENCE MICROSCOPY OF PHOTOSENSITIZER MICRODISTRIBUTIONS IN TISSUE OR CELL SAMPLES
Fluorescence microscopy has been widely used to study the microdistribution of PDT photosensitizers in cells and tissues. Qualitative imaging may be determined using conventional (epi)fluorescence microscopy, in which the whole specimen within the microscope field of view is excited and imaged simultaneously. More quantitative data can be obtained by confocal fluorescence microscopy (CFM), in which a laser beam excites the sample one point at a time and the image is formed by raster-scanning of the beam
1001
^c o 10
g -c 'ro a .t ^ c <ji
8 s 1 C CT3 O ^
O O
0.1(0
i-
0.01
0.01
0.1 1 10 100 Tissue Concentration (ug/g): Acid Extraction Method
FIGURE 11 Photosensitizer concentration in different tissues as measured by spectrofluorimetry after solubilization versus true concentration. (Adapted from Ref. 34.)
Photodynamic Therapy Dosimetry
555
FIGURE 12 Confocal fluorescence micrographs of ALA-induced PplX in gastrointestinal tissues, showing the specific microlocalization. (Courtesy of R. DaCosta and Dr. N. Marcon, Toronto, Canada.) (See color insert.)
(and or sample stage): Figure 12 shows examples of CFM in biopsied tissue samples. The signal is then proportional to the number of photosensitizer molecules within the excitation/detection voxel, apart from small background signals from fluorescence excited and detected above and below the focal plane and the contribution from autofluorescence. It is possible, in
556
Wilson et al.
principle, to make absolute measurements of the photosensitizer concentration distribution by calibrating the instrument with samples of similar optical absorption and scattering and containing known photosensitizer concentration, or by normalizing the images to the total concentration measured in a bulk sample of the tissue or call pellet (as above). However, this has not been reported much to date, and no reports achieving better than a factor of about 2 in accuracy are available. Care must be taken in handling the tissue or cell samples to avoid loss or redistribution of the photosensitizer and photobleaching. The latter can occur prior to imaging (commonly during sample sectioning from ambient light) or during the imaging from exposure to the microscope excitation light. Multiphoton CFM significantly reduces the photobleaching during imaging. 6.
FLUORESCENT OPTICAL FIBER FLUENCE RATE SENSORS
As discussed previously, the PDT efficacy is determined by the localized photosensitizer concentration, the availability of molecular oxygen, and the light fluence rate at the treatment wavelength. While reflectance probes can be used to ascertain the fluence rate close to the tissue surface, minimally invasive probes, such as optical fibers that can be inserted into the tissue, are often used. These probes need to have an isotropic response to correctly measure in an anisotropic light field. Isotropically responsing fibers were initially achieved by using a small spherical volume of a highly light-scattering material at the fiber tip. Multiple light scattering in this tip made the response, i.e., the fraction of the light that was transmitted along the fiber to a photodetector, largely independent of the direction of incidence of the light on the tip, except in the backward direction where the tip was attached to the fiber. However, these scattering probes must be at least 0.5 mm in diameter to achieve reasonable isotropy (±10-20%), since a minimal number of scattering events is required [35]. Fluorescent-based sensors, in which the scattering tip is replaced by a small volume of fluorophore (Fig. 13a), do not have this physical limitation because the fluorescence is intrinsically isotropic. Hence, if the shape of the fluorescent tip is optimized for isotropic response, scaling down in size does not change this but simply reduces the signal strength [36]. The principle is shown in Fig. 13a. Either organic fluorophores or solid-state crystals have been used in these probes. The latter have the advantage of no photobleaching, so that the responsivity (detected light/incident fluence) is independent of total fluence exposure. However, there are problems with production and mechanical stability of crystal probes. Some in vivo mea-
557
Photodynamic Therapy Dosimetry
0.!
•3
I.
0.01
CO
cp to
c 0.001
0 (b)
10 u eff
20
30
40
.-ii [cm'
FIGURE 13 Fluorescent optical fiber light fluence probes, (a) Principle showing single (upper) and multiple (lower) point sensors, (b) Investigational volume as a function of tissue optical properties. (Courtesy of N. Pomerleau, Toronto, Canada.)
surements have been performed [37,38] using a ruby microsphere, and other fluorescent crystals have been suggested [39]. These solid-state-based sensors also have fairly low fluorescent quantum yield (typically —5%) and, hence, low responsivity compared with sensors based on organic fluorophores. Various versions of fluence-rate sensors based on incorporating an organic fluorophore into a nonfluorescent matrix material have been proposed [36,39] and used in preclinical studies [37,40,41]. The fluorophore is selected to maximize the response, which is proportional to the product of its absorption coefficient at the PDT treatment wavelength and its fluorescent
558
Wilson et al.
quantum yield. The fluorescence signal is measured, after band pass filtering to block the scattered treatment light, by a standard diode and preamplifier. The matrix material is typically a polymer bonded to the distal end of the fiber, e.g., an epoxy, that is then machined to the correct shape. Monte Carlo simulations have been used to show that the physical dimensions (—200300 /xm diameter, 500 /urn length) of the fibers are much smaller than the effective ("investigational") tissue volume within which the light fluence is sampled, which depends on the tissue optical properties as shown in Fig. 13b. The responsivity of these sensors is typically —10 6 cm 2 . For fluoropores that are photobleached, the responsivity decreases with total fluence exposure, and so this must be monitored to maintain the probes in absolute calibration. Typically, the loss of signal is <5% per 100 J cm"2 exposure. The calibration can be carried out in a known light field: either in a parallel beam of known irradiance or in an integrating sphere. Since the investigational volumes of these probes have a finite size, their accuracy is biased depending on the local light fluence gradient in the tissue. The magnitude of this effect is dependent on the tissue optical properties and irradiation geometry. Thus, bias will occur if the probe is calibrated in an isotropic light field (e.g., in an integrating sphere). If the calibration is performed in an anisotropic fluence rate field, such as a liquid phantom with scattering and absorption coefficients similar to the tissue to be measured, then no bias is introduced. A recent development in this technology is a probe capable of measuring the fluence rate simultaneously at multiple points in the tissue. This is achieved by placing several sensors along the length of the optical fiber, each with a different fluorophore. The fluorophores are selected to be excited by the same (PDT treatment) wavelength, but to have distinctly different emission spectra. The total spectrum is then the sum of the individual spectra, each weighted by the local light fluence. This can be deconvolved, knowing the individual spectra, to obtain the separate fluence values. Thus, the fluence measurement position is "encoded" by the fluorescence spectra. A major advantage of these probes is that they enable volume surveillance of the fluence rate during PDT without the need to step the sensors through the tissue. The probes may be optimized by placing the fluorophore sensors with the highest responsivity at the positions of lowest anticipated fluence rate in the tissue in order to approximately equalize the signals from each sensor. An extension of this concept that is currently being investigated is to incorporate near-IR phosphors instead of (or in addition to) the fluorophores. The spectral deconvolution would be as for the fluence sensors, but now the phosphorescence lifetime is measured in each case. This determines the local
Photodynamic Therapy Dosimetry
559
tissue pO2 value, since oxygen quenches the phosphorescence and so reduces the lifetime in a deterministic way.
REFERENCES 1.
2.
3. 4. 5. 6. 7.
8.
9.
10.
11.
12.
13.
14.
15.
M Panjehpour, RE Sneed, DL Frazier, MA Barnhill, SF O'Brien, W Harb, BF Overholt. Quantification of phthalocyanine concentration in rat tissue using laser-induced fluorescence spectroscopy. Lasers Surg Med 13:23-30, 1993. AJ Durkin, R Richards-Kortum. Comparison of methods to determine chromophore concentrations from fluorescence spectra of turbid samples. Lasers Surg Med 19:75-89, 1996. WR Potter, TS Mang. Photofrin II levels by in vivo fluorescence photometry. Prog Clin Biol Res 170:177-186, 1984. AE Profio, S-S Xie, K-H Shu. Diagnosis of tumors by fluorescence: quantification of photosensitizer concentration. Proc SPIE 1203:12-18, 1990. MS Patterson. In: AJ Welch, MJC van Gemert, eds. Optical-Thermal Response of Laser-Irradiated Tissue. New York: Plenum Press, 1995, pp 333-364. JF Wu, S. Feld, P. Rava. Analytical model for extracting intrinsic fluorescence in turbid media. Appl Opt 32:3585-3595, 1993. MG Miiller, I Georgakoudi, Q Zhang, J Wu, MS Feld. Intrinsic fluorescence spectroscopy in turbid media: disentangling effects of scattering and absorption. Appl Opt 40:4633-4646, 2001. MS Patterson, BC Wilson. Mathematical model for time-resolved and frequency-domain fluorescence spectroscopy in biological tissues. Appl Opt 33: 1963-1973, 1994. CM Gardner, SL Jacques, AJ Welch. Fluorescence spectroscopy of tissue: recovery of intrinsic fluorescence from measured fluorescence. Appl Opt 35: 1780-1792, 1996. SL Jacques. Light distribution from point, line and plane sources for photochemical reactions and fluorescence in turbid biological tissues. Photochem Photobiol 67:23-32, 1998. JC Finlay, DL Conover, EL Hull, TH Foster. Porphyrin bleaching and PDTinduced spectral changes are irradiance dependent in ALA-sensitized normal rat skin in vivo. Photochem Photobiol 73:54-63, 2001. M Canpolat, JR Mourant. Monitoring photosensitizer concentration by use of a fiberoptic probe with a small source-detector separation. Appl Opt 39:65086514, 2000. RA Weersink, MS Patterson, K Diamond, S Silver, N Padgett. Noninvasive measurement of fluorophore concentration in turbid media using a simple fluorescence/reflectance ratio technique. Appl Opt 40:6389-6394, 2001. DR Doiron, JB Dunn, WL Mitchell, BK Dalton, GM Garbo, JA Warner. Fiber optic based fluorescence detection system for in vivo studies of exogenous chromophore pharmacokinetics. Proc SPIE 2396:312-322, 1995. M Sinaasappel, JCM Stenenborg. Quantification of the hematoporphyrin de-
560
16. 17. 18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
Wilson et al. rivative by fluorescence measurements using dual-wavelength excitation and dual-wavelength detection. Appl Opt 32:541-548, 1993. BW Pogue, T Hasan. Fluorophore quantification in tissue-simulating media with confocal detection. IEEE J Sel Top Quant Electron 2:959-963, 1996. BW Pogue, G Burke. Fiber-optic bundle design for quantitative fluorescence measurement from tissue. Appl Opt 37:7429-7436, 1998. TJ Farrell, RP Hawkes, MS Patterson, BC Wilson. Modeling of photosensitizer fluorescence emission and photobleaching for photodynamic therapy dosimetry. Appl Opt 37:7168-7183, 1998. DE Hyde, TJ Farrell, MS Patterson, BC Wilson. A diffusion theory model of spatially resolved fluorescence from depth-dependent fluorophore concentrations. Phys Med Biol 46:369-383, 2001. BC Wilson, TJ Farrell, MS Patterson. An optical fiber-based diffuse reflectance spectrometer for non-invasive investigation of photodynamic sensitizers in vivo. Proc SPIE 219-231, 1990. TJ Farrell, MS Patterson, B Wilson. A diffusion theory model of spatially resolved, steady-state diffuse reflectance for the noninvasive determination of tissue optical properties in vivo. Med Phys 19:879-888, 1992. G Alexandrakis, TJ Farrell, MS Patterson. Monte Carlo diffusion hybrid model for photon migration in a two-layer turbid medium in the frequency domain. Appl Opt 39:2235-2244, 2000. G Alexandrakis, TJ Farrell, MS Patterson. Accuracy of the diffusion approximation in determining the optical properties of a two-layer turbid medium. Appl Opt 37:7401-7409, 1998. RA Weersink, JE Hayward, KR Diamond, MS Patterson. Accuracy of noninvasive in vivo measurements of photosensitizer uptake based on a diffusion model of reflectance spectroscopy. Photochem Photobiol 66:326-335, 1997. JR Mourant, TM Johnson, G Los, IJ Bigio. Non-invasive measurement of chemotherapy drug concentrations in tissue: preliminary demonstrations of in vivo measurements. Phys Med Biol 44:1397-1417, 1999. JR Mourant, IJ Bigio, DA Jack, TM Johnson, HD Miller. Measuring absorption coefficients in small volumes of highly scattering media: source-detector separations for which path lengths do not depend on scattering properties. Appl Opt 36:5655-5661, 1997. JR Mourant, IJ Bigio, DA Jack, TM Johnson, HD Miller. Measurement of absorption coefficients in small volumes of highly scattering media. Proc SPIE 2979:294-299, 1997. BC Wilson, MS Patterson, L Lilge. Review Article: Implicit and explicit dosimetry in photodynamic therapy: a new paradigm. Lasers Med Sci 12:182199, 1997. BW Mcllroy. TS Mann, JS Dysart, BC Wilson. The effects of oxygenation and photosensitizer substrate binduing on the use of fluorescence photobleaching as a dose metric for photodynamic therapy. J Vib Spec 28:25-35, 2002. S Mitra. TH Foster. Photochemical oxygen consumption sensitized by a porphyrin phosphorescent probe in two model systems. Biophys J 78:2597-2605, 2000.
