An intergenic region on chromosome 13q33.3 is associated with the susceptibility to kidney disease in type 1 and 2 diabetes

Abstract

A genome-wide association (GWA) scan of the Genetics of Kidneys in Diabetes (GoKinD) collections identified four novel susceptibility loci, located on chromosomes 7p14.3, 9q21.32, 11p15.4, and 13q33.3 that were associated with nephropathy in type 1 diabetes. The recent examination of these loci in Japanese patients with type 2 diabetes further supported associations at the chromosome 13q33.3 locus. To follow up these findings, we focused on these same four loci and examined whether single nucleotide polymorphisms (SNPs) at these susceptibility loci were associated with diabetic nephropathy in the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes collection. A total of six SNPs across these loci were genotyped in 646 normoalbuminuric controls and 743 nephropathy cases of European ancestry. A significant association was identified at the 13q33.3 locus (rs9521445: OR=1.25, P=4.4×10⁻³). At this same locus, rs1411766 was also associated with type 2 diabetic nephropathy in this collection (OR=1.19, P=0.03). A meta-analysis combining this data with that from the Japanese and GoKinD collections significantly improved the strength of this association (OR=1.29 P=9.7×10⁻⁹). Additionally, we also observed an association at the 11p15.4 locus (rs451041: OR=1.21, P=0.02). Our analysis increases support that associations identified in the GoKinD collections on chromosomes 11p15.4 (near the CARS gene) and 13q33.3 (within an intergenic region between MYO16 and IRS2) are true diabetic nephropathy susceptibility loci common to both type 1 and type 2 diabetes.

INTRODUCTION

Diabetic nephropathy affects approximately 30% to 40% of all patients with either type 1 or type 2 diabetes and remains the leading cause of end-stage renal disease in the United States.^1–3 In type 2 diabetes, in particular, the burden of this devastating complication continues to grow world-wide. Yet, despite evidence that heritability contributes to its increased susceptibility, the genetic factors underlying its susceptibility remain largely unknown.^4–9

In a recent genome-wide association (GWA) scan of participants from two independent collections of the Genetics of Kidneys in Diabetes (GoKinD) study, we identified four distinct chromosomal regions that were associated with advanced nephropathy (proteinuria and end-stage renal disease) in type 1 diabetes.¹⁰ Although none of the single nucleotide polymorphisms (SNPs) identified in GoKinD achieved genome-wide significance (P<5.0×10⁻⁸), moderate levels of significance (P<1×10⁻⁵) occurred on chromosome 9q21.32 near the FERM domain containing 3 (FRMD3) gene, on chromosome 7p14.3 at the beta chimerin isoform 2/serine carboxypeptidase vitellogenic-like (CHN2/CPVL) locus, on chromosome 11p15.4 at the cysteinyl-tRNA synthetase (CARS) gene, and on chromosome 13q33.3 at an intergenic region telomeric to the myosin heavy chain Myr 8 (MYO16) gene and centromeric to the insulin receptor substrate 2 (IRS2) gene.

Evidence of association at the 9q21.32, 11p15.4, and 13q33.3 loci with severe nephropathy was also observed among Caucasian participants with more than 19,000 person-years of follow-up from the Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications (DCCT/EDIC) study.¹⁰ Replication in this prospective cohort study significantly bolsters the importance of these findings and remains the strongest evidence to date in support of the associations at these loci. Additionally, further evidence implicating the 13q33.3 locus in the risk of nephropathy was gained from a subsequent study that identified a chromosomal region at the homologous region on mouse chromosome 8 that was strongly associated with increased albuminuria in mice.¹¹

More recently, examination of SNPs at the 7p14.3, 9q21.32, 11p15.4, and 13q33.3 loci in a large collection of Japanese patients with type 2 diabetes provided additional support of the association at chromosome 13q33.3; suggesting that this region may harbor a susceptibility locus common in both type 1 and type 2 diabetic nephropathy.¹² Given the current lack of GWA data for diabetic nephropathy in type 2 diabetes, in the present study, we performed a focused evaluation of the same four loci followed-up in this type 2 cohort to investigate whether genetic variants at these loci are associated with diabetic nephropathy in the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes collection.

