VEGF Polymorphisms Are Associated with Severity of Diabetic Retinopathy

Abstract

purpose. To determine whether single nucleotide polymorphisms (SNPs) in the vascular endothelial growth factor (VEGF) gene are associated with severity of diabetic retinopathy.

methods. A case–control study was conducted in which 45 individuals with type 1 or 2 diabetes with proliferative diabetic retinopathy (PDR) and 61 individuals with type 1 or 2 diabetes without retinopathy (DWR) were genotyped for 14 SNPs in the VEGF promoter and gene.

results. Three of the promoter SNP genotypes, −160C, −152A (rs13207351), and −116A (rs1570360), showed significant independent associations with PDR, as well as the minihaplotype CAA (P = 0.00017). Two promoter haplotypes were associated with severity of retinopathy: −460C, −417T, −172C, −165C, −160C, −152A, −141A, −116A, +405C was associated with PDR (OR [95% CI] = 29.92 [3.91, 228.78], P = 1.62 × 10⁻⁵) and −460C, −2417T, −172C, −165C, −160C, −152A, −141A, −116G, +405G was associated with DWR (OR = 0.05 [0.01, 0.35], P = 0.000373). Furthermore, two haplotype-tagged (ht) SNPs, +4618 (rs735286) and +5092 (rs2146323), and five htSNP haplotypes were associated with severity of retinopathy. When the nine promoter/5′ untranslated region [UTR] and five htSNP genotypes were combined into a 14-SNP haplotype, a single haplotype, −460C, −417T, −172C, −165C, −160C, −152A, −141A, −116A, +405C, +674T, +4618C, +5092A, +9162C, +9512C was found to be significantly associated with the PDR group (OR = 18.45 [2.35, 144.67], P = 0.00622).

conclusions. A clear association was demonstrated between VEGF SNPs and severity of diabetic retinopathy. Furthermore, two of the htSNP haplotypes appear to be more generalized markers for angiogenesis, in that these have been found in prior work to be associated with neovascular age-related macular degeneration.

Diabetic retinopathy is one of the most prominent pathologic vascular complications in diabetes and is the most common cause of blind registration in the working-age group. Advanced stage, or proliferative diabetic retinopathy (PDR) is reached when abnormal growth of retinal vessels leads to neovascularization within the retina and vitreous gel, often accompanied by extensive hemorrhage and fibrosis. The most effective treatment for early PDR is pan retinal laser photocoagulation to ablate the retina and thereby reduce the angiogenic drive from proteins, such as vascular endothelial growth factor (VEGF). VEGF is massively upregulated in diabetic retinopathy and not only promotes neovascularization and migration but also increases vascular permeability and leakage. New treatments are emerging that are designed to stultify the effects of VEGF to treat the ocular complication of diabetes. Jorge et al. ¹ recently published a report of the successful use of intravitreous injections of the anti-VEGF-A monoclonal antibody, bevacizumab (Avastin; Genentech, Vacaville, CA), to treat persistent proliferative retinopathy. Pegaptanib sodium (Macugen; OSI Pharmaceuticals, Melville, NY), a pegylated anti-VEGF₁₆₅ aptamer, used to treat neovascular age-related macular degeneration, is in phase 2 clinical trials to treat diabetic macular edema and retinal vein occlusion. ²  

