Variants of the Adenosine A_2A Receptor Gene Are Protective against Proliferative Diabetic Retinopathy in Patients with Type 1 Diabetes

Abstract

Aims

The adenosine A_2A receptor (ADORA_2A) may ameliorate deleterious physiologic effects associated with tissue injury in individuals with diabetes. We explored associations between variants of the ADORA_2A gene and proliferative diabetic retinopathy (PDR) in a cohort of patients with type 1 diabetes (T1D).

Methods

The participants were from the Pittsburgh Epidemiology of Diabetes Complications prospective study of childhood-onset T1D. Stereoscopic photographs of the retinal fundus taken at baseline, then biennially, for 10 years were used to define PDR according to the modified Airlie House system. Two tagging single nucleotide polymorphisms (tSNPs; rs2236624-C/T and rs4822489-G/T) in the ADORA_2A gene were selected using the HapMap (haplotype map) reference database.

Results

A significant association was observed between SNP rs2236624 and PDR in the recessive genetic model. Participants homozygous for the T allele displayed a decreased risk of developing prevalent PDR (odds ratio, OR = 0.36; p = 0.04) and incident PDR (hazard ratio = 0.156; p = 0.009), and for all cases of PDR combined (OR = 0.23; p = 0.001). The protective effect of T allele homozygosity remained after adjusting for covariates. Similarly, for SNP rs4822489, an association between PDR and T allele homozygosity was observed following covariate adjustment (OR = 0.55; 95% CI: 0.31–0.92; p = 0.04).

Conclusion

Genetic variants of ADORA_2A offer statistically significant protection against PDR development in patients with T1D.

Introduction

Diabetic retinopathy (DR) is the cause of up to 24,000 incident cases of blindness in the USA each year, and is the most prevalent cause of blindness between the second and seventh decades of life [1]. Most individuals with diabetes eventually experience some form of DR during the course of their lives; however, a greater proportion of patients with type 1 diabetes (T1D) develop proliferative DR (PDR) than do patients with T2D, with reported estimates as high as 50% compared to 10%, respectively [2,3].

The presence of DR and/or severity of DR have been shown to be partially heritable [4,5,6]. The Family Investigation of Nephropathy and Diabetes-Eye study consisting of European Americans, African Americans and Mexican Americans estimated heritability for PDR to be about 25% [7]. Observed ethnic differences in susceptibility to DR also provide some support for the potential role of genetic factors in the etiology of DR [4,5,6,7,8,9,10].

Adenosine is a powerful physiologic mediator that modulates cellular damage and the resulting tissue injury caused by biologic stressors. In the presence of oxidative stress (OS), hydrogen peroxide is formed in the cytosol, rendering vascular endothelial cells more permeable [11], and adenosine appears to ameliorate this process. The effects of adenosine are directed by adenosine receptors (AR), with the adenosine A₁ receptor having a proinflammatory response to tissue injury, while the adenosine A₂ receptor (ADORA_2A) restricts inflammation and guards tissues from further damage [12,13]. In rats, the AR exhibit variability in function and a long-term response to ischemic brain injury [14]. Other animal studies have provided evidence that ADORA_2A protects the kidneys and the heart from ischemic reperfusion injury by reducing the generation of reactive oxygen species, thereby limiting mitochondrial damage and guarding against apoptosis [15,16,17].

