Coexpression of VEGF Receptors VEGF-R2 and Neuropilin-1 in Proliferative Diabetic Retinopathy

Abstract

purpose. To elucidate vascular endothelial growth factor (VEGF)-mediated pathogenesis of fibrovascular proliferation in diabetic retinopathy.

methods. Fibrovascular tissues were obtained at vitrectomy from 22 cases with proliferative diabetic retinopathy. The half-divided tissues were processed for reverse transcription–polymerase chain reaction (RT–PCR) analysis to examine the expression of VEGF isoforms and their receptors. Paraffin sections of the other half were used for immunohistochemistry for CD34, glial fibrillary acidic protein and VEGF, and in situ hybridization for VEGF.

results. RT–PCR analysis demonstrated the expression of VEGF receptors VEGF-R1, VEGF-R2, and neuropilin-1 in 12, 14, and 14 of 22 cases, respectively. Notably, VEGF-R2 and neuropilin-1 were simultaneously expressed in the identical 14 tissues. The isoform VEGF₁₂₁ was constitutively expressed in all the tissues examined, whereas the expression of VEGF₁₆₅ was confined to the 7 tissues that also expressed VEGF-R2 and neuropilin-1. The vascular density of fibrovascular tissues evaluated by immunohistochemistry for CD34 was significantly higher in the cases with the expression of VEGF-R2 and neuropilin-1 than in those without their expression (P < 0.01), whereas VEGF-R1 expression had no such relationship with the vascular density. The fibrovascular tissues that expressed VEGF₁₆₅ together with VEGF-R2 and neuropilin-1 were found in significantly younger patients (P < 0.01). In situ hybridization and immunohistochemical studies demonstrated that glial cells in the fibrovascular tissues express and produce VEGF.

conclusions. Coexpression of VEGF-R2 and neuropilin-1 is suggested to facilitate fibrovascular proliferation in diabetic retinopathy.

Diabetic retinopathy is a major cause of adult blindness when it progresses to the stage of proliferative diabetic retinopathy characterized by fibrovascular proliferation. However, molecular mechanisms of angiogenesis in proliferative diabetic retinopathy are not fully understood, although several growth factors including vascular endothelial growth factor (VEGF) have been suggested to be involved. ^{1 2 3 4 5 6}  

VEGF is an angiogenic mitogen secreted from various types of cells ^{7 8} and plays a major role in the angiogenesis under physiologic and pathologic conditions such as normal embryonic development, ^{9 10 11} wound healing, ¹² and solid tumor growth. ^{13 14} VEGF has five major isoforms generated by alternative splicing from a single gene (i.e., VEGF₁₂₁, VEGF₁₄₅, VEGF₁₆₅, VEGF₁₈₉, and VEGF₂₀₆). Two high-affinity tyrosine–kinase receptors for VEGF (i.e., fms-like tyrosine kinase-1, Flt-1, ¹⁵ and kinase insert domain–containing receptor, KDR ^{16 17} ) have been cloned and characterized. Flt-1 and KDR are now known as VEGF-R1 and VEGF-R2, respectively. The recent study has reported an additional receptor for VEGF (i.e., neuropilin-1 ¹⁸ ), which was originally recognized as a molecule involved in axonal guidance in the developing nervous tissue. ¹⁹ Increasing evidence indicates that the signal transduction mediated by VEGF-R2 is different from that by VEGF-R1 ^{20 21 22} and the bioactivities of VEGF differ between the isoforms. ^{23 24} Neuropilin-1 has been shown to enhance the bioactivity of VEGF₁₆₅, but not that of VEGF₁₂₁, by increasing the binding affinity of VEGF₁₆₅ to VEGF-R2. ¹⁸ Thus, neuropilin-1 is considered to function as a VEGF₁₆₅-specific receptor when coexpressed with VEGF-R2. 

Involvement of VEGF in proliferative diabetic retinopathy has been suggested by previous findings that the vitreous fluid from eyes with proliferative diabetic retinopathy contains a large amount of VEGF, ^{1 2} which is produced by the ischemic retinal cells, ^{4 5} and that the fibrovascular proliferative tissue expresses VEGF. ³ However, VEGF receptors involved in the fibrovascular proliferation have not been examined previously, and the relationship between the histopathology of the fibrovascular tissue and the expression pattern of VEGF isoforms and their receptors is a matter of concern. In addition, the cells in the fibrovascular tissue responsible for VEGF production remain unknown. 

