Introduction

Chronic complications are the major causes of morbidity and mortality for patients with type 2 diabetes. As one of the most common chronic microvascular complications, diabetic retinopathy (DR) is a leading cause of blindness in working-age adults globally1. The overall prevalence of DR is 35.4% and is higher in patients with type 1 diabetes compared to those with type 2 diabetes (77.3% vs 25.2%)2. The prevalence of DR among patients with diabetes in China varies from 11.9% to 43.1%3,4,5. Given that China has the most patients suffering from diabetes around the world6, the number of patients with DR could be quite large, highlighting what is certainly a huge public health and financial burden for the country.

The aetiology of DR is complex and remains to be fully elucidated. Large-scale prospective studies have shown that the durations of diabetes, hyperglycaemia and hypertension are the most clinically important risk factors for DR7, 8. Furthermore, multiple studies have suggested that genetic factors also play important roles in the development of DR with an estimated heritability of 25% for DR and 50% for proliferative diabetic retinopathy (PDR)9,10,11. Therefore, the elucidation of genetic susceptibility factors is helpful for revealing the pathogenesis of DR. To date, several genome-wide association studies (GWAS) in populations of different ancestries have identified some potential susceptibility loci for DR12, 13. However, only one locus, rs9896052 near GRB2, showed an association that reached the genome-wide significance level for sight-threatening DR13, and few loci have been replicated in other studies14, 15. Numerous studies have attempted to identify susceptibility loci for DR through a candidate gene approach. Multiple genes, such as VEGF, PPARG, EPO, AKR1B1, PPKCB, ACE and ICAM-1, have been suggested to be associated with DR14. However, few of these studies have been consistently replicated14, 16, 17. Recently, a new approach, whole exome sequencing, have been applied for the identification for potential susceptibility loci, and some new candidate genes for DR or PDR were reported18, 19.

As a complex genetic disorder, about 90 susceptibility loci have been identified for type 2 diabetes through GWAS to date20. Recently, a study based the Singapore Epidemiology of Eye Diseases Study indicated that participants with more type 2 diabetes genetic risk alleles had higher risk of DR21. There are also several studies that investigated the association between type 2 diabetes susceptibility genes and DR. TCF7L2 rs7903146 was reported to be associated with PDR in Caucasian patients with type 2 diabetes22, and KCNJ11 rs5219 was reported to be associated with DR in a Chinese population with type 2 diabetes23. However, most of those studies included only one locus or a few loci, and the sample size was relative small. Associations between most of the type 2 diabetes susceptibility loci and DR have not been investigated, especially in Chinese population. So, we performed the present study to investigate the effects of over 80 type 2 diabetes susceptibility loci on DR in a Chinese population with type 2 diabetes.

Results

Associations between SNPs and DR in stage 1

We first analysed the effects of these SNPs on DR in stage 1 samples. DR and diabetic kidney disease (DKD) are two important microvascular diabetes complications with a high concordance rate in patients with diabetes, and DR and DKD might share common pathogenesis. Because of the close relationship between DKD and DR, DKD could be a considerably important confounding factor when we conduct the genetic association analysis for DR. To minimize the confounding influence of DKD on the effects of SNPs on DR, we divided the participants into four groups. These four groups formed two small case-control studies for DR according to the status of DKD: patients with DR only vs control patients without DR or DKD, patients with both DR and DKD vs patients with DKD only. In both case-control studies, association between SNPs and DR was examined. A meta-analysis was done to combine the results. The distributions of SNPs among these four groups were shown in Supplementary Table 1. As shown in Table 1, with adjustment for diabetes duration, HbA1c, blood pressure and body mass index (BMI), five loci (ADAMTS9-AS2 rs4607103, WFS1 rs10010131, CDKAL1 rs7756992, VPS26A rs1802295 and IDE-KIF11-HHEX rs1111875) were significantly associated with DR, with rs7756992 showing the strongest association (OR 0.746, 95% CI 0.608–0.915, P = 0.0048 for the rs4607103 T allele; OR 1.629, 95% CI 1.019–2.606, P = 0.0416 for the rs10010131 A allele; OR 0.703, 95% CI 0.580–0.851, P = 0.0003 for the rs7756992 A allele; OR 0.673, 95% CI 0.483–0.939, P = 0.0197 for the rs1802295 T allele; and OR 0.808, 95% CI 0.656–0.995, P = 0.0448 for the rs1111875 C allele). However, on the basis of 164 independent tests, only the association between CDKAL1 rs7756992 and DR remained significant after Bonferroni correction for multiple comparisons (corrected P = 0.0492 for rs7756992).