Growing biological evidence suggests that reactive oxygen species and OS caused by long-term exposure to hyperglycemia may be responsible for some of the tissue damage associated with microvascular complications of diabetes [18,19,20]. There is evidence in bovine retinal endothelial cells that adenosine may play a role in the upregulation of expression of a gene involved in glucose transport (GLUT1) [21]. Other investigators using electroretinogram measurements were able to demonstrate that adenosine caused vasodilatation, an increase in the rod-driven b-wave, an increase in the cone-driven b-wave, followed by a decrease in the cone-driven b-wave that was mediated by ADORA_2A in the cat retina [22]. Others found that adenosine or its agonist decreased the blood flow and induced apoptosis in rat or rabbit retinae [23,24,25]. Furthermore, it has been demonstrated that OS is evident in the retinae of individuals with diabetes – a phenomenon that is likely to contribute to the development of DR [26]. ADORA_2A also appears to ameliorate the effects of hypoxia, inflammation, vasodilatation, intraocular hyperglycemia and the speed of glucose metabolism out of the cell [13,21,27]. Both human and animal studies provide evidence that adenosine [28] or its receptors [29] play a role in angiogenesis [30,31]. Specifically, ADORA_2A[32] has been shown to increase the proliferation of human retinal endothelial cells.

Based upon the multiple roles that adenosine and the ADORA_2A gene play in ameliorating the effects of biologic stressors in humans and animals, it is reasonable to consider ADORA_2A as a plausible candidate gene for increased susceptibility to or protection from the development and progression of PDR. We therefore selected tag single nucleotide polymorphisms (tSNPs) from the European ancestry (CEU) HapMap (haplotype map) data that were representative of the entire ADORA_2A gene, as described below in the Methods section. We evaluated the association between tSNPs and PDR in a well-characterized sample of patients with T1D.

Research Design and Methods

Study Population

The participants in this study were from the Pittsburgh Epidemiology of Diabetes Complications (EDC) prospective study of childhood-onset (<17 years) T1D. All participants were listed in the Children's Hospital of Pittsburgh diabetes registry. Potential participants received a letter inviting them to have a physical examination and to complete several questionnaires. There were 979 eligible participants, 658 of whom (67.2%) participated in the entire EDC evaluation process [33,34]. The current study, Genetic Basis of Diabetic Retinopathy, consists of the 496 EDC participants for whom DNA was available (75.4%).

This study adhered to the tenets of the Declaration of Helsinki, and was approved by the institutional review board at the University of Pittsburgh. Written informed consent was obtained from the participants prior to baseline data collection. The data collection methodologies for this population have been published previously [34,35,36]. Data collection relevant to the Genetic Basis of Diabetic Retinopathy included baseline and biennial stereoscopic retinal examinations, measurement of blood pressure (systolic blood pressure, diastolic blood pressure), hypertension (HTN), total cholesterol, high-density lipoprotein (HDL), low-density lipoprotein (LDL), triglycerides, glycosylated hemoglobin (GHb), body mass index (BMI), and documentation of ever-smoker status [34,35].

Determination of PDR

Dilated eye examinations and stereoscopic images of the retinal fundus were obtained for participants at baseline (1986–1988), then biennially, for those without PDR over the course of 10 years, and again at 18 years. ETDRS (Early Treatment Diabetic Retinopathy Study) fundus fields 1, 2 and 4 [37] were taken using the Zeiss Fundus camera (Carl Zeiss, Oberkochen, Germany), and diagnosis and severity grading were based on the assessment of these images by the Fundus Photography Reading Center at the University of Wisconsin using the Arlie House system. Three images were used instead of the standard 7 images since it has been previously validated that this gives acceptable results with good sensitivity and reliability [33,38]. PDR was defined as either grade ≥60 in one eye or grade <60 but with panretinal chorioretinal scars consistent with laser therapy, according to the modified Airlie House system. Baseline PDR was defined as the presence of PDR at the initial evaluation. Incident PDR was defined as PDR first diagnosed at a subsequent biennial follow-up time point.

Genotyping

Based on the European ancestry (CEU) HapMap (NCBI build 35) data (39), we identified 2 tSNPs for the ADORA_2A gene. The 2 tSNPs selected for analyses in our study (rs2236624-C/T and rs4822489-G/T) were the only SNPs in the CEU population that met the selection criteria of minor allele frequency (MAF) of ≥0.2. The tSNPs were genotyped using TaqMan allele discrimination assays (Applied Biosystems, Foster City, Calif., USA) with the ABI Prism® 7000 Sequence Detection System, and the genotype assignments were determined by 2.0 ABI software (Applied Biosystems). The cycling conditions provided by Applied Biosystems yielded poor genotype cluster separation, and an extended protocol was used and resulted in improved separation and genotype discrimination. Conditions were 2 min of activation at 50°C, denaturing at 95°C, then 50 cycles of 15 s of denaturing at 95°C, followed by annealing and extension at 58°C for 1.5 min.