In the present study, we examined the expression of VEGF receptors VEGF-R1, VEGF-R2, and neuropilin-1, together with that of VEGF isoforms, in the fibrovascular tissues excised at vitrectomy and also analyzed the histopathology of the tissues. We demonstrate for the first time that coexpression of VEGF-R2 and neuropilin-1 correlates with the degree of angiogenesis in proliferative diabetic retinopathy and that glial cells in the fibrovascular tissue express and produce VEGF. 

Methods

Fibrovascular Tissues

Fibrovascular tissues were obtained from 22 eyes of 22 patients with proliferative diabetic retinopathy during vitrectomy performed at Keio University Hospital between July 1998 and April 1999. Operative indication included persistent vitreous hemorrhage and traction retinal detachment involving or threatening the macula. All the patients gave their informed consent to our study, which followed the tenets of the Declaration of Helsinki. Clinical information of the present cases is shown in Table 1 . For histopathologic evaluation, half-divided surgical specimen of each case was fixed in 4% paraformaldehyde at 4°C immediately after removal, and embedded in paraffin for light microscopic analyses. The other half was processed for RNA extraction for reverse transcription–polymerase chain reaction (RT–PCR) analysis. 

Immunohistochemistry

Serial paraffin sections, 4-μm-thick, were mounted on 3-aminopropyl triethoxysilane–coated glasses, and were prepared for immunohistochemistry for CD34, glial fibrillary acidic protein (GFAP), and VEGF. To determine the cell types that express VEGF, serial sections were subjected to in situ hybridization for VEGF and immunohistochemistry for GFAP. For immunohistochemistry, the sections were treated with 0.3% hydrogen peroxide in methanol for 30 minutes at room temperature to block endogenous peroxidase activity. The sections for the staining of GFAP and VEGF were also treated with 0.1% trypsin for 30 minutes at 37°C. After blocking nonspecific binding with either 5% normal goat serum for CD34 and VEGF or 5% normal swine serum for GFAP, they were incubated with either mouse monoclonal antibody against CD34 (1/100 dilution; Novocastra Laboratories, Newcastle, UK), rabbit polyclonal antibodies against GFAP (1/200 dilution; DAKO A/S, Glostrup, Denmark), or rabbit polyclonal antibodies against human VEGF (1/50 dilution; Santa Cruz Biotechnology, Santa Cruz, CA) at room temperature for 60 minutes. Subsequently, the specimens were incubated at room temperature for 30 minutes with either goat antibodies against mouse immunoglobulins conjugated to peroxidase-labeled dextran polymer (no dilution; En Vision+ Mouse; DAKO, Carpinteria, CA) for CD34, peroxidase-conjugated swine antibodies against rabbit immunoglobulins (1/100 dilution; DAKO) for GFAP, or goat antibodies against rabbit immunoglobulins conjugated to peroxidase-labeled dextran polymer (no dilution; En Vision+ Rabbit; DAKO) for VEGF. Color was developed with 3,3′-diaminobenzidine tetrahydrochloride (0.2 mg/ml; Dojindo Laboratories, Kumamoto, Japan) in 0.05 M Tris–HCl (pH 7.6) containing 0.003% hydrogen peroxide, and the sections were counterstained with methyl green (Merck, Darmstadt, Germany). As a control, sections were reacted by replacing the first antibodies with nonimmune immunoglobulins (DAKO) or the VEGF antibodies preincubated with the blocking peptide for VEGF (Santa Cruz Biotechnology) according to the manufacturer’s instructions. 

Morphometric Analysis

The degree of angiogenesis was evaluated by the morphometric analysis of the tissue sections immunostained for CD34. In each section, the entire tissue was photographed to measure the whole area of the tissue, and the number of CD34-immunoreactive vessels with a distinct lumen was counted. Two of the authors independently counted the number of vessels without any clinical information on each patient, and the average number of vessels per square millimeter was calculated as vascular density (vessels per square millimeter) of the section. In each case, two sections 80- to 100-μm apart were subjected to calculation of vascular density, and the mean of these two values was used as the final vascular density for statistical analyses (Mann–Whitney U test, Spearman rank correlation, Kruskal–Wallis test, and Dunn procedure). 

