A Genome-Wide Linkage Scan for Diabetic Retinopathy Susceptibility Genes in Mexican Americans With Type 2 Diabetes From Starr County, Texas

Mexican-American families from Starr County, Texas, having two or more siblings with type 2 diabetes were eligible for the study. Diabetes classification was based on earlier National Diabetes Data Group guidelines (1979), wherein individuals currently treated for diabetes, having fasting glucose ≥140 mg/dl on more than one occasion, or having an abnormal glucose tolerance test were considered to have diabetes. A diagnosis of type 2 diabetes was excluded if age at diagnosis was <30 years, BMI was <30 kg/m², and insulin had been used continuously since diagnosis. Subjects were enrolled through the Family Blood Pressure Program, as previously described (22).

The total number of markers typed was 360, covering the 22 autosomes at an average spacing of 9.38 cM (SD = 4.13). The minimum distance between any two markers was 0.55 cM, on chromosome 18; the maximum distance was 32.97 cM, on chromosome 14.

Stereoscopic color fundus photographs of seven standard fields of each eye were scored using the Early Treatment Diabetic Retinopathy Study adaptation of the modified Airlie House classification system (23), as described previously (8). Diabetic retinopathy was classified as: none, early nonproliferative, moderate-to-severe nonproliferative, or proliferative.

Linkage analyses were performed with GeneHunter Plus, using the linear model and S_pairs, the number of pairs of alleles shared identical-by-descent by affected pedigree members (24). In addition, we used ordered-subset analysis (OSA), as implemented in the OSA program (25), to look for homogeneous subsets of families with maximal evidence of linkage to a given chromosomal region. The mean age of type 2 diabetes diagnosis within families was used to rank families for OSA. Analyses were also conducted with families ordered by mean diabetes duration, calculated as the difference between the age at diagnosis with type 2 diabetes and the age when examined for retinopathy. However, because actual ages of onset of neither diabetes nor retinopathy could be determined precisely, using diabetes duration estimated from ages of diabetes diagnosis and examination for retinopathy would compound the degree of uncertainty in the analyses. Therefore, we report results for the OSAs using age of diabetes diagnosis. The difference between the maximum LOD score in the subset of families identified by OSA and the LOD score at the same position in the full set of families was evaluated by permutation testing; a significant P value denotes a maximum OSA LOD score significantly greater than the corresponding unconditional LOD score (25). In separate analyses, individuals were considered affected if they had any diabetic retinopathy (early nonproliferative, moderate-to-severe nonproliferative, or proliferative) or more severe diabetic retinopathy (moderate-to-severe nonproliferative or proliferative).

In comparing covariates among retinopathy classes, logistic regression with generalized estimating equations was used to account for the correlations among siblings. These analyses were performed using SAS version 8 (SAS Institute, Cary, NC).

Subjects were drawn from 415 sibships identified during a study of the genetics of type 2 diabetes. No subjects were available for retinopathy examinations from 22 families (5.1%). The remaining 393 families contained from 1 to 11 offspring, with 391 (99.5%) having at least 2. A total of 794 individuals were examined for retinopathy, of whom 567 (71.4%) were affected; genotype data were available for another 791 sibship members who were not examined for retinopathy. In 100 families (25.4%), only one individual was available to be examined for retinopathy. Of the 293 families in which at least two members were examined, 16 (5.5%) had no members affected with retinopathy, 100 (34.1%) had one member affected, 143 (48.8%) had two members affected, 27 (9.2%) had three members affected, 3 (1.0%) had four members affected, and 4 (1.4%) had five members affected, yielding 282 affected pairs from 177 families. For more severe retinopathy (moderate-to-severe nonproliferative or proliferative diabetic retinopathy), 74 affected pairs from 52 families were available, whereas for proliferative diabetic retinopathy, only 8 affected pairs from 8 families were available (supplementary Table 1, which can be found in an online appendix [available at http://dx.doi.org/10.2337/db06-1373]). In our sample, families containing only pairs discordant for retinopathy contributed no linkage information.

Table 1 shows characteristics of the subjects according to retinopathy grade. Proportionately fewer male subjects than expected had no retinopathy, whereas more male subjects than expected had severe nonproliferative retinopathy. Observed numbers of male and female subjects with proliferative retinopathy (36 and 52, respectively) closely matched expected numbers (34 and 54, respectively). Systolic blood pressure and plasma cholesterol levels tended to increase with retinopathy severity, whereas BMI tended to decrease, and was highest among those with no retinopathy. Fasting blood glucose, A1C, and plasma triglyceride levels were highest in subjects with moderate-to-severe nonproliferative retinopathy. As expected, there was a clear positive relationship between duration of diabetes and the presence and severity of retinopathy. Subjects with more severe retinopathy, either proliferative or nonproliferative, were diagnosed with diabetes at markedly younger ages (46 years), on average, than those with less severe retinopathy (50 years) or no retinopathy (52 years). As expected, age at diabetes diagnosis and duration of the disease at the time of retinopathy examination were negatively correlated (ρ = −0.40, P < 0.0001).