Statistical Analysis

The data were analyzed using SAS version 9.1.3 (SAS Institute Inc., Cary, N.C., USA). Selected covariates (age, BMI, diabetes duration, GHb, HDL, LDL, total cholesterol and triglycerides) found to be significant in the univariate analysis were included in the multivariate analyses. We conducted 3 separate analyses evaluating the association of selected tSNPs with incident PDR, prevalent PDR and all PDR cases combined. Using the more conservative Bonferroni adjustment, we set the levels of significance to be 0.025 and 0.05 for highly and marginally significant p values, respectively. As far as we know this is the first study investigating the potential role of ADORA_2A in the pathophysiology of PDR.

Haploview was used to construct linkage disequilibrium plots. Genetic association analyses of PDR using additive, dominant and recessive genetic models were carried out using Plink version 1.06 [40]. The stepwise Cox proportional hazard backward regression method (significance level of 0.10), as implemented in SAS, was used to identify tSNPs associated with incident PDR while adjusting for significant covariates.

Results

A total of 161 participants (32%) developed incident PDR during the follow-up period (table 1a). There was no significant difference in duration of diabetes in those that developed incident PDR compared to those that did not (p = 0.77). In contrast, participants who developed incident PDR had a higher rate of HTN and higher mean levels of GHb, total cholesterol, LDL, triglycerides and BMI (table 1a). These findings suggest that those who did not develop PDR at the end of the follow-up period had better glycemic control and consequentially developed significantly fewer risk factors for cardiovascular disease than those who developed incident PDR.

Table 1.

Patient characteristics

a Patients who developed incident PDR compared to patients who were PDR free at the end of follow-up
	Incident PDR (n = 161)	No PDR at end of follow-up (n = 183)	p
Age, years	25.8 ± 6.9	24.8 ± 8.1	0.1992
Duration, years	17.0 ± 6.3	16.3 ± 7.1	0.7675
Male/female, n	79/82	92/91	0.9132
GHb, %	10.7 ± 1.9	9.9 ± 1.7	<0.0001
Prevalence of HTN, n	15 (9.3)	7 (3.8)	0.0435
Total cholesterol, mg/dl	190 ± 40.0	173.3 ± 32.5	<0.0001
HDL cholesterol, mg/dl	54.7 ± 12.1	55.3 ± 12.0	0.6155
LDL cholesterol, mg/dl	114.3 ± 31.4	102.2 ± 26.6	0.0004
Triglycerides, mg/dl	106.6 ± 87.0	81.1 ± 50.1	0.0012
BMI	23.7 ± 3.0	22.9 ± 3.4	0.0466
Ever smoker, n	46 (28.8)	61 (34.7)	0.1602

b Patients who had PDR at baseline compared to patients who were PDR free at baseline
	Prevalent PDR (n = 142)	No PDR at baseline (n = 183)	p
Age, years	37.2 ± 5.8	24.8 ± 8.1	<0.0001
Duration, years	25.1 ± 6.1	16.3 ± 7.1	<0.0001
Male/female, n	79/63	92/91	0.4562
GHb, %	10.3 ± 1.8	9.9 ± 1.7	0.0020
Prevalence of HTN, %	35.9	3.8	<0.0001
Total cholesterol, mg/dl	204.9 ± 45.3	173.3 ± 32.5	0.0001
HDL cholesterol, mg/dl	51.7 ± 11.3	55.3 ± 12.0	0.0058
LDL cholesterol, mg/dl	128.9 ± 38.0	102.2 ± 26.6	<0.0001
Triglycerides, mg/dl	120.8 ± 71.2	81.1 ± 50.1	<0.0001
BMI	24.2 ± 3.3	22.9 ± 3.4	0.0051
Ever smoker, n	66 (47.5)	61 (34.7)	0.9456