RESULTS

A total of 1,389 patients of European ancestry from the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes, including 743 patients with diabetic nephropathy (427 cases with microalbuminuria and 316 cases with proteinuria, of whom 59 progressed to end-stage renal disease) and 646 controls with no clinical evidence of kidney complications (persistent normoalbuminuria) were included in the present study. Baseline clinical characteristics for these patients are provided in Table 1.

Table 1.

Baseline clinical characteristics of the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes collection

Clinical characteristic	Joslin Study of Genetics of Nephropathy in Type 2 Diabetes collection
	Controls	Micro-albuminuria Cases	Proteinuria/ESRD Cases
n	646	427	316
Men/Women	320/326	302/125	228/88
Age at type 2 diabetes diagnosis (years)	44.8±8.7	44.0±8.2	42.9±8.6
Duration of type 2 diabetes (years)	14.9±7.7	16.9±8.4	18.0±8.5
Age at examination (years)	59.7±7.7	60.9±8.8	60.9±7.6
HbA1c (%)	7.81.2	7.9±1.4	8.2±1.7
BMI (kg/m2)	30.8±6.7	32.9±7.3	33.3±6.8
Cases with proteinuria/end-stage renal disease	—	—	257/59
ACR (μg/mg, median (25th and 75th percentiles))	6.8 (4.2, 11.4)	40 (31, 95)	593 (279, 1577)

Clinical characteristics are presented as mean values ± standard deviation.

HbA_1c, glycosylated hemoglobin.

Six SNPs across four loci were selected for genotyping in the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes collection.¹⁰ In the GoKinD collections, strong associations were identified at two highly correlated SNPs on chromosome 7p14.3 (rs39059 and rs39075, r²=0.96) located within the CHN2 gene and upstream CPVL. Similarly, on chromosome 11p15.4, strong associations were identified at rs739401 and rs451041 (r²=0.97) within the CARS gene. From each locus, the most strongly associated SNPs (7p14.3: rs39075 and 11p15.4: rs451041) were selected for genotyping in the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes collection. Additionally, two SNPs (rs1888747 and rs10868025) sharing partial linkage disequilibrium (LD, r²=0.81) on chromosome 9q21.32 (located upstream of FRMD3) were also strongly associated with diabetic nephropathy in the GoKinD collections and selected for genotyping. Lastly, multiple highly correlated SNPs on chromosome 13q33.3 were also identified in the GoKinD collections. Because of the extent of LD shared among these variants, only two SNPs (rs1411766 and rs9521445) with minimal LD among all other SNP pairs (r²=0.30–0.65) were included in the present study.

Recent studies have suggested that the development of diabetic nephropathy and its progression may be influenced by different genetic factors.^13,14 To test this hypothesis, we first analyzed the SNPs at chromosomes 7p14.3, 9q21.32, 11p15.4, and 13q33.3 for association with early and advanced diabetic nephropathy by stratifying cases from the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes collection based on their stage of disease. Separate analyses compared the distribution of risk alleles for each SNP among cases with either microalbuminuria or proteinuria/end-stage renal disease versus normoalbuminuric controls, respectively (Table 2). The effects at rs39075 (7p14.3), rs10868025 (9q21.32), rs451041 (11p15.4), and rs9521445 (13q33.3) were of similar magnitude among cases from this collection irrespective of their nephropathy stage. In contrast, rs1411766 on chromosome 13q33.3 was exclusively associated with the risk of early diabetic nephropathy (OR=1.34, 95% CI=1.11–1.61, P=1.9×10⁻³ for the comparison of microalbuminuric cases and OR=1.01, 95% CI=0.82–1.24, P=0.94 for the comparison of proteinuric/end-stage renal disease cases). No significant deviation from Hardy-Weinberg equilibrium was observed among the genotype distributions for the six SNPs included in the present study in either the case or control groups (P=0.09 for rs39075, P>0.34 for all other SNPs tested).

Table 2.