VEGF₁₆₅ is the most prominent member of the VEGF-A family in the eye and can be alternatively spliced into an a or b isoform, the difference being the presence at the C terminus of either exon 8 or exon 9, respectively. ³ VEGF₁₆₅a has potent angiogenic properties while VEGF₁₆₅b has antiangiogenic properties. We have demonstrated that the VEGF₁₆₅b is the predominant isoform in healthy vitreous and that insulin-like growth factor (IGF), which increases in diabetes, downregulates VEGF₁₆₅b, thereby increasing the ratio of a-to-b isoforms. ⁴ What controls splicing is still largely unknown but it is generally regulated at the 5′ end of the gene by recruitment of splicing factors to the C-terminal domain of the RNA polymerase. There can also be intronic enhancers and silencers that regulate transcription but the exact mechanism remains obscure. The VEGF gene is unusually polymorphic in the promoter and 5′ untranslated region (UTR) and several studies have focused on single nucleotide polymorphisms (SNPs) within this region and their association with angiogenic eye disease. Awata et al. ⁵ demonstrated that the rs2010963 SNP was associated with diabetic eye disease with carriage of the C allele associated with macular edema and the presence of diabetic retinopathy. Vannay et al. ⁶ found that carriage of the rs2010963 C allele or the rs833061 TT/rs2010963 CC haplotype was associated with premature infants requiring treatment for neovascularization seen in retinopathy of prematurity, demonstrating that this SNP may be important in other angiogenic eye diseases. ⁶ More recently Al-Kateb et al. ⁷ published a longitudinal data study from Canada demonstrating an association between multiple VEGF variants and the development of severe retinopathy in individuals with type 1 diabetes. 

We are interested in the role of VEGF variants in retinal angiogenesis and previously demonstrated VEGF SNP and haplotype associations with neovascular age-related macular degeneration (AMD). ⁸ In this study, we analyze the same genetic variants (nine SNPs within the promoter and 5′ UTR and five intronic SNPs spanning the VEGF gene) and demonstrate that there are specific DNA profiles that are associated with the development of diabetic retinopathy in a case-control study of European Caucasians with type 1 or 2 diabetes. Also, there were DNA profiles that were shared between exudative AMD and PDR suggesting that these are risk factors for retinal angiogenesis rather than being disease specific. 

Methods

Patient Selection

PDR cases were recruited from the Argon laser clinic at the Bristol Eye Hospital (n = 45). All cases had funduscopic evidence of diabetic retinal neovascularization. Diabetic individuals without retinopathy (DWR) were recruited from Manchester Eye Hospital (n = 61) and had no evidence of proliferative retinopathy on funduscopic examination. Duration of disease was our matching criteria, to within 7 years or less in the two groups, the ranges being 7 to 44 years for PDR (mean, 23.1 years) and 14 to 50 years for DWR (mean, 24.6 years). Age and sex ranges were similar in the two groups: PDR: 24 to 86 years (mean age, 59.6 years; 60% male); DWR: 24 to 97 years (mean age 55.3 years; 52.5% male). The type of diabetes mellitus on diagnosis was not important for the purposes of our analysis but was as follows: PDR: 38.6% type 1, 61.3% type 2; DWR: 67.2% type 1, 32.8% type 2. All study participants were Caucasians of Northern European origin. Ethics approval for the study was obtained from the North Somerset and South Bristol Research Ethics Committee and protocols conformed to the tenets of the Declaration of Helsinki, as revised in 2000. A venous blood sample was obtained from each participant after informed written consent. 

DNA Preparation

DNA was extracted by rapid salting technique and quantified by spectrophotometry (SpectraMax Plus, Molecular Devices, Wokingham, UK). ⁹ To ensure there was no selection bias between the sample populations, each DNA sample was examined for Hardy-Weinberg equilibrium using two independent polymorphic markers (in TYRP1ex2 and TP53PIN3) unrelated to diabetes, as described in several studies. ^{10 11 12}  