c All patients who developed PDR (incident and prevalent) compared to patients who were PDR free at the end of follow-up
	Any PDR (n = 306)	No PDR (n = 185)	p
Age, years	29.0 ± 7.3	24.7 ± 7.9	<0.375
Duration, years	20.8 ± 7.8	16.2 ± 7.0	<0.009
Male/female, n	159/147	94/91	0.456
GHb, %	10.5 ± 1.8	9.9 ± 1.7	<0.001
Prevalence of HTN, n	68 (90.7)	7 (9.3)	<0.001
Total cholesterol, mg/dl	204.9 ± 45.3	173.6 ± 32.5	<0.001
HDL cholesterol, mg/dl	53.4 ± 12.1	55.2 ± 11.9	0.145
LDL cholesterol, mg/dl	120.9 ± 35.2	102.5 ± 26.7	<0.001
Triglycerides, mg/dl	113.2 ± 79.5	81.7 ± 50.0	<0.001
BMI	23.9 ± 3.1	22.9 ± 3.4	0.016
Ever smoker, n	113 (64.6)	62 (35.4)	0.434

Values denote means ± SD unless otherwise specified. Figures in parentheses are percentages.

There were significant differences between participants with prevalent PDR and those who had not developed PDR at the time of their baseline evaluation (table 1b). Those with prevalent PDR tended to be older, had a longer duration of T1D and higher levels of GHb; they also had higher levels of other cardiovascular disease risk factors including total cholesterol, LDL, triglycerides and lower levels of HDL. In addition, they were heavier and had a higher prevalence of HTN. The characteristics of those who developed incident or prevalent PDR during the follow-up period versus those who did not develop PDR during the follow-up period are displayed in table 1c and are consistent with the finding reported above.

The rs2236624 tSNP is in Hardy-Weinberg equilibrium (p > 0.05), while the rs4822489 tSNP is not (p = 0.05) and shows some signs of heterozygote advantage: GG = 195, GT = 212 and TT = 84. The minor allele frequency for rs2236624 is 0.21, and for rs4822489 it is 0.39, and the r² between the SNPs is 0.36. In the recessive genetic model, comparing all participants with PDR (i.e. prevalent plus incident cases), carriers homozygous for any T allele of SNP rs2236624 had a significant protective odds ratio (OR) of 0.23 compared to carriers of any C allele (table 2). Interestingly, the protective effect of the homozygote T allele improved (OR of 0.23 vs. 0.18) following adjustment for important covariates (i.e. duration of diabetes, GHb, HTN status and LDL cholesterol (table 2). Age, gender, BMI, ever-smoker status, HDL cholesterol and triglycerides dropped out of the model. Similarly, carriers homozygous for the T allele of SNP rs4822489 had a protective OR of 0.65 compared to carriers of any G allele, and the observed protective effect of the T allele improved after adjusting for covariates (OR = 0.55) (table 2).

Table 2.

Association of ADORA SNPs with all cases of PDR under 3 genetic models

SNP (reference allele)	OR	95% CI	p
rs2236624 (T)
Recessive adjusted	0.18	0.06–0.59	0.004
Dominant adjusted	1.24	0.78–1.95	0.363
Additive adjusted	0.93	0.64–1.34	0.690
rs4822489 (T)
Recessive adjusted	0.55	0.31–0.97	0.040
Dominant adjusted	0.97	0.62–1.49	0.873
Additive adjusted	0.84	0.62–1.13	0.240

A djusted for duration of T1D, LDL, hypertension and GHb.