Photodynamic Therapy Dosimetry 31.
32.
33.
34.
35. 36. 37.
38.
39.
40.
41.
561
M Niedre, MS Patterson, BC Wilson. Direct near-infrared luminescence detection of singlet oxygen generated by photodynamic therapy in cells in vitro and tissues in vivo. Photochem Photobiol 75:382-391, 2002. BW Pogue, RD Braun, JL Lanzen, C Erickson, MW Dewhirst. Analysis of the heterogeneity of pO2 dynamics during photodynamic therapy with verteporfin. Photochem Photobiol 74:700-706, 2001. BW Pogue, JD Pitts, M-A Mycek, RD Sloboda, C Wilmot, M., JF Brandsema, JA O'Hara. In vivo NADH fluorescence monitoring as an assay for cellular damage in photodynamic therapy. Photochem Photobiol 74:817-824, 2001. L Lilge, C O'Carroll, BC Wilson. A solubilization technique for photosensitizer quantification in ex vivo tissue samples. J Photochem Photobiol B 39:229235, 1997. W Star. Light dosimetry in vivo. Phys Med Biol 42:763-787, 1997. L Lilge, T Haw, BC Wilson. Miniature isotropic optical fibre probes for quantitative light dosimetry in tissue. Phys Med Biol 38:215-230, 1993. L Lilge, K Molpus, T Hasan, BC Wilson. Light dosimetry for intraperitoneal photodynamic therapy in a murine xenograft model of human epithelial ovarian carcinoma. Photochem Photobiol 68:281-288, 1998. R Bays, G Wagnieres, D Robert, D Braichotte, JF Savary, P Monnier, H van den Bergh. Light dosimetry for photodynamic therapy in the esophagus. Lasers Surg Med 20:290-303, 1997. L Lilge, G Merberg, R Dacosta, BC Wilson. Nd3+ doped glass fluorescenttip fiber optical probes for quantitative fluence rate dosimetry in biological tissue. J Soc Opt Eng 2131:145-154, 1994. L Lilge, BC Wilson. Photodynamic therapy of intracranial tissues: A preclinical comparative study of four different photosensitizers. J Clin Laser Med Surg 16:81-91, 1998. L Lilge, MC Olivo, SW Schatz, JA MaGuire, MS Patterson, BC Wilson. The sensitivity of normal brain and intracranially implanted VX2 tumour to interstitial photodynamic therapy. Br J Cancer 73:332-343, 1996.
16 Controlled Drug Delivery in Photodynamic Therapy and Fluorescence-Based Diagnosis of Cancer Norbert Lange Universite de Lausanne, Lausanne, Switzerland
1.
INTRODUCTION
In 1998, the American Cancer Society estimated that over half a million Americans were expected to die from cancer. Nearly one-quarter of diseaseassociated mortality is related to cancer, making it the number two killer disease of Americans. With 800,000 cases, skin cancer is the most frequently diagnosed, including 40,000 cases of advanced malignant melanoma. Lung and colorectal cancer, followed by breast and uterine cancer in women and prostate cancer in men, are the leading cause of death in the United States. Statistically, in the United States one in four people will have some form of cancer at some stage in their lifetime. Thus, efficient management of these diseases remains one of the most challenging aspects of oncology care to date. 1.1 Conventional Treatment Strategies for Neoplastic Disease Conventional treatment strategies in oncology-related fields range from surgery to radiation therapy to chemotherapy. Each of these strategies has lim563
564
Lange
itations in providing the final goal: complete cure or at least satisfactory management of the disease that does not impede the quality of life. The greatest chance for complete cure is offered when therapy is performed at an early stage of disease progression. Surgical resection is often limited by the ability to identify and remove entirely the tumor and neighboring already transformed tissue areas. These remaining neoplastic cells may proliferate again and cause recurrence. Furthermore, surgery is often not effective against metastatic tumors that may have migrated from the site of the primary tumor. In some cases, effective removal of sufficient quantities of the surrounding healthy tissue ensuring the removal of all cancerous cells is limited. In the surgical removal of glioblastomas, for example, the surgeon would probably not be as aggressive as necessary in order to maintain as many functions as possible. Chemotherapy based on the systemic or regional administration of cytotoxic agents is often impeded by the lack of specificity of the drug for neoplastic tissue. Therefore, this therapy is often limited by systemic toxicity before truly therapeutic drug levels in the tumor can be achieved. Because chemotherapeutic drugs act on rapidly dividing cells, fast-growing cells of the intestinal lining and bone marrow can also be extensively damaged during treatment. Tumor cell heterogeneity and inhomogeneous blood vessel distribution in the tumor mass may result in cancerous cells that become resistant to a particular chemotherapeutic drug, leading to single- or multiple-drug resistance. On the other hand, radiation therapy can be specifically directed to the side of the tumor. It is based on the idea that noncancerous cells divide more slowly and have a better chance of DNA repair. However, this technique is limited by the potential damage to neighboring healthy tissue and, like surgery, depends on imaging devices allowing identification of the tumor site. None of these conventional therapies, whether alone or in combination, has provided complete cures for all types of cancer in all patients. 1.2
Photodynamic Therapy and Fluorescence Photodetection
An attractive and innovative alternative to these conventional therapies of cancer is based on the systemic or topical administration of a photoactive compound, which preferentially accumulates in neoplastic tissue. The interaction with light of an appropriate wavelength, results either in clear fluorescence demarcation or in destruction of tissue laden with the photoactive compound. Depending on the irradiation conditions and the properties of the photoactive substance, these modalities can be referred to as fluorescence photodetection (PD) [I] or photodynamic therapy (PDT) |2] (see Fig. 1),
Controlled Drug Delivery in Cancer
565
FIGURE 1 Schematic presentation of the principle of photodynamic therapy (PDT) and photodetection (PD). After topical or systemic administration of a photosensitizer, the dye accumulates preferentially within the tumor tissue. Upon illumination with light of an appropriate wavelength, the interaction of light with the fluorophore results in either the clear demarcation of tissue containing high amounts of the dye, or the initiation of processes that cause its destruction.
566
Lange
respectively. The recent regulatory approvals of Photofrin, benzoporphyrin derivative monoacid ring A (BPD-MA), and 5-aminolevulinic acid (ALA) as photosensitizing agents for the palliative treatment of cancers of the esophagus [3], the treatment of choroidal neovascularization (CNV) associated with age-related macular degeneration (AMD) [4,5], and the PDT of actinic keratosis (AK) [6,7] finally provided a broad acceptance of these methodologies in the medical community. On a molecular level, both PDT and PD are initiated by the excitation of the photoactive molecule to a short-lived, excited singlet state due to absorption of visible or near-infrared (near-IR) photons. Following excitation, fast radiationless relaxation processes populate the lowest excited singlet state S, (Fig. 2) of the molecule, from which it can return to its ground state S0 by radiationless processes or by emission a photon. In PD procedures this fluorescence is used to visualize diseased tissue areas with high amounts of the fluorophore using highly sensitive fluorescence imaging devices. Apart from fluorescence and radiationless deactivation, during its relaxation to the ground state, the excited molecule can undergo so-called intersystem crossing, ultimately yielding the lowest excited triplet state T,. The slower the decay of this triplet state, the more time the substance has to act on its environment via two principle mechanisms: 1.
2.
Type I photochemistry involves electron transfer between the photosensitizer and a nearby molecule, e.g., a membrane lipid or oxygen. After this redox reaction, the resultant substrate radical may react to generate peroxy radicals, subsequently undergoing radical chain reactions, or, in the case of oxygen, a reaction partner directly forms superoxide (O 2 ). In type II photochemistry, energy transfer occurs directly between the photosensitizer in its triplet state and ground state molecular oxygen, forming highly reactive singlet oxygen ('O 2 ).
