A GREM1 Gene Variant Associates with Diabetic Nephropathy

Abstract

Gremlin, a cell growth and differentiation factor, promotes the development of diabetic nephropathy in animal models, but whether GREM1 gene variants associate with diabetic nephropathy is unknown. We comprehensively screened the 5` upstream region (including the predicted promoter), all exons, intron-exon boundaries, complete untranslated regions, and the 3` region downstream of the GREM1 gene. We identified 31 unique variants, including 24 with a minor allele frequency exceeding 5%, and 9 haplotype-tagging single nucleotide polymorphisms (htSNPs). We selected one additional variant that we predicted to alter transcription factor binding. We genotyped 709 individuals with type 1 diabetes of whom 267 had nephropathy (cases) and 442 had no evidence of kidney disease (controls). Three individual SNPs significantly associated with nephropathy at the 5% level, and two remained significant after adjustment for multiple testing. Subsequently, we genotyped a replicate population comprising 597 cases and 502 controls: this population supported an association with one of the SNPs (rs1129456; P = 0.0003). Combined analysis, adjusted for recruitment center (n = 8), suggested that the T allele conferred greater odds of nephropathy (OR 1.69; 95% CI 1.36 to 2.11). In summary, the GREM1 variant rs1129456 associates with diabetic nephropathy, perhaps explaining some of the genetic susceptibility to this condition.

There is substantial epidemiologic evidence supporting genetic susceptibility to diabetic nephropathy; however, compelling replication of genetic risk factors in multiple white populations has proved difficult.¹ The development of larger carefully phenotyped case-control collections now permits appropriately designed studies to be conducted with improved statistical power.^2,3

Gremlin is implicated in several developmental pathways^4,5 and has become increasingly recognized as an important contributor to renal disease.^6–8 Gremlin influences cell growth and differentiation,⁹ particularly through bone morphogenic protein and transforming growth factor-β-mediated processes.^4,10 We originally reported elevated Gremlin mRNA in mesangial cells subjected to high extracellular glucose¹¹ and cyclic mechanical strain in vitro.¹⁰ Increased expression of GREM1 has been observed in animal models of kidney disease,¹² including in the renal cortex of rats with streptozotocin-induced diabetes and nephropathy.¹⁰ Gremlin mRNA levels correlate with serum creatinine and tubulointerstitial fibrosis in diabetic nephropathy.¹³ Gremlin is a major bone morphogenic protein antagonist,¹⁴ gremlin expression is increased during the development of diabetic nephropathy,¹⁵ and the increased glomerular basement membrane thickening and microalbuminuria associated with diabetic kidney disease are attenuated in diabetic GREM1^± mice.¹⁶ Increased levels of GREM1 expression have been observed in renal biopsies from patients with diabetic nephropathy,¹⁷ chronic allograft nephropathy,¹⁸ and GN.¹⁹

The GREM1 gene is gremlin 1 homolog, cysteine knot superfamily (Xenopus laevis), previously known as IHG-2, DRM, gremlin, DAND2, PIG2, or CKTSF1B1. We have comprehensively screened the GREM1 gene for genetic variants and investigated common single nucleotide polymorphisms (SNPs) for association with diabetic nephropathy.

Results

Individuals have been described previously in case-control association studies investigating diabetic nephropathy²; characteristics are presented in Table 1. Resequencing data were submitted to GenBank as (1) DQ100069 Gremlin 1 homolog, cysteine knot superfamily (X. laevis) (GREM1) 5′ upstream, promoter and exon 1, genomic sequence; and (2) DQ100070 Gremlin 1 homolog, cysteine knot superfamily (X. laevis) (GREM1) coding region and 3′ downstream genomic sequence. WAVE (denaturing HPLC) screening and bidirectional resequencing of 9163 bp identified 31 unique variants that were submitted to dbSNP (ss48297764 to ss48297794). Eleven SNPs were novel and 24 were observed with a minor allele frequency (MAF) >5%. The position of each SNP was mapped to the current reference sequence for the genomic contig and mRNA (Table 2). MAFs were established in 48 control individuals from the Young Hearts collection²⁰; the minimum observed MAF was 10.6% (Table 2). Linkage disequilibrium was calculated for the 23 biallelic SNPs with MAF >5% and 12 major haplotypes (with estimated frequency >2%) were identified by snphap, which jointly accounted for 80% of haplotypic variation (Figure 1). Six haplotypes were observed with MAF >5% (Figure 1b). On the basis of this haplotype structure, nine haplotype-tagging SNPs (htSNPs) were identified to comprehensively investigate all common variation in the GREM1 gene; mean r² between SNPs included in the haplotype was 0.90 (minimum r² = 0.73).

