Vitreous Levels of Vascular Endothelial Growth Factor and Stromal-DerivedFactor 1 in Patients With Diabetic Retinopathy and Cystoid Macular Edema Beforeand After Intraocular Injection of Triamcinolone

Background  Diffuse macular edema (DME) and/or aberrant neovascularization (NV)can cause vision loss in diabetic retinopathy (DR) and may be modulated bygrowth factors and chemokines. The chemokine stromal-derived factor 1 (SDF-1)is a potent stimulator of vascular endothelial growth factor (VEGF) expression,the main effector of NV, and the key inducer of vascular permeability associatedwith DME. Circulating endothelial cell precursors migrating in response toSDF-1 participate in NV.

Objective  To investigate the relationship between SDF-1 and (VEGF) in vitreousof patients with varying degrees of DR and DME before and after intraocularinjection of triamcinolone acetonide, used to treat refractory DME.

Methods  In this prospective study, 36 patients were included and observed for6 months. Vitreous VEGF and SDF-1 levels were measured by enzyme-linked immunosorbentassay in samples obtained immediately before and 1 month after injection oftriamcinolone.

Results  Both VEGF and SDF-1 were significantly higher (P<.01)in patients with proliferative DR than in patients with nonproliferative DR.Levels of SDF-1 were markedly increased in patients with DME compared withthose without DME. Vascular endothelial growth factor correlated with SDF-1levels and disease severity (r² = 0.88).

Conclusions  Triamcinolone administration resulted in dramatic reductions of VEGFand SDF-1 to nearly undetectable levels, eliminated DME, and caused regressionof active NV. Our results support a role for SDF-1 and VEGF in the pathogenesisof the adverse visual consequences of DR and suggest that the eliminationof DME with regression and/or initiation of fibrosis of NV after triamcinoloneinjection may be due to the suppression of VEGF and SDF-1.

Proliferative diabetic retinopathy (PDR), the primary cause of blindnessin the young to middle-aged population, affects many of the 7 million peoplewith diabetes in the United States.¹ The earlystage of the disease, termed nonproliferative diabetic retinopathy (NPDR),is associated with vascular permeability and can lead to vision-threateningdiffuse macular edema (DME) that is manifested as intraretinal and subretinalaccumulation of fluid. In the later stages, the nonperfused ischemic retinaproduces angiogenic growth factors, such as vascular endothelial growth factor(VEGF), which stimulate new abnormal blood vessel growth. For 3 decades, laserphotocoagulation has been the mainstay in the management of PDR.² However,laser treatment for PDR breaks down the blood retinal barrier (BRB) and cancause or worsen DME.^2,3

Thirty percent of patients with diabetes for 20 years or more have DME.¹ More than half of these patients will lose 2 or morelines of visual acuity (VA) after 2 years of follow-up evaluations.³ The Early Treatment Diabetic Retinopathy Study (ETDRS)demonstrated a significant benefit of focal laser photocoagulation for thetreatment of clinically significant DME,⁴ although24% of patients had DME that persisted at 36 months and severely affectedVA.^4,5 Surgical treatment is appropriatefor only a very small percentage of patients with DME and involves all theinherent risks, recovery time, and expense of surgery.⁶ In2 recent studies, intravitreal injection of triamcinolone acetonide reducedretinal thickening, improved BRB function, and improved VA in patients withDME.^7,8

The exact pathogenesis of DME has not been elucidated.^9,10 Severalmechanisms may be responsible, including altered BRB due to hemodynamic change,glycemic control, alterations in capillary basement membranes, and pericyteloss. The possibility that retinal-derived or blood-borne factors such asVEGF and chemokines play a role in DME should also be considered.¹¹

Chemokines participate extensively in mechanisms of leukocyte trafficking,immune surveillance, innate and adaptive immunity, and inflammation. Stromal-derivedfactor 1 (SDF-1), a member of the CXC chemokine subfamily, was initially identifiedas a bone marrow stromal cell–derived chemoattractant for hematopoieticprogenitor (CD34⁺) cells.¹² Stromal-derivedfactor 1 acts as an angiogenic agent in several model systems.¹³ TheSDF-1 receptor CXCR4 is expressed on endothelial cells, and its expressionis increased after treatment with VEGF or basic fibroblast growth factor.¹³ Animals deficient in either SDF-1 or CXCR4 have vasculardefects. Endothelial precursor cells participate in both normal and pathologicalangiogenesis, express functional CXCR4, migrate in response to SDF-1,¹⁴ and express VEGF in response to SDF-1.¹⁵