Table 1 Effects of the SNPs on DR in stage 1 samples.
Full size table

Validation of the effect of CDKAL1 rs7756992 on DR in stage 2

To further validate the effect of CDKAL1 rs7756992 on DR, we genotyped this SNP in two independent cohorts in stage 2. As shown in Table 2, in Cohort (1), with adjustment for diabetes duration, HbA1c, blood pressure and BMI, We found that rs7756992 showed similar trend as those in stage 1 samples for DR (OR 0.890, 95% CI 0.728–1.088, P = 0.26). In Cohort (2), rs7756992 showed a marginal association with DR (OR 0.874, 95% CI 0.749–1.020, P = 0.09) following adjustment for confounders. Then we conducted a meta-analysis with the fixed-effect model (P for homogeneity = 0.23), rs7756992 was significantly associated with DR, with the minor allele A conferring a lower risk of DR (OR 0.824, 95% CI 0.743–0.914, P = 2.46 × 10−4).

Table 2 Association of CDKAL1 rs7756992 with DR in Chinese patients with type 2 diabetes.
Full size table

Effects of SNPs on DR severity

Further, we tried to examine the effect of CDKAL1 rs7756992 on the disease severity of DR. From our sample population, there were 2,199 patients without DR, 709 with mild NPDR, 396 with moderate NPDR, 267 with severe NPDR, 116 with PDR and 31 patients with DR that lacked a severity assessment. However, we did not find any associations between this SNP and the severity of DR (Supplementary Table 2).

Discussion

In this study, we analysed the effects of 82 susceptibility loci for type 2 diabetes on DR in a Chinese population with type 2 diabetes. A total of 3,718 participants were recruited. On the basis of an estimated effect size of genetic loci for DR (~1.2), our samples had >90% power to detect a SNP effect with a minor allele frequency (MAF) of 0.25 and >80% power to detect a SNP effect with a MAF of 0.15 at a level of significance of 0.05. With a two-stage design, we found that CDKAL1 rs7756992 were significantly associated with DR (OR 0.824, P = 2.47 × 10−4), and the association between rs7756992 and DR remained significant after Bonferroni correction. To our knowledge, this study is the first to identify an association between this SNP and DR. Besides, we found another four loci that showed association with DR in stage 1. Although, the association between these loci and DR could not survive after Bonferroni correction, these loci might be potential candidate genes for DR and could be further investigated.

Numerous genetic studies have found that several SNPs within the CDKAL1 region are associated with type 2 diabetes among multiple ethnic populations24. rs7756992 was one of the most commonly reported SNPs in CDKAL1, with the major G allele conferring a higher risk of type 2 diabetes24. It has also been reported to be associated with impaired insulin secretion25, 26. In this study, we found that rs7756992 was associated with DR, with the minor A allele conferring a lower risk of DR. So the major G allele was the risk allele for DR too. Liu et al.27 reported that another SNP in CDKAL1, rs10946398 which was also reported to be associated with type 2 diabetes in multiple populations24, was associated with DR in 580 Chinese patients with type 2 diabetes. Mice with a beta cell-specific knockout of Cdkal1 presented decreased insulin secretion and impaired blood glucose control28. However, although some study indicated that rs7756992 was correlated with CDKAL1 protein level, the underlying mechanism has not yet been elucidated29. And functional study which investigates the role of CDKAL1 in the pathophysiological process of retinopathy has not been reported yet. Thus, how CDKAL1 rs7756992 impact the susceptibility of DR and whether it’s the causal locus of DR or not requires further investigation.

This study has some limitations. First, because the participants of this study were recruited from Shanghai and nearby regions, our findings may be specific to Chinese patients and may have inherent bias. This may explain why we did not find association between TCF7L2 rs7903146, KCNJ11 rs5219 and DR in this study. Second, lifestyle factors such as cigarette smoking and alcohol consumption were not included in the genotype-disease analysis. Whether interactions exist between lifestyle factors and these genetic variants in terms of DR remains unknown. Third, rs7756992 which was associated with DR in this study is located in the intron of CDKAL1. Thus, the relationships between rs7756992 and genes and how they modulate DR risk are largely unknown. Hence, studies in other ethnic populations are needed to further replicate our findings, and causal loci and genes should be identified to elucidate the underlying mechanism.

In summary, we identified CDKAL1 rs7756992 as a susceptibility locus for DR in a Chinese population with type 2 diabetes. Further studies are needed to replicate this finding and to elucidate the underlying mechanism.

Methods

Participants

A two-stage approach was applied for this study. In stage 1, we recruited 1,251 patients with type 2 diabetes from the Shanghai Diabetes Institute Inpatient Database of Shanghai Jiao Tong University Affiliated Sixth People’s Hospital15, 30. These patients included 313 patients with DR but no DKD, 419 patients with DKD but no DR, 281 patients with both DR and DKD, and 238 control subjects with diabetes for ≥10 years but without DR or DKD. In stage 2, two independent cohorts were recruited for replication analysis. Cohort (1) recruited a total of 993 patients with type 2 diabetes from the Shanghai Diabetic Complications Study and Shanghai Diabetes Institute Inpatient Database15, 31, and consisted of 380 patients with DR and 613 patients with diabetes for ≥5 years but without DR. Cohort (2) recruited a total of 1,474 patients with type 2 diabetes from the Shanghai Diabetes Institute Inpatient Database, including 545 patients with DR and 929 patients with diabetes for ≥5 years but without DR. The basic characteristics of the participants are shown in Tables 3 and 4.