Table 1.

Clinical characteristics of genotyped populations

Characteristic	All-Ireland Collection		U.K. Collection
	Case	Control	Case	Control
Number of individuals	267	442	597	502
Gender (% male)	57.3	40.0	59.3	44.2
Age at recruitment (years)a	48.7 ± 11.6	42.2 ± 11.7	48.5 ± 10.1	44.9 ± 11.2
Age at diagnosis (years)a	16.7 ± 8.5	15.0 ± 8.1	14.5 ± 7.6	15.6 ± 8.0
Duration of diabetes (years)a	32.0 ± 9.6	27.2 ± 9.3	33.9 ± 9.3	29.2 ± 9.1
Hemoglobin A1c (%)a	10.6 ± 1.5	9.5 ± 1.2	8.6 ± 1.7	8.2 ± 1.4

^aData are mean ± SD.

Table 2.

Reference location for submitted SNPs based on WAVE analysis and SNP confirmation^a

NCBI Unique SNP Identifier	GenBank Accession	NCBI Unique Assay Identifier ss#	Location on Reference Genomic Contig and mRNA	MAFb (%)
rs35529564	DQ100069	48297764	NT_010194.16:g0.3799002delA	Multiallelic
rs35285486	DQ100069	48297765	NT_010194.16:g0.3799426C>T	NA
rs3812934	DQ100069	48297766	NT_010194.16:g0.3799427T>C	50.0
rs1406389	DQ100069	48297767	NT_010194.16:g0.3800035A>T	15.2
rs1919364	DQ100069	48297768	NT_010194.16:g0.3800131C>G	45.6
rs35531317	DQ100069	48297769	NT_010194.16:g0.3800809C>G; NM_013372.5:c.−112C>G	NA
rs2293582c	DQ100069	48297770	NT_010194.16:g0.3800969G>A; NM_013372.5:c.−2+50G>A	24.4
rs3207357c	DQ100069	48297771	NT_010194.16:g0.3801040T>C; NM_013372.5:c.−2+121T>C	14.4
rs2293581	DQ100069	48297772	NT_010194.16:g0.3801293G>A; NM_013372.5:c.−2+374G>A	NA
rs28617440	DQ100070	48297773	NT_010194.16:g0.3811436T>C; NM_013372.5:c.−1−2012T>C	NA
rs16973303	DQ100070	48297774	NT_010194.16:g0.3811485T>G; NM_013372.5:c.−1−1963T>G	30.7
rs7178059	DQ100070	48297775	NT_010194.16:g0.3811917C>G; NM_013372.5:c.−1−1531C>G	NA
rs28433691	DQ100070	48297776	NT_010194.16:g0.3811947A>G; NM_013372.5:c.−1−1501A>G	44.6
rs35566180	DQ100070	48297777	NT_010194.16:g0.3811949delT; NM_013372.5:c.−1−1499delT	44.8
rs7182522	DQ100070	48297778	NT_010194.16:g0.3812024C>T; NM_013372.5:c.−1−1424C>T	13.0
rs34722587	DQ100070	48297779	NT_010194.16:g0.3812057T>C; NM_013372.5:c.−1−1391T>C	NA
rs35943942	DQ100070	48297780	NT_010194.16:g0.3812612insA; NM_013372.5:c.−1−835insA	32.6
rs35330276	DQ100070	48297781	NT_010194.16:g0.3812647G>A; NM_013372.5:c.−1−801G>A	20.2
rs11854391	DQ100070	48297782	NT_010194.16:g0.3813139A>T; NM_013372.5:c.−1−309A>T	33.7
rs12915554	DQ100070	48297783	NT_010194.16:g0.3814043C>A; NM_013372.5:c.*40C>A	30.2
rs33963919	DQ100070	48297784	NT_010194.16:g0.3814227C>T; NM_013372.5:c.*224C>T	31.8
rs17816260	DQ100070	48297785	NT_010194.16:g0.3814242A>C; NM_013372.5:c.*239A>C	32.6
rs3743105	DQ100070	48297786	NT_010194.16:g0.3814508T>C; NM_013372.5:c.*505T>C	31.8
rs3743104	DQ100070	48297787	NT_010194.16:g0.3814542A>G; NM_013372.5:c.*539A>G	33.0
rs34498321	DQ100070	48297788	NT_010194.16:g0.3814993T>G; NM_013372.5:c.*990T>G	NA
rs7162202	DQ100070	48297789	NT_010194.16:g0.3815012T>G; NM_013372.5:c.*1009T>G	37.5
rs17228641	DQ100070	48297790	NT_010194.16:g0.3815365G>A; NM_013372.5:c.*1362G>A	11.7
rs3743103	DQ100070	48297791	NT_010194.16:g0.3816184C>T; NM_013372.5:c.*2181C>T	45.0
rs10318c	DQ100070	48297792	NT_010194.16:g0.3816536C>T; NM_013372.5:c.*2533C>T	19.3
rs1129456	DQ100070	48297793	NT_010194.16:g0.3817224A>T; NM_013372.5:c.*3221A>T	10.6
rs7176378	DQ100070	48297794	NT_010194.16:g0.3817736T>A	27.4