RT–PCR and Sequencing Analysis

Total cellular RNA was prepared using ISOGEN (NIPPON GENE, Toyama, Japan) according to the manufacturer’s protocol. In brief, surgical materials were homogenized in 1 ml of ISOGEN, and 200 μl of chloroform was added. After centrifugation at 4°C, the aqueous phase was collected, and total RNA was precipitated with an equal volume of isopropanol. RNA was then dissolved in 10 μl of water treated with diethyl pyrocarbonate. 

We calculated the relative amount of RNA in each case by quantifying the amplified β-actin cDNA fragment because total amount of RNA extracted in each case was below the limit of the ordinary measurement with a UV photometer due to the minute size of the tissue. Two μl of the solution containing total RNA was reverse-transcribed with a First-Strand cDNA Synthesis Kit (Pharmacia Biotech, Uppsala, Sweden) at 37°C for 1 hour in a 15-μl reaction volume containing random hexadeoxynucleotide primer and Moloney Murine Leukemia Virus reverse transcriptase. A 2-μl aliquot of the reaction product was subjected to 30 cycles of PCR for amplification of β-actin cDNA. Density of the band of amplified β-actin cDNA was measured in each case, and the relative amount of total RNA extracted from each tissue was calculated. Based on the above results, we adjusted the starting amount of RNA for further RT–PCR analysis on the expression of VEGF, VEGF-R1, VEGF-R2, neuropilin-1, and β-actin. RNA was reverse-transcribed as described above, and PCR was performed at 30 cycles in a 50-μl reaction volume containing 800 nM of each primer, 100 μM of dNTP, and 5 U Taq DNA polymerase (TOYOBO, Tokyo, Japan) in a thermal controller (Mini Cycler; MJ Research, Watertown, MA). The thermal cycle was 1 minute at 94°C; 2 minutes at either 64°C (VEGF), 64°C (VEGF-R1), 63°C (VEGF-R2), 63°C (neuropilin-1), or 67°C (β-actin); and 3 minutes at 72°C, followed by 3 minutes at 72°C. The nucleotide sequences of the PCR primers were 5′-TGC CTT GCT GCT CTA CCT CC-3′ (forward, on exon 1) and 5′-TCA CCG CCT CGG CTT GTC AC-3′ (reverse, on exon 8) for VEGF; 5′-GAT GTT GAG GAA GAG GAG GAT T-3′ (forward) and 5′-AAG CTA GTT TCC TGG GGG TAT A-3′ (reverse) for VEGF-R1; 5′-GAT GTG GTT CTG AGT CCG TCT-3′ (forward) and 5′-CAT GGC TCT GCT TCT CCT TTG-3′ (reverse) for VEGF-R2; 5′-CAA CGA TAA ATG TGG CGA TAC T-3′ (forward) and 5′-TAT ACT GGG AAG AAG CTG TGA T-3′ (reverse) for neuropilin-1; 5′-TGA CGG GGT CAC CCA CAC TGT GCC CAT CTA-3′ (forward) and 5′-CTA GAA GCA TTT GCG GTG GAC GAT GGA GGG-3′ (reverse) for β-actin. The above RT–PCR analysis enabled us to discriminate each isoform of VEGF by the difference in size of each amplified DNA fragment. The expected sizes of the amplified cDNA fragments of VEGF₁₂₁, VEGF₁₄₅, VEGF₁₆₅, VEGF₁₈₉, VEGF₂₀₆, VEGF-R1, VEGF-R2, neuropilin-1, andβ -actin were 0.41, 0.48, 0.54, 0.61, 0.66, 1.1, 0.56, 0.82, and 0.66 kb, respectively. An aliquot of the PCR product was electrophoresed in a 1.5% agarose gel and stained with ethidium bromide. 

To confirm the specific amplification from the target mRNAs, the RT–PCR products were subcloned into the pBluescript KS vector (Stratagene, La Jolla, CA) and were analyzed by sequencing with fluorescent T7 primer (Amersham Pharmacia Biotech, Buckinghamshire, UK) using a Thermo Sequenase fluorescence-labeled primer cycle sequencing kit (Amersham Pharmacia Biotech) and ALF DNA sequencer II (Amersham Pharmacia Biotech). 

In Situ Hybridization

Serial paraffin sections, 4-μm-thick, were treated with proteinase K (5 μg/ml; Sigma Chemical, St. Louis, MO) in 10 mM Tris–HCl (pH 8.0), 1 mM EDTA at 37°C for 30 minutes, and post-fixed in 4% paraformaldehyde at room temperature for 10 minutes. They were then rinsed in 0.1 M phosphate buffer, incubated in 0.2 M HCl at room temperature for 10 minutes, and acetylated with 0.25% acetic anhydride in 0.1 M triethanolamine–HCl (pH 8.0) at room temperature for 10 minutes. After being washed in 0.1 M phosphate buffer, they were dehydrated in ethanol and air-dried. 

