Introduction

Decades of rigorous efforts in understanding type-2 diabetes (T2D) genetics have mirrored the disease as apparently an outcome of altered quantitative traits (Plomin et al. 2009). Dissection of heritability of such traits, especially in nondiseased individuals can reflect various intermediary disease mechanisms that are neglected otherwise. In this regard, C-peptide, a 31 amino-acid long cleavage product of insulin synthesis has been largely overlooked. Owing to a higher plasma half-life (\(\sim \)30 min) than insulin (\(\sim \)4 min), C-peptide levels are a precise and worthy measure of insulin secretion (Faber et al. 1978). Since its discovery, C-peptide has been considered as a biologically inert molecule. However today, several studies detail its independent functional activity. Comprehensive studies in animal models of diabetes and early clinical trials in type-1 diabetic patients have established that replacement of C-peptide in physiological concentrations is helpful in improving diabetes-induced functional and structural abnormalities of peripheral nerves, kidneys and brain (Wahren et al. 2002). C-peptide binds to specific G-protein coupled receptors on human cell membranes and enhances \(\hbox {H}_{4} \hbox {K}_{16}\) acetylation at promoters of ribosomal RNA genes upon cellular entry (Rigler et al. 1999; Lindahl et al. 2010). Importantly, C-peptide increases nitric oxide synthesis to relax vascular smooth muscles and restores nerve conduction velocity by enhancing \(\hbox {Na}^{+}/\hbox {K}^{+}\) ATPase activity (Zhong et al. 2004). Besides, C-peptide stimulates several transcription factors that are essential for general cellular processes (Hills and Brunskill 2009). Specific cellular functions and benefits in preventing micro-vascular complications evidently signify that the C-peptide itself has a major regulatory role independent of T2D aetiology.

To date, only two genomewide association studies (GWAS) have been conducted for C-peptide in European, Hispanic, Afro-Caribbean and Thai populations (Comuzzie et al. 2012; Hayes et al. 2013) and a meta-GWAS in Europeans (Roshandel et al. 2018), contributing 11 unique single-nucleotide polymorphism (SNP) associations. However, none of these studies have included the Indian population. Among various linguistic groups in India, Indo-Europeans are a genetically diverse population that comprises several endogamous groups contributing to mixed ethnicity and high risk for T2D and related traits (Tabassum et al. 2011; Giri et al. 2016). A lack of C-peptide genetic studies in this inherently distinct and highly susceptible population guided us to search genomewide for associated variants to find out novel biology if any. Over the last decade, genetic discoveries for physiological traits have progressed with an extra-ordinary pace. However, the understanding of biology beneath such genetic associations lags well-behind. Translation from genetics to physiology is challenging in securing mechanistic inferences to pinpoint potential biomarkers. In such a scenario, integration of epigenetic and gene regulatory information can be fruitful in pinpointing underlying driving mechanisms. Variation in DNA methylation pattern affects gene transcription by altering binding of specific methyl CpG binding proteins. Gene regulatory sites are often seats for GWAS variants that influence gene expression in a tissue-specific manner. Thus, to identify under-pinning biological processes, we also studied DNA methylation signatures at identified genetic loci and probed publicly available human tissue gene expression data for C-peptide-associated genes. Therefore, in the current study we aimed to identify variants for C-peptide in healthy Indians that may affect the trait leading to plausible biological insights into functional mechanism. This study is a sole attempt for conducting GWAS for C-peptide in Indians.

Materials and methods

Ethical approval

This study was conducted in accordance with the principles of the Helsinki Declaration and was approved by the Ethics Committee of All India Institute of Medical Sciences and CSIR-Institute of Genomics and Integrative Biology. Informed written consent was obtained from all study participants.