NCBI, National Center for Biotechnology Information; NA, not applicable.

^aThe location of SNPs are provided in relation to the reference assembly of genomic contig (NT_010194) and reference mRNA for Gremlin (NM_013372.5), which was updated on June 28, 2009.

^bMAF is based on 96 chromosomes that were resequenced where WAVE suggested a MAF > 5%.

^cPutatively functional.

Figure 1.

Linkage disequilibrium at the Grem1 locus. (a) Relationship between 23 biallelic SNPs resequenced in 48 control individuals from the Young Hearts collection. Pairwise comparisons where r² = 0 (no correlation) are highlighted in white through a gray spectrum to black, where r² = 1 (perfect correlation) for SNPs that are perfect proxies for each other. (b) Major haplotypes identified from snphap.

One of the htSNPs (rs2293582) overlapped with promoter prediction PI014696 and was predicted to generate a new binding site V$ETSF/GABP.01. An additional SNP, rs3207357, was selected for further genotyping because it was also associated with promoter prediction PI014696 and may affect transcription factor binding where T > C putatively influences the V$E2FF/E2F.01 site. In silico investigation using UTRScan²¹ (accessed July 16, 2009) revealed two elements of interest, the K-box and the GY-box. Searching the major database for microRNA information (miRBase)²² (accessed July 6, 2009) highlighted a target site for microRNA 574 (MIR574 gene) at NM_013372.5:c.*141 (Figure 2). No variants were identified in the protein coding region, and comparison of protein sequences in six species reveals 100% conservation for the 184 amino acids that comprise Homo sapiens GREM1 protein (Figure 3).

Figure 2.

Interrogation of miRBase highlighted a computationally predicted target site. The sequence of miRNA hsa-miR-574-3p is aligned to our resequenced data for the GREM1 gene with alignment across six species, demonstrating a high degree of conservation at this loci.

Figure 3.

In silico, multiple alignment of protein sequences encoded by the GREM1 gene in six species. The comparison reveals strong conservation across the amino acids comprising the H. sapiens GREM1 protein

In the Irish collection (267 cases and 442 controls) association was observed with three SNPs (rs1129456, rs3207357, rs7182522) and diabetic nephropathy at the 5% level of significance (Table 3). After adjustment for recruitment center and multiple comparisons, statistically significant association remained for two SNPs rs1129456 (P = 0.0005, P_corrected < 0.005) and the putatively functional rs3207357 (P = 0.006, P_corrected < 0.05). Genotyping in the U.K. collection (597 cases and 502 controls) supported the original association with rs1129456 (P = 0.0003). Logistic regression analysis of the combined data, with adjustment for any potential confounding effect of collection, was first used to fit a general genotypic model. Relative to the reference AA genotype, this model gave estimates of odds ratio (OR) = 1.47 [95% confidence interval (CI) 1.16 to 1.87] for the AT genotype and OR = 2.78 (95% CI 1.24 to 6.12) for the TT genotype. However, further analysis revealed that an additive model provided a better fit to the data [Akaike's information criterion (AIC) = 2252.9] than a recessive model (AIC = 2268.6), a dominant model (AIC = 2255.4), or a genotypic model (AIC 2254.7) with an approximately 70% estimated increase in risk associated with each rs1129456 T allele (OR = 1.69, 95% CI 1.36 to 2.11). Logistic regression of rs1129456 per T allele with adjustment for any potential confounding effect of collection [recruitment center (n = 8), gender, age, and duration of diabetes] showed the following results: all Ireland, OR = 1.68 (95% CI 1.17 to 2.41); United Kingdom, OR = 1.61 (95% CI 1.20 to 2.18); and heterogeneity test (P = 0.82) combined data set, OR = 1.62, (95% CI 1.29 to 2.04). Over transmission of the T allele (25 versus 16) was observed from parents to 124 affected probands (P = 0.16).

