Selective Photodynamic Therapy by Targeted Verteporfin Delivery toExperimental Choroidal Neovascularization Mediated by a Homing Peptide toVascular Endothelial Growth Factor Receptor-2 FREE

Photodynamic therapy (PDT) using verteporfin as a photosensitizer hasbeen demonstrated in large clinical trials to be an effective new treatmentfor subfoveal choroidal neovascularization (CNV) secondary to age-relatedmacular degeneration and other causes.^{1 - 4} Thepreferential occlusion of CNV after PDT as currently practiced is based onthe differences in biodistribution of the photosensitizer between CNV andretinal vessels at the time that light is applied. Preclinical work on PDTwith verteporfin (QLT PhotoTherapeutics Inc, Vancouver, British Columbia)has shown treatment-related damage to the surrounding retina, choroid, andretinal pigment epithelium (RPE).^{5 - 7} Thiscollateral damage is cumulative with repeated PDT.^{7 - 8} Modificationsto PDT, including combination with antiangiogenic therapy^{9 - 10} orwith targeted photosensitizer, may improve selectivity and vision outcomes.

Homing peptides are an emerging class of pharmaceuticals that exploitsdifferences between cell types by binding specific cell membrane receptors.^{11 - 13} The peptide targetingof photosensitizers might enable specific and enhanced retention of photosensitizerto CNV and allow more selective PDT with minimal adverse effects.

Vascular endothelial growth factor (VEGF) expression and binding ofVEGF to its kinase domain receptor (KDR/FLK1 or VEGFR-2) is an important mediatorof angiogenesis, including retinal and choroidal neovascularization.^{14 - 15} Inhibition of VEGF and VEGFR-2 preventsretinal and choroidal neovascularization.^{16 - 18} Thepresence of KDR or VEGFR-2 has been demonstrated in normal vessels but showsincreased expression in endothelial cells of neovascular tissue and is thusa potential candidate for peptide-mediated targeting of CNV.^{19 - 20} Thepeptide ATWLPPR is reported to specifically bind VEGFR-2 and completely inhibitbinding of native VEGF, thereby preventing VEGF-induced angiogenesis in vivo.²¹

We propose to use ATWLPPR as a homing peptide bound to verteporfin totarget verteporfin to CNV by binding to VEGFR-2 on CNV. Because VEGFR-2 isoverexpressed on neovascular endothelium, normal vessels should be relativelyspared and retinal cells should be unaffected after PDT using verteporfintargeted to VEGFR-2. Experiments were designed to evaluate the efficacy andselectivity of PDT with VEGFR2-targeted verteporfin in the rat laser-injurymodel of CNV.

We recovered verteporfin at a concentration of 2 mg/mL in liposomalformulation from material prepared for clinical treatments; leftovers wererefrigerated and processed within 2 weeks to ensure activity.²² Liposomalverteporfin was acidified using a 6M hydrochloric acid solution, and separationof organic (verteporfin) and aqueous (liposome) layers was achieved usingdichloromethane (CH₂Cl₂). After concentrating the solutionby means of evaporation, verteporfin was further purified by means of gravitychromatography on silica gel using an eluting solvent consisting of a 3:1ratio of CH₂Cl₂ to methanol. The verteporfin solutionwas evaporated to dryness and redissolved in dimethyl sulfoxide.