We postulated that patients with DME have activated leukocytes, microglia,Müller cells, and endothelial cells in their retina that produce cytokinesand growth factors resulting in elevated intravitreal levels of these factors.Because the expression of SDF-1 is increased during inflammation and becauseSDF-1 regulates cell trafficking and induces VEGF expression in multiple celltypes, it was particularly appealing as a candidate factor to induce DME andangiogenesis. Thus, we asked whether intravitreal levels of SDF-1 were elevatedin patients with DME and/or PDR, whether steroids could decrease SDF-1 andVEGF, and whether clinical improvement correlated with changes in vitreouslevels of these factors.

The institutional review board at the University of Florida approvedthe study protocol. Vitreous samples were obtained at the time of vitreousaspiration for treatment with triamcinolone in 48 eyes included from 33 diabeticindividuals with DME meeting the criteria of the ETDRS Report Number 12 (Table).¹⁴ Thirty-oneof the 33 patients were type 2 diabetics. Another 3 diabetic individuals (3eyes) with neovascularization of the iris (NVI) had ischemic maculopathy.Patients with NPDR and DME (21 eyes representing 15 individuals; 14 eyes classifiedas mild, 2 as moderate, 2 as severe, and 3 as very severe NPDR) had persistentDME for an average of 13 months, and 13 of the 21 eyes had undergone previousmacular laser therapy. Patients with early PDR and DME (7 eyes), active high-riskPDR and DME (13 eyes), and regressed high-risk PDR and DME (7 eyes) made upthe remaining groups. Vitreous samples from nondiabetic patients having vitrectomysurgery for macular pucker and epiretinal membrane were used as controls.Another group of patients included those with NVI (3 eyes). There were 15patients (30 eyes) who had both eyes treated. No eyes had ocular hypertension,previous periocular steroid injection, steroid drop use within 3 months, vitreoushemorrhage within 2 months, or previous vitrectomy surgery. All eyes wereobserved for 6 months or were excluded from the study.

Snellen VA was used because one clinic did not have ETDRS charts forthe study. Corrected VA was obtained on standard Snellen acuity charts andintraocular pressure measured at each monthly visit for the 6-month follow-up.Conversion of Snellen acuity to logarithm of the minimum angle of resolutionvalues was performed. All patients received intravenous fluorescein angiography(IVFA) prior to steroid injection and then monthly for the duration of thestudy. Baseline IVFA showed the amount of DME, evaluated ischemia, and measuredthe amount of diabetic retinopathy in the periphery. The postinjection IVFAwas evaluated in a subjective, unmasked manner. In 3 diabetic individuals,optical coherence tomography (OCT) was performed before and 1 month aftertriamcinolone injection to evaluate macular thickness.

All patients had complete physical examinations, laboratory tests, bloodpressure, hemoglobin A_1c, and urinalysis. Details of the medicationbeing used were compared between groups, and no significant differences wereobserved regarding use of 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA)reductase inhibitors, angiotensin-converting enzyme inhibitors, calcium channelblockers, and oral hypoglycemics, including thiazolidinediones.

Intravitreal injection of triamcinolone acetonide (Kenalog 40; Apothecon,Princeton, NJ) was offered to treat DME with NPDR, DME with PDR, and NVI.Topical proparacaine was applied to the ocular surface followed by 4% lidocaineat the injection site 3.5 mm from the limbus. Povidone iodine (5%) was appliedto the ocular surface for 2 minutes. Liquid vitreous (0.2 mL) was removedusing a 23-gauge needle, followed by injection of 4 mg (0.1 mL) of triamcinolonethrough the pars plana using a 30-gauge needle. In all patients, repeat intravitrealtaps were performed after 1 month. The vitreous aspirates that were collectedat the time of triamcinolone injection were frozen at –80°C untilanalysis.