Table 3.

Genotype counts in all-Ireland DNA collection and P values for χ² tests for trend in Irish and U.K. collections (most significant SNPs shown first)^a

SNP (bp position)	Called Genotypes	All-Ireland DNA Collection		Ptrend Valueb (corrected)	U.K. DNA Collection		Ptrend Valueb (one-sided)
		Cases Nmax = 267 Count (%)	Controls Nmax = 442 Count (%)		Cases Nmax = 597 Count (%)	Controls Nmax = 502 Count (%)
rs1129456c	AA/AT/TT	192/66/6	363/72/4	0.0005	448/132/15	406/90/5	0.0003
(3817224)		(73/25/2)	(83/16/1)	(<0.005)	(75/22/3)	(81/18/1)
rs3207357	TT/TC/CC	200/61/5	366/65/6	0.006	456/126/14	387/107/5	0.08
(3801040)		(75/23/2)	(84/15/1)	(<0.05)	(77/21/2)	(78/21/1)
rs7182522	TT/TC/CC	199/59/3	366/65/6	0.02
(3812024)		(76/23/1)	(84/15/1)	(0.2)
rs3743103	TT/TC/CC	71/134/54	148/194/81	0.08
(3816184)		(27/52/21)	(35/46/19)
rs11854391	TT/TA/AA	117/120/30	220/171/50	0.23
(3813139)		(44/45/11)	(50/39/11)
rs1406389	AA/TA/TT	160/95/10	278/152/9	0.25
(3800035)		(60/36/4)	(63/35/2)
rs7162202	TT/TG/GG	124/121/20	231/170/39	0.37
(3815012)		(47/46/7)	(52/39/9)
rs7176378	AA/AT/TT	144/104/16	249/158/27	0.52
(3817736)		(56/39/6)	(58/36/6)
rs10318	CC/CT/TT	163/90/9	275/146/12	0.64
(3816536)		(62/34 /3)	(63/34/3)
rs2293582	GG/GA/AA	180/72/10	290/126/15	0.76
(3800969)		(69/27/4)	(67/29/4)

^aSNP position in genomic base pairs is recorded in parentheses below each unique SNP identifier.

^bSignificance values presented after adjusted for recruitment center.

^cLogistic regression of rs1129456 per T allele, adjusted for recruitment center. All-Ireland collection: OR = 1.80 (95% CI 1.28 to 2.52); U.K. collection: OR = 1.62 (95% CI 1.22 to 2.16); heterogeneity test P = 0.66, combined data set adjusted for recruitment center (n = 8): OR = 1.69 (95% CI 1.36 to 2.11).

Discussion

Gremlin is a strong biologic candidate gene for involvement in the pathogenesis of diabetic nephropathy. More than 9 kb of genomic sequence encompassing the GREM1 gene were screened to identify genetic variants. The mRNA and computationally translated sequence agrees 100% with the GenBank protein NP_037504. All resequencing data were submitted to GenBank under accession numbers DQ100069 (2407 bp) and DQ100070 (6756 bp). Thirty-one SNPs were identified, mapped to the genomic (NT_010194.16) and mRNA (NM_013372.5) reference sequences, and submitted to dbSNP to facilitate further study of the GREM1 gene.