Subsequently, both photochemical processes can trigger cascades of biochemical, biophysical, immunological, and physiological reactions, finally resulting in destruction of the irradiated tissue. The therapeutic value of PDT is influenced by complex interactions of multiple parameters such as photosensitizer concentration in tissue and its localization, photosensitizer selectivity, irradiation wavelength, fluence rate, tissue optical properties, tissue metabolic conditions, delay between drug administration and irradiation, and many other factors. However, the management of tumors using PDT can cause tumor ablation within a few days. On the whole, three types of mechanisms may be involved in the rapid response to PDT. First, PDT may directly destroy the malignant cells by necrotic or apoptotic mechanisms. Second, PDT may profoundly change the morphology of the tumor vascu-
Controlled Drug Delivery in Cancer
567
o
PpIX
Energy Transfer
3 + Absorption
4
Fluorescence
PDT
Effects
•Tumor Cell Destruction •Vascular Damage •Immune and Inflammatory Effects
300
400
500
600
700
800
FIGURE 2 Basic illustration of the photophysical and photobiological process associated with PDT and PD illustrated on the basis of the absorption and fluorescence emission spectrum of protoporphyrin IX. (For detailed description, see text.)
568
Lange
lature, including blood vessel occlusion, blood vessel stasis, and/or vascular leakage. Third, PDT can cause a modulation of the immune system by the release of cytokines or other inflammatory mediators from the treated tissue. Depending on the photosensitizer, overall light dose, and experimental protocol, several cell structures have been identified as targets for PDT, among which are microtubules, membrane organelles, especially lysosomes and mitochondria, plasma membrane, and nucleus. Note that the absence of only one of the components involved in PDT (i.e., photosensitizer, light, or oxygen) does not cause any phototherapeutic effects. Thus, accumulation of a photosensitizer in body areas, distinct from the treatment site, e.g., liver or kidneys, will not impair their functionality. On the other hand, hypoxic tissues, such as the cartilage, will in turn be preserved. Although considerable effort has been made to introduce this promising treatment modality into many oncological and nononcological fields, PDT and PD suffer, like any other treatment, from several drawbacks. Major drawbacks in this kind of therapy are the lack of chemical homogeneity and stability of the photosensitizer, skin phototoxicity induced by an enhanced retention of the compound in the skin, unfavorable physicochemical properties, and low selectivity with respect to the uptake and retention in normal versus tumor tissue. The latter condition, i.e., high ratios of tumor to normal tissue for the photoactive agent, avoids or minimizes photodamage in the peritumoral tissue. In order to further increase the impact of PDT and PD on the management of cancer, qualitative improvements must be made in (1) the definition of PDT and PD protocols that yield optimal results and assessment of the potential of PDT to compete with existing treatment strategies; (2) expansion of the scope of PDT for the management of other pathologies, including rheumatoid arthritis, restenosis of arteries after angioplasty, psoriasis, warts, viral or microbial infections, and blood banking; (3) enhancement of the efficacy of PDT and PD with respect to the properties of the photosensitizing agent. Table 1 summarizes the ideal properties of such photoactive agents used for management and the improved diagnosis of malignant and nonmalignant diseases. Since the toxic effect is induced by the local activation of the photosensitizer, while normal tissue not exposed to the light is spared, PDT can be considered to be selective per se. However, due to the drawbacks related to conventional photosensitizers as mentioned above, one of the essential goals in PDT and PD in order to improve the clinical utility of these methods remains enhancement of the ratio between the concentration of the photoactive agent in the tumor and the tissue from which the tumor originated. Simple derivatization of mostly porphyrinic macrocycles into more or less lipophilic/hydrophilic molecules as compared to the lead compound with more or less amphiphilic properties has not led to a major breakthrough
C
o tj
c
-o
0
S fO
O CO
'+3 CO
S
0
2
fD
'>
>
c
CO
2
CD T3 O
D
CD T3 O
4-1
a?
w
^
CD
~ •«=
CD
4-J
X
CD
Q) C
4-J
O SI Q.
O
SI
CL T3 C CO
CD O C CD O
> a
w
CO
0)
L.
111!
o
CD
.c
_D
I-
LL
o E
O
>-> C/) "'
—
CO
c Q_ if
"a o •M
LL
5
<
O
SI
CL
Q. CD O O
-a CD C/5
Q) CO 3 CD
D
^ "co
Photodynamic he
4-"
c/>
.Q D CO CD > "4-"
O CO
O O
4-1
S^ t! C
Illli
^
1 8 CD
£
co
2
II Hill!
si
Q_
H—
O V) CD
t CD Q. O CO
it
o
"co CD T3 — i— CD
CD
'55
o
CO
t a5
° °
^O < sI- O
2 Ql
x
LLI
CO
co
Q-
O 4-<
O SI CL.
'62 _0
o o
05
I
O
CO
o
CL
Q_
570
Lange
in this field. At present, renewed efforts to improve the selectivity of photosensitive compounds specifically address typical properties of neoplastic tissue, including vascular permeability; high cell proliferation rate; lower pH values; inefficient lymphatic drainage; and overexpression of antigen receptors, growth factor receptors, and cell adhesion molecules.
2.
RATIONAL ASPECTS FOR CONTROLLED DRUG DELIVERY
A primary problem in controlled drug delivery in PDT and PD of cancer is that in most instances the differences between normal cells of origin and malignant cells are subtle and difficult to quantify. Shared properties between normal and cancer cells have further confounded the development of more specific tumor contrast agents. In normal cells, growth is regulated through a complex web of physical and chemical signals. Depending on their origin, they can be considered to be those that are continuously renewing (bone marrow, intestine), those that proliferate slowly and may regenerate in response to damage (skin, liver, lung), and those that are relatively static. In adults, the rate of cell birth is maintained to be equal to the rate of cell death, whereas in children a slightly higher cell birth rate is necessary to allow development. Cancer can be defined as the emergence of cellular clusters in which cell growth regulation has gone awry. Any cell in the body has the potential to become cancerous if it receives genetic modulations, resulting from spontaneous replication errors, chemical carcinogens, radiation, or even viral infections. Mutated cells that have lost the ability of cell growth regulation will continue to divide, forming what is called the primary tumor, often originating from renewing tissues, such as the epithelium. As they begin to multiply, angiogenic factors, growth factors, and cytokines that stimulate the formation of new blood vessels are released and tumor-specific antigens are expressed. The tumor may be benign or it may begin to invade surrounding tissues, breaking through the basement membrane. Certain cells from these malignant tumors may further mutate to the extent of gaining the ability to leave the site of the primary tumor, enter the blood stream, and invade other tissues. The metabolism of such mutated cells may be altered with respect to their normal counterparts, and indeed considerable differences in oxygen consumption, iron content, and enzymatic activity have been observed between neoplastic and normal cells. Beyond direct injection into a tumor or otherwise localized "physical delivery" of photoactive substances, the ideal strategies for controlled drug delivery are those that allow for highly selective recognition of the neoplastic
Controlled Drug Delivery in Cancer
571
tissue. Based on the above-mentioned considerations, three basic strategies can be derived (Table 2): 1. 2. 3.
3.
Active targeting of a tumor-specific functions Passive targeting by modulation of the biodistribution properties of the photosensitizer or its formulation Using tumor cell-specific metabolic deficiencies for the targeting of neoplastic cells
ACTIVE TARGETING
As can be seen from Table 2, potential sites for active targeting of neoplastic tissue can be structural membrane proteins, receptors, glycoproteins, or lipoproteins associated with abnormal cell growth. Following the concept of "magic bullets," coined by the German bacteriologist Paul Ehrlich [8], such active targeting moieties include antibodies, antibody fragments, growth factor and hormone receptor ligands (proteins, peptides, and small synthetic organic molecules), and nutrient transporters. One of the most selective interactions between biological molecules known to science is characterized by antibody binding to specific antigens. Hence, it is not surprising that, in the search for magic bullets against cancer, antibodies that specifically target tumor-associated antigens have been sought. Monoclonal antibodies (MAbs), stemming from the immunoglobuliny (IgG) class, are the most commonly used carrier system for controlled drug delivery (see Fig. 3). The intact antibody can be described as a Y-shaped, symmetrical glycoprotein, composed of two identical heavy and light chains. The tips of the Y are the so-called antigen binding sites, which are variable. Complete IgGs have a molecular weight of about 150 kD. In antibody fragments (Fabs), the complement-activating Fc region and the socalled "hinge region" are removed (Fig. 3). Such molecules, either bivalent or monovalent, are of lower molecular weight and show advantages over intact IgG molecules in terms of their tissue penetration properties. Barriers to uniform antibody penetration have been suggested to be the endothelial layer and defense fibrous stroma associated with tumors, as well as the tight packing of tumor cells and the absence of lymphatic drainage. Furthermore, deletion of the Fc region, which is responsible for "effector functions" of the immunoglobulin molecule such as complement fixation, allergic responses, and killer T-cell activation, removes the most immunogenic portion of the IgG. Small conjugates made from these fragments usually have the advantage of faster localization in the target tissue and faster renal clearance. These are desirable properties for imaging and diagnostic purposes. And-
0 o c
0 0 0
0
0
O CO
O CO
O CD CO .>
CA
CA
CA
0
0
0
O CO
O CO
CA CA CA 0 0 0
CO
CJ
'o o
«
0
Q.
o> T3 °
CL <® O)
.— 0 > c/5
33 +-< 72. '*j '72 •*-> a aa 000 a aa
CA
0 0 CO 13 ^O Q. '^_, '^_i • • Q. Q. CA
co Q. CL >
O
o
f
c
0
0
0
CA .^
C35 O) O) Q. CO CO CO CO •—O i_ i_ i_ CT <4_
Vf_
-
-
H—
.i. 0 0 .c -^ •g cr cr .t; co
X. 2 J?
CA
v^ O) CA
a,
c CO
O O O c C L Q - O c r O O ' j r O c A
•j -^ -i: ^ c c o _j -43 -43 9- -r o.
CA
"CA o +j CA
o
w
O
en
| c o
£.£l
0 CD Q O)
0 o)
o«^ E
co ^
W
CA
1| (-0
s •>
O 5 NJ
< o cc m
c o o o M—
CA
O CL 0
4-1
CA
0
.c o
O 0
CO
o
O °
o c
o 0 a. CA
O
iL O
E
E
CM CO
0 -77
0 ^ 0 ««
=
Q-
c
O "
ro
CL
8
® 0
P
O
Q. CA
O
^
Controlled Drug Delivery in Cancer
nr
573
"r
DPhe-Cys-Tyr-DTrp-Lys-Thr-Cys-Thr-OH Somatostatin Specific Peptide (MW - 1.5 kD)
FIGURE 3 Schematic representation of several different targeting moieties based on the format of IgG. IgG consists of two identical heavy chains (blue, green) and two identical light chains (red, yellow). Both are organized as domains with about 110 amino acids in each domain, interconnected by disulfide bonds. The variable domains at the distal end of both arms of the Y-shaped IgG contain the antigen binding domain. Low molecular weight antibody fragments [F(ab')2, Fab, scFv] exhibit different pharmacokinetics and enhanced penetration properties with respect to the entire antibody. Recombinant scFv consists of the variable domains linked by flexible peptide linkers. Even small targeting moieties are represented by high-affinity peptides targeting specific cell surface receptors.
