Histopathological Changes Following Photodynamic Therapy in Human Eyes

To identify histopathological changes induced by photodynamic therapy (PDT), 2 human eyes received PDT using verteporfin 1 week before enucleation for large malignant melanoma. Two light doses, 50 and 100 J/cm², were applied to unaffected chorioretinal areas and the optic disc using the standard procedure recommended for patients with age-related macular degeneration (ARMD).

Characteristic hypofluorescence following PDT was documented angiographically 1 week later. The enucleated globes were processed for standard light and electron microscopy.

The PDT-treated areas revealed uniform occlusion of the choriocapillary layer. Vascular endothelial cells were swollen, detached from the basement membrane, and showed rupture and fragmentation to complete degeneration. Capillary lumina were filled with cell debris, fibrin, and thrombocytes. Remaining intact endothelial cells seemed to generate novel vascular channels by recanalization of the obliterated lumen. The overlying retinal pigment epithelium (RPE) showed no significant alteration. Photoreceptors and optic nerve exposed to verteporfin and 100 J/cm² of light were structurally unremarkable.

Photodynamic therapy induces at a dosage used clinically in the treatment of ARMD a selective destruction of vascular endothelial cells within the choriocapillary layer. Recanalization of the physiological choroid is observed as early as 1 week following PDT. Neural structures, photoreceptors, and RPE remained intact.

The therapeutic benefit of PDT using verteporfin in the treatment of predominantly classic choroidal neovascularization (CNV) has introduced a novel strategy into the management of exudative ARMD. A prospective, randomized, double-masked clinical trial¹ demonstrated that verteporfin PDT reduces the risk of severe visual loss, preserves contrast sensitivity, and may even improve visual acuity in a subgroup of patients. The results of the Treatment of Age-Related Macular Degeneration With Photodynamic Therapy Study Group provided the rationale for the recent approval of the method by the health authorities. Meanwhile, thousands of patients with neovascular ARMD received PDT following the guidelines of the Treatment of Age-Related Macular Degeneration With Photodynamic Therapy trial.

The characteristic features observed in patients undergoing PDT seem to be well characterized: clinical exudation of serous fluid, blood, and lipids is halted; and retinal edema resolves.² By angiography, cessation of leakage from classic CNV is documented,³ which usually recurs and requires repeated treatment applications to achieve long-term absence of exudative activity.⁴ However, despite thorough clinical and angiographic evaluation, the knowledge of the photodynamic effects on human ocular structures is limited and mostly based on observation of the secondary phenomena induced by PDT than a revelation of primary mechanisms.

A histopathological evaluation from animal studies has detected changes within the physiological choroid and the RPE and laser-induced CNV.^5,6 In patients, PDT-associated findings seem to be contradictory: retinal function is maintained or improved, while homogeneous choroidal hypofluorescence seen angiographically may be consistent with choroidal nonperfusion.⁷

To identify the structural effects induced by PDT, intact retinal, RPE, and choroidal layers of human eyes that were scheduled for enucleation were exposed to verteporfin therapy. Light and electron microscopic evaluation was performed, with particular emphasis on photoreceptor, RPE, and choriocapillary alterations.

The study protocol was compiled using the Declaration of Helsinki, and approved by the institutional ethics committee. Informed consent was obtained from each patient, specifically indicating the scientific nature of the photodynamic procedure with no therapeutic relevance in respect to the underlying malignancy.

Two eyes of 2 patients received PDT. Both organs were scheduled for enucleation because of malignant melanoma of the uvea too large for conservative management. Tumors were located anteriorly so that the posterior pole was accessible ophthalmoscopically and for photosensibilization. Patient 1 was a 72-year-old man who had a large tumor reaching the ciliary body in the supranasal quadrant, with a tumor base of 35 mm radially and 36 mm at the equator and a height of 12 mm. Patient 2, an 85-year-old woman, was seen with a prominent tumor mass located superiorly at the 12-o'clock position. The tumor size was 24 mm vertically and 23 mm horizontally, and it reached 12 mm in height. Localized exudative retinal detachments were associated with the malignancies. Both eyes showed mild age-related degenerative changes of the macular area with some scattered drusen.