Modeling Prevalent and Incident PDR and Variants of the ADORA_2A Separately

Among prevalent cases and in a recessive model, carriers homozygous for the T allele (rs2236624) had a protective OR of 0.36 compared to carriers of any C allele (p = 0.04; 95% CI: 0.128–0.991). The observed protective effect was enhanced with covariate adjustment. The additive and dominant models did not reveal any association with prevalent PDR in the univariate or multivariate models for PDR (data not shown). SNP rs4822489 was not associated with prevalent PDR in the additive, dominant or recessive models (data not shown).

There were 162 incident cases of PDR during the 20-year follow-up period. The Cox proportional hazards regression showed that carriers homozygous for the T allele of SNP rs2236624 were less likely to progress to PDR compared to carriers homozygous for the C allele, resulting in a protective hazard ratio (HR) of 0.156 (95% CI: 0.040–0.63; p = 0.009); covariate adjustment significantly improved the observed protective effect from a HR of 0.156 to 0.090 (95% CI: 0.01–0.64; p = 0.016). Kaplan-Meier analysis showed that only 11% of those homozygous for the T allele progressed to PDR during follow-up compared to 48.9% of carriers of any copy of the C allele (fig. 1a). The two-group log rank χ² test showed that the observed difference in time free of PDR (i.e. time to event) was significantly different (p = 0.0027) between carriers homozygous for the T allele and carriers of any copy of the C allele.

Fig. 1.

a Kaplan-Meier curve showing follow-up time for development of PDR in 2.5-year increments by rs2236624 genotype. Blue line: T allele. Red line: C allele. b Kaplan-Meier curve showing follow-up time for development of PDR in 2.5-year increments by rs4822489 genotype. Blue line: T allele. Red line: G allele.

Similar, though not as dramatic, results were observed for rs4822489: carriers homozygous for the T allele of SNP rs4822489 were less likely than carriers of any G allele to progress to PDR during follow-up (HR = 0.67; 95% CI: 0.42–1.07; p = 0.093). Thirty-four percent of those homozygous for the T allele progressed to PDR compared to 49% among carriers of any G allele. The two-group log rank χ² test revealed marginal significance (p = 0.091) (fig. 1b).

Haplotypes were generated from the 2 SNPs with the following frequencies: TC 594 (61.36%), GC 172 (17.77%), TG 1 (0.10%) and GG 201 (20.86%). The TG haplotype was excluded from the analysis as there was only 1 individual with this haplotype. None of the haplotypes were associated with PDR (data not shown).

Discussion

The analyses of this prospective study of persons with T1D demonstrate that variants of the ADORA_2A gene confer significant protection against the development of PDR in persons with T1D. At baseline, the OR of developing PDR was only 0.23 for carriers of 2 copies of the T allele compared to carriers of any copy of the C allele. Remarkably, only 10% of the carriers of 2 copies of the T allele developed incident PDR compared to 50% for carriers of any copy of the C allele. As far as we know, this is the first study to implicate the ADORA_2A gene in PDR in individuals with childhood-onset T1D.

The reason for this decreased association for carriers homozygous for the T allele with PDR is not clear and warrants further investigation. While they appear to be protected against development of PDR, we note that the number of participants homozygous for the T allele was small (n = 24 for rs2236624), and only 2 of the 24 developed PDR during follow-up. For rs4822489, 84 were homozygotes for the protective allele, of which 20 progressed to PDR during follow-up. However, the biologic functions of the ADORA_2A gene, specifically its role in angiogenesis and protection against biologic stressors associated with diabetes, suggest that the OR and HR associated with these data are unlikely to be due solely to small sample sizes of participants homozygous for the T allele.

The ADORA_2A gene is part of the G-protein-coupled receptor super family. It is expressed in the basal ganglia, blood vessels, platelets and other tissues in the body. It encompasses 1 intron and 2 exons [41,42,43,44]. SNP rs2236624 is in linkage disequilibrium with 15 currently known SNPs. Four of these SNPs are within exons but are synonymous coding SNPs, 3 are intronic, 6 lie in the 3′>untranslated region, 1 is a nonsynonymous SNP and another is a frame shift polymorphism [45].