No SNPs were observed in protein coding region of GREM1. This is not surprising because there is a high degree of conservation in the 184 amino acids that comprise the human GREM1 protein. Several homologues were identified with 100% conservation noted between amino acids in the human protein compared with chimpanzee, dog, cow, rat, and mouse sequences. In silico evaluation of SNPs revealed two with putative functionality where they overlapped with a computationally predicted promoter. Although individual binding sites predicted in a promoter region are insufficient to indicate transcriptional function, we selected both of these SNPs for further genotyping. It is possible that other elements in the promoter region contribute to the expression of GREM1. Methylation is a key epigenetic feature of DNA that plays an important role in chromosomal integrity and regulation of gene expression. Global DNA hypermethylation from peripheral blood leukocytes of individuals with ESRD has been associated with increased mortality.²³ CpG islands are evident for GREM1,²⁴ and GREM1 expression in epithelial cells is lost through methylation in several forms of cancer²⁵ (http://matrix.ugent.be/temp/static/GREM1.html). The investigation of three CpG islands showed hypomethylation of GREM1 before and after induction of chondrogenesis, despite a decrease in expression.²⁶

In silico exploration of the extensive 3′ untranslated region (UTR) suggests several variants that may regulate GREM1. Submitting the 3′ UTR sequence to UTRscan²¹ identified three matches with two functional elements from patterns defined in the UTRsite collection. Conserved 3′ UTR sequence motifs were observed; the K-box sequence (CTGTGATT) matched once whereas the GY-box (GTCTTCC) matched twice in our resequenced data. Posttranscriptional regulation by these elements is proposed to involve the formation of RNA-RNA duplexes with complementary sequences at the 5′ end of microRNAs.²⁷ Interestingly, the K-box motif was originally identified in Notch pathway target genes encoding basic helix-loop-helix repressors, and the GY-box is common to many of the same Notch pathway target genes.²⁷ Increased expression of Notch pathway genes, concurrently with Gremlin, has been reported for diabetic nephropathy,¹⁴ and Notch has been recently suggested as a new therapeutic target for kidney disease.^28,29 MicroRNAs are relatively abundant as small RNAs that target transcripts of protein-coding genes and destabilize mRNA or cause translational repression.³⁰ Interrogation of miRBase¹⁹ highlighted a computationally predicted target site, conserved across many species, for microRNA 574 at NM_013372.5:c.*141. Possible target sites were also postulated for microRNA 556 (ATCGTGGTTATAGTCAGCTCATT, NM_013372.5:c.*1041) and microRNA 633 (TCATACCTATTAA; NM_013372.5:c.*2527). rs10318 is located within the target sequence for miR633 and rs7162202 resides 31 bp upstream of miR556. rs10318 and rs7162202 were investigated for association with diabetic nephropathy.

GREM1 has been mapped to human chromosome 15q13-q15³¹ and the sequence position in the current genome build is 15q13.3. MAFs for all common, biallelic SNPs (n = 23) were established based on 96 chromosomes in a general control population. Twelve common haplotypes were identified and the haplotype structure was then used to select nine htSNPs (mean r² = 0.90) that would survey all common variation. Comparison of our resequenced data with those available in HapMap³² (accessed July 6, 2009) reveals that HapMap is not optimal to investigate GREM1 variants because less than half of resequenced common SNPs in our study are recorded in this resource, and only one is located within the first 12.5 kb. Our results also indicate that the current version of dbSNP would have been of limited utility for linkage disequilibrium mapping because not all common haplotypes could be defined.

Standard quality control measures were good, including genotype distributions for all SNPs in Hardy-Weinberg equilibrium (P > 0.001). Phenotypic and environmental heterogeneity may contribute to individual differences within a resource.³³ Covic and colleagues have suggested that, “careful phenotyping requires a large recruitment effort but is necessary to reduce population heterogeneity, a strategy that increases the likelihood of identifying diabetic nephropathy loci”.³⁴ Population heterogeneity was minimized by carefully using comparable phenotypes in both populations, and strict inclusion criteria were used for both collections. To minimize confounding ethnic differences in allele frequencies we selected only third-generation white individuals born in the British Isles for genotyping in this study. No significant genetic heterogeneity was observed in this project; tests for heterogeneity based on genotype data between the original and replicate populations were NS (P = 0.7 adjusted for recruitment center and P = 0.8 considering the maximum potential variables).