Stromal-derived factor 1 and VEGF in the vitreous samples were quantifiedusing commercially available enzyme-linked immunosorbent assays (ELISAs) (R&DSystems, Minneapolis, Minn). The SDF-1 ELISA has a detection limit of approximately160 pg/mL, whereas the VEGF ELISA has a sensitivity of approximately 15 pg/mL.The mean concentrations of SDF-1 and VEGF in the vitreous from diabetic patientswith DME were determined and compared with those in normal control patients.

The functional activity of SDF-1 in selected vitreous samples was examinedby chemotaxis of human retinal endothelial cells using a modified blind-wellBoyden chamber (Neuroprobe, Gaithersburg, Md) assay as previously described.^16-18 Human retinal endothelialcells were isolated and cultured as previously described.¹⁹ Cellsfrom 2 separate donors were used. Vitreous samples from 3 patients showinghigh levels of both VEGF (approximately 22 to 220 μg/mL) and SDF-1 (approximately6.5 to 11.8 ng/mL) were used in triplicate to stimulate cell migration acrossa collagen-coated, polyvinyl- and pyrrolidine-free polycarbonate membrane(Neuroprobe). Serum-free medium was used as a negative control, while completegrowth medium for human retinal endothelial cells, containing numerous chemotacticfactors (in the form of 10% vol/vol fetal bovine serum) as well as fibroblastgrowth factors (as part of the normal growth medium), was used as positivecontrol. Selected cells were preincubated with AMD3100, a specific CXCR4 antagonist,prior to their being exposed to the chemotactic stimuli. The cells were given12 hours to migrate through the membrane, after which nonmigrating cells wereremoved by scraping, and the remaining cells were fixed and stained with DiffQuick(Fisher Scientific, Atlanta, Ga). Individuals masked to the identity of treatmentcounted migrating cells in at least 3 high-power fields for each well.

Statistical analysis was carried out using χ² analysisand rank analysis of data.

Intravitreal administration of triamcinolone typically appeared as awhitish suspension in the inferior portion of the vitreous cavity (Figure 1). Crystals settled preretinally in thevitreous cortex but did not interfere with vision. As shown in Figure 2 for a patient with NPDR and DME, IVFA 1 month after triamcinoloneinjection showed restoration of the BRB manifested as lack of hyperfluorescenceand DME. Postinjection IVFA showed less fluorescein leakage than on preinjectionangiograms in all 48 eyes (100%). At 3 months, 62% of eyes exhibited decreasedhyperfluorescence and DME on IVFA. At 6 months, this value was 38%.

In the group with DME and NPDR, 3 patients having OCT showed a mean ± SDbaseline central macular thickness of 487 ± 67 μm forthe eyes measured. At 1 month, there was a mean ± SD reductionof 50% to 238 ± 24 μm. Figure3 shows the OCT for the same patient shown in Figure 2. Figure 4 depictsthe IVFA of a patient with active high-risk PDR showing venous segmentation,venous loops, and hyperfluorescence from aberrant neovascularization (NV)on the retina. One month after triamcinolone injection, the IVFA shows a markedreduction of venous segmentation and loops as well as a marked decrease inhyperfluorescence from NV.

No additional macular laser was done in any group during the follow-up.All 23 eyes with active high-risk PDR and/or NVI received additional peripherallaser a mean of 3.2 months after triamcinolone injection. The 3 eyes withNVI had media opacities preventing complete laser. After injection of triamcinolone,the hyphema or vitreous haze cleared, allowing complete laser 1 month afterthe second vitreous sample was obtained.

The VA of the 48 eyes with DME improved by 1.65 lines at 1 month, 1.25lines at 3 months, and 1 line at 6 months. The regressed high-risk PDR groupdemonstrated the largest improvement in VA of 2.3 lines, 1.8 lines, and 1.6lines at 1, 3, and 6 months, respectively, after triamcinolone. The improvementin lines of VA decreased inversely to the degree of PDR. Eyes with early PDRhad 1.3 lines of improvement at 1 month, 0.9 lines at 3 months, and 0.7 linesat 6 months. Eyes with active high-risk PDR had 0.9 lines, 0.7 lines, and0.5 lines of improvement in VA at 1, 3, and 6 months, respectively. Mean VAwas unchanged in the group with NVI and macular ischemia. Measurement of therepair of the blood-ocular barrier by IVFA for all 4 groups together occurredin 100% of patients at 1 month, in 62% at 3 months, and in 38% at 6 months,with no significant difference between groups (P = .02).