Promoter SNPs

Nine SNPs were selected from a PubMed search on association studies of polymorphisms within the VEGF promoter and 5′ UTR region: C-460T (rs833061), T-417C (rs833062), C-172A, C-165T, C-160T, G-152A (rs13207351), A-141C (rs28357093), G-116A (rs1570360), and G+405C (rs2010963). All SNPs are numbered from the transcription start (GenBank accession no. M63971; http://www.ncbi.nlm.nih.gov/Genbank; provided in the public domain by the National Center for Biotechnology Information, Bethesda, MD). ¹³ PCR primers were designed to incorporate the eight promoter SNPs in one fragment (forward: 5′-GGTGAGTGAGTGTGTGCG-3′, reverse: 5′-CCGCTACCAGCCGACTTT-3′). PCR conditions are described in Churchill et al. ⁸ PCR products were purified (QIAquick PCR purification kit; Qiagen, Crawley, UK), and sequenced (Genetic Research Instrumentation, Ltd. [GRI], Braintree, UK). Sequence data were checked visually (Chromas ver. 1.45; Technelysium Pty Ltd., Tewantin, QLD, Australia), genotypes were determined (Sequencher ver. 4.01 software; GeneCodes, Ann Arbor, MI), with VEGF data from GenBank (AL136131) as a reference sequence. Genotyping for the +405 polymorphism (rs2010963) was by PCR-RLFP analysis as previously described. ⁸ A subset of samples were sequenced to verify the PCR-RFLP analysis. 

Haplotype-Tagged SNP Selection

SNPs were selected from data obtained from the University of Washington and the Fred Hutchinson Cancer Research Center (UW-FHCRC) Variation Discovery Resource (SeattleSNPs) Database, a collaborative online resource of SNP data, developed by the UW-FHCRC (available at http://pga.gs.washington.edu/). ¹⁴ Using the SeattleSNP haplotype data (haplotype frequency >3%), we selected five htSNPs: +674 (rs1413711), +4618 (rs735286), +5092 (rs2146323), +9162 (rs3025021), and +9512 (rs3025024) using the SNPtagger Program, ^{8 15} as SNPs representative of 19 different polymorphic sites. Genotype data were obtained by using allele-specific PCR techniques as described in Churchill et al. ⁸ To ensure accuracy of allele-specific results, a randomized selection of sample PCRs were sequenced. The resultant data were collated, the samples genotyped, and the frequencies analyzed. 

Haplotype and Statistical Analysis

Individual haplotypes and their estimated population frequencies were inferred by using the PHASE program, ver. 2.1, ^{16 17} with all parameters set at the default values. Statistical analysis was performed using commercial software (SPSS ver. 11.5.0; SPSS UK Ltd., Woking, UK), Epi Info 6 v6.04d software (Centers for Disease Control and Prevention, Atlanta, GA), and SISA binomial. ¹⁸ Pearson’s χ² test and/or the Fisher exact test were used to compare the PDR and DWR groups for possible associations between SNP genotype and allele frequency and disease state. Statistical significance was assumed at corrected P < 0.05. Odds ratios were also calculated, adjusted by the Haldane correction when necessary. Because of the multiple analysis conducted, the Bonferroni correction was applied to produce corrected probabilities. 

Results

Analysis of the TYRP1ex2 and TP53 observed genotypes shows the study population was in Hardy-Weinberg equilibrium for all alleles investigated. No significant difference between the PDR and DWR group polymorphism distributions suggests that no selection bias was present (data not shown). Samples selected for sequencing produced genotypes in agreement with those observed via allele-specific PCR. 

Promoter and 5′-UTR Data

Individual SNP Genotype Analysis.

Three of the promoter SNPs, −160, −152 (rs13207351), and −116 (rs1570360), showed significant independent associations (Table 1A) . 

The C allele at −160 was significantly associated with the PDR group (P = 0.0008) as was the CC genotype (P = 0.0003). The presence of the T allele was associated with the DWR group. 

The A allele at −152 was significantly associated with the PDR group (P = 0.0353) as was the AA genotype (P = 0.0022). Carriage of the G allele (GG or GA genotypes) was associated with the DWR group. 

The A allele at −116 was significantly associated with the PDR group (P = 1.16 × 10⁻⁷), as was the AA genotype alone (P = 3.23 × 10⁻⁶). Carriage of the G allele and the GG genotype were associated with the DWR group. 

Haplotype Analysis.