Study population

The present GWAS was executed in two-stages, discovery phase and replication phase. The study participants are part of INdian DIabetes Consortium (INdian DIabetes Consortium 2011) and were collected from Diabetes Awareness Camps piloted in and around Delhi. The samples served as control individuals (normoglycemic) in T2D GWAS previously done in our laboratory (Tabassum et al. 2013). Sample details and criteria for selection have been outlined in a flow diagram (figure 1). The Indo-European ethnicity was defined by an individual’s birth place and place of origin of the parents if residing in northern part of Indian subcontinent. Anthropometric and clinical characteristics of study participants are provided in (table 1 in electronic supplementary material at http://www.ias.ac.in/jgenet/). Plasma C-peptide levels were measured by electrochemiluminescence immunoassay using Elecsys 2010 (Roche Diagnostics). The Hispanic study population is a part of VIVA LA FAMILIA study recruited between 2000 and 2005 in Houston, Texas, USA and comprises 815 children and adolescents from 263 Hispanic families between the ages of 4 and 19 years (Butte et al. 2006; Comuzzie et al. 2012).

Fig. 1
figure 1

Flow diagram of study participants. Sample numbers for various phases of the study and their criteria for selection are outlined.

Full size image

Discovery phase

Illumina Human 610-Quad BeadChips (Illumina, San Diego, USA) were used for the genomewide scan. The GenCall algorithm was employed to ascertain genotype calls (GenomeStudio, Illumina). A stringent quality control (QC) check for samples and SNPs was executed. Samples with low genotype call rate (\({\,<\,}95\%\)), heterozygosity and discordant sex were rejected. Identity-by-descent check was implemented to remove related or duplicate samples (\(\hbox {pi}\_\hbox {hat} {\,>\,} 0.1875\) for relatedness and \(\hbox {pi}\_\hbox {hat} {\,>\,} 0.98\) for duplication). SNP calls with missing rate \({\,>\,}5\%\), \(\hbox {MAF} {\,<\,} 0.01\) were excluded. SNPs with MAF 0.01–0.05 and Hardy Weinberg equilibrium (HWE) \(P{\,<}10^{-4}\) were removed. Also SNPs with MAF \({\,>\,}0.5\) and HWE \(P{\,<}10^{-7}\) were discarded. Linkage disequilibrium (LD) pruning of SNPs was done with genotyped autosomal SNPs applying the –indep-pairwise option of PLINK with a window size of 50 SNPs, step size of 5 and \(r^{2}{\,<\,}0.2\). Population stratification was checked through principal component analysis using HapMap Phase III population, CEU, CHB and YRI and the outliers were removed. Finally, a total of 552,011 autosomal SNPs were retained after stringent QC for association testing in 877 healthy normoglycemic individuals. C-peptide values were inverse normalized exercising an inbuilt R command (www.r-project.org). Association of QC-passed SNPs with C-peptide was tested by linear regression analysis under additive model adjusting for age, sex, body-mass index (BMI) and first three principal components using PLINK. Any deviation of observed P values from the P values expected under a null hypothesis was checked by constructing a quantile–quantile plot (QQ plot). The observed \(-\hbox {log}_{10}\) adjusted P values were plotted against expected \(-\hbox {log}_{10}~P\) values. Median \(\chi ^{2}\) statistics was used to compute genomic inflation factor \(\lambda \) using PLINK. QQ and Manhattan plots were constructed using R (www.r-project.org).

Validation phase and meta-analysis

Associated SNPs with discovery phase P value \({\,<}10^{-3}\) were selected for replication in an independent sample set using Illumina Golden Gate assay. SNPs with genotype call \({\,<\,}90\%\), genotype confidence score \({\,<\,}0.25\), GenTrans score \({\,<\,} 0.60\), cluster separation score \({\,<\,}0.40\), \(\hbox {MAF} {\,<\,} 0.01\) and HWE \(P {\,<\,} 1 \times 10^{-5}\) were removed. Also, samples with call rate \({\,<\,}90\%\) were excluded. Finally, a set of 1350 SNPs were tested for association in an independent set of 1829 healthy normoglycemic individuals. QC-passed SNPs were assessed for association with inverse normalized values of C-peptide through linear regression model adjusting for age, sex, BMI and first three principal components. An inverse variance method was used to meta-analyse discovery and replication phase results under a fixed effect model by METAL (www.sph.umich.edu/csg/abecasis/Metal).