Photodynamic therapy using verteporfin was applied 7 days before the scheduled enucleation. Retinochoroidal areas without involvement by the tumor or the accompanying serous detachment were selected by ophthalmoscopy and angiography. The procedure was performed according to the approved treatment recommendations for patients with ARMD. Verteporfin was administered intravenously by infusion of 6 mg of drug per square meter of body surface area over 10 minutes. Photoactivation was started 5 minutes after the completion of the administration, with light at a wavelength of 689 nm delivered by a diode laser and a laser-link slitlamp (Coherent Inc, Palo Alto, Calif). Two areas per eye were treated subsequently, a first one with a total light dose of 50 J/cm² followed by a second one with application of a light dose of 100 J/cm², both at an irradiance constant of 600 mW/cm². Exposures are shown schematically in Figure 1A and B. In eye 1, the 100-J/cm² spot had a diameter of 5000 µm and was positioned superiorly to the optic nerve, including the disc (Figure 1A). The 50-J/cm² spot with a diameter of 3000 µm was directed onto the retina below the optic nerve head. In eye 2, the area receiving 50 J/cm² was located central to the temporal inferior vascular arcade, while the 100-J/cm² spot was placed below the vascular arcade (Figure 1B).

Visual acuity was examined before, 1 day after, and 1 week after PDT, and was unchanged, with 20/400 in patient 1 and 20/100 in patient 2. No additional field defects were described by the patients. Ophthalmoscopically, no change in the clinical appearance of the retina, the RPE, or the choroid was noted; treated areas were indistinguishable from the surrounding retina. Fluorescein angiography (FA) and indocyanine green angiography (ICGA) were performed to localize and quantify the vascular effects before and 1 week after PDT.

Immediately after surgical removal, the globes were fixed in a combination of 4% paraformaldehyde and 1% glutaraldehyde in a 0.1M phosphate buffer for 5 days and processed for routine paraffin embedding. A macroscopic examination and transillumination were performed, and the tumor dimensions were measured. Angiographic images were used to locate PDT-treated areas before dissection of the posterior pole. Globe 1 was bisected along the laser spots and processed for light and transmission electron microscopy. Globe 2 was reembedded from paraffin into epoxy resin for electron microscopy.

For light microscopy, 8-µm-thick paraffin sections were stained with hematoxylin-eosin and periodic acid–Schiff. For electron microscopy, tissue specimens containing the choroid and retina of the PDT-treated areas and untreated control areas were postfixed in 2% buffered osmium tetroxide and embedded in epoxy resin (Epon) according to the standard method. Semithin sections (1 µm) were stained with toluidine blue O; ultrathin sections (0.2 µm) were stained with uranyl acetate–lead citrate and examined with an electron microscope (LEO 906E; Zeiss, Oberkochen, Germany).

By FA, light-exposed areas revealed homogeneous hypofluorescence characteristic for PDT, with no difference in fluorescence intensity between the 2 different light doses. Indocyanine green angiography provided a more detailed image of the choroidal vasculature and the extent of vaso-occlusion. In patient 1, the image taken 1 week after PDT was partially obscured by the bullous serous detachment associated with the tumor located in the nasal periphery. However, a round hypofluorescent area with a sharp demarcation during early ICGA was documented in the 50- and the 100-J/cm² spot (Figure 2A). Retinal perfusion was not compromised with the higher-dose treatment, including the optic nerve. In late ICGA, hypofluorescence was still present, with scattered areas of leakage from larger choroidal vessels (Figure 2B). No pathologic exudation was seen from the vasculature of the optic nerve.

In patient 2, the central spot that illuminated with 50 J/cm² of light showed absence of the choriocapillary background fluorescence but maintenance of the vascular pattern of medium- and larger-caliber vessels without any reduction in fluorescent demarcation of these vessels during early ICGA (Figure 2C). The area that had received 100 J/cm² of light demonstrated choriocapillary loss combined with a marked decrease in the overall density of choroidal vessels. Late-phase ICGA showed the persistence of hypofluorescence in both areas, with exudation from the margins and the remaining patent vessels within the lesions (Figure 2D).

