Lack of association of genetic variation in chromosome region 15q14-22.1 with type 2 diabetes in a Japanese population

Background Chromosome 15q14-22.1 has been linked to type 2 diabetes (T2D) and its related traits in Japanese and other populations. The presence of T2D disease susceptibility variant(s) was assessed in the 21.8 Mb region between D15S118 and D15S117 in a Japanese population using a region-wide case-control association test. Methods A two-stage association test was performed using Japanese subjects: The discovery panel (Stage 1) used 372 cases and 360 controls, while an independent replication panel (Stage 2) used 532 cases and 530 controls. A total of 1,317 evenly-spaced, common SNP markers with minor allele frequencies > 0.10 were typed for each stage. Captured genetic variation was examined in HapMap JPT SNPs, and a haplotype-based association test was performed. Results SNP2140 (rs2412747) (C/T) in intron 33 of the ubiquitin protein ligase E3 component n-recognin 1 (UBR1) gene was selected as a landmark SNP based on repeated significant associations in Stage 1 and Stage 2. However, the marginal p value (p = 0.0043 in the allelic test, OR = 1.26, 95% CI = 1.07–1.48 for combined samples) was weak in a single locus or haplotype-based association test. We failed to find any significant SNPs after correcting for multiple testing. Conclusion The two-stage association test did not reveal a strong association between T2D and any common variants on chromosome 15q14-22.1 in 1,794 Japanese subjects. A further association test with a larger sample size and denser SNP markers is required to confirm these observations.


Background
Insulin resistance in muscle, adipose, and liver tissues and impaired insulin secretion from pancreatic β cells contribute to the pathogenesis of type 2 diabetes (T2D) [1]. Analyses using a traditional candidate gene approach and, more recently, genome-wide association studies (GWAS) [2][3][4][5][6][7], have generated a large amount of data that have confirmed 11 genes with replicated association with T2D in Caucasians [8]. However, the assessment in Japanese T2D patients, who characteristically have a lower body mass index (BMI) and a lower fasting insulin level than Caucasians [9], has not been largely performed. These differences suggest that Japanese individuals with T2D might have a different genetic background from other populations, and susceptibility variant(s) or gene(s) for the development of T2D in Japanese can only be identified with genetic assessment. Because candidate regions supported by replicated linkage signals in Japanese and other populations are expected to map T2D susceptibility gene(s), a region-wide association test was used as an alternative affective approach with the availability of unbiased reliability, timeliness and cost efficiency. Among a number of proposed candidate regions for T2D and its related traits [10], the chromosome 15q14-22.1 candidate region is supported by replicated linkage signals of non-overlapping samples from Japanese subjects. In three studies of genome-wide linkage scan of Japanese subjects [11][12][13], chromosome 15q14-22.1 showed significant evidence of linkage with T2D (LOD = 2.41) [12] and a maximum LOD score (MLS) of 3.91 for early onset T2D [11] (Table 1). This region overlaps candidate region found in other populations, including Mexican Americans [14] and Pima Indians [15], findings that were replicated in a subsequent study of Mexican Americans [16] (Table  1). This candidate region supported by replicated linkage signals is expected to contain susceptibility gene(s). In the absence of convincing comprehensive association data on chromosome 15q, the challenge remains to identify the disease susceptibility variant(s) or gene(s) that definitively contribute to T2D. We focused on chromosome 15q14-22.1 at 32.6-51.2 cM as the susceptibility region in Japanese.
In this study, a region-wide two-stage association test was used to perform a comprehensive examination of genetic variant(s), using 1,317 common SNPs, in a Japanese population comprised of 1,794 unrelated cases and controls.

Subjects
This study utilized samples obtained from a total of 1,794 Japanese subjects: 904 T2D cases and 890 controls. The majority of T2D patients was recruited from the outpatient clinic of Tokushima University Hospital, Kyoto Prefectural University Hospital and their affiliated hospitals. All T2D diagnoses were based on the 1985 World Health Organization criteria, and patients were clinically defined as having gradual adult onset of the disease with medication. Patients with clinical criteria for monogenic forms of diabetes were excluded. The clinical characteristics of these samples are presented in Table 2. The controls were recruited from healthy adult members of the general population, who were intensively checked for a negative family history of diabetes, normal HbA 1c levels of < 5.8%, the absence of other diseases, and Japanese ancestry based on Genomic DNA was prepared from peripheral blood leukocytes or EB-virus-immortalized B lymphoblasts using a standard protocol. The study was approved by the Ethical Committee for Human Genome and Gene Research at the University of Tokushima in accordance with the tenets of the Declaration of Helsinki. Informed consent was obtained from all subjects prior to blood sampling, and all personal information and samples were anonymously analyzed for genotyping.