574
Lange
body fragments can be derived from selective cleavage of the whole antibody by pepsin and papain. Besides these antibody fragments, the most commonly used Fabs are recombinant single-chain antibody fragments (scFc), in which the resting fragments of the heavy and light chain are linked by flexible linkers [9,10). Such engineered proteins can also be humanized by replacing the murine framework regions with human framework regions. These carrying moieties, either MAbs, MAb fragments, or even smaller selective targeting peptides, can than be conjugated either directly to the photoactive substance or via intermediated linkers, including dextran [11,12], polyglutamic acid (PGA) [13,14], polyvinyl alcohols (PVA) [15], and polyethylene glycol (PEG) [16] (Fig. 4). In both cases of coupling procedure, the resulting dye-targeting moiety conjugate should fulfill several requirements with respect to its PDT and PD efficacy: • • •
The targeting unit will be selective for the tumor cells independent of their localization, with little or no recognition of healthy tissue. The conjugate will have the ability to bind to or be taken up by the targeted cells. The complex can carry the photoactive substance to the target cells without being disintegrated along the way.
FIGURE 4 Photosensitizers or fluorescent dyes covalently bound to targeting agents might help to improve PDT and PD of neoplastic tissue.
Controlled Drug Delivery in Cancer
• •
•
3.1
575
The photoactive substance will not or will only minimally impair the binding activity of the targeting unit. The targeting unit will not or will only minimally influence the biophysical properties of the photoactive substance, including fluorescence and singlet oxygen quantum yield. The conjugate will have the ability to be cleared from the body rapidly without impairing other body functions or inducing unwanted immunogenic responses.
Antibody-Mediated Drug Delivery in PDT and PD
Since Gold and Freedman discovered the presence of carcinoembyronic antigens (CEAs) [17,18] expressed in colon cancer tissue in 1966, it became an obvious call to involve agents specifically addressing such cellular targets in the management of cancer. However, only 17 years the pioneering work by Mew et al. [19] introduced the combination of photoactive substances and targeting moieties to the field of PDT and PD. Although it is widely accepted that the use of MAb as carrier moiety is one of the major breakthroughs further increasing the impact of PDT and PD in the management of cancer, most studies using this approach remained at the stage of in vitro testing of the immunoconjugates. In the few cases where in vitro and in vivo studies were performed with the same compound, the extrapolation of in vitro results to in vivo experiments was difficult to establish. In the first work with MAb-photosensitizer conjugates [19], hematoporphyrin (HP) was coupled to M-l, a MAb against the DBA/J2 myosarcoma. In this study, largely unpurified MAb, extracted from peritoneal ascites fluid, was conjugated via nonspecific carbdiimid coupling to give a 30:1 HP/Mab ratio. In vitro, this preparation, probably a mixture of HP and HP aggregates absorbed on MAb and mouse serum albumin, induced 95% killing of tumor cells. This was shown to be superior to controls treated with HP, Mab, or light alone. In vivo, however, the enhanced treatment efficacy was only minimal. In another in vitro study, carried out by the same group [20] using a similar HP-Mab conjugate directed against a leukemiaassociated antigen (CAMAL), phototoxicity increased by two orders of magnitude in comparison with free HP. In order to avoid an impairment of the antibody binding affinity and to enhance the molar photosensitizer/antibody ratio, early in the development of such compounds intermediary linkers were used to couple photoactive compounds to the targeting moiety [11]. Since then, photosensitizer/Mab ratios as high as 36:1 have been achieved [11]. However, most of the early works in this research field suffer from poor conjugate characterization and
576
Lange
purification. Furthermore, the charge of the complex may have a role with respect to the pharmacokinetics, biodistribution, binding activity, and internalization capacity of the complex [21]. In a recent work, del Governatore et al. addressed this specific problem in in vitro studies [22]. They coupled the photosensitizer chlorin e6 via a poly-L-lysine (pi) linker and a pl-succinyl linker (pl-sc) to a 17.1 MAb to give positively and negatively charged conjugates having a 4:5 (pi) and 8:9 (pl-sc) photosensitizer/MAb ratio, respectively. When cells from a human colorectal adenocarcinoma with a positive expression of antigens, recognized by the 17.1 MAb, were exposed to the conjugates, the uptake of the cationic (pi) form was found to be 38-fold enhanced as compared with the free photosensitizer, while only a 16-fold increase was observed using the anionic (pl-sc) conjugate. In subsequent PDT studies, the cationic photosensitizer-MAb conjugate was found to be more phototoxic than the anionic form. Compared with the free chlorin e6 and a nonspecific IgG-photosensitizer conjugate, the positively charged 17.1chlorin e6 conjugate was found to be 6.5 and 4.5 times more phototoxic, respectively, when cells were irradiated with a light dose of 3 J/cm2 at 660 nm. The same irradiation conditions resulted only in a 30% cell death when nontarget ovarian cancer cells (OVCAR-5) were exposed to the Mab 17.1photosensitizer conjugate, whereas for target HT29 cells 90% cell death was observed. Although enhanced photosensitizer/Mab [23] ratios are highly desirable to maximize the PDT efficacy of the immunoconjugate for PDT and for PD purposes, high photosensitizer/Mab ratios may result in physicochemical processes, such as self-quenching of the fluorescence, that counteract the use of coupling a photoactive compound to a targeting moiety. For example, when M-35 antibodies were directly conjugated to fluorescein via fluorescein isothiocyanate coupling, immunoconjugates with molar ratios between 4 and 24 were obtained [24]. In vitro characterization of these conjugates showed that highest fluorescence quantum yields were observed with conjugates of a molar photosensitizer/MAb ratio ranging from 10 to 14, whereas higher conjugation ratios showed decreased fluorescence, probably due to exciplex formation upon excitation. Furthermore, the binding activities of the MAb-fluorescein conjugates as a function of MAb/fluorescein ratio were investigated in vitro, using immobilized CEA. A small decrease in binding activity and increase in nonspecific binding was observed when CEA was exposed to Mabs labeled with more than 19 fluorescein molecules. Although the impact of high fluorescein loading on the biological activity was small in vitro, the picture changed completely when nude mice bearing CEA-expressing human colon carcinoma xenografts were injected with these conjugates. Twenty-four hours after administration, 38% and 43% of the injected dose (ID) was found per gram tumor, when unlabeled and
Controlled Drug Delivery in Cancer
577
fluorescein-labeled antibodies (1:4 ratio) were administrated, respectively. Only 31% ID per gram tumor was found in the tumors when higher Mab/ fluorescein ratios of 1:10 were used, further decreasing to 6% ID per gram tumor for 1:14 ratios. These results indicate that the in vivo evaluation of new conjugates should be performed in the early phase of the drug development process for immunophotodetection or immunopho to therapy. Thus, its not rational to design highly sophisticated conjugates with good photophysical properties and high binding activities similar to the native antibody in vitro, without optimization of their biodistribution, binding activities, and selectivity in vivo. Unfortunately, most studies addressing the biodistribution of conjugates of MAb-photosensitizer conjugates and their biological activities with respect to binding affinity and/or phototoxicity have been carried out are mostly in vitro [24-29]. However, in the few available in vivo studies, depending on the MAb-photosensitizer conjugate and the administration route, T/N ratios as high as 70:1 were reported (Table 3) [25]. Schultes et al. [26] reported a 60- to 130-fold increase in phototoxicity when phthalocyanines were coupled to a B43.13 antibody against the CA125 antigen expressed in abundance in ovarian cancer cells. No improvement in the in vitro phototoxic properties was reported when benzoporphyrin derivative (BPD) was coupled via a PVA linker to antibodies [30]. In vitro [28] as well as in vivo studies [31] have shown that the phototoxicity of MAb-photosensitizer conjugates can be reduced as compared with the free photosensitizer even if highly favorable T/N ratios were achieved. This can be attributed to ineffective binding [31] or poor internalization properties of some MAb-photosensitizer conjugates [28,32]. With respect to the biodistribution, there are at least four parameters that may affect the efficacy of the MAb-mediated PDT. These are: 1. 2. 3. 4.
A maximal concentration of the photoactive compound in the tumor A maximal T/N ratio Delivery of the photosensitizer to cellular or tumoral compartments, susceptible to PDT Homogeneous distribution of the targeting complex within the tumor tissue
Most authors focus on (1) and (2), while point (4) is insufficiently addressed in most papers on the improved drug delivery using MAbs as targeting moiety in PDT. Experimentally, the most convenient way to directly investigate the parameters listed above is by radiolabeling of the conjugate. Prior to conjugation to the photoactive substance, antibodies have to be labeled
v^
E
DC
'c _ CD
<^-
t—
CM
CO
o ID
r-
CO CM
CM co
CO
CM CO
a>
CM
oo
CO
CO
CM
C CD
S o
E
*- J2
"Z. ^
^
CD Q. X LU
H Q. • CD X 0 CD X "> CD ^ .^
C CO CD
D
__^
f
0 ^ o ° -r-
0 O
E 0
^ >— c 0 "^ ^ CO ^
0 — CM CD CM "-'
. . _0
CO "o m D CO >
C
0
'E
E
3
CJ
< °
o CD
o
0 T3 O
.C Q. O 0
>
a
> >
0
CO
0 _0
_0
•4-J
^
^
CC ^
^
O
-
"CD
^f
^ 0 ^ CO *". 0 ~
Z t- LO ro oo co r~
>
> >
>
>
CD
00 o o
0 o
0 o
CJ
CD =
E
. . _0
E
- E ' oo — ^
^^ _0
Je 3 ^r 3 'Jr ^ 15 /— 0 (— r- 0 § > _§ ^ _§ > o ro
e
1 OT CM 5 A CO O) CO ^ CD r- Z
CD CO
ir
^ i> ^
^
r- CO
Z
>
>
C CD
c
E 13 I
s
iZ
I
0
_i
0 CO
^ .C
LO CM
CD
|—
— > CD
<^ LU O
<^[
O
Lf) CM
r- O < CJ O CO
"c W
^ LU < ^ 0) . ~*f LU ^~i ^" ^l* CO ^i LO LU CO CO < CJ
CD _^ -0 V, **f 0 ^M» CD r— 111 0 LJJ ^ CJ CJ I
LT) CM
<£ O
"O 0 0 CD _O T3 CD DC
o
Q_ c/)
_0
T3
*^
o> c
+
1
1 +
1
1
+
+
4-
CO
c in Lf) CO
« ~O if,
co
|g
CM
U 0
<M
CO 00 ^" ^t CQ LU
'-P ^ <
LD
; 00 CO
^ UJ 4CD Lf) ^
r--
Lf5 CO
to co
O) CM <J ^~ D_
CM ^CD ^ ID O
o c: CQ co
H ^
c
Fluoresce
CD
CO "1 uj CD ^ "° LU Q
E o
Fluorochr
^
_f£ c
0 C
0
r^
'c 0
0
C
C
CD
c
£c
o —
CD
CJ
CD >
CO >• < >
0
|^^^ Q. S
O
D
CD
C
c
CO
^$ O 00 Q 0 >QL "0 — ^ QQ ^
"^
(j
CO DU X 9= 1< E
Fluoresce
o
T3 0 C 0 >
o
LU CD
Lf) CM
O
O 0 c 'c ro o _0 ro
0
- _C
'ECD 1
3 .Q 'i_
v
1 CD C
4—'
^O
+
_c
monoel
O
ZndDPhth
E
CO
0 .—
~ CD O
CD O
c _o
o CD E ^: 0 £ > o Hro > 0
-a
c ^ t-
D
T tetras
CD
.—
^f"
"CD CO O
O ±r
•4-1
5 o
uma
CD
c
Q.