In globe 1, PDT-treated areas were identified superior and inferior to the optic nerve. Complete occlusion of the choriocapillary vascular lumina was noted in these areas. Vascular channels were blocked by swollen endothelial cells that enclosed red blood cells and fibrin. In some segments, the lumen was completely obliterated, while others demonstrated a residual small opening. Occlusion was seen throughout the dimensions of the treatment spot and reached a depth of approximately 150 µm below the Bruch membrane. Small deeper vessels were also clogged, while larger choroidal vessels showed open lumina with intact, not deformed, erythrocytes. Areas adjacent to the photosensitized spot exhibited a regular choroidal vascular structure with open capillary channels and larger choroidal vessels.

Intact choroidal sections taken from control areas appeared less condensed and thicker than PDT-treated sections. Retinal pigment epithelial cells were attached and showed only minor age-related changes, with vacuolization in the treatment zone and in control areas. Photoreceptors were artificially detached in some areas, but did not show any structural changes. Capillaries of the optic nerve head were open and lined with intact endothelial cells. Neural structures, such as ganglion cells and axons, showed no evidence of structural alteration, histologic signs related to ischemia, or infiltration of the nerve.

Choroidal vascular changes seen in globe 2 were identical to findings observed in globe 1. An occlusion of capillaries of the choroid was seen only in PDT-exposed areas (Figure 3A). Endothelial cells were swollen, detached from the basement membrane, and partially ruptured. Large-caliber vessels appeared to be regular and perfused. However, the RPE and the Bruch membrane were unchanged. Untreated areas showed vascular channels with intact endothelial lining (Figure 3B).

Photodynamic therapy–treated areas showed occlusion and degenerative changes of the choriocapillaris (Figure 4A, B, and D, Figure 5A-F, and Figure 6B-D) compared with untreated areas with wide open lumina, several rows of loosely organized erythrocytes, and intact fenestrated endothelia of the capillaries (Figure 4C and Figure 4E, Figure 5G, and Figure 6A). Vascular alterations in the treated areas comprised swelling of endothelial cells and a few pericytes leading to narrowed, compressed, capillary lumina with densely packed deformed erythrocytes immured between the endothelial lining (Figure 4D and Figure 6B). Alterations further comprised endothelial shrinkage, ruptures, fragmentations, and the detachment of cells from their basement membrane up to complete degeneration of the endothelial lining, leaving denuded vascular basement membranes (Figure 5A and Figure 1B and Figure 6C). Finally, complete occlusion of capillary lumina by fibrin, platelets, and cellular debris derived from endothelial cells, macrophages, and granulocytes was frequently observed ( Figure 4A and Figure 4B and Figure 5C and D). Erythrocytes in the affected capillaries were often degenerative, appearing as ghost cells. Granulocytes accumulated in the lumina of affected capillaries and macrophages seemed to gather in the periphery of degenerated vessels. Extravasation of erythrocytes and migration of inflammatory cells into the extracellular space could often be observed (Figure 5B). In case of advanced endothelial degeneration, the original vessel outline was occasionally delineated by erythrocyte configurations or by electron-dense blood-plasma accumulations ( Figure 6C and D). Remaining intact endothelial cells appeared to reorganize into smaller capillaries within the degenerated original vessel outline, forming a novel vascular lumen (Figure 6D).

Degenerative changes of deeper (outer) choroidal vessels with extravasation of red blood cells was observed in the 100-J/cm², but not in the 50-J/cm², treated spots, where those vessels appeared essentially normal with wide open lumina (Figure 4A and Figure 4B and Figure 5E). Capillaries of the optic nerve head consistently appeared intact after exposure to 100 J/cm² of light.

The RPE appeared mostly intact over large areas and showed focal vacuolar degeneration in the 100-J/cm² treated areas only; partly, bullous separation of individual pigment epithelial cells from the Bruch membrane could be observed. The overlying photoreceptor layer appeared intact in all PDT-treated areas (Figure 5F). Figure 4A and Figure 4D and Figure 5C, Figure 5D, and Figure 5E show areas treated with 100 J/cm² and, therefore, represent rather extensive vacuolization of the RPE cytoplasm and separation from the Bruch membrane.Figure 6, depicting patient 2, is not quite representative because of reembedding from paraffin and a rather poor preservation of the ultrastructure. However, slight alterations of RPE cells, such as intracellular vacuolization indicative of intracellular edema, were also evident in the areas treated with 50 J/cm² (Figure 7). Retinal pigment epithelial cells appeared otherwise normal with regard to their plasma membrane, cytoplasmic organelles, nuclei, and melanin granules.