Association testing of SNP markers and genotyping
At the onset of this study, there was no sufficient SNPs database, such as HapMap. Thus, we constructed a database of common Japanese SNPs based on the genotyping results of 45 unrelated Japanese control subjects (23 males and 22 females) in a whole genome in collabora-  [18]. Out of 70,099 common (MAF > 0.15) gene-centric and intergenic SNPs made available by this project, we provisionally selected 1,450 evenly-spaced SNPs on chromosome 15q between two sequence-tagged site (STS) markers of D15S118 and D15S117.
The initial selection was based on the following criteria: 1) optimal suitability for the design of TaqMan highthroughput genotyping assays; 2) dense and evenlyspaced SNP coverage in both coding and intergenic regions; 3) definition of the gene location in a region between 10 kb upstream of the transcription start site and 10 kb downstream of the final exon; 4) a distance < 10 kb between adjacent SNPs; and 5) common SNPs with a MAF > 0. 10. The criterion of a MAF > 0.10 was imposed subsequent to the selection of SNPs with a MAF > 0.15.
SNPs were assayed using TaqMan Universal PCR Master-Mix (no UNG; ABI) as previously described [19][20][21]. The genotyping results were obtained in an auto-call mode, after which two researchers independently assessed the genotyping data. When the ratio of undetermined results to the total samples was greater than 2% (8/384 including four negative controls) or negative controls without DNA template were incorrectly detected, the genotyping was repeated. Genotyping results were compared to those obtained by direct sequencing as previously described [20] with perfect agreement between the two methods.

Two-stage design of the association test
We performed a two-stage association test on 1,794 samples by randomly assigning them to two independent panels. In the first test, all of the SNPs were genotyped in Stage 1. Those exhibiting significant allelic or genotypic association (p < 0.05) were further examined in Stage 2. In each stage, the association was evaluated with two types of χ 2 test (allelic or genotypic) by 2 × 2 and 2 × 3 contingency tables for the status of cases or controls. Finally, the association was evaluated using the combined samples from both Stages.
The power of this case-control test was calculated based on MAF, a type 1 error rate (false-positive rate), and sample size with the PS program [22,23]. The standard Bonferroni's correction was used to evaluate the false-positive rate [24]. In addition, we used the false discovery rate (FDR) approach [25] as implemented with the FDR control program in the R language [26]. A logistic regression analysis to adjust for age, gender, and BMI was carried out using the SPSS program (ver.12, SPSS Japan Inc., Tokyo, Japan).

Searching for novel SNPs and mutational screening of the UBR1 gene
The 47 coding exons, relevant intron-exon boundaries, and 5' and 3' UTR of UBR1 gene were screened for putative novel Japanese variants by direct sequencing the genomic DNA of 48 individuals (24 cases and 24 controls). All PCR products were treated with ExoSAP-IT (GE Healthcare, NJ, USA) and sequenced using BigDye Terminator Cycle Sequencing Kit, ver1.1 (ABI) in both directions on a 3730xl automated sequencer (ABI) according to the manufacturer's standard protocol.

Estimation of linkage disequilibrium (LD) analysis for haplotype inference
Pairwise LD coefficients of |D'| and r 2 were assessed in Stage 1 samples with SNPAlyze ver 5.

The amount of captured genetic variation based on HapMap JPT SNPs
HapMap JPT SNPs corresponding to 21.8 Mb were investigated [28]. Information on 29,728 SNPs within the target region was downloaded from the HapMap website (release 22) [29]. The density of the 1,317 SNPs (the finally selected SNPs from the 1,450 pre-HapMap SNPs) per 300-kb bin was compared against those in the Hap-Map JPT SNPs. In addition, we estimated the captured tag SNPs as proxies at r 2 > 0.7, 0.8, or 0.9 using Tagger program [30].