Figure 1 shows results for the unconditional linkage analyses for retinopathy of any severity, whereas Table 2 shows all unconditional LOD scores >1.00 for any retinopathy and for more severe retinopathy (moderate-to-severe nonproliferative or proliferative diabetic retinopathy). For retinopathy of any severity, the highest LOD scores occurred on chromosomes 3 (LOD score of 2.41 at 117.0 cM) and 12 (2.47 at 15.5); no other unconditional LOD scores exceeded 1.24 (chromosome 1, 45.3 cM). Although the LOD scores on chromosomes 3 and 12 did not reach the level of chromosome-wide significance, they do provide suggestive evidence of linkage (26).

Table 3 shows results of the OSAs with families ranked according to average age of type 2 diabetes diagnosis. For retinopathy of any degree, subsets of families yielded significantly increased LOD scores on chromosomes 12, 15, 18, and 20. For chromosomes 12, 15, and 20, maximum LOD scores occurred with families ranked from highest to lowest average age of diabetes diagnosis. On chromosome 12, the maximum OSA LOD score was 4.47 at 13.2 cM, compared with an unconditional LOD score of 2.47 (P = 0.018). On chromosome 15, the maximum OSA LOD score was 3.65 at 100.6 cM, compared with an unconditional LOD score of 0.99 (P = 0.030). On chromosome 20, the maximum OSA LOD score was 2.67 at 54.1 cM, compared with an unconditional LOD score of 0.00 (P = 0.004). On chromosome 18, the peak OSA LOD score was obtained with families ranked from lowest to highest average age of diabetes diagnosis (OSA LOD = 1.90 at 99.0 cM, unconditional LOD = 0.06, P = 0.033).

The maximum unconditional LOD score for more severe retinopathy was 1.40 on chromosome 3 at 117.0 cM, the same location at which a LOD score of 2.41 was obtained for retinopathy of any degree. Another peak unconditional LOD score above 1.0 (1.29) for more severe retinopathy occurred on chromosome 3 at 9.4 cM. LOD scores >1.0 for more severe retinopathy also occurred at the distal end of chromosome 2 (1.11 at 260.6 cM) and on chromosome 12 (1.03 at 100.5 cM). OSAs of more severe retinopathy, again ranking families by average age of type 2 diabetes diagnosis, yielded significantly increased LOD scores on chromosomes 5 (OSA LOD of 2.53 at 11.2 cM vs. unconditional LOD of 0.15, P = 0.013), 6 (2.28 at 30.6 vs. 0.62, P = 0.041), and 19 (2.21 at 100.6 vs. 0.28, P = 0.037).

Although evidence has been accumulating that genetic factors can influence either the occurrence or severity of diabetic retinopathy, it may be difficult to separate the genetics of diabetes from the genetics of its complications. However, it is important that we make the effort because although treating the underlying disease may ameliorate its complications, treating the complications may be just as important in reducing the personal, social, and economic burdens of the disease. Understanding the genetic factors that either contribute to the development of retinopathy or increase its severity may allow us to move toward treatment of the underlying biology of the condition, rather than relying on palliative treatments, such as laser photocoagulation, that are aimed at its symptoms.

Our study represents a step toward this goal. Our results are consistent with evidence from earlier studies of familial aggregation (5–8) and with prior linkage (21) and association (9) studies, suggesting that genes that influence the risk of diabetic retinopathy exist. Furthermore, our findings provide evidence that such genes may be distinguishable from those that influence the risk of diabetes itself: none of the LOD scores >2.0 reported here for retinopathy coincided with any LOD score peaks for type 2 diabetes in this same population (data not shown). This may have implications for all diabetic retinopathy, if the same genes that influence risk of retinopathy in type 2 diabetes also affect retinopathy associated with other forms of diabetes, such as type 1 diabetes and maturity-onset diabetes of the young.

Even though our study represents the largest linkage analysis of diabetic retinopathy yet reported, none of the unconditional LOD scores reached the level of genome-wide significance (26). Nonetheless, our findings are strongly suggestive of linkage between retinopathy and several chromosomal regions, particularly in view of the presence of several strong candidate genes in chromosomal regions implicated by our analyses (Table 4; citations refer to supplementary Table 2). Considering both unconditional and OSAs, the strongest evidence of linkage with retinopathy involved the proximal end of chromosome 12. The unconditional LOD score for any retinopathy of 2.47 at 15.5 cM was the highest unconditional score on any chromosome, whereas the peak LOD score of 4.47 at 13.2 cM was the highest score for any of the OSAs; the region under these peaks encompasses ∼24 cM. As shown in Table 4, several candidate genes potentially involved with either type 2 diabetes, diabetic retinopathy, or both occurred in this region, including WNT5B, TULP3, and GNB3. At least two genes associated with hypertension, WNK1 and SCNN1A, also occurred in this region. The OLR1 gene has been associated with hypertensive vascular damage and may be involved in choroidal neovascularization in age-related macular degeneration, suggesting that it could play a similar role in proliferative diabetic retinopathy.