In-silico replication in Hispanic population-VIVA LA FAMILIA study

For in-silico replication of identified signals, association data on C-peptide was requested from the VIVA LA FAMILIA Study which identifies genetic variants influencing paediatric obesity and its comorbidities in Hispanic population (Butte et al. 2006; Comuzzie et al. 2012). The data were obtained for 815 children and adolescents of 263 Hispanic families (Comuzzie et al. 2012). The C-peptide association in Hispanics was adjusted for age, sex and BMI. We meta-analysed by combining summary statistics of our lead signals of C-peptide GWAS following a sample size based analysis by METAL. An overall z-statistic and P value were calculated from a weighted sum of the individual statistics. Weights were taken proportional to the square-root of the number of individuals examined in each sample and selected such that the squared weights sum to 1.0. The z-statistic summarized the magnitude and the direction of effect relative to the reference allele and all studies are aligned to the same reference allele.

Fig. 2
figure 2

Manhattan plot of associated P values for C-peptide in Indians. Lead variants showing the strongest association after meta-analysis are shown in italics. The \(-{\log }_{10}~P\) values for association of directly genotyped SNPs are plotted as a function of genomic position (National Center for Biotechnology Information Build 37). P values were determined using linear regression adjusted for age, sex, BMI, PC1, PC2 and PC3 in discovery phase analysis.

Full size image

DNA methylation analysis

We studied DNA methylation signatures in peripheral blood using Infinium HumanMethylation450 K BeadChips of 233 normoglycemic Indians considered in discovery phase of GWAS. Data generation has been described previously (Giri et al. 2017). Sample QC involved extreme missingness, sex disparity checks and samples with failed bisulphite conversion (samples having intensity three SD away from mean intensity for C1, C2, C3 and C4 probes). CpGs with bead count \({\,<\,}3\) in 5% of samples and detection P value \({\,>\,}0.01\) for \({\,<\,}1\%\) of samples were excluded. CpGs in sex chromosomes (X and Y), established cross-hybridization probes and polymorphic CpGs were also removed. CpGs with 100% call rate in all the samples have only been considered for analysis. Methylation outlier value for CpGs was fixed by fixMeth-Outliers command in minifi. Methylation data were regressed for confounders such as cell composition, age, sex, BMI, bisulphite conversion efficiency and plate number. Methylation data were extracted for identified genetic variants and tested for SNP-CpG association using linear regression model in PLINK.

Power calculation

We calculated the power of study using Quanto software (http://biostats.usc.edu/Quanto.html) under an additive genetic model for a range of allele frequencies (0.01–0.5). Two-tailed test at significance level of 0.05 with effect size ranging from 0.0001 to 1.514 was used for calculating power. For final power analysis, MAF for combined samples and calculated effect size from meta-analysis was used (figure 1 in electronic supplementary material).