During the study period, intraocular pressure was higher than 21 mmHg in 16 (33%) of the 51 eyes and was normalized with topical antiglaucomatousmedication. After 6 months of follow-up, 12 (23.5%) of the 51 eyes were stillreceiving topical antiglaucomatous medication. Three (6%) of the 51 eyes developedglaucomatous damage to the optic nerve. One eye required glaucoma surgeryto control pressure.

One patient with active high-risk PDR experienced a traction/rhegmatogenousretinal detachment 4 months after triamcinolone injection and 2 months afterextensive peripheral laser treatment. Subsequent retinal reattachment surgerywas successful with ultimate improvement of visual acuity to 20/30 at 6 months.None of the eyes developed endophthalmitis.

Systemic factors such as duration of diabetes, hypertension, and hemoglobinA1c were documented for each group. A trend of longer duration of diabetes,hypertension, and higher hemoglobin A _1c was found in the eyeswith PDR or NVI, compared with those with NPDR; however, these differenceswere not statistically significant (P = .02).

The vitreous concentration of SDF-1 was increased in diabetic subjectswith DME (Figure 5A), especially inactive high-risk PDR, where VEGF levels were markedly elevated (Figure 5B) and SDF-1 uniformly decreased after steroids in all groupswith DME (Figure 5A). Intravitreal levelsof VEGF increased with the severity of the disease. Vascular endothelial growthfactor levels uniformly decreased after steroid treatment in all groups (Figure 5B). Vascular endothelial growth factorlevels correlated significantly with SDF-1 levels pretreatment (r² = 0.88) (Figure5C) but not posttreatment (r² = 0.20)(Figure 5C). Patients with NVI respondedwith resolution of their NV after treatment with triamcinolone. All patientshad a reduction of DME and resolution of macular cystic changes upon subjectiveclinical examination at 1 month. Pretreatment and posttreatment OCT was performedon 3 of the patients. Patients with active high-risk PDR also had regressionof angiogenesis. In control subjects, SDF-1 and VEGF levels were below thelimit of detection.

Vitreous from patients with high SDF-1 levels as shown by ELISA alsodemonstrated functional activity in the chemotaxis assay with human retinalendothelial cells (Figure 6). All 3vitreous samples tested induced significant chemotaxis in retinal endothelialcells, at least as effective as the positive control. Preincubation of thecells with the selective CXCR4 antagonist AMD3100 resulted in significantreduction of chemotaxis to levels nearly comparable with unstimulated (negativecontrol) cells.

In this study, triamcinolone injection reduced DME at 1 month and causedregression of preretinal NV and NVI. Intraocular triamcinolone injection inpatients with active high-risk PDR and DME allowed peripheral laser to beused without worsening DME. Eyes with NVI and media opacities had subsequentclearing of vitreous haze, allowing extensive peripheral laser treatment andpreventing neovascular glaucoma. The clinical response seen in this studyis similar to that previously reported.^7,8 Previously,vitreous VEGF levels were found to be elevated in patients with DME. Unlikethe authors of any study to date, however, we measured changes in VEGF andSDF-1 prior to and following intervention with triamcinolone and correlatedthe clinical improvement with subsequent changes in levels of these factors.Thus, this study is the first to demonstrate an association between the therapeuticbenefit of triamcinolone and dramatic reductions in vitreous VEGF and SDF-1levels.

