Introduction

Type 2 diabetes is a metabolic disorder characterised by impaired insulin secretion and action that ultimately lead to chronic hyperglycaemia. Chronic hyperglycaemia is, in turn, associated with increased risk of progression to microvascular complications (nephropathy and retinopathy) of diabetes [1]. About 35% of patients with type 2 diabetes develop diabetic nephropathy (DN) [2]. Diabetes is also the most common cause of end-stage renal disease (ESRD) and the need for dialysis or kidney transplantation, which are associated with increased mortality [3, 4]. Several studies have demonstrated the preventive effect of controlling blood glucose on developing microvascular complications of diabetes [5–7].

DN is characterised by structural damage to the kidney, which results in leakage of albumin in urine. Both increased micro- (AER 20–200 μg/min or albumin/creatinine ratio >2.5 and >3.5 mg/mmol in men and women, respectively) and macroalbuminuria (AER >200 μg/min) are strongly associated with risk of morbidity and mortality from cardiovascular diseases [8, 9]. The GFR describes the flow rate of filtered fluid by the kidney and is used for assessment of renal function. Since direct measurements of GFR are tedious, the estimated GFR (eGFR) is often predicted from serum creatinine (S-creatinine), age, sex and ethnicity [10]. Clustering of DN within ethnic groups and families [11] indicates existing genetic predisposition to renal pathology [11]. Genome-wide associated studies (GWASs) have been shown to be an unbiased approach to identify genetic susceptibility loci for a number of diseases. Currently there are more than 60 common genetic variants that have been associated with type 2 diabetes and/or glucose or insulin levels using GWASs [12]. However, whether these variants also associated with the development of complications of diabetes is not established. In this study we explored whether variants influencing type 2 diabetes and/or glycaemic traits were also associated with the development of nephropathy in patients with type 2 diabetes.

Methods

Participants

Scania Diabetes Registry

The Scania Diabetes Registry (SDR) was initiated in Malmö (south of Sweden) in 1996; the majority of patients regularly attended the Department of Endocrinology, Skåne University Hospital. The aim of the study was to find factors associationed with the development of complications using biomarkers and genetic markers [13]. Microalbuminuria was defined as at least two out of three consecutive measurements with AER ≥ 20 < 200 μg/min. Macroalbuminuria was defined as at least one measurement with AER ≥200 μg/min or ≥300 mg/24 h. ESRD was defined as eGFR ≤15 ml/min or dialysis or kidney transplantation. S-creatinine was determined using an enzymatic colorimetric method (Cobas NPU04998; Roche Diagnostic, Basel, Switzerland). Urine albumin was determined using immunonephelometry (Beckman Instruments, Brea, CA, USA) until 1998 and thereafter using an immunoturbimetric method (Beckman Coulter; Beckman Instruments, Brea, CA, USA) [14]. eGFR was estimated using the MDRD-4 formula. Nephropathy was defined as micro- or macroalbuminuria and/or ESRD.

Steno type 2 diabetes cohort

In the Steno cohort patients with type 2 diabetes (n = 345) were included as either cases with DN or controls without nephropathy collected as part of a European case–control study collaboration [15, 16]. The study design included nephropathy with inclusion criteria of albuminuria >300 mg/l and presence of diabetic retinopathy to ensure that albuminuria was the consequence of DN rather than a non-diabetic glomerulopathy [17]. Urinary albumin was determined using an enzyme immunoassay. S-creatinine was determined using a modified Jaffe’s method.

Genetics of Diabetes Audit and Research in Tayside Scotland

The Genetics of Diabetes Audit and Research in Tayside Scotland (Go-DARTS) database includes prescription and biochemistry information and clinical phenotypes of all patients with diabetes within Tayside, Scotland, from 1992 onwards. The Go-DARTS study is a joint initiative of the Department of Medicine and the Medicines Monitoring Unit (MEMO) at the University of Dundee. A total of 3,800 patients with type 2 diabetes from the Go-DARTS cohort were genotyped on the Affymetrix 6.0 single nucleotide polymorphism (SNP) array (Affymetrix, High Wycombe, UK) and linked with the Go-DARTS database. DN was defined as macroalbuminuria (albumin–creatinine ratio [ACR] ≥25 mg/mmol for men and ACR ≥35 for women) [18].