Statistically significant association was observed with three SNPs, two of which (rs1129456 and rs3207357) maintained significance after adjustment for multiple testing. rs1129456 is located in the 3′ UTR, 177 bp upstream of the poly(A) signal. The variants adjacent to rs1129456 were both genotyped as htSNPs; however, allele frequencies were similar in case and control groups for these SNPs. The associated SNP rs1129456 was correlated (81%) with the putatively functional SNP rs3207357; however, there were no statistically significant differences observed between replicate case and control groups for this variant. It is possible that a larger replicate collection would support the original association with rs3207357; however, our data suggest that this is not the primary functional variant associated with diabetic nephropathy in type 1 diabetes. It is plausible that rs1129456 is in linkage disequilibrium with more distal regulatory elements for the GREM1 gene. The initial association detected at rs1129456 with the Irish collection was replicated in a U.K. population with similar phenotypic criteria. A similar, significant trend was observed in both collections, and combining data (adjusted for recruitment center, gender, age at recruitment, and duration of diabetes) generated a highly significant test for trend (P = 0.00004) and an approximately 60% estimated increase in risk associated with the rs1129456 T allele (OR = 1.62, 95% CI 1.29 to 2.04). Additionally, we have performed a family-based test of association that is not sensitive to population substructure. Although this family collection was underpowered (n = 124 trios) to identify an association, the results do support our case-control findings and reveal an overtransmission of the risk allele (T allele = 61%).

Genetic variants in the GREM1 gene were recently examined in 39 Mexican-American families with type 2 diabetes in which less than 100 individuals had albuminuria.³⁵ No association was reported with urinary albumin:creatinine ratio; however, a weak association was observed with estimated GFR (SNP14, P = 0.01; SNP16, P = 0.049).³⁵ On the basis of their resequencing data (Dr. Farook Thameem, personal communication), neither of these two variants were observed during our screening phase. Interestingly, these variants are adjacent to the SNP (rs1129456) that we observed to be associated with diabetic nephropathy. It should be noted that the base position of variants reported by Thameem and colleagues³⁵ are different from those observed in our study and validated in dbSNP. To the best of our knowledge, no other group has published results evaluating variants in the GREM1 gene with kidney disease phenotypes.

The GREM1 gene initiates and maintains important activities during development and disease processes. However, GREM1 does not work in isolation and it is possible that screening related genes may provide further meaningful associations with kidney disease. We have examined the proximal genetic region surrounding GREM1, but it is possible that this gene is regulated by the interaction of several more distal control regions.³⁶ For example, genetic variation in the LMBR1 gene directly regulates expression of SHH, which is located approximately 1 Mb distant on chromosome 7q36.³⁷ Adjacent to GREM1 on chromosome 15q13.3 is the FMN1 gene, which is transcribed in the reverse orientation (http://www.ncbi.nlm.nih.gov/mapview). Several groups have described the tissue-specific co-regulation of GREM1 and FMN1, despite their diverse structure and functions.^14,36 GREM1 and FMN1 are expressed in similar patterns during kidney organogenesis and both have important functions during renal development.^7,36,38 In addition to FMN1, GREM1 also interacts with SLIT1 and SLIT2 in a glycosylation-dependent manner,³⁹ and GREM1 binds to YWHAH in vivo and in vitro.⁴⁰

In conclusion, GREM1 is implicated in the pathogenesis of diabetic nephropathy, and our research has revealed a novel association between the GREM1 gene and diabetic nephropathy. This genetic association was subsequently replicated in a larger collection of individuals with similar case-control phenotypic criteria. We have determined that the T allele of rs1129456 SNP is a significant risk allele for diabetic nephropathy.