The retinal capillaries appeared generally normal in areas treated with either 50 or 100 J/cm² (Figure 8).

Photodynamic treatments using verteporfin with the procedure recommended for patients with neovascular ARMD were performed in human eyes. The treatment effect was controlled by angiographic imaging, and the areas of interest were identified and subsequently evaluated by light and electron microscopic histologic features, with the aim to qualify structural changes induced by PDT.

The primary effect of PDT seems to be damage of the choriocapillary endothelium, as indicated by swelling, fragmentation, detachment from its basement membrane, and degeneration; however, recanalization of occluded choriocapillaries seems to occur within a short interval. Significant alterations of the RPE and the neural retina do not occur.

Vascular endothelial damage is the major hallmark of photodynamic tissue effects and results from the direct interaction of a singlet of oxygen with the lipids of endothelial cytoplasmic membranes.⁸ Additional intracellular photo-oxidation causes a rearrangement of the cytoskeleton, with shrinkage of the endothelial cells and exposure of the basement membrane, an intensive thrombogenic stimulus.⁹ Activated platelets release vasoactive mediators (eg, thromboxane, tumor necrosis factor α, and histamine).⁸ Blood flow stasis is the result of endothelial swelling, erythrocyte sludging, platelet adhesion, vasoconstriction, and increased vascular permeability. The release of cytokines, such as interleukin 1β, interleukin 2, and tumor necrosis factor α, from attracted macrophages further facilitates vessel closure.¹⁰ The mechanistic sequelae of these phototoxic effects were seen histologically (Figure 5 and Figure 6): vascular endothelial membranes were discontinuous, and cells were fragmented. Cellular swelling and detachment from the basement membrane were common features. Clots consisted of densely packed red blood cells, thrombocytes, and fibrin. Macrophages were found, consistent with an inflammatory stimulus. None of the described mechanisms is selective for neovascular endothelial cells. Hence, intensive damage to physiological vascular structures should be expected.

In animal studies, the only source of histologic features to date, closure of choriocapillaris following verteporfin therapy, was proved by light and electron microscopy.^5,6 A direct dose-response relationship was found with capillary occlusion at lower doses, while higher light doses invariably led to alteration of larger choroidal vessels and RPE.⁵ The photothrombotic efficacy of verteporfin clearly depends on the drug and light dose applied, with larger-caliber vessels occluded at an increasing dosage.¹¹ Accordingly, occlusive effects were also found in deeper layers and in vessels with larger lumina in areas exposed to 100 J/cm² of light in human eyes. Occlusive choroidal effects are not specific for verteporfin or liposomal preparations, but have also been described with various other compounds with hydrophilic or amphiphilic character, such as purlytin, lutetium texaphyrin, ATX-S10 (Na), and mono-L-aspartyl chlorin e₆.^12-15

Intensive controversy was raised by the characteristic hypofluorescence seen regularly after PDT. Hypofluorescence is most pronounced in FA performed 1 week after PDT. Based on the homogeneity and the intensity of hypofluorescence in an angiographic modality using a low wavelength, like FA, masking of the underlying fluorescence was suspected (eg, by hemoglobin or an altered RPE).¹⁶ The fact that hypofluorescence covers precisely the area of the treatment spot and resolves slowly during the following weeks suggests transient choroidal nonperfusion rather than masking—which was substantiated with correlation of angiographic and histopathological features. The consistency of histologic results and ICGA imaging is striking because by ICGA, choriocapillary shutdown was seen in the 50-J/cm² treated spot (Figure 3A) and disappearance of larger vessels occurred in the 100-J/cm² treated spot (which was supported by the respective extent of vascular obliteration in histopathological features). Indocyanine green angiography may, therefore, be recommended as a useful clinical tool for documentation of photodynamic vascular effects.^7,17,18