In all diabetic cohorts (SDR, Steno and Go-DARTS) controls were defined having normoalbuminuria.

Malmö Preventive Project

The Malmö Preventive Project (MPP) is a population-based study from the city of Malmö in southern Sweden. The study started with 33,346 participants in total, of which 25,677 eligible persons participated in a health screening visit during 1974–1992 aiming at investigating large strata of the adult population to find high-risk individuals for preventive interventions. Blood samples were collected for measurements of fasting blood glucose and lipid concentrations. S-creatinine was determined with Jaffe’s alkaline picrate method [19]. During 2002–2006, 17,284 persons participated in the re-screening visit and blood samples were taken for genetic analyses [20]. In our study we included 15,066 non-diabetic individuals with available baseline S-creatinine data, of them 2,536 with available S-creatinine data at follow-up.

Botnia Study

The Botnia study is a family-based study from the western coast of Finland that started in 1990, aiming to identify genes that increase susceptibility to type 2 diabetes [21]. In the present study, we included 2,247 non-diabetic participants from the Botnia study with available eGFR measurements.

For all studies the protocols were approved by local ethics committees, and informed consent was obtained from all participants.

Assessment of renal function

Morning ACR and/or overnight AER were used for classification of nephropathy in patients with type 2 diabetes. In addition, as a marker of renal function, we used eGFR calculated using the MDRD-4 formula: eGFR = 186 × S-creatinine−1.54 × Age−0.203 × (1.210 if black) × (0.742 if female sex) [16, 22].

Genotyping

We genotyped 28 known SNPs associated with type 2 diabetes/glycaemic traits published during 2006–2010 [23–27] in 2,229 patients with type 2 diabetes from SDR using the Mass Extend Mass ARRAY system (Sequenom, San Diego, CA, USA). Out of 28 SNPs 26 were in Hardy–Weinberg equilibrium with the call rate ranging from 90% to 100% (electronic supplementary material [ESM] Table 1). Replication of the HMGA2 variants SNP rs1531343 in the Steno, MPP and Botnia cohorts was performed using the TaqMan allelic discrimination assay with an ABI 7900HT sequence detection system (Applied Biosystems, Foster City, CA, USA). In Go-DARTS the SNP rs1531343 was directly genotyped on the Affymetrix 6.0 SNP array.

Statistical analysis

The OR for developing nephropathy was analysed using logistic regression, adjusted for sex and diabetes duration as well as with and without HbA1c. Univariate linear regression was used to study the association of SNP rs1531343 (genotypic additive model) with eGFR adjusted for sex, diabetes duration and HbA1c in patients with type 2 diabetes (SDR and Steno) and with log e transformed eGFR (log e eGFR) adjusted for age, sex and BMI in non-diabetic individuals (MPP and Botnia). HbA1c in SDR was calculated as mean value of HbA1c during a mean follow-up period of 10.9 years. Diabetes duration was calculated from the age at onset of diabetes until development of nephropathy for cases and from age at onset of diabetes until last visit for controls. Bonferroni correction for multiple comparisons was used with an adjusted p level = 0.0018 (0.05/28). Statistical analyses were carried out using Statistical Package for the Social Sciences version 17.0 (SPSS, Chicago, IL, USA). Fixed-effects meta-analysis was performed by Meta-Analysis Package for R (Metafor 1.6-0; http://www.metafor-project.org/). Power calculations were performed using Quanto (Quanto Version 1.2.4; http://hydra.usc.edu/gxe).