Triamcinolone is an intermediate-acting corticosteroid suspension, traditionallyused as a periocular injection to treat ocular inflammatory diseases suchas cystoid macular edema, scleritis, and uveitis.^20,21 Intravitrealinjection of triamcinolone has been used for regression of iris, preretinal,and choroidal NV^22,23 and to reducethe incidence of proliferative vitreoretinopathy after complicated posteriorsegment surgical procedures.²⁴ Triamcinolonehas the physical advantage of being less soluble acting as a local depot.The adverse effects of ocular corticosteroids are well known, including intraocularpressure elevation,²⁵ cataractogenesis,²⁰ and potential cytotoxicity on ocular structures suchas photoreceptors and retinal pigment epithelial cells.²⁶ Triamcinoloneis believed to be safe and nontoxic at levels up to about 1 mg/mL when administeredintraocularly²⁷ or in in vitro conditions.Intraocular crystalline triamcinolone (25 mg) has been used to treat DME withoutretinal toxicity.²⁸ Other trials have shownsimilar improvement of DME caused by diabetes with implanted sustained-releasesteroid vehicles.^29,30

Further investigations are needed to clarify the ocular interactionbetween SDF-1 and VEGF, as well as the role of SDF-1 in the pathogenesis ofDME. The simplest explanation for our results is that SDF-1 and VEGF participatein the aberrant NV observed in patients with diabetic retinopathy and DME.This hypothesis is supported by published data on the requirement of SDF-1in vascular development,^31-34 theangiogenic potential of SDF-1 in several adult model systems,^35-37 andfunctional interactions between SDF-1 and VEGF.^15,38,39 Weshowed that vitreous samples containing high SDF-1 levels induced migrationin cultured human retinal endothelial cells. Furthermore, the migration-inducingactivity of those same samples was almost completely abrogated by a selectiveinhibitor of the SDF-1 receptor, despite the presence of high concentrationsof VEGF. Endothelial cells, including those from developing brain and heart,and endothelial precursor cells express CXCR4 and migrate in response to SDF-1.¹⁴ In an in vivo animal model of ischemic NV, SDF-1augmented endothelial precursor cell recruitment.¹⁴ Increasedhoming of CD34⁺ progenitor cells to the liver after a stressfulinsult is also dependent on SDF-1.⁴⁰

Stromal-derived factor 1 stimulates VEGF expression in endothelial cells.³⁸ Furthermore, VEGF has been shown to up-regulate CXCR4on cultured endothelial cells.³⁹ Thus, a synergisticinteraction between these factors is likely occurring within the diabeticeyes, with SDF-1 promoting VEGF production and VEGF in turn enhancing theresponsiveness of the endothelial cells to SDF-1 through increased surfaceexpression of CXCR4. We are the first to demonstrate a link between reductionof vitreous levels of VEGF and SDF-1 and clinical improvement of DME, as wellas regression of retinal NV and NVI after injection with triamcinolone. Themechanism by which decreased levels of VEGF and SDF-1 may mediate this clinicalimprovement is currently under investigation, but possible mechanisms includealtering the BRB integrity, altering progenitor cell behavior, or affectingleukocytes and macrophage trafficking and modulation of tight junction andcell adhesion proteins. There is no published information available concerningthe regulation of SDF-1 expression by corticosteroids. However, these potentanti-inflammatory agents are known to inhibit the production of other chemokinesby both transcriptional and posttranscriptional mechanism.^41-43

The coincident reduction of intravitreal levels of both SDF-1 and VEGFand the clinical improvement by triamcinolone indicates that these moleculesmay contribute to the maintenance of the phenotype. The development of agentstargeting SDF-1 and VEGF, and their respective receptors, may offer alternativetherapeutic approaches to treating visual defects associated with diabeticretinopathy.

Correspondence: Maria B. Grant, MD, Departmentof Pharmacology and Therapeutics, University of Florida, Box 100267, Gainesville,FL 32610-0267 (grantma@pharmacology.ufl.edu).

Submitted for Publication: November 21, 2003;final revision received May 17, 2004; accepted May 26, 2004.

Financial Disclosure: None.

Funding/Support: This study was supported bygrants EY-012601 and EY-007739 from the National Institutes of Health, Bethesda,Md, and grant 4-2000-847 from the Juvenile Diabetes Foundation, New York,NY.

Previous Presentation: A preliminary reporton this research was presented at the Annual Meeting of the Association forResearch in Vision and Ophthalmology; May 8, 2003; Fort Lauderdale, Fla.