Previous studies linking photosensitizers to antibodies suggested thatderivatives of polyvinyl alcohol (PVA; molecular weight, 10 000-11 000Da; Sigma-Aldrich Corp; St Louis, Mo) provide suitable carriers for photosensitizerswithout jeopardizing the biological activity of the photosensitizer.²³ The procedure for loading verteporfin on PVA hasbeen described elsewhere.^{23 - 24} Briefly,PVA was modified with 2-fluoro-1-methyl pyridinium toluene-4-sulfonate (Sigma-AldrichCorp) and 1,6-hexanedimanine (Sigma-Aldrich Corp) to produce side chains containingterminal-free amino groups. Conjugation of modified PVA with verteporfin wasaffected by reacting it with a 25-fold molar excess of verteporfin in thepresence of carbodiimide as coupling agent in dimethyl sulfoxide. Carrierconjugates were analyzed by means of high-performance liquid chromatographyusing a column with ultrasphere consisting of 5-µm optical density,250 × 4.6 mm; a solvent system consisting of solution A (500 mL eachof 1% [NH4]₂SO₄ and CH₃CN and 50 mL of CH₃COOH) and solution B (500 mL of 1% [NH4]₂ SO₄ andC₄H₈O and 50 mL of CH₃COOH 50mL); a flowrate of 1.7 mL/min; and a gradient of solutions A-B of 60%:40% for 5 minutes,then a starting gradient flow from 40% solution B to 70% solution B in 20minutes and staying at 70% solution B for 5 minutes before returning to 40%solution B. Verteporfin-PVA eluted at 8 to 10 minutes, whereas unconjugatedverteporfin eluted at 18 to 19 minutes. The molecular weight of verteporfin-PVAwas determined by mass spectrometry and found to be approximately 28 kDa (verteporfin-PVAratio, approximately 28:1).

Before binding to the homing peptide, thiol groups were introduced toverteporfin-PVA using 3-mercaptopropionic acid (Sigma-Aldrich Corp) in acetatebuffer (pH, 5.5). Coupling of verteporfin-PVA to the targeting peptide (ATWLPPR;molecular weight, 840 Da synthesized to our specifications by Anaspec, SanDiego, Calif) was performed using sulfo-m-maleimidobenzoyl-N-hydroxysulfosuccinimide ester as a heterobifunctionalcross-linking reagent in carbonate buffer (pH, 8.5). Products were separatedby means of high-performance liquid chromatography with the solvent systemof solution A, 100% water, and 0.1% trifluoroacetic acid and solution B, 100%acetonitrile and 0.1% trifluoroacetic acid (gradient starting at solutionsA-B, 80%:20% and going to 80% solution B in 45 minutes). The molecular weightof VEGFR-2–targeted verteporfin was determined by mass spectrometryand found to be approximately 30 kDa. For the remainder of this report, thephotosensitizer dose will be expressed in verteporfin-equivalents (in milligramsdivided by the square of the body surface in square meters) as determinedby spectrofluorometry using a verteporfin calibration curve. Briefly, a calibrationcurve correlating concentration vs spectral emission of liposomal verteporfinwas constructed. Spectral emission of a sample of VEGFR-2–targeted verteporfinwas determined and used to extrapolate from the calibration curve its verteporfincontent.

All intermediates (free verteporfin, verteporfin-PVA, and VEGFR-2–targetedverteporfin) were found to have the same excitation and emission spectra asthe liposomal verteporfin formulation as determined by spectrofluorometryand to preserve an equivalent in vitro photosensitizing activity (tested inhuman umbilical vein endothelial cells) as determined by the tetrazolium saltMMT assay.²⁵

The rat laser-injury model of CNV was modified from earlier reportsand used in our laboratory for PDT.^{26 - 28} Adultmale pigmented rats (Brown-Norway; Charles River Laboratories, Wilmington,Mass) were used in the study, and all procedures were conducted in accordancewith the Association for Research in Vision and Ophthalmology Statement forthe Use of Animals in Research and the guidelines of the Massachusetts Eyeand Ear Infirmary (Boston) Animal Care Committee. The rats were anesthetizedfor all procedures with an intramuscular injection of 0.2 mL of a 50:50 mixtureof ketamine hydrochloride (20 mg/mL) and xylazine hydrochloride (100 mg/mL)(both from Phoenix Pharmaceutical Inc, St Joseph, Mo). For killing, a mixtureconsisting of pentobarbital sodium, 390 mg/mL; propylene glycol; ethanol;and water (Fatal Plus, 650 mg/kg; Vortech Pharmaceuticals, Dearborn, Mich)was given intraperitoneally.