Table 3 Clinical characteristics of the participants in stage 1.
Full size table
Table 4 Clinical characteristics of the participants in stage 2.
Full size table

This study was approved by the Institutional Review Board of Shanghai Jiao Tong University Affiliated Sixth People’s Hospital. All experiments were performed in accordance with relevant guidelines and regulations. Written informed consent was obtained from each participant.

Clinical assessment

A diagnosis of type 2 diabetes was based on the World Health Organization guidelines (1999)32. DR was diagnosed by fundus photography or a history of panretinal photocoagulation (scatter laser) treatment. Fundus photography was performed with a 45-degree 6.3-megapixel digital nonmydriatic camera (Canon CR6-45NM, Lake Success, NY, USA) according to a standardised protocol at the Department of Ophthalmology, Shanghai Jiao Tong University Affiliated Sixth People’s Hospital. Retinopathy was graded according to the International Classification of Diabetic Retinopathy as follows: mild non-proliferative DR (NPDR), moderate NPDR, severe NPDR, or PDR33. For each patient, both eyes were examined, and the more severely affected eye was used to classify the DR. For the definition of DR, a subject with any DR was considered as a DR case. The 24-h albumin excretion rate (AER) and estimated glomerular filtration rate (eGFR) were applied to assess DKD. The AER was measured over 3 consecutive days, and the mean value was recorded for each patient. eGFR was calculated using a formula developed by the Modification of Diet in Renal Disease study group with adjustment for Chinese ethnicity: 186 × (serum creatinine in mmol/L × 0.0113) – 1.154 × (age in years) – 0.203 × (0.742 if female) × (1.233 if Chinese)34. Patients with AER ≥ 30 mg/24 h or eGFR < 90 mL/min/1.73 m2 were diagnosed with DKD. Besides, patients with history of other renal diseases were excluded in the enrollment of study participants. Glycaemic control was evaluated by measuring haemoglobin A1c (HbA1c) levels. Anthropometric parameters, blood pressure and lipid profile data were also collected for each participant.

Single-nucleotide polymorphism (SNP) selection and genotyping

In stage 1, 88 SNPs which were previously reported to be associated with type 2 diabetes by GWAS or large-scale association studies were genotyped, including rs340874, rs780094, rs243021, rs998451, rs7593730, rs3923113, rs13389219, rs16856187, rs7578326, rs2943641, rs7612463, rs831571, rs11708067, rs4402960, rs16861329, rs6815464, rs459193, rs4457053, rs9470794, rs1535500, rs2191349, rs1799884, rs917793, rs6467136, rs10229583, rs791595, rs972283, rs516946, rs515071, rs896854, rs7041847, rs17584499, rs13292136, rs2796441, rs11787792, rs10906115, rs1802295, rs12571751, rs10886471, rs231362, rs2237892, rs1552224, rs10751301, rs1387153, rs10842994, rs1531343, rs7957197, rs9552911, rs1359790, rs7403531, rs7172432, rs1436955, rs7178572, rs7177055, rs11634397, rs2028299, rs8042680, rs7202877, rs17797882, rs16955379, rs391300, rs312457, rs13342232, rs4430796, rs12970134, rs10401969, rs3786897, rs6017317, rs4812829, rs5945326, rs864745, rs10811661, rs2641348, rs7578597, rs1801282, rs4607103, rs7651090, rs10010131, rs7756992, rs10946398, rs13266634, rs12779790, rs1111875, rs7903146, rs5219, rs7961581, rs8050136, and rs1092393120, 35. The SNPs that showed associations with DR were genotyped in stage 2 samples. All of the SNPs were genotyped using a MassARRAY iPLEX system (MassARRAY Compact Analyser, Sequenom, San Diego, CA, USA). The genotyping data underwent a series of quality control checks. The concordance rate based on 100 duplicates was greater than 99% for all SNPs. Sample call rate was greater than 85% for all samples. The Hardy–Weinberg equilibrium test was performed before the statistical analysis (a two-tailed P value < 0.01 was considered statistically significant), and rs1552224 and rs10946398 were excluded. rs780094 was exclude due to call rate <90%. Another 3 SNPs (rs7957197, rs998451 and rs11708067) were rare in our population (minor allele frequency < 0.0015) and were excluded from statistical analyses.

Statistical analysis

Logistic regression was used to compare the difference in genotype distributions between patients with or without DR under an additive model with adjustment for confounders using PLINK (v1.07)36; odds ratios (ORs) with 95% confidence intervals (CIs) are presented with reference to the minor allele. Combined ORs from different studies were calculated using a Comprehensive Meta-Analysis (v2.2.057) with a fixed- or random-effect model after testing for heterogeneity. The test for homogeneity was assessed with the Cochran Q test. The effects of SNPs on the level of DR severity were analysed by trend analysis using SAS 9.3 (SAS Institute, Cary, NC, USA). Bonferroni correction was applied to adjust for multiple comparisons. Statistical significance was defined as a two-tailed P value < 0.05.