Results

Clinical characteristics of the study participants are shown in Table 1. In the SDR, 947 (42%) patients progressed to nephropathy during a 10.2-year period. In the Steno study, there were 172 (50%) nephropathy cases with mean diabetes duration of 15.5 years. In the Go-DARTS cohort, 114 (15%) patients progressed to nephropathy during a 10.8-year period.

Table 1 Clinical characteristics of the studied cohorts
Full size table

The prevalence and/or incidence of DN in men were significantly higher than in women (Table 1). In all studies, diabetic patients with nephropathy had higher HbA1c (SDR, 7.1% [65.6 mmol/mol] vs 6.6% [60.47 mmol/mol], p = 4.9 × 10−24; Steno, 9.1% [76.6 mmol/mol] vs 8.6% [71.5 mmol/mol], p = 0.009; Go-DARTS, 8.0% [62.53 mmol/mol] vs 7.5% [56.07 mmol/mol], p = 0.0002) and in two of the studies patients with nephropathy were younger at onset of diabetes (SDR, 53.6 ± 12 vs 54.6 ± 12.1 years, p = 0.021; Steno, 45.7 ± 12.4 vs 47.5 ± 9.3 years, p = 0.124) compared with those without nephropathy.

In the SDR, out of the 28 genetic loci analysed, the HMGA2 SNP rs1531343 was associated with increased risk of developing nephropathy after correction for multiple testing (ESM Table 2). The frequency of the minor C-allele of HMGA2 rs1531343 was significantly higher in patients with type 2 diabetes who had nephropathy (10.2% vs 6.9%, p = 9.8 × 10−5) as compared with those who did not. This was translated into a 1.5-fold increased risk of developing nephropathy (95% CI, 1.20, 1.87, p = 0.00035) adjusted for sex, diabetes duration and HbA1c. To replicate this finding, we validated this association in the additional available diabetic Steno and Go-DARTS cohorts. Although these cohorts had less power, similarly to SDR we observed that the frequency of the minor C-allele of HMGA2 rs1531343 was significantly higher in Steno (11.3% vs 6.6%, p = 0.031) and also tended to be higher in Go-DARTS (12.7% vs 10.9%, p = 0.41) patients with type 2 diabetes who had nephropathy as compared with those who did not. The combined analysis of SDR, Steno and Go-DARTS cohorts showed that carriers of the C-allele had a 1.45-fold increased risk of DN (95% CI, 1.20, 1.75, p = 0.00010) (Table 2). Additionally, we also explored whether HMGA2 rs1531343 would influence eGFR in patients with type 2 diabetes. We found that the C-allele of rs1531343 was also associated with lower eGFR (β ± SEM: −3.7 ± 1.2 ml/min, p = 0.002) (Table 3).

Table 2 Effect of HMGA2 rs1531343 on nephropathy in patients with type 2 diabetes
Full size table
Table 3 Effect of HMGA2 rs1531343 on eGFR
Full size table

Next, we analysed 15,066 non-diabetic participants from the MPP study with available information on eGFR (Table 3). Notably, in the MPP the C-allele was associated with lower eGFR at baseline (β ± SEM: −0.007 ± 0.003 ml/min, p = 0.038) and remained lower after a 6.4-year follow-up period (β ± SEM: −0.021 ± 0.008 ml/min, p = 0.007). Similarly, in the non-diabetic participants of the Botnia study the risk C-allele was associated with lower eGFR (β ± SEM: −0.024 ± 0.012 ml/min, p = 0.040). Combined analysis of the MPP and Botnia studies strengthened this association (β ± SEM: −0.008 ± 0.003 ml/min, p = 0.006) (Table 3).

Discussion

In this study, we demonstrated that the risk C-allele of the common variant (rs1531343) in the HMGA2 loci predisposing to type 2 diabetes was also associated with progression to nephropathy in patients with type 2 diabetes and decline in renal function measured with eGFR in non-diabetic individuals.

