Regulation of alternative splicing in human obesity loci

Abstract

Objective

Genome-wide association studies (GWAS) have identified multiple obesity susceptibility loci, yet the mechanisms by which these loci influence obesity remain unclear. We hypothesized that alternative splicing could contribute to obesity by regulating the transcriptomic and proteomic diversity of genes in these loci.

Methods

Seventy-two of the 136 genes at the 13 obesity loci encoded multiple protein isoforms based on database search. Thus, we analyzed alternative splicing of these genes in adipose tissue samples from the Metabolic Syndrome in Men (METSIM) population-based study, and from two weight loss intervention studies (surgical and very low calorie diet).

Results

Alternative splicing was confirmed in 11 genes with PCR capillary electrophoresis in human subcutaneous adipose tissue. Interestingly, we observed differential splicing of TRA2B, BAG6 and MSH5 between lean normoglycemic and overweight type 2 diabetic individuals. Of these genes, we detected fat depot dependent splicing of TRA2B and BAG6, and weight loss induced regulation of MSH5 splicing in the intervention studies. Finally, BMI was a major determinant of TRA2B, BAG6 and MSH5 splicing in the combined data.

Conclusions

Our study provides evidence for alternative splicing in obesity loci, suggesting that alternative splicing at least in part mediates the obesity-associated risk in these loci.

Clinical trial number

ISRCTN67529475

INTRODUCTION

Genome wide association studies (GWAS) increased the number of loci associated with susceptibility to obesity. Polymorphisms at the obesity loci can possibly affect alternative splicing resulting in functionally different protein isoforms. Additionally, expression of several RNA processing genes is reduced in individuals with obesity (1), suggesting link between dysregulation of alternative splicing and obesity.

Alternative splicing of genes is an essential regulatory mechanism responsible for the generation of transcript and protein diversity. Aberrant splicing is increasingly recognized in many forms of diseases and has been linked to obesity, insulin resistance and cholesterol homeostasis (2, 3). Interestingly, rare gene variants in subjects with obesity are enriched in splicing regulatory regions indicating that genetic regulation of splicing may associate with obesity (3).

We hypothesized that alternative splicing of genes at obesity loci may contribute to obesity. To investigate this hypothesis, we investigated 136 genes at 13 of the 97 known obesity loci (4, 5, 6). Of these genes, 72 had multiple protein isoforms, and 25 of these resulted from alternative splicing of cassette exons. These genes were screened for alternative splicing in adipose tissue samples in three different study cohorts: a population-based Metabolic Syndrome in Men (METSIM) study and two weight loss intervention studies: Kuopio Obesity Surgery Study (KOBS) and very low calorie diet study (VLCD) (Fig. 1).

Fig. 1.

Analysis pipeline to identify alternatively spliced variants in the obesity loci. The genes were selected based on the known splicing events. Splice variants including 5′ and 3′ splicing events as well as cassette exons bigger that 600bp were excluded from the analysis due to the limitation of the method used.

METHODS

Study subjects

Subcutaneous adipose tissue samples from 46 men were obtained from the METSIM study (7). Two independent follow-up weight loss study groups were used to examine the effect of weight loss on alternative splicing. First, subcutaneous and visceral adipose tissue samples were taken during the Rouxen-Y gastric bypass surgery from a total of 108 participants with severe obesity of the KOBS study. Additionally, subcutaneous fat biopsies were collected a year after the bariatric surgery (7). Second, subcutaneous adipose tissue samples were obtained from 32 non-diabetic participants of a 7-week VLCD intervention study followed by a 24-week weight maintenance period. The tissue biopsies were collected at all visits (baseline, 7 weeks and 24 weeks) (8). Basic characteristics of the study participants are given in Table 1. The study protocols were approved by the Ethics Committee of Northern Savo Hospital District and carried out in the accordance with Helsinki Declaration.

Table 1.

Characteristics of the study groups.