We analyzed a minihaplotype for the SNPs −160, −152, and −116, and found that −160C/−152A/−116A was significantly associated with PDR (P = 0.00017) and that the haplotype −160T/−152G/−116G was significantly associated with DWR (P = 0.00618). 

PHASE predicted 29 possible extended haplotypes for the promoter and 5′ UTR, of which two haplotypes were significant (Table 1B) . The haplotype −460C/−417T/−172C/−165C/−160C/−152A/−141A/−116A/+405C was significantly associated with PDR (P = 1.62 × 10⁻⁵). The second significant haplotype, −460C/−2417T/−172C/−165C/−160C/−152A/−141A/−116G/+405G, was significantly associated with DWR (P = 0.000373). 

The difference between these two haplotypes lies in the last two SNPs with −116A/+405C associated with PDR and −116G/+405G associated with DWR. When we analyzed these two SNPs in isolation as a minihaplotype, we confirmed that −116A/+405C was associated with PDR (P = 2.07 × 10⁻⁶; Table 1B ) and that both −116G/+405G and −116G/+405C were significantly associated with DWR. 

Haplotype- Tagged SNP Data

Individual SNP Genotype Analysis.

Two of the haplotype-tagged (ht) SNPs, +4618 (rs735286) and +5092 (rs2146323), showed significant independent associations (Table 2A) :

•  
  Carriage of the T allele at +4618 was associated with the PDR group (P = 0.0016). The CC genotype is associated with the DWR group (P = 0.0016).
•  
  Carriage of the C allele at +5092 was associated with the PDR group (P = 0.0009). Overall carriage of the A allele was similar in both groups, but the AA genotype was significantly associated with the DWR group (P = 0.0009).

No significant associations were seen at the other three htSNP loci. 

Haplotype Analysis.

PHASE predicted 21 different haplotypes from the combined genotypes of the five htSNPs (+674, +4618, +5092, +9162, and +9512). Five haplotypes were significant after correction for multiple analysis (Table 2B) . The haplotypes CTCCT, CTCTT, and TCACC were associated with the PDR group. The haplotypes CTCTC and TCACT were associated with the DWR group. A control group was also screened to look at the incidence of these five haplotypes in the general population when diabetes and age-related macular degeneration were excluded. 

Combined Promoter/5′ UTR/htSNP Haplotype Data.

When the promoter/5′ UTR and htSNP genotypes were combined into a 14-SNP haplotype (representing 28 VEGF SNPs overall), PHASE predicted 85 haplotypes (data not shown). A single haplotype, −460C, −417T, −172C, −165C, −160C, −152A, −141A, −116A, +405C, +674T/+4618C, +5092A, +9162C, +9512C, was found to be significantly associated with the PDR group after correction for multiple analyses (OR [95% CI] = 18.45 [2.35, 144.67], corrected P = 0.00622). 

Discussion

Our purpose was to determine whether a VEGF SNP profile was associated with proliferative diabetic retinopathy. By making use of a combination of promoter SNPs, htSNPs, and the strong linkage disequilibrium (LD) that exists across the VEGF gene, we were able to provide a comprehensive coverage of the gene relatively efficiently. We found three promoter SNPs, −160, −152 (rs13207351), and −116 (rs1570360), associated with PDR, both independently and when combined as a three-SNP haplotype. Analyzing the mechanisms that underlie these associations is complex, since there have been relatively few individual functional SNP studies. Moreover, the high degree of LD means that the association may lie within the LD bin rather than with the individual SNP. ⁸  

The −160CC genotype is significantly associated with PDR (OR = 10.5 [2.3, 47.7] P = 0.0003). The expected population frequency of the −160C allele is high (allele frequency, 0.94) so the lack of TT genotypes in both PDR and DWR groups is to be expected. Although the functional effect of this isolated SNP has not been studied, there appears to be a modifying influence on the retinopathy when one allele in the genotype is T. 