3 C
O
E
^ o Q. o CD t~ o
N
C
O
J3
_o
0
"c CO < CO CJ 03
<
Controlled Drug Delivery in Cancer
579
with i25I or I31 I using the iodogen or similar methods [24,25,28,29,33]. These techniques enable, for instance, the comparison of the binding activities of the MAb-photosensitizer conjugate with free MAbs, the determination of MAb/photosensitizer ratios, as well as the quantitative evaluation of the accumulation of the immunoconjugate in the target tissue. In experimental studies, most authors generally consider the tumor to muscle ratio. However, these ratios may not be relevant for potential clinical applications, since muscles are rarely the origin site or primary site of disseminated tumors. As mentioned above, antibodies have also been considered as carrier moieties for fluorochromes, enabling improved detection of early cancerous disease or superficial tumors using modern fluorescence imaging techniques. In such cases, several parameters should be taken into account with respect to the fluorescent label, including the detection and excitation efficacy and the native absorption, scattering, and fluorescence properties of the tissue [34]. One drawback of using fluorochromes for cancer detection is the relatively short mean pathway of light in tissue, which would give poor definition of deeply buried tumors. While the possibility to perform any kind of imaging could be reduced for such lesions, detection would still be possible through considerable thickness of tissue. First experiments with fluorescence-labeled antibodies were carried out using fluorescein as a fluorescent label. This specific dye has a high molar extinction coefficient at 490 nm, corresponding to one of the principle emission lines of an Ar-ion laser (488 nm). It has a high fluorescence quantum yield and a low toxicity in humans [35]. However, the excitation at 488 nm also produces a strong native autofluorescence in tissues [36], causing a significant background signal, which must then be subtracted using complex algorithms. Moreover, the fluorescence quantum yield of fluorescein is strongly pH sensitive having its maximum at pH 8, sharply decreasing below pH 6. In consequence, several research groups [25,33,37,38] started to use alternative fluorescence markers, thus circumventing the above-mentioned drawbacks. Several sulfoindocyanine dyes have advantageous optical properties, such as an excitation (absorption) maximum in the red and a fluorescence emission in the far red and near-infrared, where CCD cameras have spectral sensitivity peaking. At such wavelengths, the intrinsic tissue fluorescence is negligible and scattering of both excitation and emission light is minimal. Moreover, such dyes are insensitive to pH over a broad range, and self-quenching [24] is less important. Antibodies labeled with Cy5 were used for in vivo tumor detection using both mouse tumor models and human tumors xenografted in nude mice [25,33,37,38]. Folli et al. [25] tested the serum lifetime of iodinated and Cy5-labeled antibodies and found the Cy5 conjugates to have a similar circulation half-time to that of unconjugated antibodies. In contrast, and-
580
Lange
bodies labeled with high amounts of fluorescein were found to have dramatically shorter whole-body half-lives with a reduced ability to localize in tumors. Other derivatives, including Cy5.5 and Cy7, were tested for the efficacy of tumor visualization when coupled to specific monoclonal antibodies [37-39]. However, the presently available conjugates have shown significant limitations in vivo. Upon injection, some of these immunoconjugates were nonspecifically deposited in other organs [37-39]: First in the liver and spleen followed by an increasing fluorescence in the gall bladder [33,39], and second, in the intestines and colon, which were found to be weakly fluorescent for several days [39]. Despite the promising results obtained with conjugates between MAbs and photoactive substances both in vitro and in vivo, they have received only minimal clinical testing for PDT as well as for PD purposes [36,40,41]. Only three pilot studies have been carried out so far. Since then all efforts for clinical testing have been stopped, presumably due to several disadvantages of macromolecular carrier systems, including: 1. 2. 3. 4. 5.
Limited availability disposability of sufficient amounts of MAb Potential risk of systemic toxicity Complicated synthesis and purification procedures Loss of photochemical and/or biological activities Low bioavailability, caused by the large size of such conjugates barely penetrating biological barriers
In humans, a long time is required for a significant portion of circulating blood to access a small tumor, while the circulating half-time of antibodies is relatively short. Thus, the majority of the antibody will be cleared from the body before it has accessed the tumor tissue, and only small amounts of the conjugate will accumulate within the tumor when maintaining acceptably low exposure of patients to the drug. This problem becomes evident when comparing the data obtained in mice and in humans using practically the same MAb-fluorescein conjugates for photodetection purposes [24,36]. In a clinical pilot study for photodiagnostic purposes, a total of six patients were injected with 4.5-9 mg of MAb equivalents directed against CEA in colon cancer. The anti-CEA MAbs (CGP44290) were loaded with 10 and 14 molecules of fluorescein per radiolabeled MAb, respectively. In this study, photodiagnosis was performed 24 hr after injection ex vivo on surgical resections for all patients. In addition, one patient underwent an in vivo fluorescence diagnosis during rectosigmoidoscopy. All tumors were found to be positively stained by the targeting conjugates, but the fluorescence was found to be heterogeneously distributed within the tumor mass. Although 10 times more fluorescence was found in the tumor as compared to the healthy surrounding mucosa, only
Controlled Drug Delivery in Cancer
581
0.059% of the ID was found to accumulate per gram tumor. In contrast, 31 % ID/g tumor was found in mice bearing human colon carcinoma for the same MAb-fluorescein conjugate, indicating the huge difference between preclinical experiments performed on small animals and clinical studies in humans. In gynecology, Zn(II)phthalocyanine, coupled via an avidin-biotin complex to the OC125 antibody, was tested in women with ovarian cancer, positively proven for the expression of the CA125 antigen [40,41]. In these studies, the MAb-photosensitizer conjugate was applied 72 hr prior to irradiation of three women. In the first report [40], tumor devitalization was tested by histopathological analysis of biopsies taken 6 hr after irradiation in the course of the operation procedure. The first evidence of cell death was provided by the destruction of the mitochondrial membrane as confirmed by transmission electron microscopy. Furthermore, mitochondria swelling as well as an increased density of mitochondria was found in the irradiated areas. In a recent update of this study, a controlled trial involving 62 patients suffering from ovarian cancer was carried out [42]. All patients underwent surgical resection of the visual tumor nodules. After tumor debulking, 31 patients were intraperitoneally instilled with MAb-photosensitizer conjugates (2 mg B4313) 72 hr prior to operation and were locally irradiated with 5 J/cm2 at 675 nm. Subsequently, the course of these patients was compared to the control group treated by standard surgery. The survival was found to be 90% after 30 months in the PDT group, whereas only 78% of the control group survived up to this point. However, after 60 months both groups showed essentially the same survival rate (40% for the PDT group and 37% for the control group). Furthermore, 23% of women who underwent PDT showed subelius and 6.8% anastomosis following treatment. 3.2
Antibody Fragment-Mediated Drug Delivery in PDT and PD
In spite of their early promise, MAb conjugates were largely unsuccessful as targeting agents in chemo- as well as in photodynamic therapy. However, since it takes about one decade to develop a new therapeutic agent into a viable commercial product, the design of most antibodies was quite primitive and not conceived by modern molecular biology. Nevertheless, it must be considered that several intrinsic properties of intact antibodies will further delay their use in daily clinical practice. In IgG, the activity of antigen binding is determined by the sequences and conformation of the amino acids of the six complementary determining regions (CDRs) that are located on the light and heavy chains of the variable portion (Fv) of the antibody (see Fig. 3). Once Fv is bound to a specific
582
Lange
antigen, the sequences in the constant region (Fc) will trigger numerous immune functions. Due to their high molecular weight of approximately 150 kD, intact antibodies exhibit a long circulation half-life, which can lead to bone marrow exposure. Furthermore, interactions with Fc receptors located on normal tissue can alter the biodistribution of the MAb and potentially harm the patient. The most important drawback for the use of MAbs in PDT ad PD is, however, their poor bioavailability. Their large size limits the ability to diffuse from vasculature into a tumor, which can be attributed to the high hydrostatic pressure resulting from the disordered tumor blood vessels and the lack of draining lymphatics [43,44]. The resulting heterogeneous distribution of a photoactive substance within the target tissue will subsequently impede the efficacy of photodynamic treatment of the entire tumor mass because the range of action of [O2 was estimated to be in the order of 0.05 fjim [45]. However, instead of targeting tumor cells, one interesting approach worth noting in this context is the targeting of the tumor-associated vasculature. Because PDT is known to induce photothrombic effects, the targeting of vascular endothelial cells by MAb-photosensitizer conjugates with subsequent irradiation can result in occlusion of the blood vessel network. In consequence, tumor cells that depend on the nutrition and oxygen supplied by these vessels will become unable to survive. Another approach to overcoming the problem of poor bioavailability of intact immunoglobulins is to use small antibody fragments, thereby increasing their ability of tumor penetration and reducing the circulation halftime without losing the affinity for the targeted antigen. The smallest fragments, containing only the variable region of the light chain (VL) and the variable region of the heavy chain (VH), have a molecular weight of approximately 30 kD. In vivo, these molecules display rapid, biphasic pharmacokinetics with an equilibration phase of between 2 and 12 min and an elimination phase of typically 1.5-4 hr [46]. The most pertinent way to produce antibody fragments is to derive them from intact antibodies via pepsin cleavage to give bivalent antigen-binding fragments [F(ab') 2 ] (see Fig. 3). Hasan et al. [47] have investigated the influence of the charge on the intraperitoneal (IP) biodistribution of chlorin e6 coupled to a F(ab') 2 derived from the OC125 antibody recognizing the CA125 antigen expressed in 85% of nonmucinous epithelial ovarian cancer [48]. Comparable to similar work by the same group [22], chlorin e6 was specifically coupled via pi and pl-sc linkers to OC125-F(ab') 2 fragments. Subsequently, their binding activity and in vivo biodistribution was compared to that of the free F(ab') 2 . Moreover, the photosensitizer was compared to a nonrelevant rabbit Ig-photosensitizer conjugate, respectively. While nonspecific mouse IgG showed no specific binding in enzyme-linked immunosorbent assay (ELISA), cat-
Controlled Drug Delivery in Cancer
583