Results

Under the null distribution, QQ plots indicated a good agreement (genomic inflation factor \(\lambda = 1\), mean \(\chi ^{2} = 0.97\)) (figure 2 in electronic supplementary material). In discovery phase, the strongest association was observed for HDAC4 variant, rs1796447 (\(P = 2.85 \times 10^{-6}\)) (figure 2). This association receded in meta-analysis (\(P = 0.002\)). Two other variants residing on chromosome 20 presented zinc-finger protein 831 (ZNF831) as a second lead signal in discovery phase (rs3026620, \(P = 4.17 \times 10^{-6}\); rs6128532, \(P = 3.49 \times 10^{-5}\)). The association with ZNF831 was sustained by two variants, rs6128525 and rs2064865 in meta-analysis (\(P = 7.51 \times 10^{-5}\) and \(P = 6.11 \times 10^{-4}\) respectively). Besides, a region FHIT marked by three variants on chromosome 3, featured strong association in discovery phase but lost when meta-analysed (rs17684830, \(P = 7.32 \times 10^{-6}\); rs6414610, \(P = 1.54 \times 10^{-5}\); rs7373003, \(P = 5.52 \times 10^{-5}\) in discovery phase). Also, GABRA6 locus on chromosome 5 followed a similar trend for association to C-peptide levels with significant discovery phase P values of \(8.25 \times 10^{-6}\) and \(2.31 \times 10^{-5}\) for SNPs rs4454083 and rs1444740, respectively. In meta-analysis, GABRA6 displayed nominal significance (rs4454083, \(P = 0.007\)). A region on chromosome 10 mapping to nonprotein coding RPS3AP5 locus was observed as another robustly associated signal in discovery phase (rs11201130, \(P = 7.6 \times 10^{-6}\); rs12783251, \(P = 2.21 \times 10^{-5}\)). This association remained significant in meta-analysis with rs11201130 (\(P = 5.38 \times 10^{-4}\)).

In meta-analysis combining discovery and replication phases, the strongest association was observed at a novel locus, long noncoding RNA (lncRNA) gene, AC073333.8 on chromosome 7 (two variants rs1674809 and rs818488) (table 1). Another novel lncRNA gene, RP1-209A6.1 surfaced as lead signal after meta-analysing discovery and replication phases. Besides, two other variants, an intergenic variant rs10138141 in ETS2 and an exonic variant in SLC15A5 featured suggestive association (table 1).

An in silico replication and meta-analysis of summary statistics of three lead signals (rs1013841, rs9460790 and rs1527014) in Hispanic children and adolescents did not improve the association status (table 2). All three variants retained nominal association with C-peptide after meta-analysis (\(P {\,<\,} 0.01\)), however the directionality of association of variants rs9460790 and rs1527014 was observed to be reversed in Hispanics. Heterogeneity was observed to be more than 70% across populations for all three variants (table 2). Also, we evaluated the status of previously known GWAS variants in our study. For C-peptide, we did not observe any association for reported GWAS locus, GCKR (table 2 in electronic supplementary material). To attest functional relevance of identified variants, we studied and integrated whole genome DNA methylation data from peripheral blood in Indians. We found that certain genetic variants within ETS2 and ZNF831 loci were strongly associated with variable DNA methylation pattern at multiple genic CpG sites in Indians (table 3). Further, while investigating the publicly available human tissue gene expression data (The GTEx Consortium 2015), we noticed that all associated genes (except for RP1-209A6.1) were expressed in human adipose tissue (figure 3 in electronic supplementary material). Moreover, we observed that all the six genes showed considerable expression in lungs, in particular.

Table 1 SNPs showing association with C-peptide levels at meta-analysis P value \({<}10^{-3}\).
Full size table
Table 2 In silico replication and meta-analysis of novel SNPs rs1013841, rs9460790 and rs1527014 in Hispanic population.
Full size table
Table 3 Meth-QTL analysis for novel variants in 233 Indians who have been genotyped in discovery phase.
Full size table

