Photodynamic parameters in the chick chorioallantoic membrane (CAM) bioassay for topically applied photosensitizers

The relative efﬁcacy of Photofrin w -based photodynamic therapy ( PDT ) has been compared with that of the second-generation photosen-sitizers 5-aminolevulinic acid ( ALA ) , sulfonated chloro-aluminum phthalocyanine ( AlPcS n ) , benzoporphyrin derivative monoacid ring A ( BPD-MA ) , and lutetium texaphyrin ( Lutex ) . PDT-induced vascular damage in the chick chorioallantoic membrane ( CAM ) is measured following topical application of the photosensitizers. In order to make meaningful comparisons, care is taken to keep treatment variables the same. These include light dose ( 5 and 10 J / cm 2 ) , power density ( 33 and 100 mW / cm 2 ) , and drug uptake time ( 30 and 90 min ) . The drug dose ranges from 0.1 m g / cm 2 for BPD to 5000 m g / cm 2 for ALA. Results are also analyzed statistically according to CAM vessel type ( arterioles versus venules ) , vessel diameter, and vessel development ( embryonic age ) . For each photosensitizer, the order of importance for the various PDT parameters is found to be unique. The differences between the sensitizers are most likely due to variation in biophysical and biochemical characteristics


Introduction
Photodynamic therapy (PDT), also called photochemotherapy, is a new modality for the treatment of cancer [1][2][3] and a variety of nononcologic conditions [3][4][5][6][7].PDT is based on administering a photosensitizer, which is preferentially retained in tumor (and other proliferating) tissue.When the ratio of photosensitizer concentration present in tumor and surrounding normal tissue is optimal, the tumor area is exposed to light, at a wavelength coinciding with an absorption peak of the photosensitizer (630-800 nm).Photoexcitation of the sensitizer leads to formation of short-lived reactive oxygen intermediaries that are cytotoxic and cause tumor regression, either directly by cell inactivation [8] and/ or indirectly by destruction of the tumor vascular microcirculation [9,10].
To date, only Photofrin w , a mixture of porphyrins of partially known composition [11], has received US FDA approval for use in PDT of some selected neoplastic diseases [1].Unfortunately, Photofrin absorbs weakly at 630 nm and, moreover, is characterized by prolonged cutaneous phototoxicity.Second-generation photosensitizers for improved photochemotherapy are under development and are in various stages of preclinical and clinical trials [1,3].
It is often difficult to compare the in vivo PDT efficacy of different sensitizers.This is because widely dissimilar testing conditions are used, such as different animals, tumor type/ location, route of drug administration, time interval between drug and light application, irradiation conditions (light dose and power density), and treatment endpoint.Clearly, it is of interest to compare PDT efficacy of different photosensitizers under identical protocols, using Photofrin as the common reference material.
Because of the importance of the vascular pathway in PDT [9], we have chosen the chick embryo chorioallantoic membrane (CAM) to study vascular effects.The CAM bioassay has two distinct advantages over other animal models.First, it contains a dense capillary plexus supplied by allantoic blood vessels, which allows noninvasive study of in vivo microvasculature and blood circulation together with observation and documentation by video-microscopy.Secondly, it is relatively easy to collect statistically significant data in a fast and efficient way.
For some gynecological [6,39], urological [40], and dermatological [41,42] diseases, topical/intravesical application of sensitizer is preferred.This can be simulated by topical application of drug on the CAM.PDT-induced vascular damage in the CAM using different sensitizers was studied with respect to treatment variables and in vivo conditions.

Photosensitizers
The following compounds were used without further purification.Stock solutions were prepared in accordance with the manufacturers' specifications.
Benzoporphyrin derivative monoacid, BPD-MA (QLT, Vancouver); the liposomal powder was reconstituted in H 2 O, 2 mg/ml.Working solutions were prepared by diluting the stock preparation with 5% dextrose.
Photofrin, AlPcS n and Lutex were stored frozen at y208C in 1 ml aliquots, used once after thawing; BPD was stored at 48C for a maximum of three weeks.Working solutions were freshly prepared by dilution with the appropriate solvent.

CAM preparation
The CAM as an in vivo bioassay for PDT has been described previously [43][44][45][46][47] and was adopted here with minor modifications.Briefly, fertilized chicken eggs were disinfected and transferred to a hatching incubator set at 388C and 60% humidity.At day four of embryo age (EA 4), 5 ml of albumin was aspired with a 20-gauge needle, through a hole drilled at the narrow apex, to create a false air sac.At EA 7, a 20 mm diameter window was cut into the shell and covered with a sterile Petri dish; incubation was continued in a static incubator.