Concise Methods

Participants

Ethical approval was obtained from the appropriate research ethics committees and written informed consent was provided by all participants. Genomic DNA was obtained from white participants who were recruited as part of an all-Ireland case-control collection² specifically designed to investigate genetic risk factors for diabetic nephropathy. Individuals with overt nephropathy (n = 267) had type 1 diabetes at least 10 years before the onset of proteinuria (>0.5 g/24 h) with hypertension (blood pressure >135/85 mmHg and/or treatment with antihypertensive agents) developing on or after the onset of proteinuria. Individuals in the control group (n = 442) had a long duration of diabetes (>15 years), were not prescribed antihypertensive medication, and demonstrated no evidence of microalbuminuria on repeated testing. Type 1 diabetes was diagnosed where age of onset was <35 years and insulin was required from diagnosis. White individuals recruited to the Warren 3 and the Genetics of Kidneys in Diabetes (GoKinD) U.K. collections (cases, n = 597; controls, n = 502) were used as a replicate population.² Parents of Warren 3 probands (n = 124) were used to assess transmission of alleles to affected offspring.⁴¹ The Warren 3/GoKinD U.K. collections were established as a collaborative resource comprising DNA and clinical information from adults having long-duration type 1 diabetes with or without kidney disease and, where possible, their parents. Briefly, all individuals were third generation born in the United Kingdom with similar recruitment criteria to those described for the Irish study, except that type 1 diabetes was diagnosed <31 years of age. Forty-eight control individuals selected for resequencing have been used previously⁴² and were obtained from the Young Hearts collection.²⁰

Screening

Genomic DNA sequence was characterized from H. sapiens chromosome 15 genomic contig NT_010194.16, gi:37540936. Using AmpliTaq Gold (Applied Biosystems, Warrington, United Kingdom), 9163 bp were amplified and screened for the variants using WAVE (dHPLC) technology (Transgenomic, Omaha, NE) on 15 case and 15 control individuals. Variants were confirmed by bidirectional capillary sequencing of representative samples on an ABI 3730 Genetic Analyser (Applied Biosystems). Allele frequencies were determined by resequencing in 48 individuals from the Young Hearts collection²⁰ for those SNPs where WAVE mutations suggested a MAF > 5%. Resequencing data were submitted to GenBank⁴³ and unique identifiers were obtained for all variants from dbSNP.⁴⁴

Selection of SNPs and Genotyping

Haplotype distributions were estimated for all SNPs with a MAF > 10% using snphap.⁴⁵ Haplotype frequencies were then input to Stata release 8 (http://www.stata.com), and the htsearch command in the htSNP2 package was used to select tag SNPs from among all SNPs for which the MAFs exceeded 10% and genotype distributions were in Hardy-Weinberg equilibrium.⁴⁶ The htsearch command performs an exhaustive search of all subsets of SNPs to identify the subset of tag SNPs that best represents all SNPs according to the r² criterion. All SNPs with P_corrected < 0.05 were selected for further genotyping (Table 3). In silico analysis was performed to assess the putative functionality of proximal genomic variants using Transcription Element Search Software (http://www.cbil.upenn.edu/tess), MatInspector Professional (http://genomatrix.de/cgi-bin/matinspector_prof/mat_fam.pl), UTRScan,²¹ and miRBase.²² SNPs were genotyped in the Irish case-control collection using MassARRAY iPLEX (Sequenom, San Diego, CA), TaqMan (Applied Biosystems, Foster City, CA), and sequencing technologies.

Statistical Analysis

Genotype frequencies were assessed for deviation from Hardy-Weinberg equilibrium (P < 0.001) using a χ² goodness-of-fit test. The extent of linkage disequilibrium between pairs of SNPs was evaluated using r² and visualized in Haploview.⁴⁷ Genotype distributions in cases and controls were compared using the χ² test for trend adjusted for the recruitment center, with the level of statistical significance set at 5%. Permutation (n = 100,000 simulations) adjustment for multiple testing was performed in Plink.⁴⁸ Logistic regression was used to analyze the combined data from the Irish and U.K. collections with a term to represent the collection included in the logistic model.⁴⁹ This approach enabled recessive, dominant, additive, and general genotypic models to be fitted to the data, and AIC was used where the lowest AIC value identified the best-fitting model. Logistic regression of statistically significant, replicated SNPs was considered with adjustment for potential confounding effects [recruitment center (n = 8), gender, age and duration of diabetes]. The threshold for replication was set at P < 0.05 in a one-sided test to detect an effect observed in the same direction as the original result. Where association was supported by the replicate population, the transmission disequilibrium test was used to investigate transmission of alleles from heterozygous parents to an affected offspring (n = 124).⁵⁰ The size of the Irish diabetic nephropathy case-control collection provided approximately 90% power to detect an OR of 1.75 for the genotyped htSNPs (MAF > 10%) and >80% power to detect the same OR for the nongenotyped SNPs (MAF > 10% and minimum r² = 0.73).

Disclosures

None.