Our observations are in line with a recently reported GWAS for kidney function (eGFR) [28]. The study included 67,093 participants of European ancestry from 20 predominantly population-based studies to identify new susceptibility loci for reduced renal function. The risk C-allele of SNP HMGA2 rs1531343 was also negatively associated with eGFR (p = 0.039) in this study, supporting our findings (https://intramural.nhlbi.nih.gov/labs/CF/Pages/CKDGenConsortium.aspx, accessed 08/08/2012) [28]. Notably, in a GWAS for DN in African-Americans, another variant (rs2358944), located 57 kb downstream of the HMGA2 rs1531343, was one of the strongest signals associated with diabetic ESRD, but also to a lesser extent with non-diabetic ESRD [29].

The high mobility group (HMG) protein family includes HMGA, HMGB and HMGN [30]. HMGA proteins are transcribed from two genes, HMGA1 and HMGA2 [31]. HMGA proteins are found in high levels in human embryonic stem cells and derived embryoid bodies in vitro [32]. HMGA2 is ascribed an important role in the control of stem-cell development and proliferation, and also high levels of HMGA2 is found in many benign, as well as malignant, tumours [33, 34]. HMG proteins are the most abundant non-histone DNA-binding chromatin factors in the eukaryotic nucleus involved in up- and downregulation of several genes [33].

Although the role of HMGA2 in diabetes or DN is still not fully understood, it was demonstrated that overexpression of HMGA2 was associated with formation of micropolycystic kidney suggesting involvement of HMGA2 in kidney development [35].

Diabetic kidney abnormalities are characterised mainly by hypertrophy, thickening of the glomerular basement membrane, tubular atrophy and interstitial fibrosis, which have been shown to lead to increased loss of renal function [33]. Recently, epithelial–mesenchymal transition (EMT) has been suggested as a novel pathway playing a key role in the pathogenesis of tubulointerstitial fibrosis in DN and blockage of EMT has been shown to delay progression to DN [33, 34]. Of note, transcriptomic analyses of TGF-β-stimulated EMT in NAMRU mouse mammary gland (NMuMG) cells identified HMGA2 as a main target involved in the control of epithelial differentiation [36]. TGF-β is a pro-sclerotic cytokine strongly associated with the development of fibrosis in DN [34]. The induction of cytokines, chemokines and growth factors, particularly TGFβ1 and connective tissue growth factor (CTGF), leads to glomerular basement membrane thickening, podocyte injury and mesangial matrix expansion. Thus changes in HMGA2 expression may ultimately increase the risk of development of irreversible glomerular sclerosis and tubulointerstitial fibrosis involved in the pathogenesis of DN. However, as for many genetic association studies, the observed phenotypic effect could be caused by one or several genetic variants in LD with the genotyped variant rather than the genotyped SNP itself. Thus, we cannot rule out that the effect is mediated by cis- or trans-regulatory genetic elements other than the HMGA2 gene.

Although in this study we had 96% power to detect an association of HMGA2 rs1531343 with nephropathy, the available replication cohorts were less powered. Also, the original discovery cohort still had limited power to detect an effect of other genetic loci with smaller effect sizes. Thus, we cannot exclude weaker associations and false-negative results. Larger studies will be needed to detect or exclude a role of these SNPs in DN. Additionally, given that long-term glycaemia and duration of diabetes are major determinants of progression to nephropathy, glycaemic control and longer duration of diabetes should be clearly applied, if available, to define patients without nephropathy in the design of future studies.

In conclusion, we demonstrate that an HMGA2 variant seems to be associated with increased risk of developing nephropathy in patients with type 2 diabetes and with lower eGFR in both diabetic and non-diabetic individuals and thus could represent a common denominator in the pathogenesis of type 2 diabetes and kidney complications. However, the biology of most genetic determinants influencing type 2 diabetes/glycaemic traits is most likely different from those involved in progression to complications of diabetes.