Sensitizer application
At EA 12 or 13, when the CAM had matured, 30 ml of sensitizer solution was applied topically in an area demarcated by a Teflon O-ring (6 mm i.d., 1.4 mm annular width) placed aseptically on a well-vascularized site of the CAM.This defined a 30 mm 2 region of interest on the CAM.Sterile, blackened Petri dishes were placed over the eggshell opening and all further manipulations were performed in subdued light.Prior to irradiation, the sensitizer solution inside the ring area was removed with a pipette and the area was washed twice with 150 ml PBS.This ensured that incident radiation was absorbed by the sensitizer actually taken up by the CAM tissue, without being attenuated through a layer of residual sensitizer solution on the membrane.

Irradiation parameters
Irradiation wavelengths were 630 nm (with Photofrin and ALA), 675 nm (with phthalocyanine), 690 nm (with BPD), and 729 nm (with Lutex).CW radiation at 630 and 675 nm was delivered by an argon-laser-pumped-dye laser, model CR599 (Coherent, Palo Alto, CA) and at 690 and 729 nm by stacked-diode lasers (Lawrence Livermore Laboratories, Livermore, CA).The laser light was transmitted through a 400 mm multimode fiber, terminated with an uncoated microlens (Miravant Medical Technologies, Santa Barbara, CA), which expanded the beam to cover the whole ring area (60 mm 2 , including the annular width).Laser output was determined using a Coherent model 210 power meter.Eggs without photosensitizer were irradiated at each wavelength to determine at which light dose (J/cm 2 ) damage occurred.All irradiation conditions were well below these light doses.The power density was 33 mW/cm 2 (with 150 or 300 s irradiation time) or 100 mW/cm 2 (with 50 or 100 s irradiation time).With both power densities, the same total light doses (5 and 10 J/cm 2 ) were obtained, corresponding to energies incident on the CAM surface inside the ring area (30 mm 2 ) of 1.5 and 3 J.

PDT
Preliminary experiments were carried out for each sensitizer by visual inspection, to determine the threshold drug dose that causes visible vascular damage in the region of interest at the lower light dose (5 J/cm 2 ).Groups of five eggs were treated as follows: controls (no drug, no light), dark toxicity (drug only), light sensitivity (irradiation only), and 24 experimental groups using a combination of treatment variables.These included: two light doses (5 and 10 J/cm 2 ) delivered at two time points (30 and 90 min) after drug application, two power densities (33 and 100 mW/cm 2 ), and three drug dosages.Thus, at least 2=2=2=3=5s120 eggs were used for each drug.For selected combinations of treatment variables, the experiments were repeated using groups of five eggs at embryo age EA 12 or EA 13, different from that of the corresponding treatment group.After irradiation, eggs were covered with blackened Petri dishes and returned to the incubator.

Video-microscopic documentation
Immediately prior to drug application and also at one hour post-irradiation, the CAM area inside the O-ring was videotaped with a CCD color camera (Sony, model DXC-101) mounted on a stereomicroscope (Olympus, model SZH), using oblique illumination provided by a fiber-optic light guide (Volpi, Switzerland).Final magnification at the TV monitor was 70=.At a later date, videotapes were analyzed, in a double-blind fashion, for quantitation of PDT-induced damage to the CAM microcirculation.

Damage assessment
The CAM consists of an external layer, derived from the chorion ectoderm, a mesodermal layer containing the extraembryonic vascular network, and an internal layer derived from the allantoic endoderm [43,44].In the present study we have refined the classification of PDT-induced vascular damage in the CAM mesoderm [45][46][47][48].Careful analysis of video images allowed determination of additional gradations in vascular response.'Slight damage' was assigned a score 1 or 1.2, 'moderate damage' a score 1.6 or 2, and 'severe damage' a score 2.5 or 3. Damage was also graded according to vessel diameter (d).We assigned the following (arbitrary) weight factors to the damage score: 0 for 'zero-order' capillaries (d-20 mm); 1 for 'order-1' venules or arterioles (df40 mm); 2 for 'order-2' vessels (df70 mm); and 3 for 'order-3' vessels (d)100 mm).Typically, the CAM area (f30 mm 2 ) inside the ring contained about 75 order-1 blood vessels, 25 order-2 vessels and 10 order-3 vessels, with a slight preponderance of arterioles [43].
For each egg, the weighted damage score was denoted as the highest value of the product 'damage score=vessel-order weight factor', observed in the vasculature of that egg.We have also used this classification to differentiate between the weighted damage scores observed for arterioles and venules.In Table 2 we present the damage classification used in this study (scale 0-9).
In order to utilize only the 'linear' portion of the sigmoidshaped curve depicting damage score versus treatment variable, the slowly varying portion of the curve near the PDT threshold (with weighted damage scores 0-2) was eliminated by shifting the curve such that scores -2 became zero.For the high-damage portion of the curve (scores 6-9), treatment variables were chosen such that stasis and hemorrhage of 'order-3' vessels, associated with weighted damage scores of 7.5 and 9 respectively, were avoided.The remaining scores 2-6 (Table 2, bold face values) were considered to behave in a linear fashion when comparing treatment variables for a given sensitizer.
The mean damage"standard error of the mean (SEM) was calculated and the 'effective' damage scores were obtained by subtracting a constant value of two from the mean damage.The downshifted scores defined the 'effective' damage scale, 0-4, used in Figs.2-4.