ionic F(ab')2/chlorin e6 conjugates did and the anionic form even increased binding activity in comparison with the free OC125 F(ab')2. After IP administration of these substances to tumor-bearing mice, free chlorin e6 showed the highest mean T/N ratio as compared with both the anionic and cationic F(ab')2-photosensitizer conjugates 24 hr after application. However, this high T/N ratio for the free photosensitizer is strongly influenced by the high T/N ratio observed in the kidneys, which are one of the major elimination routes for smaller proteins, thus showing rather small T/N ratios for the immunoconjugates. The highest tumor fluorescence was attained using the cationic F(ab')2-chlorin e6 conjugate. Fluorescence was found to be about three times higher than that of the anionic chlorin e6-F(ab')2 or the free chlorin e6. The results of this study indicate that, due to the high amounts of chlorin e6 delivered by the cationic immunoconjugate per gram of tumor, cationic charges stimulate the endocytosis and the lysosomal degradation of the OC125-pl F(ab')2 chlorin e6 conjugate. In the past, much of the screening for tumor-homing molecules was based on the immunization of mice and the screening for monoclonals in cultured pools of antibody-producing cells. In 1988, methodologies were developed to produce active antibody fragments in Escherichia coli [49,50]. The pioneering work of McCafferty and colleagues [51] a few years later allowed the development of large, combinatorial libraries of antibody fragments on the surface of bacteriophages for the screening of more efficient targeting moieties. In such, so-called phage display libraries, the interaction force between VL and VH is very low. Thus, engineering of a covalent link between VH and VL becomes necessary to obtain a stable molecule. The most common method for this purpose is to use a flexible linker of 15-20 residues to join the two domains, giving the resultant single-chain Fv fragment or scFv (see Fig. 3) [52]. Nowadays, these recombinant antibodies and antibody fragments represent over 30% of all biological proteins undergoing clinical trials. This boom is mostly due to modern phage display techniques. For the production of human antibody fragments, the VH and VL domain of human antibodies are polymerase chain reaction (PCR)-cloned in a filamentous phage DNA and displayed on the surface coat of phage. Then antigen-specific phage is selected by in vitro panning against the target antigen of interest. The advantage of such methodologies is in the isolation of high-affinity human antibodies to many different antigens within a few weeks. Selections performed in large libraries consisting of about 10'° different clones yielded scFv with KD as low as 2 X 10~ l() M [53]. In recent work, Ramjiawan and coworkers coupled Cy5 to a recombinant scFv expressed in E. coli for in vivo imaging of mice bearing subcutaneously human melanoma tumor cells [54]. The NovoMab-G2-scFv used in this work recognizes tumor antigens expressed in a variety of human
584
Lange
tumors, including melanoma, breast cancer, colon adenocarcinoma, squamous cell carcinoma, lung cancer, and prostate cancer. Whole-body imaging after illumination at 670 nm of mice injected with the scFv-Cy5 complex resulted in a strong fluorescence at the tumor site two hours after administration. Neither free Cy5 nor an irrelevant antibody fragment conjugated to Cy5 gave marked tumor fluorescence in this study. The strongest fluorescence surpassing the tumor fluorescence was observed in the kidneys in all cases. As expected for small scFv, the fluorescence was found to decrease rapidly, with a mean half-life of approximately 8 hr. Following the approach of targeting the tumor vasculature, the research group of Dario Neri has developed several recombinant scFv's directed against the extra domain B (ED-B) of the fibronectin isoform B (B-FN), which is found in abundance in vessels of neoplastic tissue [55,56]. Antibody fragments isolated from phage display libraries were labeled with different infrared dyes and tested in vitro and in vivo. Affinity measurements for nonlabeled scFv against immobilized fibronectin analogs using surface plasmon resonance techniques have shown KD values in the order of intact antibodies ( 1 . 1 nM for scFv(CGS-2) and 0.54 nM for scFV(L19) [57]). In some scFv, such high affinities were impeded after labeling with Cy7, presumably due to preferential labeling of amino groups in the antigen binding site [56]. Sulfoindocyanine dye labeled scFv's have been tested in several experimental animal models for angiogenic research, including tumor-bearing mice, the chick's chorioallantoic membrane model, and the rabbit's corneal pocket assay. In tumor-bearing mice, a strong and selective accumulation of the scFv(L19)-dye conjugate, but not of the scFv(HyHEL-10)-Cy7, was observed after IV administration (see Fig. 5). Using the scFv(CGS-2)/ Cy7 conjugate, tumor imaging was possible 10 min postinjection, lasting for at least 2 days [55]. For this conjugate, 2% ID/g tumor was localized in the tumor mass 24 hr after administration. Intact antibodies are polyvalent molecules. This provides a significant increase in functional affinity due to simultaneous binding to two or more target epitopes. As mentioned above, a major limitation of scFv molecules is their monovalent nature. Logically, one of the easiest approaches to further improve the binding affinity of scFv to its target antigen is the multimerization of the targeting moiety. Such an attempt will further reduce the elimination of the antibody fragment through renal routes [38,39,47,55,56]. Neri and colleagues addressed this specific problem by introducing an amphiphatic helix containing a cysteine residue at the C terminus to allow two scFv's for dermerization. While scFv(CSG-2) conjugates showed no increase in affinity, a second conjugate [scFv(CSG-l)] showed dramatically improved pharmacokinetic properties. This "diabody" exhibited good stability against
Controlled Drug Delivery in Cancer
585
FIGURE 5 Fluorescence images of nude mice, bearing a subcutaneously grafted tumor, recorded at three time points after injection of Cy7-labeled scFv(L19) and nonspecific scFv(HyHEL-IO). T = tumor site. (From Ref. 56.)
enzymatic proteolysis and low tumor uptake while retaining its binding activities to tumors. Pharmacokinetically, the targeting of neovascular structures is a "onecompartment" problem in which the main input parameters are vascular permeability and interstitial pressure. Furthermore, this approach overcomes the problem of producing specific targeting agents for each cancer, thereby generalizing the problem. However, tumor development is a highly complex process in which angiogenesis is generally preceded by an avascular phase, where the primary tumor is still localized. Superficial and microinvasive tumors as well as carcinoma in situ often lack the presence of a developed neovasculature. Thus, the targeting of angiogenesis-associated antigens by antibody fragment-photosensitizer conjugates can only help to manage or detect advanced tumors. Besides tumor development, angiogenesis is a process that is associated with many other human diseases, such as choroidal
586
Lange
neovascularization (CNV) associated with age-related macular degeneration, diabetic retinopathy, cardiovascular diseases, psoriasis, and rheumatoid arthritis. Thus, the targeting of neovessels in such disorders with scFv-photosensitizer conjugates and subsequent irradiation of the site of the lesion can improve current treatment strategies. Using their scFv(L19), Neri et al. recently demonstrated selective occlusion of neovessels in a rabbit cornea micropocket angiogenesis assay, when coupling chlorin e6 to this targeting moiety [58]. 3.3
Miscellaneous Receptor-Mediated Drug Delivery in PDT and PD
Due to the huge progress that has been made in biological and biochemical research related to cancer during the last decade, many of the processes in the development and growth of cancer cells as well as factors in the microenvironment that are required for tumor growth have been unraveled. This expansion of knowledge has identified a number of potential targets other than antigens for an improved drug delivery in the management of cancer. Worth mentioning in this context are those factors that are involved in angiogenesis, including cell adhesion molecules (CAMs) such as integrins [59-62] and growth factor receptors [63,64] such as vascular endothelial growth factor (VEGF) receptor [65,66] and basic fibroblast growth factor (bFGF) receptor [67]. Furthermore, receptors overexpressed in certain types of tumors, such as epidermal growth factor receptor (EGF-R) [68,69], estrogen receptor (ER) [70], and somatostatin receptors (sst) [71], have known internalizing ligands that can be used as targeting moiety for improved drug delivery. It is beyond the scope of this chapter to list all possible targets for cancer treatment. Instead, it will focus on those targets with potential use in PDT and PD. Recently, scientists involved in the field of biomedical optics have used new targeting agents to enhance the selectivity of photoactive compounds as required for PDT and PD. Although small, synthetic peptides have by far smaller affinities to receptors than antibodies or antibody fragments for their corresponding antigen, they have been proposed as targeting moieties. It has been demonstrated that such molecules, homing specifically on receptors and CAMs, can be conjugated to chelating agents without losing their affinity to the target molecule [72]. Some of the peptides consisting of fewer than 10 amino acids have shown significantly low IC50 values, e.g., octreotide (see Fig. 3) for the sst2 receptor (IC 5() = 0.4-2.1 nM) [73]. Small peptides have several advantages over large biomolecules, including ease of synthesis of a variety of compounds for combinatorial screening, reproducibility of high-purity substances, diffusiveness to solid tumors, absence of
Controlled Drug Delivery in Cancer
587
immunogenic responses, and the ability to incorporate pharmacokinetic modifying functional groups. Octeotride, a cyclic, eight-amino-acid peptide, that mimics somatostatin, has been recently coupled to an indocynanine green (ICG) analog for infrared fluorescence photodetection in rats bearing sst2 expressing pancreatic acinar carcinoma [74]. In this study, tumor visualization was already possible 10 min after intravenous administration of both free ICG and the peptide ICG-peptide conjugate. While ICG retention in the tumor was found to be only transitional, the octeotride-dye conjugate was observable for at least 24 hr in the target tissue. However, relatively high amounts of the targeting contrast agent were also observed in the pancreas and adrenal glands, known for increased sst2 expression, as well as in the kidneys, the main elimination channels for small peptides. A similar approach was followed by another research group with the objective of screening a large number of dye-labeled peptides in a highly parallel automatic nanosynthesis [75]. The methodology presented in this paper was applied in the preparation of different vasoactive intestinal peptide (VlP)-derived peptides and allows parallel screening of approximately 560 compounds. The corresponding receptors for this peptide have been detected in various tumors [76,77]. Cell surface receptors are proteins, which are typically composed of several binding domains, which contain at least one specific binding site. Besides their role in cell signaling processes, they are responsible for the delivery and uptake of cell substrates. Thus, receptors may be subdivided in two basic classes: (1) the "cargo"-type receptors, which deliver metabolic substrates, nutrients, and minerals to the cell; and (2) receptors engaged in signaling, triggering different intracellular events, when bound to a specific ligand. The latter class includes receptors with intrinsic tyrosine kinase activity, such as the EGF-R [68,69] and the insulin receptor. Receptors in this class respond to circulating stimuli to provoke the cell to undergo a wide array of shifts in metabolism that ultimately leads to growth and/or differentiation of individual cells. Gijsens et al. proposed using EGF as the targeting carrier for photosensitizer [78,79]. EGF-R was found in elevated levels in many tumors, such as squamous cell carcinoma, melanoma, glioma, bladder cancer, and breast cancer [68,69]. They coupled CM A via dextran, PVA, and human serum albumin (HSA) to EGF and investigated the influence of these linkers on the binding activity. Moreover, the cellular uptake and the production of reactive oxygen species (ROS) was studied and compared with the free CMA. Nonspecific coupling to the targeting unit presumably was the reason that the CMA-EGF conjugates showed smaller binding affinities in vitro as compared with free EGF. Using HSA as linker was found to impair the