Information of selected SNPs
One STS marker (D15S118) was located 3.9 Mb proximal to D15S994, which exhibited the highest evidence for linkage in the Japanese subjects with a LOD = 3.91. The second STS marker (D15S117) located 9.0 Mb distal to D15S119, was linked to T2D susceptibility in Mexican Americans with a LOD = 4.0 (Table 1) (Figure 1), although they were registered in the public database [32] without information on allele fre-quency in the Japanese population. Figure 1 shows only one gap in 43. 6-43.9 Mb in this study, and the presence of a rare density of SNPs (< 10) was observed, particularly in the intergenic region and between 42. 1-44.8 Mb.
This approach is more efficient if tag SNPs based on the HapMap JPT r 2 structure are used to find the susceptibility variant [28]. We estimated the captured tag SNPs as proxies, meaning that the SNP showed a strong correlation with one or more SNPs at r 2 > 0.7, 0.8, or 0.9. The density of most tag SNPs was < 10 SNPs per 300-kb bin (Figure 1). Taken together, the coverage of common SNPs variation by our chosen SNP set was rather balanced in both genecentric and intergenic regions, and the average number of common tag SNPs in this study was 30% of those in the HapMap JPT (r 2 > 0.8).
SNP marker distribution in the target region

Selection of a landmark SNP
In Stage 1 association test, 112 SNPs out of 1,317 (8.5%) showed significant associations (χ 2 test, p < 0.05) in the allelic (Figure 2A) or genotypic test. In Stage 2, 11 SNPs out of 112 yielded significant associations (p < 0.05) in the allelic ( Figure 2B) or genotypic test. However, only one SNP, namely SNP2140 (rs2412747), showed a replicated association for the allelic test (p < 0.05) in both stages ( Figure 2C). These results suggest that our association test is underpowered. None of the other SNPs showed replicated association for genotypic frequency (Additional file 1).
When the raw data from Stage 1 and Stage 2 were combined, SNP2140 in intron 33 of the ubiquitin protein ligase E3 component n-recognin 1 (UBR1) gene exhibited the nominal significant association with T2D in both the allelic (p = 0.0043, OR = 1.26, 95% CI = 1.07-1.48) (Figure 2C) and genotypic (p = 0.0081) tests. However, a logistic regression analysis revealed that the association between SNP2140 and T2D was not statistically significant after adjustment for age, gender, and BMI. In addition, when the standard Bonferroni's correction for multiple testing was applied (corrected for 1,317 SNPs in the allelic and genotypic tests), the association with SNP2140 was no longer significant. Moreover, FDR for 112 SNPs in the Stage 2 association test was not significant at a threshold value of 0.1 [25].
The results of association for the top 20 Stage 1-associations using the Armitage trend test are shown in Table 3. Based on the combined p values, three SNPs (SNP3347: rs1386166, SNP3096: rs156787, and SNP2072: rs936216) showed a smaller p value than the landmark SNP2140 in the allelic test (Table 3). However, none of the three SNPs showed significant association in Stage 2, although sample size and power to detect susceptibility SNPs was larger than in Stage 1. We believe that the SNP with replicated significance in both Stage 1 and Stage 2 is regarded as the more likely SNP for T2D association. Unfortunately, none of 112 SNPs reached the statistical confidence level in our sample population.
We next searched for novel SNPs or mutation in the UBR1 gene. Two novel non-synonymous SNPs (M1359I in exon 37 and R1712H in exon 47) were identified by resequencing the UBR1 gene in 48 samples. The SNPs were not common (MAF = 0.02 and 0.01) and subsequent typing of all of the samples revealed no significance for these two variants in the allelic test (p = 0.17 and 0.96 for combined samples). To confirm our data, it will be necessary to test association using a larger number of independent samples.

Fine LD mapping and a haplotype-based association test
Haplotype-based association tests assessed as a multilocus test can be more sensitive for the detection of association than assessment by a single-locus association test.