The pupils were dilated with 5% phenylephrine hydrochloride and 0.8%tropicamide, and 4 to 6 photocoagulation lesions using a Coherent 920 argondye laser (100-µm spot size; 0.1-second duration; 630 nm and 120-160mW) (Coherent Medical Laser, Palo Alto, Calif) were delivered between theretinal vessels in a peripapillary distribution in each fundus using a slitlampdelivery system and a cover glass as a contact lens. Production of a bubbleat the time of laser confirmed the rupture of the Bruch membrane. The presenceof CNV was confirmed by fluorescein angiography using a TRC-50VT camera (Topcon,Paramus, NJ) with images captured on IMAGEnet for Windows system (Topcon)after an injection of 1 mL of 1% fluorescein sodium (Alcon, Fort Worth, Tex).A choroidal neovascular membrane was defined as present if early hyperfluorescencewith late leakage was present at the site of the inducing laser injury aspreviously described.²⁸

Verteporfin conjugates were tested in eyes with experimental CNV andin normal eyes. For targeted PDT, verteporfin-PVA was bound to the peptideATWLPPR. The untargeted photosensitizer conjugate, verteporfin-PVA, was chosenas a control because both molecules have comparable molecular weights (28and 30 kDa, respectively). The results with the targeted and control conjugateswere compared with previous results in the rat model with PDT using verteporfin(718 Da) in liposomal formulation, which is the therapy currently in clinicaluse.²⁸

Angiographies with VEGFR-2–targeted verteporfin and verteporfin-PVAwere performed using a standard fundus camera but with verteporfin-specificfilters, with the excitation spectral band centered at 580 nm and fluorescencedetection at 695 nm. Maximal gain settings were required. Photosensitizerdoses were given via tail-vein injection, and angiography was performed ata drug dose of 12 mg/m². Conversion to body surface area (in squaremeters) from weight (in kilograms) was made using a nomogram developed byGilpin.²⁹ Relative fluorescence intensitieswere determined by visual analysis of the angiograms.

Photodynamic therapy was performed on experimental CNV and areas ofnormal choroid and retina. Laser light of 689 nm was administered using adiode laser (Coherent Medical Laser, Palo Alto, Calif) delivered via a slitlampadapter (Laserlink; Coherent Medical Laser). Laser power at the focal planewas measured with a power meter (Coherent Fieldmaster; Coherent, Auburn, Calif).The laser spot size was set at 750 µm and was confirmed using a micrometer,and the irradiance used was 600 mW/cm², which was delivered for17, 42, or 83 seconds to achieve total energy doses of 10, 25, or 50 J/cm², respectively. Activating light fluences were based on previous dosimetryestablished in the rat model of CNV for verteporfin PDT.²⁸

Fluorescein angiograms were performed at 24 hours after treatment. Closureof CNV was defined by absence of leakage from CNV compared with the baselineangiogram as previously described. All angiograms were graded in masked fashion(as to dose and photosensitizer) by 2 experienced graders (E.S.G. and J.W.M.).

Eyes were enucleated and the lens and anterior segment were removed.The remaining eyecups were placed in a fixative containing 2.5% glutaraldehydeand 2% formaldehyde in 0.1M cacodylate buffer (pH, 7.4) at 4°C overnight.Tissue samples were then postfixed in 2% osmium tetroxide, dehydrated in agraded ethanol series, and embedded in epoxy resin. For light microscopy,1-µm sections were stained with 1% toluidine blue in 1% borate bufferand examined with a Zeiss photomicroscope (Axiophot, Oberkochen, Germany).For electron microscopy, sections were stained with a saturated aqueous uranylacetate solution and Sato lead stain. Sections were viewed with a transmissionelectron microscope (Philips CM 10; Royal Philips Electronics, Eindhoven,the Netherlands).

Grading of sections was performed in a masked fashion (as to dose andphotosensitizer) by an experienced grader (N.A.M.) using the following histologicalgrading scheme for PDT effects on normal choroid and retina modified fromKramer et al⁶ : grade 1 indicates damage inthe RPE and photoreceptors, with occasional pyknosis in the outer nuclearlayer (ONL), with or without choriocapillaris damage; grade 2, choriocapillarisclosure, RPE and photoreceptor damage, and 10% to 20% pyknosis in the ONL;grade 3, grade 2 with less than 50% pyknosis in the ONL; grade 4, grade 3with greater than 50% pyknosis in the ONL; and grade 5, grade 4 with damageto large choroidal vessels or retinal vessels or inner retinal layers.