588
Lange
binding to EGF-R-expressing cells less and to deliver five times more CMA to the intracellular space as compared with the free photosensitizer. The rationale of using insulin as a targeting unit for improved drug delivery of photosensitizers is the overexpression of the insulin receptor in human breast cancer. Up to 20 times higher insulin receptor expression in breast cancer cells was observed than in nonmalignant human breast cells [80,81]. Chlorin e6-insulin conjugates were reported to be approximately 100 times more phototoxic than free photosensitizer. Furthermore, lower light doses were needed with the targeting agent. This high phototoxicity was attributed to the favorable intracellular distribution of the photosensitizer following insulin-mediated endocytosis [82,83]. However, the preferential accumulation of the photosensitizer within the environment of the nucleus bear the potential risk of mutagenicity of the insulin-mediated PDT, since 'O2 may damage the DNA closely positioned near the nuclear membrane. Both approaches using tyrosine kinase-activating mediators focus on the targeted intracellular delivery of photoactive compounds, an important aspect with respect to the efficacy of the photodynamic treatment, which has been recently reviewed by Sobolev and colleagues [84]. However, these studies have not addressed the principle parameters of PDT which are probably more critical for PDT efficacy such as biodistribution, T/N ratios, and photosensitizer distribution within nodular tumors. The other category of cell surface receptors, the "cargo" receptor family, overexpressed in tumors and neoplastic tissue responds to the high metabolic activity of proliferating cells. Examples of these include the lowdensity lipoprotein (LDL) receptor and the transferrin (Tf) receptor, which serves as a docking site for iron-laden Tf [85,86]. The use of LDL as a drug delivery vehicle can be traced back to Gal et al. in the early 1980s [85]. The core of these blood plasma proteins consists of fatty acid-esterified cholesterol surrounded by a monolayer of lipids and free cholesterol for most tissues. Many neoplastic cells, including many of the most aggressive tumors, show enhanced uptake of LDL, which can presumably be attributed to the enhanced requirement for structural cholesterol and steroid-derived products in highly proliferative tissue [87]. In PDT, this specific property of rapidly dividing cells can be exploited by both direct coupling of a photosensitizer to LDLs or administration of liposomal or free photosensitizers, which upon injection become associated with LDL. The latter and mostly used approach, however, can be considered as passive targeting and will be discussed later in this chapter. The first indication of the targeted delivery of photosensitizer to tumor cells via (LDL)-(LDL-R) interaction was given, when BPD-MA was preincubated with lipoproteins and applied to tumor-bearing mice [88,89]. Fast pharmacokinetics were observed for all compounds in this study. As com-
Controlled Drug Delivery in Cancer
589
pared to the free BPD-MA, both HDL and LDL complexed BPD-MA was found to be more preferentially accumulated in the tumor and exhibited a high phototoxic potential 3 hr postadministration. Higher retention of the LDL precomplexed photosensitizer, as compared with the HDL analog, was concluded from PDT treatments 8 hr after administration, where only the LDL/BPD-MA showed significant photodamage [89]. In a more recent report, chlorin e6 was selectively conjugated via lysine-carbodiimide coupling to the apoprotein of LDL [90]. With an optimal chlorin e6-to-LDL ratio of 50:1, a four- to fivefold increase of photosensitizer uptake was observed in vitro, using a human retinoblastoma cell line. Phototoxicity studies in the same cell line showed an eightfold enhancement of cell killing using the LDL-chlorin e6 conjugate when compared with the free photosensitizer. Macrophages have also been found to accumulate LDL-photosensitizer conjugates to a high extent [91]. These cells act as mediators of foreignbody response to other cells that regulate the immune system. Consequently, they have been used for indirect targeted PDT and PD. Macrophages that have undergone phagocytosis can deliver selectively high amounts of LDLphotosensitizer conjugates to malignant cells via scavenger receptor mechanisms. Highly specific uptake of maleyated bovine serum albumin (BSA) was also observed in J774 murin macrophage-like cells [92]. 4.
PASSIVE DRUG DELIVERY IN PDT AND PD
In contrast to active tumor targeting, where the photoactive compound is covalently bound to a carrier substance that recognizes tumor-specific functions, in passive targeting only the pharmacokinetic properties of the photosensitizer or the corresponding formulation decide on the in vivo biodistribution upon systemic administration. Again, one of the principle targets for such passive drug delivery will be the intrinsic properties of the tumor-associated vasculature. Increasingly, unequivocal evidence has shown that solid tumors are characterized by hyperpermeability of the vasculature [43,44,93-97] and the immaturity of the lymph drainage [93,97]. Furthermore, blood flow velocity may be significantly slower in tumors in contrast with normal tissue. The tumor vascular system consists of both vessels from the preexisting organ or tissue and new vessels stimulated to grow by the release of angiogenic factors [43,44,98]. Among other things, the distribution of nutrients and drugs through the target tissue depends on the length, diameter, number, and geometry of the available blood vessels. The differences in permeability and lymphatic drainage in tumor tissue can produce significant tumor targeting of high molecular weight compounds and particulate carriers, which have a long plasma circulation half-life. These
590
Lange
observations led Maeda and coworkers to coin the term "enhanced permeability and retention" (EPR) for passive drug targeting of polymer conjugates (Fig. 6). Many macromolecules, such as albumin, antibodies, blood plasma proteins, and polymers, accumulate passively within solid tumors. Observations that liposomes and particles with a size of 200-600 nm [96], and many small differences of uptake of macromolecular weighting in the range between 10 and 800 kD are indicative that passive tumor targeting by the EPR effect can be accomplished using polymeric and micellar carriers. After IV administration, polymers that do not bind to blood plasma proteins display a plasma clearance that is primarily governed by the rate of liver uptake and clearance by the kidneys (see Fig. 6b). However, the biodistribution of photoactive substances essentially depends on their ability to bind to serum components, including lipoproteins and albumin. Pharmacological concepts provide several strategies that can considerably alter the pharmacokinetic properties of substances to be used for PDT and PD, including encapsulation in liposomes and derivatization of the lead structures resulting in more lipophilic, hydrophilic, or amphiphilic compounds. In the early 1980s, researchers began a more or less systematic search for improved tumor photosensitizers. Design criteria for such second-generation photosensitizers were built up with respect to purity, toxicity, stability, photophysical parameters, and solubility (Table 1). Presumably due to the high singlet oxygen quantum yield of the porphyrin skeleton, most of the second-generation photosensitizers described belong to this family (Fig. 7). For improved passive delivery of photoactive substances, the porphyrin system on its own provides 12 positions that can potentially be substituted by sulfonic acid, carboxylic acids, hydroxyl, quaternary ammonium salts, carbonyl substituents, and so forth to give a nearly countless number of possible derivatives of the lead structure. Furthermore, the porphyrin cycle can be oxidized, extended, and/or a central ion may be introduced to alter the pharmacological as well as the photophysical properties of the molecule. Similar derivatization possibilities apply for other singlet oxygen-generating compounds, including perylenequinones, triarylmethane dyes, phenothiazines, and hypericin derivatives. The resulting derivatives can be classified according to their solubility and charge into: 1. 2. 3.
Hydrophobic photosensitizers: reduced solubility in water and alcohol; no charged peripherical substituent Hydrophilic photosensitizers: soluble in water and alcohol; three or more charged peripherical substituents Amphiphilic photosensitizers: soluble in water and alcohols; two or fewer charged peripherical substituents
Controlled Drug Delivery in Cancer
591
This part of the chapter will focus only on such photosensitizers and corresponding formulations, which directly or indirectly tend to use the EPR effect for the delivery of photoactive compound to the target tissue. In particular, this applies to hydrophobic and, to a lesser extent, to amphiphilic photosensitizers which, due to their reduced water solubility, must be injected in a suitable delivery vehicle (Table 4). The porphyrin as well as the phthalocyanine skeletons are essentially hydrophobic. Consequently, derivatives of these two classes, having nonpolar substituents, have to be administered in a pharmaceutically acceptable form. The most common way for highly or moderately lipophilic compounds is to encapsulate the photosensitizer in liposomes or to prepare emulsions using emulsifiers such as ethoxy castor oil, Cremaphor-EL (CR), or polyoxyethylsorbitan monooleate (Tween) [99-125]. Liposomes are built up from lipids that contain two regions: (1) a polar hydrophilic head group (usually a phosphate group) and (2) a nonpolar, lipophilic tail, most likely consisting of a hydrocarbon chain with 14-18 carbon atoms. Highly hydrophobic compounds are generally carried in the lipid bilayer of liposomes, while hydrophilic drugs are entrapped inside the liposomal sphere. Jori and coworkers were the first to study the impact of liposomes on the biodistribution of Hp following IP administration [99]. The tumor uptake of Hp encapsulated in unilamellar liposomes of dipalmitoylphosphatidylcholine (DPPC) in MS-2 fibrosarcoma bearing mice was found to be slower as compared with the free drug. However, the final tumor concentration of liposomal-formulated Hp was approximately twice as high, while uptake in normal skin was lower than with the free Hp. Similar experimental conditions were used with another photosensitizing agent, Zn(II)phthalocyanine (ZnPC), by the same group [100]. Phthalocyanine and naphthalocyanine (Nc) dyes are highly hydrophobic and belong formally to the family of tetraazaporphyrins, with four benzenoid or naphthaloid rings attached to the tetrapyrrolic skeleton, respectively. In this study, maximal tumor concentrations of 0.6 ju-g/g tissue at 24 hr after IP application were reported, while for the same drug IV administration resulted in 0.3 /xg/g tumor tissue using approximately fourfold lower drug doses [101]. For IP administration, the maximal tumor/muscle ratio was found to be 7.5 at 24 hr postinjection, which was twice as high as for IV injection. For both administration routes, never more than 0.1 /Ltg/g tissue was observed in the skin between 1 and 168 hr postadministration. The highest T/N ratio with this photosensitizer was observed when ZnPc was delivered in a liposomal formulation consisting of a mixture of palmitoyloleoylphosphatylcholine (POPC) and dioleoylphosphatylcholine (9:1) instead of DPPC liposomes [102]. Approximately 10 times more of the photosensitizer was found in the tumor than in the normal tissue 48 hr after IV administration in Meth-A sarcoma-bearing mice.