The −152AA genotype is significantly associated with PDR (OR = 3.5 [1.5, 7.7], P = 0.0022). Stevens et al. ¹⁹ published a very low frequency of 0.017 for the −152A allele. In our ethnically matched control population we found a higher allele frequency of 0.66. In our study there appeared to be a modifying influence on the retinopathy when one allele in the genotype was G. The functional significance of this change is unknown, but when studied as part of a five-SNP promoter haplotype, the two haplotypes with −152A allele showed higher basal promoter activity in vitro than did the haplotype containing the −152G allele. ¹⁹  

The −116 SNP showed the strongest association in the study, with the AA genotype significantly more prevalent in the PDR group (OR = 7.9 [3.1, 19.9], P = 3.23 × 10⁻⁶). There have been several studies investigating the role of this SNP in angiogenesis with conflicting results. Del Bo et al. ²⁰ demonstrated that the −116 GG genotype was protective against the development of type 1 diabetes and associated with a delayed onset in the Finnish population. Lambrechts et al. ²¹ showed that presence of the −116A allele led to decreased VEGF transcription in vitro and was associated with lower total VEGF serum levels. In contrast, however, they showed that the combined haplotypes −1540A/−116A/+405G and −1540A/−116G/+405G were both associated with lower VEGF serum levels. 

Given the high LD in the VEGF gene, there is evidence of conserved haplotypes in the promoter and 5′ UTR. We found that the promoter SNPs −160C, −152A, and −116A were associated with PDR when combined as a three-SNP haplotype. 

When we studied the larger nine-SNP promoter/5′ UTR haplotype we found the profile associated with PDR (−460C, −417T, −172C, −165C, −160C, −152A, −141A, −116A, +405C) differed from that associated with DWR (−460C,−417T, −172C, −165C, −160C, −152A, −141A, −116G, +405G) by the last two SNPs only. Several studies have concluded that if the polymorphisms −116 and +405 are analyzed together, there are three major haplotypes: GC, GG, and AG. The AC haplotype is very rare (<1%) due to tight LD. ^{5 19 22} We analyzed our groups and found the −116A/+405C haplotype prevalent among diabetic patients with proliferative disease: 17 of 45 PDR were obligate AC haplotypes, in that the genotypes were either AA/C* or A*/CC compared with only 2 of 61 DWR. The two SNP AC haplotype was significantly associated with PDR (OR = 18.26 [4.16, 80.25], corrected P = 2.07 × 10⁻⁶). In contrast, both −116G/+405G and −116G/+405C were significantly associated with DWR highlighting the controversial role of the +405 polymorphism. 

Awata et al. ⁵ demonstrated an association between the +405 C allele and increased VEGF transcription, diabetic macular edema, and the presence of diabetic retinopathy (but not the severity) in a Japanese population. Errera et al ²³ (2007) analyzed 167 PDR versus 334 DWR (type 2 diabetic individuals) and found no association with +405 C allele as a dominant model, but an association with CC and PDR as a recessive model. Liinamaa ²⁴ and Ray et al. ²⁵ did not find an association with +405C allele and PDR in a mix of type 1 and 2 diabetic Caucasians. Watson et al. ²⁶ found no association with the +405 genotype and diabetic retinopathy in a Caucasian sample but showed highest levels of VEGF in individuals carrying the +405GG genotype. In contrast, Lambrechts et al. ²¹ showed that +405G allele decreased VEGF translation in vitro, probably due to a less favorable secondary structure of IRES-B sequence resulting in less L-VEGF thereby reducing the reservoir for cleavage and production of shorter VEGFs (e.g.,VEGF₁₆₅). We did not find an independent association between the +405 SNP and PDR using a dominant or recessive model, suggesting that the +405C allele may only be influential when associated with −116A. 