Single-strand sense and antisense digoxigenin-labeled RNA probes were generated by in vitro transcription of the cDNA with T3 or T7 RNA polymerase using the DIG RNA Labeling Kit (Boehringer Mannheim, Mannheim, Germany) following the protocol from the manufacturer. Template DNA was a 517-bp cDNA encoding human VEGF₁₂₁, which was cloned in pBluescript KS vector. This cDNA clone was kindly provided by Herbert A. Weich (Gesellschaft für Biotechnologische Forschung, Braunschweig, Germany). 

Hybridization with the digoxigenin-labeled RNA probes was performed at 50°C for 16 hours in 40 μl of buffer containing 50% formamide, 10 mM Tris–HCl (pH 7.6), 0.2 μg/μl tRNA, 1× Denhardt’s solution (0.02% Ficoll, 0.02% bovine serum albumin, 0.02% polyvinylpyrrolidone), 10% dextran sulfate, 600 mM NaCl, 0.25% sodium dodecyl sulfate (SDS) and 1 mM EDTA. After hybridization, the sections were washed in a buffer containing 50% formamide and 2× SSC at 50°C for 30 minutes, followed by digestion with ribonuclease A (10 μg/ml; Sigma Chemical) at 37°C for 30 minutes. After being washed in 2× SSC at 50°C for 20 minutes and twice in 0.2× SSC at 50°C for 20 minutes, they were treated with 0.3% hydrogen peroxide and 0.1% sodium azide in distilled water for 30 minutes at room temperature to block endogenous peroxidase activity. After blocking nonspecific binding with 10% normal horse serum, they were incubated with mouse anti-digoxigenin antibody (1/750 dilution; Boehringer Mannheim) at room temperature for 90 minutes, then incubated with biotinylated horse antibodies against mouse immunoglobulins (1/200 dilution; Vector Laboratories, Burlingame, CA) for 30 minutes, and, finally, reacted with a solution containing the complex of avidin (1/100 dilution; DAKO A/S) and biotinylated horseradish peroxidase (1/100 dilution; DAKO A/S) for 30 minutes. Color was developed with 3,3′-diaminobenzidine tetrahydrochloride (0.2 mg/ml) in 0.05 M Tris–HCl (pH 7.6) containing 0.003% hydrogen peroxide, and the sections were counterstained with methyl green. 

Results

Vascular Density of Fibrovascular Tissues

The degree of angiogenesis in fibrovascular tissues was different among the patients (Fig. 1) , and, thus, the vascular density was measured by morphometric analysis of the sections immunostained for CD34. Vascular structures were clearly detected in the immunostained sections as shown in Figure 1 , and the vascular density of the tissue was readily evaluated. The vascular density of each case is shown in Table 1 . 

Expression of VEGF Isoforms and Their Receptors

Expression of VEGF isoforms and their receptors VEGF-R1, VEGF-R2, and neuropilin-1 in the fibrovascular tissue was examined by RT–PCR. As shown in Figure 2 , VEGF₁₂₁ of 0.41 kb was constitutively expressed in all the tissues examined, whereas the expression of VEGF₁₆₅ of 0.54 kb was detected in 7 of 22 patients (32%). On the other hand, VEGF₁₄₅, VEGF₁₈₉, and VEGF₂₀₆ were not expressed in any except one patient (patient 2), who expressed VEGF₁₈₉ of 0.61 kb. As for the receptors, VEGF-R1, VEGF-R2, and neuropilin-1 expression was observed in 12 of 22 (55%) patients, 14 of 22 (64%) patients, and 14 of 22 (64%) patients, respectively. It is notable that VEGF-R2 and neuropilin-1 were simultaneously expressed in the same 14 patients and that all 7 patients with VEGF₁₆₅ expression also expressed VEGF-R2 and neuropilin-1. The specific amplification from the target mRNA was confirmed by sequencing the amplified DNA fragments (data not shown). 