Association results at associated SNPs identified in GoKinD in the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes collection

Locus				Risk Allele Frequencies, P-values, and ORs for the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes Collection
SNP	Chr.	Position (Mb)	Nearest Gene(s)	Risk Allele (Non-risk Allele)*	Controls (n=646)	Micro. Cases (n=427)	Prot./ESRD Cases (n=316)	Controls vs. Micro.		Controls vs. Prot./ESRD		Controls vs. All Cases
								P-value	OR (95% CI)	P-value	OR (95% CI)	P-value	OR (95% CI)
rs39075	7p14.3	29.2	CPVL/CHN2	G(A)	0.63	0.60	0.60	0.23	0.89 (0.74–1.07)	0.33	0.90 (0.74–1.11)	0.19	0.90 (0.77–1.05)
rs1888747	9q21.32	85.3	FRMD3	G(C)	0.68	0.70	0.65	0.15	1.15 (0.95–1.40)	0.21	0.88 (0.71–1.08)	0.92	1.01 (0.86–1.19)
rs10868025	9q21.32	85.4	FRMD3	A(G)	0.60	0.61	0.59	0.63	1.05 (0.87–1.27)	0.60	0.95 (0.77–1.16)	0.90	0.99 (0.84–1.16)
rs451041	11p15.4	3.0	CARS	A(G)	0.48	0.52	0.52	0.06	1.19 (1.00–1.42)	0.03	1.25 (1.02–1.52)	0.02	1.21 (1.03–1.41)
rs1411766	13q33.3	109.1	no gene	A(G)	0.35	0.41	0.35	1.9×10−3	1.34 (1.11–1.61)	0.94	1.01 (0.82–1.24)	0.03	1.19 (1.01–1.40)
rs9521445	13q33.3	109.1	no gene	A(C)	0.50	0.55	0.56	0.02	1.23 (1.03–1.47)	0.02	1.26 (1.04–1.53)	4.4×10−3	1.25 (1.07–1.46)

^*Risk allele reported in the GoKinD collections. ESRD, end-stage renal disease.

In the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes collection, P-values and odds ratios (ORs) were calculated using stratified additive tests of association (adjusting for gender) using the Cochran-Mantel-Haenszel test procedure. SNP positions and gene annotations are in reference to NCBI Build 36.1.

Next, we assessed associations between SNPs across these four distinct susceptibility loci and diabetic nephropathy in the entire Joslin Study of Genetics of Nephropathy in Type 2 Diabetes collection (646 control and 743 case subjects). Following adjustment for multiple testing, rs9521445 (OR=1.25, 95% CI=1.07–1.46, P=4.4×10⁻³) on chromosome 13q33.3 showed significant evidence of association among the samples in this collection (Table 2). At this same locus, rs1411766 was also associated with type 2 diabetic nephropathy (OR=1.19, 95%CI=1.01–1.40, P=0.03). This SNP, however, does not achieve statistical significance when a conservative Bonferroni correction is applied (Bonferroni-corrected P=0.18). Additionally, we also observed an association at the 11p15.4 locus (rs451041: OR=1.21, 95% CI=1.03–1.41, P=0.02). Similarly, after accounting for the number of SNPs analyzed in the present study, rs451041 was not significant (Bonferroni-corrected P=0.12) in this collection. Lastly, we did not observe any support for associations at the 7p14.3 and 9q21.32 loci in the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes collection. Adjustments for HbA_1c, BMI, and duration of type 2 diabetes did not appreciably alter the evidence of these associations (data not shown).

The proportion of males and females differed in the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes collection between the case and control groups. While we adjusted for this difference in the analysis, further investigation of these data by sex suggests that effect modification may be present only at the 9q21.32 locus (data not shown). As no association with diabetic nephropathy was observed at this locus, and no modification was observed for other loci in either strata, no further analyses were performed.

DISCUSSION

The recent examination of four loci from a GWA scan of diabetic nephropathy in type 1 diabetes¹⁰ in Japanese patients with type 2 diabetes¹² suggests that associations on chromosome 13q33.3 confer increased susceptibility to nephropathy in both type 1 and type 2 diabetes and across different ethnic groups. To follow-up on these findings, we focused on the same four loci investigated in this type 2 cohort and examined whether SNPs at susceptibility loci on chromosomes 7p14.3, 9q21.32, 11p15.4, and 13q33.3 were associated with diabetic nephropathy in the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes collection.

