Polymorphisms in igf-Binding Protein 1 Are Associated With Impaired Renal Function in Type 2 Diabetes Free

A population-based, shared-care electronic diabetes management system encompassing patients cared for in both primary and secondary care settings has existed in Salford, U.K., since 1992. This system enables long-term audit and follow-up of patient outcomes throughout the conurbation. Currently, 8,300 diabetic patients are registered from a total population of 220,000. The nature and goals of this study were explained to patients attending routine follow-up in both primary and secondary care settings; they were then invited to give consent to donate samples of DNA (from whole blood), serum, and plasma for inclusion in the Salford Diabetes Archive. The 732 samples from the collection were chosen because their baseline characteristics matched those of the background type 2 diabetes population based on age, sex, and duration of diabetes. Approval for these studies was obtained from the Local Research Ethics Committee.

Nephropathy was defined, using the National Kidney Foundation guidelines, as including any or all of the following: microalbuminuria (albumin-to-creatinine ratio [ACR] >3.5 in men and >2.5 in women), macroalbuminuria (ACR >30), and renal impairment (creatinine clearance <60 ml/min plus serum creatinine >1.4 mg/dl in men or >1.2 mg/dl in women). Data were excluded from the analysis for four male patients who had undergone renal transplant. No patients included in this study were undergoing renal replacement therapy (dialysis). The actual number of observations included in each analysis was further reduced by genotyping failures or by missing phenotype data.

DNA was extracted from whole blood using the phenol/chloroform method. After being normalized in nuclease-free water to 50 ng/μl, 15-ng working samples were dispensed onto 384-well reaction plates using a Cybi-Well 96-well simultaneous pipettor (Cybio, Maidstone, U.K.) and air-dried before being stored at 4°C.

SNPs located throughout the IGFBP1 locus were selected from electronic sources, including Sequenom RealSNP.com (http://www.realsnp.com/), HapMAP (http://www.hapmap.org/), and EBI Ensembl (www.ensembl.org). All had unique ref-SNP identifiers on the National Center for Biotechnology Information’s single nucleotide polymorphism database (http://www.ncbi.nlm.nih.gov/SNP/). These are shown in the correct context in Fig. 1. Emphasis was placed on those SNPs with available genotype frequency data and those in exon and promoter regions of the gene. SNP genotyping assays are detailed in Table 1.

All SNPs were genotyped using ABI Taqman end point PCR, for which assays were procured using the Assays-by-Design service (Applied Biosystems, Warrington, U.K.). Genotyping was carried out using 15 ng of dried genomic DNA in a 10-μl reaction volume, comprised of a 5-μl Universal Master Mix (Applied Biosystems), 0.25 μl assay mix (40×), 0.04 μl BioTaq (0.5 units/μl; Bioline, London, U.K.), and 4.71 μl nuclease-free water. All reactions were optimized and run at an annealing temperature of 58°C or 60°C for 35–45 cycles on a DNA Engine Dyad Thermal Cycler (MJ Research). Sample reporter fluorescence was measured using an ABI Prism 7700 Sequence Analyzer.

Helixtree software version 2.4.2 (Golden Helix, Bozeman, MT) was used to verify that genotype frequencies were in Hardy-Weinberg equilibrium and to estimate linkage disequilibrium (LD) and haplotype frequencies. Haplotype association analysis was carried out using a stochastic expectation-maximization algorithm, implemented in the THESIAS software package developed by Tregouet and Tiret (28) and available at http://ecgene.net/genecanvas/. Genotype and phenotype data were collated using intercooled Stata version 8.21 (Stata, College Station, TX) and analyzed for association using logistic regression. Odds ratios (ORs) were used when describing the strength of genotype-to-phenotype associations. The glomerular filtration rate (GFR) was estimated from serum creatinine using the formula developed by Levey et al. (29) for the Modification of Diet in Renal Disease study.

The IGFBP1 SNPs genotyped and found to be in Hardy-Weinberg equilibrium using Helixtree are shown with their genotype frequencies in Table 2. Nine SNPs were polymorphic, while the remainder (C-1058T, A-830T, C1084G, and A3299G) were monomorphic and were omitted.

Logistic regression revealed that the minor alleles of G-575A, T555C, A643G, and A4403G were all associated with ORs that were significantly protective against nephropathy when corrected for sex (Table 3). Age and BMI were not confounders, and their inclusion in the regression model did not substantially alter the regression coefficients. The prevalence of nephropathy was slightly associated with BMI (OR 1.06 [CI 1.02–1.11], P = 0.002), but this association was sex-specific due to the higher mean BMI in women (OR 0.004 [−0.003 to 0.01], P = 0.257 in men; 0.02 [0.01–0.02], P = 0.002 in women). BMI was strongly associated with female sex (2.84 [1.8–3.9], P < 0.001) but not with genotype for any SNP (Table 4).