Relationship of the Expression of VEGF Isoforms and Their Receptors with the Vascular Density

Although there was no significant difference in vascular density between the fibrovascular tissues with VEGF-R1 expression (87.4 ± 62.4 vessels/mm², mean ± SD) and those without its expression (84.3 ± 73.8 vessels/mm²; P = 0.74, Fig. 3A ), the vascular density of the tissues with the expression of both VEGF-R2 and neuropilin-1 (114.8 ± 61.8 vessels/mm²) was significantly higher than those without their expression (35.6 ± 38.3 vessels/mm²; P < 0.01, Fig. 3B ). Because the vascular density was higher in younger patients, showing a reverse correlation with age (P < 0.01, Fig. 3C ), the statistical analysis with age restriction was performed to exclude the confounding effect of age on vascular density. The age of 14 patients with the coexpression of VEGF-R2 and neuropilin-1 and that of 8 patients without the coexpression ranged from 28 to 66 (47.2 ± 12.0, mean ± SD) and 49 to 76 (61.5 ± 7.6) years, respectively, and showed a statistically significant difference (P < 0.01). The age was restricted from 49 years (the lowest age of the 8 patients without the coexpression) to 66 years (the highest age of the 14 patients with the coexpression). Each of the age-restricted groups with or without the coexpression consisted of 7 cases, the age of which ranged from 49 to 66 (56.7 ± 5.9) or 49 to 66 (59.4 ± 5.2) years, respectively. Comparison between these two age-restricted groups demonstrated that the VEGF-R2 and neuropilin-1 coexpression itself made a statistically significant difference in vascular density (101.2 ± 62.5 versus 30.1 ± 37.9 vessels/mm²; P < 0.01, Fig. 3D ). 

We also evaluated the significance of the expression of VEGF₁₆₅, the bioactivity of which is enhanced by the coexpression of VEGF-R2 and neuropilin-1. ¹¹ Because the 7 patients with VEGF₁₆₅ expression also expressed VEGF-R2 and neuropilin-1, the 22 patients could be classified into three groups as follows: Group I, 7 patients who expressed not only VEGF-R2 and neuropilin-1 but also VEGF₁₆₅, VEGF₁₆₅ (+), VEGF-R2 (+), neuropilin-1 (+); group II, 7 patients who expressed VEGF-R2 and neuropilin-1 but not VEGF₁₆₅, VEGF₁₆₅ (−), VEGF-R2 (+), or neuropilin-1 (+); and group III, 8 patients who expressed neither VEGF-R2, neuropilin-1, nor VEGF₁₆₅, VEGF₁₆₅ (−), VEGF-R2 (−), neuropilin-1 (−). Table 2 shows the mean ± SD of vascular density and clinical factors in these groups. As shown in Figure 4A , the vascular density was not different between group I and group II (120.8 ± 73.4 versus 108.8 ± 53.0 vessels/mm²) and was significantly higher in both groups I and II than in group III (35.6 ± 38.3 vessels/mm²; P < 0.05, P < 0.05). 

To ascertain the influence of age on the expression of VEGF₁₆₅, the age was compared among groups I, II and III. As shown in Figure 4B , the age of group I (38.6 ± 9.4 years) tended to be lower than that of group II (55.9 ± 7.1 years) and was significantly lower than that of group III (61.5 ± 7.6 years; P < 0.01). 

To see the influence of preoperative panretinal photocoagulation on the expression of VEGF₁₆₅ and its receptors VEGF-R2 and neuropilin-1 in fibrovascular tissues, the total amount of the therapy was compared among groups I, II, and III. However, the difference was not statistically significant (P = 0.43, Table 2 ). 

Other clinical factors including duration of diabetes and HbA1c were also compared among groups I, II, and III. However, no difference was shown in duration of diabetes (P = 0.69) or in HbA1c (P = 0.55, Table 2 ), and, thus, the vascular density was not confounded by these clinical factors or the preoperative laser therapy. 

Tissue Localization of VEGF

In situ hybridization was performed to identify the VEGF-expressing cells in fibrovascular tissues. As shown in Figure 5A , VEGF mRNA was detected with the antisense probe mainly in the cells located at the margin of the tissues and frequently in the cluster-forming cells, whereas the sense probe gave only a background signal (Fig. 5B) . Immunohistochemistry for GFAP on the serial section revealed that the cells expressing VEGF mRNA were also positive for GFAP (Fig. 5C) , indicating that they are glial cells. No or negligible immunostaining was found with nonimmune antibodies (Fig. 5D) . GFAP-positive cells were found in all the fibrovascular tissues examined, although such cells in each tissue varied in number. VEGF protein was immunolocalized not only to the glial cells but also to the endothelial cells (Fig. 5E) . The positive immunostaining for VEGF was abolished with the antibodies absorbed with the antigen (Fig. 5F) . 