Choice of treatment variables
In addition to the standard treatment variables (a) drug type, (b) drug dose, (c) light dose and (d) drug uptake time, commonly used in PDT, we have added four more treatment variables, in an effort to more fully elucidate the mechanism(s) of topical PDT.These are (e) power density (less commonly studied in PDT), and three additional variables particularly suited to the CAM model: (f) vessel type (i.e., arterioles and venules), (g) vessel diameter, and (h) embryo age (i.e., vascular development).
(a) The compounds used were second-generation photosensitizers currently under active investigation [1,12,13].They have different solubilities, span a wide range of photophysical and biochemical properties, and are likely to localize in different subcellular targets.
(b) For topical application on the CAM, three drug dosages were chosen that caused damage in the 'linear' region of the dose-response curve.The selected doses (in units of mg/cm 2 ) were BPD 0.1, 0.25, 0.5; Photofrin 1, 2.5, 5; AlPcS n 1, 2.5, 5; Lutex 5, 10, 25, and ALA 1000, 2500, 5000.It should be noted that ALA requires intracellular conversion to PpIX before becoming photoreactive.Topically applied ALA in humans does not produce a measurable change in the normal systemic levels of porphyrin even when applied at concentrations as high as 200 mg/cm 2 [49], even though localized fluorescence and PDT effects are seen in nearly 100% of the test cases [20].Measured PpIX concentrations following intravenous [50] or intraperitoneal [51] administration indicate conversion efficiencies in healthy tissues of less than 1%.
(c) Light doses required to elicit vascular response in the CAM model were lower [41] than in other animal systems, because there are no losses due to absorption and scattering of radiation by overlying tissue.We have chosen doses of 5 and 10 J/cm 2 , which are about ten times lower than typical light doses used in animal and clinical PDT.
(d) A wide range of drug uptake times is used in PDT.One distinguishes between 'fast'-and 'slow'-acting photosensitizers, having characteristic elimination half-lives (t 1/2 ) in plasma.For example, for lipophilic Photofrin t 1/2 s8 h [52], for PpIX t 1/2 s2 h [41], for hydrophilic AlPcS n t 1/2 s1.5 h [52], whereas for liposomally encapsulated BPD t 1/2 s20 min [4].However, with topical application on the CAM, less difference is expected than with systemic administration, since uptake through the extremely thin ectoderm is dominated by diffusion, which is of the same order of magnitude for molecules having similar molecular weights.For the liposomal preparations of BPD, faster diffusion may occur through the ectoderm CAM membrane and, subsequently, through the endothelial cells of blood vessel walls.In order to monitor the time course of uptake for all sensitizers, we have chosen uniform drug-uptake times of 30 and 90 min prior to PDT treatment.
(e) Power densities have been varied to study PDTinduced oxygen depletion [53,54], tissue response to PDT [37,55] or hyperthermia effects [56].We have used two values (33 and 100 mW/cm 2 ) and have kept the light dose constant by varying the irradiation time.
(f) Because of the difficulty in determining whether arterioles or venules constitute primary targets, vessel type has been the subject of only few PDT studies; the main findings have been arteriole constriction and venule leakage [57].
(g) Vessel diameters in the CAM range from 20 to 120 mm [43], which encompasses typical diameters of neovasculature in tumors.The vessel diameters have been considered implicitly in the determination of weighted damage scores (Table 2).
(h) The CAM model is characterized by a window of time for experimentation between EA 10, when the CAM is complete, and EA 17, when the embryo becomes immune competent.It is possible to perform vascular PDT from EA 10 onward.However, we selected EA 12 and 13 to make the results relevant for the tumor-bearing CAM.In CAM tumor studies, tumor cells are implanted on EA 8 and the eggs are returned to the incubator until EA 12 or 13, when the tumors are of sufficient size for PDT [58].They can then be monitored for two more days to determine the PDT efficacy.It should be noted that the CAM is a dynamic system and between EA 12 and 13 the extraembryonic tissues increase in weight by about 15% (from an average of 4.16 g to 4.78 g) [59].The increase in membrane and/or in vessel-wall thickness may influence drug diffusion across the membrane and increase endothelial cell resistance to PDT-induced damage.The arterial tree is static and vasodilated while aggregates of red blood cells continue to pass through the venous tree.For PDT the CAM was treated with 0.5 mg/cm 2 BPD for 30 min prior to being irradiated with 690 nm light (33 mW/cm 2 , 10 J/cm 2 ).