To test the temporospatial localization of targeted and control verteporfinconjugates in the retinal and choroidal circulations and CNV, angiographywas performed using targeted verteporfin and verteporfin-PVA at a dose of12 mg/m². At the earliest time points captured (5-10 seconds) afterdrug injection, fluorescence of targeted and control verteporfin conjugateswas already noted in the choroidal and retinal circulation. Fluorescence intensitypeaked in the retinal circulation at approximately 30 to 60 seconds and thencleared from retinal circulation (20-21 minutes), followed by the choroidalcirculation (35-40 minutes). The drug was retained within the CNV with peakintensity of fluorescence seen at 1 hour (Figure 1). Verteporfin-PVA exhibited the same temporospatial localization.In contrast, liposomal verteporfin in CNV peaked at 15 to 20 minutes.²⁸

Since angiography demonstrated accumulation of targeted verteporfinand verteporfin-PVA in CNV at 1 hour after drug administration, this was thetime selected for light application for PDT. Photodynamic therapy was performedusing targeted verteporfin or verteporfin-PVA on 27 or 32 areas of CNV, respectively.Angiographic closure of CNV was assessed at 24 hours after PDT and was definedas an absence of leakage from the CNV compared with the pretreatment fluoresceinangiogram. The angiograms were graded with masking as to light, drug dose,and photosensitizer used. Twenty-four hours after PDT with verteporfin-PVA,occluded CNV typically showed a circle of hypofluorescence in the early framescorresponding to the treatment spot, with late leakage from injured RPE andchoriocapillaris usually originating at the rim of the treatment spot (Figure 2). Open CNV showed early hyperfluorescencein the area of CNV and late leakage. The same pattern was observed with PDTwith verteporfin or with targeted verteporfin irradiated at 15 to 20 minutes(optimal timing for verteporfin but suboptimal timing of treatment for targetedverteporfin). In contrast, CNV treated with PDT with targeted verteporfinand activating light applied at 1 hour showed no early hypofluorescence andonly rarely any late hyperfluorescence or leakage. In addition, occluded CNVdid not show any hyperfluorescence or late leakage in the area of CNV, consistentwith occlusion of CNV (Figure 3).

Effective CNV closure was demonstrated by fluorescein angiography withtargeted verteporfin and verteporfin-PVA at all tested photosensitizer doses. Table 1 and the histogram (Figure 4) summarize the effect of PDT on CNV, using different photosensitizerand light doses. Although the total number of treated CNVs was small, 100%closure of CNV was achieved using targeted verteporfin with drug doses aslow as 3 mg/m² and a light dose of 10 J/cm². Similarly,for verteporfin-PVA, 100% closure was achieved using 3 mg/m² anda light dose of 25 J/cm².

All lesions defined as closed, regardless of whether targeted verteporfinor verteporfin-PVA was used and regardless of the light energy dose, sharedsimilar histological features. Figure 5 showsa section of a CNV membrane treated with 4.5 mg/m² of targetedverteporfin and 10 J/cm² at 24 hours after PDT. This lesion wasgraded angiographically as closed. Vessels within the CNV showed vacuolizationof endothelial cells and occlusion with platelets, fibrin, and erythrocytes.Extravasated erythrocytes were noted, and macrophages were seen within andaround the treated CNV complex. Proliferating RPE cells can also be seen surroundingthe CNV complex. Gross disruption of the outer retina, RPE, and Bruch membranewas generally attributed to the laser injury inducing the CNV.