Our data, as well as that from Maeda et al. (Supplemental Table 1), were unable to support associations on chromosomes 7p14.3 and 9q21.32 in type 2 diabetic nephropathy, suggesting that the increased susceptibility due to these loci may be exclusive to type 1 diabetes. While further studies are needed to confirm this hypothesis, evidence of the specificity of the 9q21.32 locus in type 1 diabetic nephropathy in the DCCT/EDIC collection supports this notion.¹⁰

Variants on chromosomes 11p15.4 and 13q33.3 were associated with disease in our analysis of the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes collection. Associations with diabetic nephropathy at both of these regions have now been identified in multiple independent collections. The 13q33.3 locus was first reported to contribute to increased susceptibility of nephropathy in a genome-wide linkage scan of 296 end-stage renal disease affected sibling pairs enriched for non-diabetic nephropathy (logarithm of odds (LOD) = 1.02 at 13q33.3 near D13S796).¹⁵ Stronger evidence of linkage was subsequently found at this same location among 563 African-American affected sibling pairs with all-cause end-stage renal disease, including 267 sibling pairs with type 2 diabetes (LOD = 1.72 near D13S796).¹⁶ We later reported evidence of association at both the 11p15.4 and 13q33.3 loci in our GWA scan of 885 type 1 diabetic controls with normoalbuminuria and 820 type 1 diabetic cases with advanced diabetic nephropathy (284 with proteinuria and 536 with end-stage renal disease) from the GoKinD collections and among 132 cases with severe nephropathy (proteinuria or ESRD) from the DCCT/EDIC study.¹⁰

More recently, in a study that included 3,433 Japanese patients with type 2 diabetes from four independent populations, including 1,903 normoalbuminuric controls and 1,530 advanced diabetic nephropathy cases, Maeda et al.¹² reported associations on chromosome 13q33.3 with the increased risk of type 2 diabetic nephropathy. Their findings for rs1411766 are consistent with associations observed in both the GoKinD collections and in the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes collection. A meta-analysis combining data for this variant from both type 2 collections improved evidence of association at this locus (OR=1.22, 95% CI=1.07–1.33, P=3.8×10⁻⁴) (Table 3). Importantly, combined data from all three studies reached genome-wide significance (OR=1.29, 95% CI=1.16–1.38, P=9.7×10⁻⁹) (Table 3).

Table 3.

Meta-analysis of SNPs across the 7p14.3, 9q21.32, 11p15.4, and 13q33.3 loci in the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes, Japanese Type 2 Diabetes Diabetic Nephropathy, and GoKinD Collections

Locus					Meta-analysis of the Joslin and Japanese Collections				Meta-analysis of the Joslin, Japanese, and GoKinD Collections
SNP	Chr.	Position (Mb)	Nearest Gene(s)	Risk Allele (Non-risk Allele)	P-value	OR	Q-statistic P-value	I2	P-value	OR	Q-statistic P-value	I2
rs39075	7p14.3	29.2	CPVL/CHN2	G(A)	0.06	0.93 (0.84, 1.01)	0.68	0.00	0.36	1.03 (0.96, 1.10)	<1.0×10−4	92.50
rs1888747	9q21.32	85.3	FRMD3	G(C)	0.54	1.03 (0.93, 1.12)	0.75	0.00	1.2×10−3	1.14 (1.01, 1.19)	5.0×10−4	86.92
rs10868025	9q21.32	85.4	FRMD3	A(G)	0.53	1.03 (0.94, 1.13)	0.55	0.00	8.4×10−4	1.14 (1.01, 1.17)	6.0×10−4	86.51
rs451041	11p15.4	3.0	CARS	A(G)	0.03	1.11 (1.01, 1.18)	0.19	41.53	7.9×10−6	1.19 (1.07, 1.25)	0.02	76.17
rs1411766	13q33.3	109.1	no gene	A(G)	3.8×10−4	1.22 (1.07, 1.33)	0.61	0.00	9.7×10−9	1.29 (1.16, 1.38)	0.27	23.75
rs9521445	13q33.3	109.1	no gene	A(C)	0.01	1.11 (1.03, 1.21)	0.08	68.32	4.8×10−6	1.18 (1.10, 1.27)	5.7×10−3	80.67

Meta-analyses were performed for the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes, Japanese type 2 diabetes diabetic nephropathy¹², and GoKinD¹⁰ collections as indicated above. P-values and odds ratios (ORs) were calculated using the Cochran-Mantel-Haenszel method with fixed-effects models Heterogeneity across these collections was evaluated using a Q-statistic P-value. The I² statistic was calculated to quantify the proportion of the total variation due to heterogeneity. SNP positions and gene annotations are in reference to NCBI Build 36.1.