Discussion

The present study attempted to identify genomewide common variants that govern plasma C-peptide in Indians. Our results highlight polygenic nature of C-peptide inheritance. It is well-established that the genetic variance in complex diseases/traits is mainly contributed by multiple SNPs with smaller effects that may often be missed out due to stringent GWAS P value thresholds and multiple testing corrections (Yang et al. 2010; Fransen et al. 2015). In view of a limited sample size (\(n = 2706\)), we identified multiple such variants that fail to attain GWAS significance but are associated nominally with C-peptide in Indians. Trait-associated GWAS SNPs are highly enriched in gene regulatory regions (Nica et al. 2010). To understand biological rationale for our observed associations, we studied DNA methylation marks and explored publicly available gene regulatory data (gene expression, histone active marks, repressive marks, DNase I HS sites, transcription factor binding sites etc.). Associated loci generally harbour several noncoding variants whose signal spans several genes. Multiple lines of evidence indicate noncoding RNA transcription as a proxy for functional activity of such putative genomic regions. Seemingly, in the present study, we obtained lead-associated variants in noncoding RNA genes, AC073333.8, RP1-209A6.1 and RPS3AP5. AC073333.8, represented by two variants, surfaced as the strongest signal for association to C-peptide in Indians. Both variants reside in a highly active chromatin region (DNase I hypersensitive) of a novel long noncoding RNA, AC073333.8 that is exonic antisense to a basic leucine zipper and W2 domain 2 protein, BZW2. AC073333.8 lncRNA is plausibly a regulator of BZW2 gene that needs to be validated experimentally. Previously, BZW2 has been documented for association to Alzheimer’s disease (Sherva et al. 2013) but never with plasma C-peptide levels. BZW2 has been earlier reported to get significantly upregulated upon administration of insulin analogue in murine colon cancer model (Hvid et al. 2012). C-peptide is well-known to regulate the availability of monomeric insulin in physiological processes (Jörnvall et al. 2010). Thus, during hyperinsulinaemia, variation in C-peptide levels may tend to alter AC073333.8 activity and hence BZW2 function.

Additionally, two other noncoding genes, RP1-209A6.1 and RPS3AP5 cropped up for the first time in association with C-peptide. Earlier, no association of any disease/trait has been observed for lncRNA RP1-209A6.1. However, RPS3AP5, a ribosomal protein pseudogene has been originally reported to be a GWAS locus for childhood obesity in Hispanic population (Comuzzie et al. 2012).

Besides noncoding genes, other protein gene regulators, ZNF831 and ETS2, also appeared as strong genetic variants and simultaneously as robust meth-QTLs in our study. The two variants that mark ZNF831 reside in a polycomb-repressed chromatin region. ZNF831 has previously been associated with hypertension and blood pressure at genomewide significance (International Consortium for Blood Pressure Genome-Wide Association Studies et al. 2001; Levy et al. 2009). Additionally, C-peptide has been well-elucidated to be significantly associated with hypertension in various populations (Chen et al. 1995; Purohit and Mathur 2013). Moreover, ETS2, a transcription factor that regulates developmental and apoptotic genes, has been formerly associated to genomewide significance with BMI (Locke et al. 2015). Higher BMI results in drastically higher plasma C-peptide levels (Chan et al. 2004). Also, reports have demonstrated that C-peptide upon cellular entry stimulates the activation of key transcription factors (Wahren et al. 2002). Thus, these studies reasonably justify the observed associations in our study. Further, we obtained an exonic nonsynonymous variant in a key solute carrier family protein, SLC15A5 associated with C-peptide in Indians. SLC15 is a family of membrane transporters known for key role in cellular uptake of peptides from dietary protein digestion, ultrafiltration or brain–blood efflux (Alexander et al. 2015). Prior studies have documented genomewide association of SLC15A5 with visceral fat (Fox et al. 2012). Incidentally, C-peptide concentrations are seen to predict and increase visceral fat adiposity in nondiabetic individuals (Seidell et al. 1990; Tong et al. 2005). Hence, the presence of alternate alleles of SLC15A5-associated variant and modulation of C-peptide levels may be accountable for visceral adiposity in Indians to some extent.

Interestingly, we observed that all the six genes are moderately expressed in human lung tissue. C-peptide has essentially been demonstrated to alter the balance between pro-inflammatory and anti-inflammatory signalling in lungs to ameliorate inflammatory response after a haemorrhagic shock in male Wistar rats (Chima et al. 2011). The findings are indicative of lungs being alternatively important organs for C-peptide physiology other than pancreas and adipose tissue.

To summarize, the present study endeavoured to explore the heritability of C-peptide in Indians, a highly susceptible population for metabolic disorders. The GWAS figured several variants robustly associated with C-peptide that fail to be genomewide significant. The observations fortify the nature of polygenic traits. The study is unique to investigate and secure biological basis to observed genetic associations for varied C-peptide levels in Indians.