Statistical analysis
For each sensitizer, statistical analysis of damage scores was performed using a forward stepwise regression analysis to ascertain how well the treatment variables determine vascular damage.In forward stepwise regression, the first treatment variable entered into the regression equation is that with the strongest (positive or negative) correlation with vascular damage.The percentage of variance in vascular damage, which is accounted for by that treatment, is then calculated.The remaining treatment variables are added to the regression equation in successive steps.The stepwise regression algorithm selects the order of entry into the equation such that the treatment that adds the most additional explaining power to the existing regression is added next.

Results
Fig. 1 depicts a representative CAM before (a) and after PDT (b), illustrating damage in blood vessels of various diameters.Damage scores for arterioles and venules were measured separately.Table 3 shows the distribution of maximal damage seen for the combined data from all photosensitizers when sorted by vessel type and diameter.In 68% of the CAMs we observed equal damage of arterioles and venules.In the 220 CAMs in which the arteriole and venule damage was unequal, the maximum damage occurred three times more frequently in arterioles (24%) than in venules (8%).This finding is in agreement with rodent studies, where larger PDT effects were observed in arteries as opposed to veins [57,60].Breakdown of the distribution of maximal damage into photosensitizer groups gave similar patterns for each photosensitizer, with the exception of BPD.With BPD 15% of the CAMs had more venous maximal damage, 9% had more arterial maximal damage, and 76% had equal arterial and venous damage.
The size of a vessel appears to be more important than its type (arteriole or venule) as a predictor of its potential for undergoing PDT damage.Table 3 clearly shows that 'order-2 vessels' are the most likely to be damaged.Although they   represent only 23% of the vessels present in the average region of interest [43], they account for 88% of the maximally damaged vessels.It should be noted that for a maximal damage score in any one vessel of 'order-n', many other vessels of that order or of 'order ny1' may have received the same (or a lower) damage score.It is noteworthy that of all 'order-3' vessels, only seven arterioles, representing seven CAMs, contributed to the maximum damage scores (i.e., about 1% of all CAMs).For the remainder of the analysis, arterioles and venules were grouped together and the higher of the two damage scores was recorded.Embryo age (EA) was an important variable only for Photofrin, where it explained 9% of the total variance in damage.For simplicity, we have combined corresponding data obtained at different embryo ages EA 12 and EA 13.
Table 4 shows the contributions of four treatment variables:drug dose, light dose, power density, and drug uptake time.Since each row of the Table displays the results of a forward stepwise regression analysis, the numbers in a row are order dependent.At each step, the percentage of vascular damage attributed to a particular treatment variable was calculated, starting with the largest, i.e., the most important treatment variable for that sensitizer.Thus the largest number in a row is the percentage of damage with that sensitizer accounted for by that treatment variable.The second-largest number in a row is the percentage of additional variance accounted for by the addition of that treatment variable to the regression equation, and so forth.
The bar graphs, Figs.2-4, include the two or three treatment variables of primary importance to the damage score (Table 4, bold face).The remaining variables, which are of secondary importance, were ignored.An exception to this rule was the inclusion of drug dose as a determining treatment variable, even if not justified by the analysis (as in the case of Photofrin), in order to allow drug dosages to be plotted on the abscissa in a uniform fashion.Figs.2-4 present the mean 'effective' damage "SEM, induced in the CAM vasculature by the five photosensitizers as a function of drug dose and of one or two of the most influential treatment variables.
The number of eggs analyzed per group depended on the number of treatment variables considered to be important; i.e., 10 eggs per group for Photofrin and ALA (Fig. 2) and 20 eggs per group for AlPcS n , BPD, and Lutex (Figs. 3 and  4).
In Fig. 5, the percentage of variance accounting for the vascular damage explained by an individual treatment variable is proportional to the height of the corresponding section in the stacked bar graph.For each sensitizer, the total percentage that could be explained using all four treatment variables is given by the bar height and ranged from 11% for BPD to 54% for Lutex.