	METSIM			Kuopio Obesity Surgery Study			Very low calorie diet
	NGT n= 26	T2DM n= 20	p value	Baseline n=108	1 year Follow-up n=81	p value	Baseline n=32	7 wks VLCD (1) n=32	24 wks WM (2) n=32	p value (1)	p value (2)
Sex male/female	26/0	20/0		34/74	26/55		9/23	9/23	9/23
Age years	56.5 ± 5.5	56.7 ± 5.3	0.902	47.1±9.1	47.6±9.0		48.8 ± 8.7	49.0 ± 8.8	49.5 ± 8.8
Body mass index kg/m2	23.1 ± 1.4	26.1 ± 3.9	0.003	45.0±6.3	34.6±5.8	2×10−37	34.7 ± 2.7	29.9 ± 2.4	30.3 ± 2.8	3×10−24	2×10−16
Fasting plasma glucose mmol/l	5.5 ± 0.3	7.1 ± 1.0	1×10−6	6.7±1.7	5.6±0.9	2×10−8	6.1 ± 0.8	5.6 ± 0.6	5.7 ± 0.6	1×10−4	6×10−5
Fasting insulin pmol/l	27.6 ± 8.8	62.9 ± 47.3	0.004	135.7±135.5	71.4±67.4	3×10−10	82.8±46.2	40.8 ± 19.2	51.0 ± 27.0	3×10−6	4×10−6
Total cholesterol mmol/l	5.3 ± 1.0	5.7 ± 1.1	0.215	4.4 ± 1.1	4.4 ± 0.9	0.291	5.1 ± 0.7	4.2 ± 0.6	5.0 ± 0.6	7×10−10	0.481
HDL cholesterol mmol/l	1.5 ± 0.4	1.7 ± 0.6	0.217	1.1 ± 0.3	1.4 ± 0.4	3×10−19	1.4 ± 0.3	1.3 ± 0.3	1.6 ± 0.4	4×10−3	1×10−4
Triglycerides mmol/l	1.0 ± 0.3	1.8 ± 1.7	0.045	1.7 ± 0.8	1.1 ± 0.5	6×10−11	1.4 ± 0.6	0.9 ± 0.3	1.1 ± 0.4	4×10−5	5×10−4
Fasting free fatty acids mmol/l	0.33 ± 0.18	0.51 ± 0.23	0.007	0.64 ± 0.23	0.51 ± 0.18	1×10−4	0.48 ± 0.13	0.72 ± 0.23	0.45 ±0.20	1×10−6	0.304

Data are expressed as mean ± SD. Data of the same individuals at different time points were compared using paired t test. p values <0.05 are shown in boldface. VLCD, very low calorie diet; WM, weight maintenance

Splicing analysis

PCR-capillary electrophoresis was used to examine the relative ratio of splice variants (3100 DNA Genetic Analyzer, ABI). Quantification of peak areas was performed with Peak Scanner Software v1.0 (ABI). Primers, shown in Table S1, spanning the regions around alternatively spliced regions allow for detection of cassette exons but not alternative 3′/5′ exons.

Statistical analysis

Data are presented as mean±SD. Data from the same individuals at different time points were compared with the two-tailed paired samples t-test. Non-parametric Spearman’s correlation test was used to assess correlations with metabolic traits. Body mass index (BMI), insulin, glucose and free fatty acid (FFA) levels were logarithmically transformed to obtain a normal distribution. The main effects of BMI, fasting insulin, fasting glucose, fasting FFA and age on TRA2B, BAG6 and MSH5 splicing were evaluated by linear regression model.

RESULTS

Selection of candidate genes in obesity loci

One-hundred and thirty-six genes in the 13 known obesity loci were selected (4, 5). All genes in the same haplotype block with the most associated (lowest p-value) obesity single nucleotide polymorphism (SNP) were considered as potentially alternatively spliced genes. Out of these 136 genes, 72 genes had known isoforms derived from alternative splicing, based on the UniProt database (Table S2). From these, we selected 25 genes with known exon cassettes within the genes for PCR capillary electrophoresis analysis. Eleven alternatively spliced genes were detected in adipose tissue. Our analysis pipeline is shown in Fig. 1.

Alternative splicing in lean normoglycemic and overweight type 2 diabetic subjects

We first investigated the subcutaneous fat splicing profile of these 11 genes, using samples from the METSIM study. Subcutaneous fat distribution of TRA2B var2 was 28.2% lower (p=3×10⁻⁵), BAG6 varX7 was 54.1% higher (p=2×10⁻⁵) and MSH5 var1 was 24.4% lower (p=0.004) in overweight individuals with type 2 diabetes compared to lean normoglycemic individuals (Fig. 2a). These 3 genes were selected for further analysis (Fig. 1).

Fig. 2.

(a) The relative proportion of TRA2B var2, BAG6 varX7 and MSH5 var1 in the subcutaneous fat of lean normoglycemic individuals (white bars, lean NGT) and overweight type 2 diabetic (black bars, overweight T2DM) individuals in the METSIM study. Mean±SD shown. Comparison made using t- test. (b) Difference in expression levels of TRA2B, BAG6 and MSH5 splice variants between subcutaneous fat and visceral fat in the KOBS study. Comparison made using paired t- test. Total gene expression of TRA2B, BAG6 and MSH5 did not differ between fat depots (data not shown) (c) Change in expression level of MSH5 var1 during the year after gastric bypass surgery (white bar, KOBS study) and during a 7 week very low calorie diet followed by a 24-week weight-maintenance (WM) period (black bar, VLCD study) in subcutaneous fat. Comparison made using paired t- test. *p< 0.05, **p< 0.01, ***p<0.001. No change in total gene expression of TRA2B, BAG6 and MSH5 was detected in response to weight loss (data not shown).

Splicing in subcutaneous and visceral fat

Splicing profiles of TRA2B, BAG6 and MSH5 genes were compared between subcutaneous and visceral fat from 108 participants of the KOBS study. We found a 3.3% lower proportion of TRA2B var2 (p=0.011) and a 32.8% lower proportion of BAG6 varX7 (p=5×10⁻¹²) in visceral fat compared to subcutaneous fat. The splicing pattern of MSH5 did not differ between fat depots (Fig. 2b).