Discussion

The present study demonstrated for the first time that VEGF receptors VEGF-R2 and neuropilin-1 are simultaneously expressed in the highly vascularized fibrovascular tissue in proliferative diabetic retinopathy. Our data also showed that highly vascularized tissues were more commonly found in younger patients, compatible with a general finding that young diabetic patients develop severe retinopathy. However, through the statistical analysis with age restriction, we could still obtain the significant relationship, indicating that the coexpression of VEGF-R2 and neuropilin-1, independent of the age, influenced the vascular density of the fibrovascular tissue. Importantly, the expression of VEGF-R1 had no such relationship with the vascular density. Thus, our present data suggest that the coexpression of VEGF-R2 and neuropilin-1 regulates the vascular density that probably represents the activity of proliferative diabetic retinopathy. This is compatible with the recent studies demonstrating that mitogenic activity of VEGF is mediated by binding to VEGF-R2 rather than VEGF-R1 ²⁰ and that neuropilin-1 functions as a coreceptor of VEGF-R2 to enhance the signal transduction from VEGF-R2. ¹⁸  

In the present study, the isoform VEGF₁₂₁ was constitutively expressed in the fibrovascular tissues, which varied in vascularity. Thus, this isoform may not be a molecule controlling the activity of fibrovascular proliferation. On the other hand, the expression of VEGF₁₆₅ was confined to some of the patients with highly vascularized tissue that coexpressed VEGF-R2 and neuropilin-1. It has been reported that VEGF₁₆₅ is a more potent endothelial cell mitogen than VEGF₁₂₁ ^{23 24} and that neuropilin-1 is an isoform-specific receptor that enhances the bioactivity of VEGF₁₆₅ through VEGF-R2. ¹⁸ Taking these findings into consideration, it is conceivable that the expression of VEGF₁₆₅ in the fibrovascular tissue that expresses VEGF-R2 and neuropilin-1 facilitates angiogenesis by means of self-sufficient interaction between the ligand and the receptors inside the growing tissue. However, the present study did not show any difference in the vascular density between group I, VEGF₁₆₅ (+), VEGF-R2 (+), neuropilin-1 (+) and group II, VEGF₁₆₅ (−), VEGF-R2 (+), neuropilin-1 (+) (120.8 ± 73.4 versus 108.8 ± 53.0 vessels/mm²). VEGF-mediated angiogenesis is for the most part hypoxia-induced ^{25 26} and ischemic retinal cells in diabetic retinopathy ^{4 5} secrete the soluble forms of VEGF that can act remotely in the closed space of the eye, resulting in an intraocular pathologic angiogenic process such as preretinal fibrovascular proliferation and iris neovascularization. Thus, the fibrovascular tissue can readily receive VEGF₁₆₅ secreted from the ischemic retina once the tissue coexpresses VEGF-R2 and neuropilin-1. This idea supports our result showing that the fibrovascular tissues with the coexpression of VEGF-R2 and neuropilin-1 were highly vascularized despite the absence of VEGF₁₆₅ expression inside the tissue. 

Our study showed that the isoform VEGF₁₆₅ was expressed inside the fibrovascular tissues of significantly younger patients (group I). We clinically encounter proliferative diabetic retinopathy, which takes a rapid course especially in young patients. This age distribution leads to the idea that the production of VEGF₁₆₅ inside the fibrovascular tissue with its receptors may contribute to rapid growth of the tissue in a self-proliferation system. 

Panretinal photocoagulation is a well-established therapy to reduce angiogenesis in diabetic retinopathy. However, the present study showed that the amount of the therapy had no relationship to the expression of VEGF-R2, neuropilin-1, or VEGF₁₆₅ in the fibrovascular tissue. Thus, other mechanisms, such as the destruction of VEGF-producing cells in the retina, are suggested to mediate the effect of the therapy. 