At this stage of the investigation, the reasons for the differences in the phenotypes associated with the 2 tSNPs in our study are not clear and require further exploration. Investigators have reported that ADORA_2A receptor activation may decrease vascular endothelial growth factor (VEGF) and endothelial cell production in rodent pheochromocytoma cells [46]. Similar modulation of VEGF may occur in humans and, thus, the possibility of influencing the development and/or the progression of PDR.

Myriad physiological processes may be affected by adenosine and its receptors in vivo since it is found in the vasculature of all organs and is a byproduct of adenosine triphosphate metabolism. Variation in ADORA_2A expression and the ability of the receptor to bind with adenosine may have an impact on its ability to attenuate the negative effects of OS and other biologic homeostatic processes in ways that have yet to be elucidated. To date adenosine has been linked to retinal VEGF production [32,47] and regulation of angiogenesis via VEGF production [32,48]. It is also suspected to play a role in interleukin-8, mast cell and insulin-like-growth-factor-induced angiogenesis [49].

ADORA_2A stimulation has been linked to psychomotor depression, sleep induction, immune suppression and vasodilatation, while its suppression has been linked to HTN, aggression and inflammation, alleviation of symptoms of ethanol withdrawal and amelioration of neurotoxicity [44]. There is evidence that ADORA_2A activation decreases the expression of VEGF, a primary mitogen associated with the development of DR [46]. Experimental models have also shown that the ADORA_2A receptor plays a role in glucose transport, vasodilatation, hypoxia, and inflammation, resolution of inflammation and prevention of apoptosis [13,16,21,50,51].

This study evaluated only the ADORA_2A gene, which is one of 4 members of the adenosine receptor family. ADORA₁ plays a significant role in the proinflammatory responses of the cell, while ADORA_2A and ADORA_2B play significant roles in limiting the cell response to inflammation, thereby producing an antiinflammatory response. ADORA₃ plays a role in the cell's response to ischemia [13]. The potential role of the remaining 3 genes of the adenosine receptor family in PDR susceptibility deserves further investigation.

The majority (98%) of the participants enrolled in this study of T1D were of European ancestry. As a result, the generalizability of these findings awaits confirmation in other ancestral population groups. The fact that the mean diabetes duration was 19 years at the time of entry into the study limited our ability to explore the natural history of DR in reference to specific genotypes. However, because 85% of the cohort had some form of DR at the time of entry into the study, we were able to explore the effect of genetic variation on the progression to PDR. It would be beneficial to explore the pathophysiologic role of the ADORA_2A gene in patients with T2D.

Replication of our observed associations in other cohorts, fully evaluating this segment of the gene for underlying variability, as well as functional characterization of this variability regarding receptor function or mRNA stability are potential areas of further investigation. Based on the HapMap populations, there is a high frequency of the T allele in populations of Asian ancestry (29.8% in Han Chinese from Beijing China, CHB, and 22.1% in Japanese from Tokyo, JPT) and Western and Northern European ancestry (CEU; 23.5%). In contrast, it is far less common in the HapMap Yoruba sample from Ibadan, Nigeria (YRI; 0.9%). These observations suggest that a more comprehensive documentation of the global distribution of the genetic variation of the ADORA_2A gene may contribute to the understanding of how this gene influences differential susceptibility to visual complications of diabetes, particularly the growing evidence that DR is less prevalent in West African individuals with diabetes than in Caucasians [52].

In summary, findings from this study should spark interest in all of the AR as potential modifiers in diabetes complications. If replicated, this association of ADORA_2A with PDR could lead to additional studies of ADORA_2A and its role in humans with DR and other microvascular complications of diabetes. Illustratively, the potential therapeutic effects of adenosine in ischemic retinal diseases have been suggested [22].