The second-highest unconditional LOD score for any retinopathy (2.41) occurred on chromosome 3 at 117 cM, with a one-LOD support region covering nearly 21 cM. No significantly higher score in this region was found with OSA. Within the core of this region, at least two genes known to be involved in retinal diseases occur, PROS1 and ARL6, as well as two others, ROBO2 and IMPG2, that are involved in retinal development or function. Another gene involved in retinal development, MITF, is proximal to the core region, whereas GUCA1C, involved in retinal photoreceptor activity, is distal to it. It should be noted that this region differs from the region on chromosome 3 that produced a peak LOD score of 1.36 for retinopathy in Pima Indians (21).

Because age at diabetes diagnosis is likely to be inversely correlated with the duration of the disease in those examined for retinopathy, maximum LOD scores obtained when ranking families from higher to lower age at diagnosis may implicate variants that contribute to more rapid development of retinopathy. Conversely, maximum LOD scores obtained when ranking families from lower to higher mean ages of diabetes diagnosis may implicate variants that are associated with a milder course of development of retinal damage.

On chromosome 15, OSA yielded a maximum LOD score of 3.65 at 100.6 cM (P = 0.031); the highest unconditional LOD score in this region was 1.16 at 108.3 cM. IGF1R and RGMA are among the potential candidates involved in retinal biology that occur in or near this region. The IDDM3 gene is near this region, although the gene itself has not yet been identified (27).

On chromosome 20, OSA for any retinopathy produced a peak LOD score of 2.67 at 54.1 cM, significantly different (P = 0.004) from the unconditional LOD score of 0 at this position. Several genes in the region are known to be involved in retinal biology or retinal disease, including KCNS1 and the transglutaminases TGM2 and TGM3. Several other genes in the region may be associated with various aspects of insulin resistance or type 2 diabetes, including E2F1, ASIP, and HNF4A, the gene for type 1 maturity-onset diabetes of the young. The putative NIDDM3 locus is distal to the implicated region (28–31).

The OSAs for more severe retinopathy produced LOD scores >2.00 on chromosomes 5 (2.53 at 11.2 cM), 6 (2.22 at 104.7), and 19 (2.21 at 100.6). Within the implicated region of chromosome 5, we identified no obvious candidate genes for retinopathy, although several (GDNF, MCDR3, and SLC1A3) occur somewhat distal to this region (32–34). On chromosome 6, candidates in the implicated region include EDN1 and GMNN. On chromosome 19, the peak occurred at the distal end of the chromosome. Within 20 cM of the end, however, is PRPF31, the gene for autosomal dominant retinitis pigmentosa type 11; also in this region is FIZ1, which interacts with NRL, a gene involved in another form of retinitis pigmentosa.

That the strongest evidence for genetic linkage was found in analyses of retinopathy of any severity is interesting because previous analyses in this population indicated that more severe retinopathy, but not retinopathy per se, showed familial aggregation (8). These seemingly contradictory findings may be attributable to the low statistical power of nonparametric linkage analyses relative to that of the association analyses used to assess familial aggregation (35); the set of subjects having any degree of retinopathy is much larger than the subsets with either more severe nonproliferative or proliferative retinopathy, or proliferative retinopathy alone. Also, there is a distinction between testing whether a phenotype shows familial aggregation overall, and testing whether certain markers are shared more frequently by individuals with a given phenotype. Heritability estimates for quantitative traits have been found to show little correlation with measures of association in genomic scans (36), suggesting that measures of familial aggregation of disease could well be poorly correlated with measures of linkage. In addition, the very high prevalence of retinopathy in our family sample (∼70%) may help obscure evidence of familial aggregation of overall retinopathy, but have less effect on tests of marker sharing, especially among more homogeneous subsets of families.

In summary, both unconditional and ordered-subset linkage analysis identified several regions possibly harboring retinopathy susceptibility loci. Strong candidate genes, many of them specific to the retina or associated with other retinal pathology, were identified in most of these regions, strengthening the case that genes can affect susceptibility to diabetic retinopathy. Such genes, however, may have pathological effects in the retina only when the underlying pathology of diabetes is present. Inasmuch as most morbidity and mortality from diabetes are attributable to its complications, analyses that focus on the complications of diabetes, as here, may be an important adjunct to studies of the genetics of diabetes.

This study was supported in part by funding from the National Eye Institute (EY12386) and the National Heart Lung and Blood Institute (HL54504 and HL054481).

We express our appreciation to Hilda Guerra and the staff in Starr County, who collected data, and to those who have kindly participated in these studies.