Our understanding of VEGF biology is further complicated by the recent discovery that VEGF exists as different isoforms VEGF_xxx and VEGF_xxxb, with the b isoforms possessing antiangiogenic properties. ³ Although investigators have measured total VEGF, they have not distinguished between the isoforms that may explain some of the apparently conflicting results. We have demonstrated that VEGF₁₆₅b is the predominant isoform in the healthy eye and that the ratio of VEGF₁₆₅b to VEGF₁₆₅ shifts significantly in favor of VEGF₁₆₅ in PDR. ⁴ The control of splicing is currently not understood, but it is conceivable that the −116A allele may favor splicing to produce the VEGF₁₆₅ isoform (despite an overall reduction in VEGF transcription), and when coupled with the +405C allele, L-VEGF production increases thus creating a large reservoir for VEGF₁₆₅ production and a proangiogenic environment. Because it is the ratio of VEGF_xxx to VEGF_xxxb isoforms rather than the level of total VEGF that is important, many of these studies must be repeated to gain a full understanding of the underlying processes. 

Analysis of the VEGF gene revealed that two of the htSNPs +4618 (rs735286) and +5092 (rs2146323) showed significant associations with PDR (Table 1) . Both SNPs lie within intron 2, are in the same LD bin, and have been shown to be tagged to SNPs at +6166 and +7150. ^{8 27} The +4618 polymorphism is located in intron 2, 585 bp 5′ of exon 3, and analysis (Transplorer, ²⁸ TFSearch ²⁹ ) suggest that this region is rich in potential transcription factor-binding sites. In particular, a putative myeloid zinc finger protein (MZF-1) binding site is located 8 bp 3′ of the +4618 polymorphism. Transcription factors such as MZF-1 bind directly to DNA and control transcription of the gene, ³⁰ thus the short distance between this putative binding site and +4618 may mean that the polymorphism has some influence on VEGF transcription. Watson et al. ²⁶ also implicated MZF-1 in their analysis of the +405 polymorphism. Alternatively, +4618 may influence exon splicing by altering splicing factor recognition sequences, such as ESEs (exonic splicing enhancers). Analysis with ESE Finder ^{31 32} suggests the +4618T risk allele creates an additional SRp55 ESE site and increases the potential of a SC35 ESE site from 4.18 to 5.57 (threshold, 2.38). Such splicing factors could influence downstream alternative splicing of VEGF from antiangiogenic to proangiogenic isoforms. The +5092 (rs2146323) polymorphism is located 111-bp 5′ of exon 3. It is relatively distant from predicted transcription factor binding sites (5′, 306 bp to second MZF-1 site; 3′, 80 bp from PAX-4 and 212 bp from GATA-2 sites), and there are no changes in putative ESE sites from either of the +5092 alleles, nor any predicted alterations to exon splicing making it a less attractive SNP for functional analysis. Moreover, a recent Finnish case–control study found no association between the +5092 (rs2146323) polymorphism and severity of diabetic retinopathy. ²⁴ It is perhaps more likely that these two SNPs serve to highlight an as-yet-unknown variant. Further evidence of this is provided by Al-Kateb et al. ⁷ in a North American study of individuals with type 1 diabetes, who found in a multivariate analysis that severe diabetic retinopathy was associated with both +5092 (P = 0.011) and a second SNP, rs3025010 (P = 0.0052), within the same LD bin. This parallels previous association studies of VEGF polymorphisms in AMD, ^{8 27} where strong LD was observed across the VEGF gene region. 