In the study by Maeda et al., rs9521445, a second variant on 13q33.3 modestly correlated with rs1411766 (r²=0.38 in individuals of European ancestry, r²=0.09 in individuals of Japanese ancestry), was not found to be associated with disease. A sub-analysis of Japanese patients receiving chronic hemodialysis, however, revealed this SNP to be associated primarily with end-stage renal disease in this population (OR=1.45, P=0.02). In the present analysis, rs9521445 was significantly associated with diabetic nephropathy in the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes collection. Both the magnitude and direction of this association are consistent with that observed in GoKinD (Supplemental Table 1). A meta-analysis combing our type 2 data with that from GoKinD further strengthens evidence of this association (OR=1.32, 95% CI=1.19–1.45, P=7.2×10⁻⁸) (Supplemental Table 2).

To explore the possibility that two potentially independent variants underlie the association signal at the 13q33.3 locus, we performed conditional analyses of rs1411766 and rs9521445 in the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes and GoKinD collections. In both collections, the association signals observed for rs1411766 were attenuated after conditioning on rs9521445. Similarly, conditional analysis revealed that rs1411766 could account for most of the association of rs9521445, suggesting that these two signals are not independent.

Although both 13q33.3 SNPs were of similar strength in the GoKinD collections, rs9521445 was more strongly associated with diabetic nephropathy in the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes collection than rs1411766, suggesting that allelic heterogeneity at this locus may contribute differently to the risk of disease among patients with type 1 or type 2 diabetes and, based on data presented by Maeda et al., these variants may also differentially impact disease susceptibility among individuals with diverse ancestral/ethnic backgrounds. Together, these results further implicate this locus in the genetic susceptibility of diabetic nephropathy. It remain difficult, however, to speculate whether variants at 13q33.3 differentially affect initiation of diabetic nephropathy and/or its progression to the advanced stages of disease and whether the contribution of 13q33.3 to the risk of diabetic nephropathy differs in type 1 and type 2 diabetes.

Variants on at the 13q33.3 locus localize to an intergenic region approximately 384 kb telomeric to MYO16 and 120 kb centromeric to IRS2 and, therefore, identifying the disease genes that underlie these associations is likely to be particularly challenging. Recent studies by Pomerantz et al.¹⁷ (using chromosome confirmation capture, or 3C, at a highly-replicated colorectal cancer locus on chromosome 8q24) and Visel et al. ¹⁸ (through the targeted deletion of a coronary artery disease risk interval in mice), however, highlight successful approaches in mapping genes located in genomic regions devoid of protein coding genes. Similar studies will likely be necessary to disentangle the relationship between variants at this locus and increased susceptibility to diabetic nephropathy.

Additionally, support at this locus from two SNPs in loose LD is consistent with the presence of multiple signals in this region. As recently described by Dickson et al., multiple significant, independent associations at a single locus can be due to ‘synthetic association’ caused by rare genetic variants.¹⁹ Further study is needed to resolve this question and to determine if multiple low frequency variants, perhaps in MYO16 or IRS2, underlie the observed associations with diabetic nephropathy.