Glial cells are known to migrate from the retina to the preretinal tissues of various pathologic conditions such as idiopathic epiretinal membrane, ²⁷ proliferative vitreoretinopathy, ²⁸ and proliferative diabetic retinopathy ²⁹ and are suggested to be involved in the pathogenesis. During normal development of retinal vasculature, VEGF-secreting glial cells in the retina play a key role in the formation of new vessels. ^{30 31} Our present immunohistochemical and in situ hybridization studies demonstrated for the first time that glial cells inside the fibrovascular tissue express and produce VEGF. In addition, VEGF protein was immunolocalized to the endothelial cells in which VEGF mRNA expression was not detected by in situ hybridization. These suggest the possibility that VEGF secreted from glial cells reached its receptors in the endothelial cells. Furthermore, the isoform VEGF₁₆₅ expression inside the fibrovascular tissue of group I patients, which was demonstrated by our RT–PCR analysis, must be attributed to glial cells inside the tissue. In these young patients, VEGF₁₆₅ produced by glial cells inside the fibrovascular tissues could activate the isoform-specific signaling via VEGF-R2 and neuropilin-1 in endothelial cells. Thus, rapid progression of retinopathy in young diabetic patients may be explained by this glial cell–mediated self-proliferation mechanism. Further studies are necessary to substantiate this hypothesis. 

 Submitted for publication August 5, 1999; revised December 17, 1999; accepted January 11, 2000.

 Commercial relationships policy: N.

 

Table 1.

 

Clinical Information of the Present Cases and Results of Morphometry and RT-PCR on the Fibrovascular Tissues

Case	Age (years)	Sex	Type of Diabetes	Duration of Diabetes (years)	HbA1c (%)	Total Amount of Preoperative Panretinal Photocoagulation (burns)	Vascular Density (vessels/mm2)	Expression of VEGF Isoforms and Their Receptors
																VEGF121	VEGF165	VEGF- R1	VEGF- R2	Neuropilin-1
1	51	F	NIDDM	20	7.3	3200	45.5	+	+	+	+	+
2	41	M	NIDDM	3	9.8	0	81.7	+	+	+	+	+
3	66	M	NIDDM	10	6.3	2800	53.2	+	−	+	+	+
4	62	M	NIDDM	27	7.6	700	5.9	+	−	−	−	−
5	62	M	NIDDM	25	9.7	2900	43.0	+	−	+	+	+
6	47	M	IDDM	17	9.7	2800	183.0	+	−	+	+	+
7	60	M	NIDDM	10	6.2	2200	32.1	+	−	+	−	−
8	49	M	NIDDM	20	7.5	1600	201.3	+	+	−	+	+
9	29	F	IDDM	20	7.4	3100	55.9	+	+	+	+	+
10	61	M	NIDDM	27	6.5	2100	7.7	+	−	−	−	−
11	76	M	IDDM	45	7.2	0	73.9	+	−	−	−	−
12	58	M	NIDDM	25	10.3	2300	137.3	+	−	+	+	+
13	31	M	IDDM	20	8.4	0	77.5	+	+	+	+	+
14	61	M	NIDDM	20	6.7	200	12.2	+	−	−	−	−
15	61	F	NIDDM	10	8.4	1700	110.8	+	−	−	−	−
16	49	F	NIDDM	1	8.5	1200	37.2	+	−	+	−	−
17	41	F	NIDDM	8	5.1	2100	157.2	+	+	−	+	+
18	56	M	NIDDM	4	7.6	800	75.3	+	−	+	+	+
19	66	F	NIDDM	10	7.1	1500	4.9	+	−	−	−	−
20	47	M	NIDDM	20	7.2	1900	116.7	+	−	−	+	+
21	55	M	NIDDM	17	7.4	0	152.8	+	−	−	+	+
22	28	M	IDDM	7	5.3	200	226.8	+	+	+	+	+

 VEGF, vascular endothelial growth factor; IDDM, insulin-dependent diabetes mellitus; NIDDM, non-insulin–dependent diabetes mellitus.

Figure 1.

 

Histology of the fibrovascular tissues in proliferative diabetic retinopathy. Two representative cases are presented. (A, B) Highly vascularized fibrovascular tissue (157.2 vessels/mm²) from patient 17. (C, D) Sparsely vascularized tissue (32.1 vessels/mm²) from patient 7. (A, C) Hematoxylin–eosin stain; (B, D) immunohistochemistry for CD34 counterstained with methyl green. Scale bars, 200 μm.

Figure 2.

 

RT–PCR for the expression of VEGF isoforms and their receptors (VEGF-R1, VEGF-R2, and neuropilin-1) in the fibrovascular tissue. Patients who express VEGF₁₆₅, VEGF-R1, VEGF-R2, or neuropilin-1 are indicated with “+” under each panel. Note that VEGF-R2 and neuropilin-1 are expressed in the identical cases.