We found three htSNP risk haplotypes associated with PDR and two htSNP haplotypes associated with DWR. Note that the two htSNPs that held independent allele associations did not maintain this association in the haplotype analysis, not unlike the +405 SNP, suggesting that haplotype analysis may be more informative than individual SNP associations. One of the at-risk haplotypes, CTCTT, was found to have approximately the same frequency in nondiabetic individuals from the general population, which may suggest that this haplotype is less important in the pathogenesis of proliferative retinopathy. One of the protective haplotypes, CTCTC, however, was also more common in the nondiabetic general population, perhaps adding weight to the beneficial effect of possessing this haplotype. Two of the risk haplotypes (CTCCT and TCACC) were also found to be risk haplotypes in a study on neovascular age-related macular degeneration (AMD), ⁸ which indicates that these may be more widely associated with the process of angiogenesis than disease specific. We feel that this is an area that merits further study in other ocular and nonocular angiogenic disease. The TCACT haplotype associated with DWR also positively correlated with the control group in our AMD study, ⁸ suggesting that this haplotype is more generally protective against pathologic angiogenesis. A single combined promoter and 5′ UTR/htSNP haplotype (CTCCCAAAC/TCACC) was significantly associated with the PDR group after correction for multiple analyses (OR = 18.45 [2.35, 144.67], corrected P = 0.00622) and not surprisingly combines the previously noted promoter haplotype and one of the risk htSNP haplotypes. This differs from the risk promoter haplotype (TTCCCGAAC) seen in neovascular AMD at two positions only (−460 and−152). This difference is further reduced when one considers that these two SNPs have been found to be in complete LD as either TG or CA and should therefore be considered as one. ³³ Of note, the −116A, +405C combination holds true in both AMD and PDR studies, suggesting a wider role in angiogenic disease. Functional studies will allow further investigation into the role of this haplotype in VEGF production and, more important, how this influences the ratio of VEGF_xxx to VEGF_xxxb isoforms. Such studies will advance our understanding of the mechanisms underlying not only proliferative diabetic retinopathy but potentially other retinal angiogenic diseases. 

With the great increase in SNP association studies, DNA profiling is likely to become a popular way to determine risk of both ocular and nonocular disease. At present there is a patent on the use of a VEGF genotype information in the diagnosis of diabetic retinopathy, ³⁴ despite the wide range and often conflicting findings of VEGF polymorphic association studies in diabetic patients. ^{24 25 33} Early PDR can be treated with reasonable success, but it is virtually asymptomatic in the early stages, and detection relies on individuals attending regular retinal monitoring programs. Given that not all diabetic individuals progress to PDR, a large number of individuals may be attending monitoring unnecessarily. Although we would caution against the development of a predictive genetic test on the results of a single study, we can foresee that area will expand in an attempt to direct our ever-diminishing healthcare resources more effectively to those who need it most. 

 Supported by the Underwood Trust.

 Submitted for publication October 29, 2007; revised March 10 and April 15, 2008; accepted June 16, 2008.

 Disclosure: A . J. Churchill, None; J.G. Carter, None; C. Ramsden, None; S . J. Turner, None; A. Yeung, None; P . E.C. Brenchley, None; D . W. Ray, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Table 1.

 

Significant VEGF Associations between Diabetic Individuals with and without Proliferative Retinopathy

A. Three Significant Individual VEGF SNP Associations
SNP	Status	Genotype					Significant at-Risk Genotype/Allele
				CC	CT	TT		OR (95% CI)	P
−160C>T	PDR*	43 (95.6)	2 (4.4)	0	CC	10.5 (2.3 ,47.7)	0.0003
	DWR, †	41 (67.2)	20 (32.8)	0	C	8.63 (1.96 ,37.95)	0.0008

SNP	Status	Genotype					Significant at-Risk Genotype/Allele
				AA	AG	GG		OR (95% CI)	P
−152A>G	PDR*	29 (64.4)	7 (15.6)	9 (20)	AA	3.5 (1.5 ,7.7)	0.0022
	DWR, †	21 (34.4)	29 (47.5)	11 (18)	A	1.87 (1.04 ,3.35)	0.0353

SNP	Status	Genotype					Significant at-Risk Genotype/Allele
				AA	AG	GG		OR (95% CI)	P
−116A>G	PDR*	26 (57.8)	15 (33.3)	4 (8.9)	AA	7.9 (3.1 ,19.9)	<0.0001
	DWR, †	9 (14.8)	28 (45.9)	24 (39.3)	A	4.8 (2.6 ,8.8)	<0.0001