Discussion
The photosensitizers chosen for this study ranged from predominantly lipophilic (Photofrin), through amphiphilic (BPD) to hydrophilic (ALA, AlPcS n , and Lutex).This influenced uptake characteristics, since the degree of hydrophilicity of a photosensitizer and the clearance kinetics from blood plasma determine the rate of accumulation in the CAM vasculature.For example, Fig. 5  ence of uptake time for Photofrin.This may be due to the fact that Photofrin is taken up more slowly [52] than the other sensitizers used in this study.In fact, for the relatively short uptake times (up to 90 min) in the present study, the uptake kinetics was rate limiting.As a result, vascular damage depended little on the actual drug dose, which ranged from 'low' (1 mg/cm 2 ) through 'medium' (2.5 mg/cm 2 ) to 'high' (5 mg/cm 2 ).Similarly, uptake time was important for ALA.The accumulation of PpIX, presumably in the mitochondria of endothelial cells [61] of the CAM vasculature, required additional time after cellular uptake of ALA for the intracellular conversion to PpIX.Dependencies on the remaining treatment variables can be understood in a similar fashion [62].The power density was significant only with Lutex, which may be related [34] to a higher sensitivity to heat generation, noting that hyperthermia is known to play a role in PDT [33,47,56,63].
Based on the statistical analysis of five regression equations, we can rank the treatment variables according to their importance in explaining the observed vascular damage, as listed in Table 4: It is apparent from the preceding summary that the photosensitizers tested vary widely in their response to treatment variables, including those commonly investigated; e.g., drug and light dose and drug-light time interval.An important finding is that for Lutex, treatment variables that have not been investigated extensively (e.g., power density) may play a larger role than previously thought.For BPD, the treatment variables considered were capable of accounting for only a small percentage of the variance in damage.This may point to a special role being played by its amphiphilicity or by its liposomal encapsulation [12,13].Additional work needs to be done before the implications of these findings for a better understanding of the mechanisms of PDT become clear.However, these results do establish that different treatment variables, including some lesser-studied ones such as power density, are more important for certain sensitizers than for others.
This study was focused on the vascular effects of topically applied photosensitizers in the non-tumor-bearing CAM.Because of similar size and developmental characteristics of the neovasculature in the CAM and in tumors, our results have relevance to in vivo PDT.A separate investigation is under way in which sensitizers are administered systemically in the chick embryo.This is relevant for clinical PDT mediated by i.p.-or i.v.-administered agents.Future studies will examine PDT of the tumor-bearing CAM, which will allow us to compare the importance of cellular versus vascular pathways of PDT in a controlled model system that approximates the in vivo situation more closely.

Fig. 1 .
Fig. 1.Video-micrographs of CAM vasculature before (a) and after (b) BPD-mediated PDT.Major arterial (A) and venous (V) vessels are indicated.In (b) the arrows denote points of hemorrhage and the arrowheads indicate the location of the empty large vein.The arterial tree is static and vasodilated while aggregates of red blood cells continue to pass through the venous tree.For PDT the CAM was treated with 0.5 mg/cm 2 BPD for 30 min prior to being irradiated with 690 nm light (33 mW/cm 2 , 10 J/cm 2 ).

Fig. 4 .
Fig. 4. Maximal 'effective' damage in CAM vasculature with power density as dominant treatment variable for Lutex-based PDT.

Fig. 5 .
Fig. 5. Percentage of variance accounting for vascular damage, following PDT mediated by different photosensitizers, as explained by five treatment variables.

Table 1
In vivo PDT efficacy, relative to that of Photofrin at 630 nm, for selected second-generation sensitizers (data are on a per weight basis and are not corrected for absorbance) Thursday Dec 23 09:04 AM StyleTag --Journal: JPB (J.Photochem.Photobiol.B: Biol.)Article: 7888

Table 4
Percentage of damage of CAM blood vessels explained by the treatment variables (from stepwise regression).Treatment variables retained in Figs.2-4to explain topical PDT damage are given in bold face