Treatment selectivity was investigated by performing PDT in normal retinaand choroid, using a qualitative assessment of angiographic findings afterPDT and the histological grading scheme previously described.⁶ Photodynamictherapy for normal retina and choroid using verteporfin-PVA at a dose of 2.5or 4.5 mg/m² and irradiation 1 hour after drug injection usinga fluence of 25 or 50 J/cm² gave similar results. Fluorescein angiography24 hours after PDT showed hypofluorescence in the area of treatment in theearly phase of the angiogram with late hyperfluorescence (Figure 6A and B). Histological examination revealed occlusion ofthe choriocapillaris with RPE necrosis and pyknosis of the ONL ranging fromoccasional to less than 10%, with mild vacuolization and disarray of the innerand outer segments (Figure 7). Theinner retinal layers and larger choroidal vessels, however, showed no damage,and the lesions were classified as grade 1 damage according to the publishedscheme.

Photodynamic therapy for normal retina and choroid using targeted verteporfinat a dose of 3 or 4.5 mg/m² and irradiation 1 hour after drug injectionusing a fluence of 10, 25, or 50 J/cm² showed even milder effectsthan the PDT using verteporfin-PVA. Fluorescein angiography performed 24 hoursafter PDT for normal retina and choroid using targeted verteporfin with lightapplied at 1 hour after photosensitizer injection showed no change in thearea of treatment (Figure 6C andD). Lesions were difficult to find by light microscopy, and marker lesionswere used to ensure localization. The retina appeared normal in almost allrespects, with all retinal layers appearing similar to control areas at alldoses of targeted verteporfin and all light fluences with negligible effecton RPE. A few pyknotic nuclei were seen in the ONL at all tested doses, alwaysless than 5%, and usually only a few per field (Figure 8). All lesions were classified as grade 1, but showed muchless damage than the typical grade 1 lesion because there was minimal effecton the RPE and virtually none on the photoreceptors. Closure of the choriocapillariswas the single consistent marker of PDT damage. Electron microscopy (Figure 9) confirmed the occlusion of thechoriocapillaris by platelets, leukocytes, erythrocytes, and occasionallyclumps of fibrin. The choriocapillaris endothelium was usually damaged andoften missing entirely. No extravasation of cells or fibrin was seen. ThisPDT effect contrasted with the condition of the RPE adjacent to the closedchoriocapillaris. No necrotic RPE cells were seen; most RPE cells had normal-appearingmitochondria and intact basal infoldings. Intracellular vacuoles were occasionallyseen. Rare pyknosis was observed in the ONL, but the inner segment mitochondriaappeared normal. The outer segments showed some disorganization but not muchvacuolization. No changes were seen in the cells and capillaries in the innerretina.

Photodynamic therapy with verteporfin as currently practiced has relativeselectivity for CNV based on specific treatment variables, including drug,light dose, and timing of light application.^{1 - 2 ,5 - 6 ,30} However,a wealth of experimental data in tumor animal models indicates that smallmolecules, which do not bind specifically to a tumor marker, discriminatepoorly between tumor and normal tissues in vivo.^{31 - 32} Similarly,preclinical studies have shown that PDT using untargeted verteporfin for CNVand normal retina and choroid causes some damage to normal structures. Theuse of homing vehicles, such as a recombinant monoclonal antibody or a targetingpeptide to VEGFR-2 with the ability to selectively target neovascular endothelium,should be useful in improving PDT outcomes and expanding its applications.

Our choice of a peptide as a homing vehicle was multifactorial and includedthe ease of synthesis and modification, the lack of tissue cross-reactivity,a minimized immunological reaction, a low cost of production, and, most important,the potential incorporation of multiple targeting peptides to the photosensitizer-carriercomplex. This last factor might allow one to target different molecular markersexpressed by the pathological tissue to achieve even higher levels of specificityby directing the cytotoxic agent through a distinct routing path to the desiredcell or subcellular compartment.^{33 - 36}