Patients with nephropathy accounted for 11.8% of female and 24.6% of male subjects included in the analysis and had a mean age of 64.0 ± 9.7 years compared with 64.5 ± 11.1 years for those without nephropathy. Overall, 76% of patients with nephropathy were male. The mean ACR (± SD) was higher in the nephropathic group: 24.2 ± 68.3 vs. 1.0 ± 0.7 units (P = 0.0008 by t test). The mean GFR (in 2003) in the nephropathic versus the nonnephropathic subgroup was 53.4 ± 27.4 vs. 69.4 ± 17.5 ml/min.

The associations between genotype and continuous physiological parameters related to diabetic nephropathy were also examined. No reno-protective allele was found to be associated with GFR, serum creatinine, or ACR. This may have been because of the relatively small proportion of this sample (19.9%) formed by cases of nephropathy, so that little association was seen between genotype and continuous outcome variables due to a strong statistical dilution of effect. This is demonstrated in Fig. 2, in which it can be seen that genotype-dependent variations in the skewness of the mean ACR values did indeed vary in a manner consistent with poorer renal function in the major allele homozygotes. For the purposes of statistical analysis, this effect was swamped by the large standard deviation ranges caused by the heterogeneity of the sample as a whole. The median values remained stable, resulting in departures from normal distribution in a genotype-dependent manner (Fig. 2).

There was very strong LD among the eight SNPs from the IGFBP1 gene. Pairwise LD analysis using Helixtree revealed that G-575A, T555C, A643G, and A4403G were in close LD with each other (R > 0.9), A5881G formed a slightly looser LD pair with A3963G (R = 0.8), and T2877C and A3963G were in relatively loose LD with the tightly clustered SNPs (Table 5). Six haplotypes accounted for almost 97% of the variation, which meant that only six SNPs were required to classify all subjects into their haplotype categories. SNPs A5881G and C4332T added no additional information to SNPs G-575A and T555C; therefore, because the former have lower frequencies, we excluded them from our analysis. The haplotype resulting from having the major allele for all six SNPs had a frequency of 0.57 and was the reference haplotype. For the analysis of haplotypes associated with nephropathy, having the haplotype of all the major alleles was associated with the highest risk. Further scrutiny of the haplotype effects suggested that SNPs A4403G and A643G were sufficient to explain the haplotype effects. The haplotype resulting from having the major allele (A) for both SNPs had a frequency of 0.58, whereas the haplotype resulting from the minor alleles (G) of both SNPs had a frequency of 0.38 and conferred an OR for nephropathy of 0.66 compared with the major allele haplotype; this finding was consistent with the protective effects of the SNPs when taken individually (0.66² = 0.44). Intermediate haplotypes where only one minor allele was present accounted for only 3.9% of individuals and had no significant effect on nephropathy prevalence.

In a representative sample of type 2 diabetic patients, we found that four SNPs in IGFBP1 were associated with a significantly reduced prevalence of diabetic nephropathy. These are the first such reno-protective genetic markers in the IGF system to be reported. A common haplotype (frequency = 0.377) formed by the reno-protective alleles of two of these SNPs was also associated with protection against nephropathy. The relation between IGFBP1 polymorphisms and the ACR was also consistent with a reno-protective effect.

The minor alleles of four IGFBP1 SNPs (G-575A, T555C, A643G, and A4403G) were associated with decreased prevalence of nephropathy. This association was confirmed by haplotype analysis, which also suggested that two of the four SNPs (A643G and A4403G) were of particular importance. Of these SNPs, three are nonexonic (G-575A, T555C, and A643G) and therefore are unlikely to affect the eventual function of the IGFBP-1 protein, whereas the fourth is exonic and may well affect protein function by virtue of its location. It is more likely that any direct effect of the nonexonic SNPs on the overall effectiveness of IGFBP-1 would be exerted by the regulation of gene expression. It is possible that this could occur by altering the behavior of one or more of the many regulatory factor recognition motifs that are situated near the SNPs (see Fig. 3).