Similar to rs9521445 on chromosome 13q33.3, rs451041 on chromosome 11p15.4 was not associated with type 2 diabetic nephropathy in Japanese patients (Supplemental Table 1). Associations at this variant were consistent in both the GoKinD and Joslin Study of Genetics of Nephropathy in Type 2 Diabetes collections. A meta-analysis of these data improved evidence of this association with diabetic nephropathy (OR=1.31, 95% CI=1.16–1.43, P=6.7×10⁻⁷) (Supplemental Table 2). While our study provides further evidence that the 13q33.3 region is a common susceptibility locus for diabetic nephropathy in both type 1 and type 2 diabetes, it also suggests a similar role for variants at the chromosome 11p15.4 locus. The associations at this region map to intron 4 of the CARS gene, a regulator of intracellular cysteine concentrations and protein biosynthesis.²⁰

Although associations on chromosome 11p15.4 and 13q33.3 in the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes collection further implicate these regions in the pathogenesis of diabetic nephropathy, we acknowledge some limitations of the our study. First, although none of the SNPs selected for examination in this collection achieved genome-wide significance in GoKinD, they represent the leading loci identified in the only GWA study of diabetic nephropathy published to date. Their examination by both Maeda et al. and in the current study constitutes a pragmatic strategy to investigate whether the top findings from this GWA scan are common susceptibility loci in both type 1 and type 2 diabetic nephropathy. It does not exclude the possibility that SNPs at other loci beyond those investigated in these studies harbor true disease susceptibility variants.

Second, all patients from the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes included in this study were self-identified as being of Caucasian heritage. Because of the paucity of SNP data available in this collection, the substructure of these samples has not been assessed. Residual confounding as a result of undetected population stratification could potentially lead to biases affecting the magnitude and significance of the observed associations.

Third, the present study is a focused evaluation of four loci identified in a GWA scan of the GoKinD type 1 diabetic nephropathy collection and recently examined in a large cohort of Japanese patients with type 2 diabetic nephropathy. In our study, there is greater than 90% power to detect similar effects as those observed in GoKinD (OR≥1.30) based on a P<0.05/6=8.3×10⁻³ threshold. It should be noted that the clinical subphenotypes of our collection produce some uncertainty in the estimation of this effect and its interpretation. Moreover, we note that applying the ‘Winners Curse’ correction to the effect sizes at each of the four loci would result in an over-estimation of the power of this study for a fixed sample size.

Despite these limitations, our analysis increases support that associations identified in the GoKinD collections on chromosomes 11p15.4 (near the CARS gene) and 13q33.3 (within an intergenic region between MYO16 and IRS2) are true diabetic nephropathy susceptibility loci common to both type 1 and type 2 diabetes. Further characterization of these regions is necessary to define both the predisposing variants and the disease genes underlying these associations.

METHODS

Study Patients

The Joslin Clinic is a large center for the treatment of patients of all ages with diabetes, as well as for the treatment of late diabetic complications. Approximately three-fourths of the approximately 16,000 patients under the care of the Joslin Clinic have type 2 diabetes. The clinic population is approximately 90% Caucasian and includes patients from all social strata. The majority of patients reside in eastern Massachusetts and generally come to the clinic within a few years of their diagnosis. Many remain with the clinic for decades, receiving integrated care from endocrinologists, nephrologists and ophthalmologists. Computerized medical records, together with laboratory measurements, including routine determinations of urinary albumin to creatinine ratio (ACR) measurements, are available for all Joslin Clinic patients.

In the 1990s and 2000s, we recruited 2,571 patients for the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes. Patients were recruited in two intervals between 1991 and 1996 (n=571), and between 2002 and 2008 (n=2,000). Patients with microalbuminuria and proteinuria were identified in the Joslin Clinic laboratory database and invited to participate in this study through a special examination. In parallel with the selection and examination of cases with diabetic nephropathy, a similar number of controls with normoalbuminuria was selected and invited for examination.

The study protocol and informed consent procedures were approved by the Committee on Human Subjects of the Joslin Diabetes Center.

Diagnosis of Type 2 Diabetes, Definition of Nephropathy, and Patient Examinations

Type 2 diabetes was considered if medical record reviews indicated that hyperglycemia was diagnosed after age 35 and patients did not require insulin therapy for at least for 2 years following their diagnosis. The computer databases of the Joslin Clinic retain all measurements of urinary albumin, urinary creatinine, and their ratio (albumin to creatinine ratio, ACR) obtained for clinical and research use since 1990 and provided the basis for diagnosing normoalbuminuria, microalbuminuria, or proteinuria in all clinic patients. Methods for measuring ACR and defining normoalbuminuria, microalbuminuria, or proteinuria were described previously.²¹

Study recruiters examined eligible patients during their clinic visits. After informing patients about the study and obtaining their written consent to participate, the recruiters administered a structured interview that solicited the history of diabetes and its treatment, other health problems and medications, measured seated blood pressure (two measurements, taken 5 minutes apart), and obtained blood samples for DNA and biochemical determinations (stored at −85° C). The questionnaire data were supplemented with information retrieved from the Joslin Clinic medical records and laboratory database to exclude patients with non-diabetic renal disease.