Angiography demonstrated peak localization of the targeted verteporfinand verteporfin-PVA to CNV by 1 hour after intravenous administration, withboth drugs clearing from the CNV by 2 hours. In contrast, liposomal verteporfinreaches peak intensity in rat CNV at 15 to 20 minutes after intravenous administrationand clears within 30 minutes.²⁸ Targeted verteporfinand verteporfin-PVA are relatively large molecules (28-30 kDa compared with718 Da for unbound verteporfin), and this probably accounts for the laterand more prolonged localization to CNV. To lentino and colleagues³⁷ demonstrated localization of fluoresceinated dextransand antibodies in experimental CNV after intravenous administration and acorrelation with molecular weight and radius. Presumably, molecules of thissize are able to exit the vascular space through fenestrations in the choriocapillarisand CNV, but are less readily cleared than are smaller molecules. Thus itis not surprising that PDT using both larger molecules, targeted verteporfinand verteporfin-PVA, showed greater efficacy for CNV closure with lower drugand light doses than those seen with PDT using unbound verteporfin. One hundredpercent closure was achieved with as little as 3 mg/m² and 10 J/cm² for targeted verteporfin and 25 J/cm² for verteporfin-PVA,although even with 3 or 6 mg/m² and 25 J/cm², the closurerate ranges from 83% to 92%.²⁸

Although the efficacy of CNV closure was similar for PDT using targetedverteporfin and verteporfin-PVA, the drugs differed somewhat in their selectivity.Angiography after PDT of normal retina and choroid using verteporfin-PVA demonstratedearly hypofluorescence and late leakage 24 hours after PDT, and results ofthe histological examination showed grade 1 lesions with RPE necrosis, mildpyknosis of the photoreceptor nuclei, and vacuolization and disarray of theinner and outer segments. Although this damage is still mild, it is no betterthan PDT with standard verteporfin. One can speculate that the verteporfin-PVAcan still leak through the CNV without binding to the endothelium, reachingthe extravascular space and clearing slowly. In contrast, PDT for normal retinaand choroid using targeted verteporfin did not show any angiographic hypofluorescenceor late leakage 24 hours after PDT, and results of the histological examinationshowed minimal effect on the RPE and no injury to photoreceptors. Althoughthese lesions were formally classified as grade 1 lesions, the observed damagewas substantially less. Presumably, the VEGFR-2–targeted verteporfinlocalizes to CNV on the basis of size in a manner similar to verteporfin-PVA,but can then bind VEGFR-2 receptors expressed on neovascular endothelium.This binding could lead to increased efficacy of CNV closure and increasedselectivity, because the photosensitizer would be sequestered at the endotheliumand would spare the RPE and photoreceptors.

An additional advantage to PDT using VEGFR-2–targeted verteporfinmight be the combination of an antiangiogenic effect with PDT. Previous studiesfrom our group^{9 - 10} have shownthat combining antiangiogenic agents with PDT causes a selective increasedcytotoxicity to neovascular endothelial cells in vitro and in vivo. The VEGFR-2–targetingpeptide used in the present study has been shown by Binetruy-Tournaire andcolleagues²¹ to completely inhibit VEGF-inducedangiogenesis. Thus, VEGFR-2–targeted verteporfin has the potential toexert an antiangiogenic potentiating effect before its photoactivation.

We showed that the efficacy of PDT can be enhanced by conjugating thephotosensitizer to PVA. In addition, selectivity of PDT can be enhanced bytargeting the photosensitizer to VEGFR-2. If these preliminary findings aresubstantiated in primate models of CNV, clinical studies may be warrantedto determine whether vision outcomes can be improved. Our results also highlightthe utility of designing peptide photosensitizer conjugates as vehicles forregulating the distribution of photosensitizer to CNV to maximize their selectivityin PDT. In the future, other candidate homing molecules may be identifiedwith even greater specificity for neovascular endothelium. These preliminaryresults indicate that targeted PDT for CNV is feasible and may offer a qualitativeimprovement in current clinical therapies.

Correspondence: Joan W. Miller, MD, Angiogenesis and Laser ResearchLaboratory, Retina Service, Massachusetts Eye and Ear Infirmary, 243 CharlesSt, Boston, MA 02114 (jwmiller@meei.harvard.edu).

Submitted for publication August 12, 2002; final revision received October23, 2003; accepted January 30, 2004.

This study was supported by the Foundation Fighting Blindness, OwingsMills, Md (Drs Renno, Terada, and Miller), and the Iacocca Foundation, Boston,Mass (Dr Renno).