Differential alternative splicing in response to weight loss

The splicing profiles of TRA2B, BAG6 and MSH5 were assessed in subcutaneous fat before and after gastric bypass obesity surgery (KOBS study) and dietary weight loss (VLCD study). We detected a 12.4% increase in the proportion of MSH5 var1 after surgery-induced weight loss (p=0.002) and a 24.8% increase in the proportion of MSH5 var1 (p=0.005) after diet induced weight loss (Fig. 2c). We did not detect any difference in splice variant distribution of TRA2B and BAG6 before and after weight loss (data not shown).

Determinants of TRA2B, BAG6 and MSH5 splicing in adipose tissue

To interrogate the determinants of alternative splicing, we pooled the data of the subcutaneous adipose tissue biopsies from the METSIM and KOBS studies (n=154). TRA2B splicing was associated with BMI, fasting insulin and FFA levels in univariate analysis (p=4×10⁻⁴, p=0.004 and p=0.001 respectively) whereas forward stepwise multivariate analysis indicated that only BMI was an independent determinant of TRA2B splicing (p=0.001). BMI associated with BAG6 and MSH5 splicing in univariate analysis (p=0.001, p=0.034 respectively) and was an independent predictor of BAG6 and MSH5 splicing profile in forward stepwise multivariate analysis (p=0.002, p=0.015 respectively) (Table S3).

DISCUSSION

The functional relevance of most obesity-associated SNPs remains unknown. SNPs in obesity loci could alter splicing of regional genes, resulting in changes in proteome composition. The current study provides a systematic evaluation for alternative splicing at the 13 GWAS loci.

Our main finding was that 3 genes at two obesity loci were alternatively spliced in type 2 diabetic overweight individuals and at least one of them, MSH5, was also regulated by weight loss and associated with BMI, suggesting functional significance of alternative splicing. The weight-loss-associated effect on alternative splicing of MSH5 demonstrates that obesity risk loci can also be regulated by weight loss, suggesting a potential gene-environment interaction in the regulation of alternative splicing. MSH5 is involved in DNA damage repair (9), and polymorphisms of MSH5 have been associated with cancer, immune diseases and reproductive disorders. The role of MSH5 in obesity remains unknown.

Interestingly, MSH5 and BAG6 genes are localized on chromosome 6p21, suggesting that splicing of genes in the HLA chromosomal region, known to be strongly regulated by splicing, might contribute to obesity.

Additionally, we observed differential splicing of TRA2B and BAG6 between visceral and subcutaneous fat that could contribute to functional differences between adipose tissue depots. TRA2B is a known RNA binding protein that regulates mRNA splicing in a concentration-dependent manner. Importantly, exclusion of exon 2 form TRA2B final mRNA regulates expression of TRA2B itself (10). Our observation that BMI is an independent predictor of TRA2B splicing in subcutaneous fat is in line with a previous report demonstrating reduced expression of TRA2B in individuals with obesity (1). BAG6 plays a role in apoptosis and gene regulation, in addition to regulating protein synthesis and degradation. The function of different BAG6 isoforms remains elusive, but recent studies have demonstrated different sub-cellular expression of BAG6 protein isoforms, with exclusive nuclear expression of the isoform encoded by transcript variant BAG6 varX7 (11). Our analysis indicates BMI as the strongest determinant of BAG6 splicing in subcutaneous fat. A mechanism on how the aberrant splicing of BAG6 contributes to obesity remains to be determined.

It is important to point out the limitations of our study. PCR capillary electrophoresis is not suitable for detection of alternative first and last exons. However, this method allows for efficient screening of subtle changes in splicing of cassette exons and direct comparison of splicing pattern across different tissues. Additionally, the small sample size used in this study does not provide enough statistical power to test an association between the genotypes and splicing of selected candidate genes. However, our study design, which utilized the power of paired tissue samples, allowed us to demonstrate statistically significant tissue-dependent splicing for TRA2B and BAG6, in addition to weight loss-dependent changes in MSH5 splicing in two independent intervention studies. Due to limitations of the method we were not able to study all 97 known obesity loci.

CONCLUSION

Our findings imply that obesity-associated SNPs may act through regulation of splicing and could underlie the pathogenesis of obesity in individuals carrying these risk SNPs. This study revealed the need of utilizing whole transcriptome RNA sequencing to detect all splicing variants in obesity loci in a larger sample size with a longitudinal setting.

What is already known about this subject?

• There are 97 known loci associated with obesity.

• The functional significance remains unknown for the most of the obesity-associated SNPs.

• Aberrant splicing has been linked with metabolic diseases.

What does this study add?

• Alternative splicing was detected in the 13 GWAS obesity loci.

• Differential splicing was observed between overweight type 2 diabetic and lean normoglycemic individuals.

• Change in alternative splicing was observed in response to weight loss.

Abbreviations

BMI

body mass index

HOMA-IR

homeostasis model assessment of insulin resistance

KOBS

Kuopio Obesity Surgery study

METSIM

Metabolic Syndrome in Men study

VLCD

very low calorie diet