Discussion
We applied a region-wide association test as an alternative to GWAS. This approach has been effective in the identification of candidate susceptibility genes in the target regions for T2D [19,20] and rheumatoid arthritis [33,34] in the Japanese population. We hypothesized that there was a high probability of detecting disease susceptibility genes with the use of evenly-spaced common SNP markers in the target region defined by replicated linkage evidence. We could not find any strong variant with singlelocus association test or haplotype analysis in 1,794 Japanese subjects, although one SNP, SNP2140 (rs2412747), in intron 33 of the UBR1 gene showed nominal significance (p = 0.0043 for combined samples).
UBR1 encodes an E3 ubiquitin ligase of the N-end rule pathway, a conserved proteolytic pathway of the ubiquitin system whose substrates include proteins with destabilizing N-terminal residues. Recent data suggest that in Johanson-Blizzard syndrome (OMIM 243800), the pancreas exhibits pancreatic exocrine insufficiency and does not express UBR1 [35]. Ubr1 knockout mice exhibit decreased body weight or adipose tissue and an exocrine pancreatic insufficiency [35,36]. In addition, UBR1 mRNA is elevated in the atrophic muscles of diabetic rats [37]. At present, however, the pathophysiological mechanism by which the UBR1 influences T2D is unknown.
Genetic association tests aim to detect true positive associations between a trait and genetic polymorphisms while eliminating false positives. The results of this study have been evaluated within the framework of several criteria. Accordingly, the statistical power (with a MAF of 0.30) in Stage 1 or Stage 2 showed 22-60% or 29-76%, respectively, at detecting the OR of 1.2-1.4 at a significance level of 0.05 ( Figure 3A, B). Although the power using the whole sample size was estimated as 45-93%, it was not sufficient to detect an association with T2D ( Figure 3C). To achieve 60-80% power (with a MAF of 0.30 and OR of 1.2), 1,400-2,200 each for case and control samples (almost twice as many as those in each stage of this study) is needed. Recent GWAS analyzed more than one thousand cases and controls in Stage 1 and up to more than 10,000 samples in total [2][3][4][5][6][7], as recommended [38]. Replication studies with thousands of subjects are a priority to determine the consistency of our observations of the UBR1 gene, because it is predicted that true variant(s) have only a modest effect on T2D.
Our samples had the major weakness of an age difference between the case and control subjects (63.0 vs. 41.2). After logistic regression analysis, the association with SNP2140 did not remain statistically significant after adjustment for age and gender. Because aging is an important factor in diabetes development, it is necessary to con- firm the result of our association test using age-matched control subjects. In selecting the target region for association assessment, the initial linkage studies might overestimate the relevance of chromosome 15q as a candidate region in the Japanese population. The linkage of chromosome 15q14-22.1 to T2D and its related traits have been replicated in the Japanese at a significant level (LOD = 3.91). However, this LOD score was observed in cases with young onset (< 45 years of age) T2D, which is not reflected in our cases (average age = 63.0).
Though SNPs markers were placed, on average, every 9,169 bp, which is comparable in density to a genomewide association test using a 500 kb SNP set (~6 kb), it might be necessary to use denser SNPs with shorter intervals. Even though tag SNPs were not available at the onset of this study, tag SNPs with LD information and a highdensity SNP map from the HapMap data would afford useful resources.
Based on these results, we propose that the discovery of a landmark SNP, SNP2140, in the UBR1 gene was statistically marginal and did not reach the confidence level of this study. Other SNPs did not exhibit a strong effect on the risk of T2D within the target region, despite suggestive replicated evidence for linkage in the Japanese population. In addition, a replication study is of prime importance for confirmation of these results.

Conclusion
The two-stage association test on chromosome 15q14-22.1 failed to detect an association between T2D and SNPs in the target region. These observations could uncover new insights on the susceptibility variant(s) on chromosome 15q that might be used as a guide for further association tests, including replication and/or meta-analysis. Replication studies in other independent populations are a priority if the consistency of our observations is to be determined.

Competing interests
The author(s) declare that they have no competing interests.

Authors' contributions
YY carried out genotyping, sequencing, the association test, and LD analysis, and contributed to experimental design and manuscript writing. MM contributed to manuscript design and manuscript writing. DO, KN, and YF performed genotyping, performed the association test and LD analysis. TT, PK, and KK contributed to the development of statistical genetics concepts and maintained storage of the DNA samples. NN, TY, EI, HS, and NY supplied T2D DNA samples. HI and MI contributed to project design. All authors read and approved the final manuscript.