IGFBP-1 is expressed in and secreted by only three tissues in human adults: the liver, the decidua, and the kidney. In the kidney, this expression is localized to the cortical and outer medullary reaches of the convoluted tubules and the loop of Henlé (30). Although the role and function of IGFBP-1 have been extensively evaluated in the liver and decidua, comparatively little is known about the function of IGFBP-1 in the kidney. The onset of human type 1 diabetes is marked by a fall in serum IGF-I and IGFBP-3 levels, but increases in serum IGF-II and IGFBP-2 and -1 levels. One of the first manifestations of diabetic nephropathy is renal hypertrophy, characterized by glomerular hypertrophy, which is thought to be driven by IGF-I. This is ultimately followed by obstruction of the glomerular capillary lumen and failure of glomerular filtration (31).

IGFBPs have also been implicated in the development of nephropathy in human and rodent models (3,4,23–26,30,32,33). Flyvbjerg et al. (33) and Landau et al. (31) reported that the growth hormone, the principle biological regulator of IGF-I secretion, enhances diabetes-related renal IGFBP-1 upregulation at the mRNA and protein levels and exacerbates diabetic renal hypertrophy in STZ rats. (It should be noted that this study reports data only from type 2 diabetes, in whom the relation between the growth hormone and IGF systems is more convoluted than in type 1 diabetes [in which the primary defect is insulin deficiency] and involves complex dysregulation of the IGF/IGFBP system that may result in altered IGF-I bioavailability.)

Unlike all other IGFBPs, IGFBP-1 lies at the hub of a complex web of metabolic interactions, with its gene expression and circulating concentration being acutely regulated by many facets of metabolic status, such as circulating glucose and hypoxia. The most important regulatory effects appear to be stimulation by cAMP/glucocorticoids, glucose via USF-1, hypoxia via HIF-1α, and inhibition by insulin (17–22). The effect of insulin on IGFBP1 expression is dominant, although the insulin response elements, which are activated by the binding HNF-3β and Foxa-2, are also required as accessory factor binding sites for full glucocorticoid-mediated gene induction.

The promoter region SNP G-575A would initially seem to be well suited for a plausible functional role, being located in the 5′-flanking region of the gene. However, this particular region of the promoter is sparse in consensus regulatory motifs, with the majority being within 200 bp of the transcription start site (22,34–36) (Fig. 3). It is equally likely that it is merely in close LD with one or more SNPs that affect IGFBP1 expression or protein function.

The SNPs with a more plausible functional role are A643G and T555C, which are both located in intron 1 and fall immediately to each side of the furthest downstream of the three motifs of the hypoxia-responsive HIF-1α binding element (19). Also nearby are binding motifs for USF-1 and p300, which is an important cofactor in the binding of HNF-3 and HIF-1α (22,35,36). Binding of HIF-1α and USF-1 results in increased IGFBP-1 expression (19,20,37). USF1 expression and activity are highly sensitive to glucose concentration, and this gene has been implicated in pathologies resulting from hyperglycemia (37).

The coding SNP A4403G, which causes an amino acid substitution from isoleucine to methionine at position 253, exhibited the greatest association with diabetic nephropathy in our study and is the only nonsynonymous SNP thus far reported in IGFBP-1. This SNP would most likely affect some aspect of the posttranslational function of the IGFBP-1 protein. This polymorphism lies very close to the RGD motif in the COOH-terminal of the protein (residues 246–248). The RGD motif is crucially important in extracellular matrix interactions mediated by integrin-α₅β₁ binding, which facilitates targeted ligand delivery and whose disruption might inhibit one of IGFBP-1’s most important IGF-potentiating functions. This would fit extremely well with the hypothesis that IGFBP-1 ordinarily favors nephropathy through the increased accumulation and activity of IGF-I, and it is tempting to speculate that this SNP might ablate this effect and in so doing act in a reno-protective manner. Other mechanisms by which the functional behavior of the protein might be altered include changes in protein stability either through altered phosphorylation and/or proteolysis. This SNP is located in the 4th exon, well away from potential phosphorylation sites but close to the end of the thyroglobulin-1 domain (38). This domain is thought to be important in the regulation of proteolysis, one of the key nongenetic determinants of IGFBP bioactivity, greatly reducing affinity for IGFs, increasing the free ligand concentration in the pericellular space, and facilitating receptor binding. In addition, the entire COOH-terminal region of the protein is thought to be important in conferring high IGF-binding affinity on IGFBP-1 when the protein is folded (14).

In conclusion, this is the first study to report an association between polymorphisms in the IGFBP1 gene and decreased risk of diabetic nephropathy in a representative population with type 2 diabetes. There is abundant evidence from other studies to implicate IGFBP-1 in the development of diabetic nephropathy, making our findings biologically plausible. This study raises important questions regarding the role of IGFBP-1 in the pathogenesis of diabetic nephropathy that merit further investigation.

This work was supported by Salford Royal Hospitals National Health Service Trust Research and Development Directorate.