How is obvious choriocapillary occlusion compatible with maintenance of visual function? Histologic features reveal structural integrity of overlying photoreceptors after an interval of hypothetical hypoxia for as long as 1 week. Microperimetry, a sensitive test of photoreceptor function, demonstrates improvement of retinal sensitivity within the treated CNV area in 80% of eyes at 4 weeks.¹⁸ The oxygen supply provided by patent large vessels might guarantee photoreceptor survival. A more convincing argument is the observation that progression of occlusion is extremely slow. Short-term follow-up by ICGA demonstrated that choroidal perfusion is not compromised at all for several hours after PDT. Choroidal darkening only occurs during the following days and reaches its maximal intensity after 3 days to 1 week.¹⁹ Apparently, photoreceptors and RPE are better capable of tolerating a prolonged reduction in oxygen supply than an immediate choroidal thrombosis, which is usually associated with significant visual loss.^20,21 A relative reduction in photoreceptor function, however, defined as a transient visual disturbance, was reported by 18% of the PDT-treated patients in the Treatment of Age-Related Macular Degeneration With Photodynamic Therapy trial.¹ Although the progression of the disease was halted successfully, patients lost a mean of 2 lines of visual acuity during follow-up, which might reflect a residual choriocapillary alteration.¹

Recanalization was found in multiple areas of the primarily occluded choriocapillary. Slow progression of photothrombosis over 1 week in concordance with increasing recanalization during the same interval might be a competing mechanism and in summary reduce the extent of ischemia. Recanalization is illustrated histologically by the formation of novel lumina within previously occluded channels. Electron microscopy in experimental models showed duplication of vascular basement membranes indicative of recanalization.²² An intraluminal reorganization would lead to a reduplication of the vascular wall, potentially with restoration of the vascular barrier. Recanalization processes might play an important role in the observed change in the biological features of the CNV complex as well. Cessation of leakage from classic CNV is the prominent angiographic feature in clinical PDT.^1-4 Typically, CNV nets are still delineated in early FA during follow-up, but become progressively silent with resolution of leakage activity. Histopathological features reveal a possible mechanism of barrier stabilization in those with treated CNV with recanalization. Persistent CNV without extravasation would not further compromise retinal integrity. Such an involutional stabilization is the therapeutic rationale of various interventions, such as radiotherapy, antiangiogenesis, and transpupillary thermotherapy, which do not eradicate CNV but decrease the exudative activity.^23-25 Spontaneously regressed CNV histologically exhibits persistent neovascular channels, however, with lack of exudation due to a restored barrier function either due to RPE or endothelial reproliferation.²⁶

Recanalization is most likely an important mechanism of CNV recurrence, particularly because recurrent "growth" and enlargement of CNV occurs at a faster rate than de novo occurrence of CNV.⁴ Choroidal ischemia induced by PDT might stimulate growth factor expression (eg, vascular endothelial growth factor), which is recognized as a major stimulator of CNV.^27,28 Repair of choroidal damage might be a challenge in elderly eyes because choroidal density and integrity were compromised with age.²⁹ A combination therapy of PDT with antiangiogenic agents might, therefore, be promising in respect to CNV recurrence but problematic in respect to choroidal regeneration.

Although verteporfin PDT of CNV does not spare the physiological choroidal capillaries, retinal vessels and neural structures, including photoreceptors and the optic nerve, remained intact even if exposed to a light dose twice as high as conventionally used. Moderate RPE changes could be found. Increased vacuolization was documented, which was partially within normal age-related limits, at 50 J/cm² in our patients, who primarily had age-related macular changes. Whether this damage was a direct consequence of the PDT effect or developed secondarily due to choriocapillary occlusion for as long as 1 week is difficult to analyze. Determining the histologic features immediately or shortly after photosensitization is obviously not an option considering the risk of widespread phototoxic damage during the surgical procedure at a time of high sensitizer retention. Retinal pigment epithelial cells also demonstrate a high regenerative potential following structural damage.^30,31

The histopathological findings of this study identify important mechanisms of PDT in human eyes. The extent of vaso-occlusion with concomitant thrombosis of a normal choriocapillaris is documented, as are active repair mechanisms. In future clinical applications of PDT, the application of a low light dose should be considered to allow for appropriate choroidal regeneration in between treatment intervals and to include choroidal perfusion as a factor for retreatment indication.

The Wellman Laboratories of Photomedicine, Massachusetts General Hospital, Boston, and Dr Schmidt-Erfurth are holders of a patent on the use of verteporfin and have a proprietary interest under the guidelines of Harvard Medical School, Boston.

Corresponding author: Ursula Schmidt-Erfurth, MD, Department of Ophthalmology, University Eye Hospital Luebeck, Ratzeburger Allee 160, D-23538 Luebeck, Germany (e-mail: uschmidterfurth@ophtha.mu-luebeck.de).