B. Significant VEGF SNP Haplotype Associations
	Status	With Haplotype	Without Haplotype	Pearson χ2 OR (95% CI)	Corrected P , ‡
−160C/−152A/−116A	PDR, §	55 (61.1%)	35 (38.9%)	3.34 (1.89 ,5.91)	0.0001695
	DWR, ∥	39 (32.0%)	83 (68.0%)
−160T/−152G/−116G	PDR, §	0	90 (100%)	0.080 (0.010 ,0.062)	0.006181
	DWR, ∥	14 (11.5%)	108 (88.5%)
−460C/−417T/−172C/−165C/−160C/	PDR, §	17 (18.9%)	73 (81.1%)	29.92 (3.91 ,229)	1.62 × 10−5
−152A/−141A/−116A/+405C	DWR, ∥	0	122 (100%)
−460C/−417T/−172C/−165C/−160C/	PDR, §	0	90 (100%)	0.05 (0.01 ,0.35)	0.000373
−152A/−141A/−116G/+405G	DWR, ∥	23 (18.9%)	99 (81.1%)
−116A/+405C	PDR, §	21 (23.3%)	69 (76.7%)	18.26 (4.16 ,80.25)	2.07 × 10−4
	DWR, ∥	2 (1.6%)	120 (98.4%)

*  PDR, n = 45.

†  DWR, n = 61.

‡  Probabilities corrected for multiple analysis (n = 27).

§  PDR, n = 90.

∥  DWR, n = 122.

Table 2.

 

Significant VEGF htSNP Associations between Diabetic Individuals with and without Proliferative Retinopathy

A. Two Significant Individual VEGF htSNP Associations
SNP	Status	Genotype					Significant Genotype/Alleles
				CC	CT	TT		OR (95% CI)	P
+4618C>T	PDR*	10 (22.2)	35 (77.8)	0	CC	0.26 (0.11 ,0.61)	0.0016
	DWR, †	32 (52.5)	27 (44.3)	2 (3.3)	T	3.86 (1.63 ,9.16)	0.0016

SNP	Status	Genotype					Significant Genotype/Alleles
				CC	CA	AA		OR (95% CI)	P
+5092C>A	PDR*	19 (42.2)	26 (57.8)	0	AA	0.076 (0.0096 ,0.60)	0.0009
	DWR, †	26 (42.6)	22 (36.1)	13 (21.3)	C	13.14 (1.66 ,103.98)	0.0009

B. Five Significant VEGF htSNP Haplotype Associations, ‡
Haplotype	PDR n = 90 (%)	DWR n = 122 (%)	OR (95% CI)	P , §	Nondiabetic Individuals, ∥ n = 188 (%)
CTCTT	12 (13.3)	3 (2.5)	6.10 (1.67 ,22.33)	0.0462	23 (12.2)
CTCCT, ¶	19 (21.1)	4 (3.3)	7.89 (2.58 ,24.14)	0.000773	1 (0.5)
TCACC, ¶	26 (28.9)	5 (4.1)	9.51 (3.48 ,25.96)	9.31 × 10−6	4 (2.1)
TCACT, ¶	0	29 (23.8)	0.03 (0.00 ,0.26)	1.35 × 10−5	46 (24.5)
CTCTC	0	15 (12.3)	0.07 (0.01 ,0.57)	0.0105	15 (8.0)

*  PDR, n = 45.

†  DWR, n = 61.

‡  htSNP haplotype order: +674/+4618/+5092/+9162/+9512.

§  Corrected for multiple analyses (n = 21).

∥  Controls from general population that do not have diabetes or age-related macular degeneration used in previous associated studies (see Churchill et al. ⁸ ).

¶  The haplotype for neovascular age-related macular degeneration (see Churchill et al. ⁸ ) was also shown to be significant, suggesting that it may be a significant risk factor for other types of ocular angiogenesis

The authors thank Helen Lovell for assistance in some of the preliminary work for this project and Julia Escardo, Cheryl Lee, and Denize Atan for assistance in the collection of blood samples from study volunteers.