592
Lange
Controlled Drug Delivery in Cancer
593
10
20
15 FIGURE 7 Schematic representation of the porphyrin skeleton, which is the basic structure for most photosensitizers.
However, using other rodent animal models for the investigation of the biodistribution of DPPC-delivered ZnPC qualifies the promising results observed in murine models. Only slight differences in fluorescence intensities in tumor tissue and blood vessels were reported for rats with isogeneic mammary carcinoma [103], and golden hamsters with chemically induced tumors showed tumor/muscle ratios of 1.29 at 72 hr after injection of 0.2 mg/kg body weight [104]. The second group of tetraazaporphyrins, Nc's, are even more hydrophobic than PC'S due to their additional four benzenoid substituents. However, the extension of the conjugated TT system results in a very strong absorption in the 750- to 800-nm region of the spectrum. One of the simplest of these compounds, Zn(II)naphthalocyanine (ZnNc), and three tetrasubstituted derivatives (see Table 4) formulated in DPPC liposomes were the subject of biodistribution and pharmacokinetic studies of Whorle et al. [105] and Shopova et al. [106]. Chemically induced rhabdomyosarcoma in golden hamsters accumulated about 0.7 /Ag/g tissue 24 hr postinjection and the corresponding tumor/muscle ratio was found to be 14. Male C57/BL mice implanted with Lewis lung carcinoma were used to compare the biodistribution of three ZnNc derivatives [106]. Following IP ad-
FIGURE 6 Schematic representation of models for accumulation of low, middle, and high molecular weight targeting agents at tumor and normal tissues after intravenous injection following the EPR concept.
Lange
IV* LL.
c o
ro
"5 E UJ
Z '"" .2 x 03 *~
CNi-CMCNr-CNCNCNCNCNCNCNO)
CDCOCOOO
CD t-
CD r-
CNI'3'
LOCOI^O^
CM C\i (<
IT) 00 CN IT) O) -^
«-OOO<-<-CN<-'-
^ T< <
<<
<-:
*- CD
o
E =
w
-o o
CD
<
> > > > > > > > > > > > Q .
> > > > > > > > >
E 0
o
_
Q. '~\
CD T3 0
-Q
^
T3 CD
CD ^
tj CD CD
"Q) Q
Qw \ 0 O O)
Ef
t i O O O O O O O O O O O O
CD CD CO C/5 3 3 o o
0 CD CD CD CO CO C/) CO 3 3 3 3
o 3
o o o
^
>
' c n n 3
cc^E^^IJE^cE^cE^E^^ieE <EScc^^2E«:EDCQ:;
**""~
•—
w
C D C D C D C D C D C D ( D C D C D C D 0 C D OOC/3C/3C/3C/5C/)C/)C/)C/)C/5C/)C/) 3 3 3 3 3 3 3 3 3 3 3 3
^
^E
LO
LO LT)
LO LO LO LO LO LO CO
LO
r--ococoLOcocr>oooococor^o
OLOCMLOOOOOO
O O ^ O O O O O O O O O O L O 00
L O O O O L O L O r - ^ - ^
c
o
Q.
E o
CJ CD .>
E ® GO
4-J
^
O 03
C/) ^
o
o
S>
^~ 6)
CO Q_
O5 LU
O Q
o ^
Q-D fl^
c
.^
0 Q
\U
o
CD CD
JJ
^CJ OCJ CJ CJ OCJ °r CJ 0 O 0 U ^Q_ D_ Q_ Q_ Q_ Q- Q_ TO Q- CD Q0 n D C ^ Q - Q C D - Q - C C Q - i r Q - D C O Q - O)Q- $ D - D C ^ CC Q- DC L U Q F Q O F C J Q U CJQQCJQQCJQOQCJQ-Q
(A
M— O
1
ti _5 —
?
3 o
Q.
H
c
.2
o
.2 DO
CD >
"D
'4-1
Tf
uj
m H
CJ CD
2
o Q-
X
00
o
0 < r
~O "O i o o io 3 3 CQOQC
X
< < 2
i i i 1 1 1 1 1 1 r" r~ (^ QJ [^ o i x x i r o i o oi o oi o oi o io ;rn : 3 3 3 3 3 3 3— QOQCQOQDQCQm 2
^goooooooooo |_
i
O
O
O
O
O
O
O
O
O
^»
c —
c £5
C3 '^
c ~
O(/3(/)cn-C^
^ ^
O
^ D _ C D C D C D C D C D C D C D C D C D 0 Q _
ooCDCUOOCDOCDCDCDOCDl
> O O O 0 0 ? C C
i^^^^^cococo
o <-
LO CO CO CO CO 0 0 0 0 o
0 <- *•* * 0 0 0 0
r^ oo
CM CO CM CM
o
c" c .E 6
o E
^ 5.
> ^co
o" CL
Q. CD
c
O
0
("•• CD ^t ^ <
*~ Z
<£ <£
LO ^ 00
z z z r^ co
t- LO
r» o o LO «- to
< z £2 £1
Q
CL Q)
CO
a 0 c
CD
-C
c o_ S CD
N Q_CJ
Q. a Q. a a
a > > > > >
> >
-Q E
.— •* CD ^f *™1
o E > >
°1
^ -^_ IM O £ c c co
2 co
Q o i-
CL
CQ CL o ^ . , 1—
a. O
Jp. O C r1-
N ^
3 "
-. CD CD C C —
.E
E D
o
0) CO
Q) CO
D
D
O 0
„ cB
CD
•(->
CO 3
CO E
0) CD
CD
CD
CO D
CO 3
CO D
CO D
O O 0 O
^ ^ 20 ro Is ^ s s ^^s
"o 'c >5 •£,o CD 0 c o ••• -—i CJ
Si >-D
s1
"CD — > O -C >" ^ "co
Is CO
•| g.
ss ss> CM CM CN LO r r LO r- r- CM t- CM 00 00 O O 0 0 O O o o
^. CD
%.
CO 0
"o
CJ
CD
tCD
CO
a. o c
CD
o
.c o
CO
c
o
o"- CO LO
•a m
Q_
^o
VL/
CO
+ O
0 0
0 a. a. CL Q- CC CL CL CL Q U Q Q Q
0 CL
U CL CL Q
CJ CL CL Q
o O
CJ U O U 0 CL CL _l CL CL CL
Q n Q QQ Q _i Q
Q- O
Q a_
(TJ
DC LU 0 CL
CD C CD
C
c
'E N
Q- .E 2 "5 c
Q
CD CD
^ i
CD" ^~ *J ^ c !5 H ^ "o. o o ra 3 X ''"' LU o a §E S- w o ro ^ '•M a a 2 ^ ">- o
gl <S
CD
w
CD O
O ^
^ CO 4—1
Q- c ? > CM" B l ^
§ Q.
S o '<£ c:
> CD X £
-i^ ^° .— JO > ^ >• ^ \ ° CD (D
£'•5 CJ fco '5 2 o >_o 2£ D
Q.
IT
"D < C 4--
^_
O O O) -C 00 ^^ — C Q. ^ ..- i_
O **-** r^j^ 5 2 CL 6 ^ -n S: CD C o (^ 0 o > '^ o -^ .c :- 0 cB n >- CD T3 "5
O ~
X
Q- !5 O *•• DC ^2
Q. 0
•?* 6 ^
0 0 0 O
Z Z z 'K n c c c c c 1- 1- N N N N N
zz
CD
0 0
CL CL c c
o o
CL CL C C
(0
uT uT
0 0 0 0 CL CL c c C C
CL a.
NNNNNNNN
.. -C O CO Q. CD
osome yiphos
Q" O CD L ,J_J
C CO CO V
c
>
=5 LUc"
"CD CD" a .c
JC Q.
—
c <^
CJ CD
Q. ^ W o
'c
fN ^ m •v 0
O)
Q -c _o
-f
N 2 CO
.2 cj •~ °r
E °
o
C
~^~ O) ^ D) O O
C>
£
— -S ^ a.
0 ^
LO LO LO LO LO O CM «- CM CN CM CN CM LO O O O O O
h aJ f—
c
'•3 ? CD CD S 0) CD CD CD CO C/J CO CO CO CO CO D D D D 15 0 0 £ ° O 0 I 5 ^2
JSl Q
M—
'55 3
a CD
£
'_J O
LU
co fp "n C ~r CD 2 N 4-1
"?rt
'*""'
•—^ o .- c o •Q S £
cB S c > ^ co < O o 0 Z
^ C CL N
'o ™ .2 0
CD 2
J2 "a CL
596
Lange
ministration of 0.26 mg/kg body weight of each photoactive compound, tumor/skin ratios were similar for all sensitizers, ranging from 3.0 (ZnPc) to 4.0 (ZnMA4). This indicates that presumably the nature of the delivery vehicle rather than the nature of the photosensitizer influences the targeting properties of these drugs. The biodistribution of bis(diisobutyloctadecylsiloxy)silicon-2,3-naphthalocyanine (isoBosinc) encapsulated in DPPC liposomes was tested in three different animal models [107-109]. At comparable isoBosinc doses of 0.5 mg/kg body weight for both tumor-bearing mice models, controversial results were reported. While tumor/muscle ratios for female BALB7c mice with MS-2 fibrosarcoma were found to be as high as 17 [107], for C57BL/ 6 mice carrying a B16 melanoma neoplasm, this value was reported to be only 1.1 [109], although approximately the same amount of photosensitizer was found inside the tumor (—0.35 yu-g/g tissue). Applied in emulsions using Tween as surfactant, about 0.6 yu,g/g tumor tissue of isoBosinc was found 24 hr after injection [108]. However, using this drug formulation, skin uptake with 0.42 jjLg/g was very high. Interestingly, serum clearance was rapid in all three cases ranging from 85% [107] to 95% [109] at 24 hr postinjection. Several other hydrophobic and amphiphilic photoactive compounds incorporated in liposomes were studied with respect to their biodistribution and pharmacokinetics. Tumor/muscle ratios and tumor/skin ratios of 16.7 and 3.4, respectively, were found when tert-w-propylporphyvene encapsulated in DPPC liposomes was injected into female BALB/c mice bearing a MS-2 fibrosarcoma [110]. The highest tumor/muscle ratio was obtained with tetraphenylporphyrin (TPP) and the correspo