In addition to using historical ACR measurements to identify cases and controls for enrollment into the study, ACR measurements obtained at least 2 years after their enrollment were used for final classification of patients as either cases with microalbuminuria or proteinuria or controls with normoalbuminuria. Cases with microalbuminuria were defined by median ACR values between 20 and 300μg/mg in multiple follow-up measurements. Cases with proteinuria or end-stage renal disease were defined by median ACR values >300μg/mg in multiple follow-up measurements or the development of end-stage renal disease (requiring dialysis or renal transplant therapy), respectively. Minimally, cases with microalbuminuria or proteinuria were defined based on ACR measurements meeting these criteria in two of the last three measurements taken at least one month apart. Controls with normoalbuminuria were defined by median value of ACR <20μg/mg in multiple follow-up measurements. A minimum of two of the last three measurements taken at least one month apart were required to be <20ug/mg. If a third measurement was required, a value <40μg/mg was necessary for inclusion.

Of the 2,571 patients recruited into the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes as of 2008, 1,400 patients (751 with diabetic nephropathy, including 431 with microalbuminuria and 320 with proteinuria and 649 with no clinical evidence of kidney complications and persistent normoalbuminuria despite more than 5 years duration of type 2 diabetes) who identified themselves as Caucasian, had DNA extracted, had type 2 diabetes diagnosed between 35 and 64 years of age, were between 40 and 69 years of age at the time of examination were included in the current study. The remaining 1,171 patients were not included because they either identified themselves as non-Caucasians or did not provide ethnic information (n=623), were Caucasians and normoalbuminuric but had less than 5 years duration of type 2 diabetes (n=43), or were Caucasians but did not have DNA extracted at the time of initiation of the present study (n=505). Sixty-one of the 320 patients with proteinuria have progressed to end-stage renal disease since their enrollment in the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes.

DNA from all 1,400 subjects was isolated from buffy coat specimens (leukocytes) using a standard phenol-chloroform extraction procedure.

Genotyping

Six SNPs across the four loci identified in the GoKinD collections were selected for inclusion in the present study; including rs39075 on chromosome 7p14.3, rs1888747 and rs10868025 on chromosome 9q21.32, rs451041 on chromosome 11p15.4, and rs1411766 and rs9521445 on chromosome 13q33.3. All six SNPs were genotyped using Taqman (Applied Biosystems, Foster City, CA) technology by the Genetics Core of the Diabetes and Endocrinology Research Center at the Joslin Diabetes Center in accordance with the manufacturer’s protocols.

A total of 11 samples had missing data for three or more SNPs and were not included in our analysis. Among the remaining 1,389 samples (743 with diabetic nephropathy, including 427 with microalbuminuria and 316 with proteinuria, and 646 with persistent normoalbuminuria), the overall genotype success rate was >98% for all SNPs, with the exception of rs10868025 which had a success rate of 92%. The concordance rate among duplicate samples was 100% for all six genotyped SNPs.

Statistical Analysis

Deviation from Hardy-Weinberg equilibrium was assessed for using a genotypic chi-squared test. All SNPs were analyzed using stratified additive tests of association, adjusting for gender, using the Cochran-Mantel-Haenszel test procedure as implemented in PLINK.²² The effects of gender, HbA_1c, BMI, and duration of type 2 diabetes on these associations were assessed using logistic models. Combined meta-analyses of the Joslin Study of Genetics of Nephropathy in Type 2 Diabetes, Japanese type 2 diabetes diabetic nephropathy collection¹², and GoKinD¹⁰ collections were performed using the Cochran-Mantel-Haenszel method with fixed-effects models. Heterogeneity across these collections was evaluated using a Q-statistic P-value. The I² statistic was calculated to quantify the proportion of the total variation due to heterogeneity.