Figure 3.

 

Relationship between the vascular density and the expression of VEGF-R1, VEGF-R2, and neuropilin-1. (A) The vascular density shows no significant difference between the tissues with VEGF-R1 expression and those without its expression (P = 0.74, Mann–Whitney U test). (B) The vascular density of the tissues with the coexpression of VEGF-R2 and neuropilin-1 is significantly higher than that without their coexpression (P < 0.01, Mann–Whitney U test). (C) There is a reverse correlation between the age of cases and the vascular density of fibrovascular tissues (P < 0.01, Spearman rank correlation). (D) Analysis after age restriction shows the significant difference in vascular density between the patients with the coexpression of VEGF-R2 and neuropilin-1 and those without the coexpression (P < 0.01, Mann–Whitney U test).

Table 2.

 

Comparison of Vascular Density and Clinical Factors with the Expression of VEGF₁₆₅, VEGF-R2 and Neuropilin-1

	Group I*	Group II, †	Group III, ‡
Vascular density (vessels/mm2)	120.8 ± 73.4	108.8 ± 53.0	35.6 ± 38.3	P < 0.01
Age (years)	38.6 ± 9.4	55.9 ± 7.1	61.5 ± 7.6	P < 0.01
Total amount of preoperative panretinal photocoagulation (burns)	1460 ± 1410	1930 ± 1130	1200 ± 830	P = 0.43
HbA1c (%)	7.3 ± 1.7	7.9 ± 1.2	7.3 ± 0.8	P = 0.55
Duration of diabetes (years)	14.0 ± 7.6	16.9 ± 7.7	18.8 ± 14.0	P = 0.69

 Differences were examined by Kruskal-Wallis test. As for vascular density and age, Dunn procedure was subsequently performed as a post-hoc test as shown in Fig. 4 .

*  Group I: Seven cases that expressed VEGF₁₆₅ (+), VEGF-R2 (+), and neuropilin-1 (+).

†  Group II: Seven cases that expressed VEGF₁₆₅ (−), VEGF-R2 (+), and neuropilin-1 (+).

‡  Group III: Eight cases that expressed VEGF₁₆₅ (−), VEGF-R2 (−), and neuropilin-1 (−).

Figure 4.

 

Difference in vascular density and age among group I[ VEGF₁₆₅ (+), VEGF-R2 (+), and neuropilin-1 (+)]; group II [VEGF₁₆₅ (−), VEGF-R2 (+), and neuropilin-1 (+)]; and group III [VEGF₁₆₅ (−), VEGF-R2 (−), and neuropilin-1 (−)]. Significant differences are shown in vascular density (A) between groups I and III (P < 0.05, Dunn procedure) and between groups II and III (P < 0.05, Dunn procedure), and in age (B) between groups I and III (P < 0.01, Dunn procedure). Differences were examined by Kruskal–Wallis test and subsequent Dunn procedure as a post-hoc test.

Figure 5.

 

In situ hybridization for VEGF and immunohistochemistry for GFAP and VEGF in the serial sections of the fibrovascular tissue from patient 17: VEGF₁₆₅ (+), VEGF-R2 (+), neuropilin-1 (+). (A, B) In situ hybridization for VEGF. Note that the cells located at the margin of the tissue and the cluster-forming cells (arrows) are labeled with the antisense probe (A), whereas only a background signal is shown with the sense probe (B). (C) Immunohistochemistry for GFAP. Note that the cells expressing VEGF mRNA (arrows in A) are also positive for GFAP (arrows). (D) Negative control with nonimmune immunoglobulins. (E) Immunohistochemistry for VEGF. VEGF protein is immunolocalized not only to the GFAP-positive cells (large arrows) but also to the endothelial cells (small double arrows). (F) Immunostaining with the VEGF antibodies absorbed with the antigen. Methyl green counterstain for all. Scale bars, 100 μm.

We thank Yuichirou Akiyama and Hitoshi Abe for their excellent technical assistance, Toru Takebayashi and Yuji Nishiwaki (Department of Preventive Medicine and Public Health, Keio University School of Medicine, Tokyo, Japan) for their advice on the statistical analyses, and Shingo Kato (Department of Microbiology, Keio University School of Medicine, Tokyo, Japan) for his assistance in RT–PCR analysis.