Introduction

The insulin receptor (INSR) exists in two protein isoforms arising from inclusion (INSR-A) or skipping (INSR-B) of exon 11. Expression of the INSR gene variants depends on the tissue type and stage of tissue development. The INSR-A isoform is predominantly expressed in fetal cells and plays a role in fetal development, whereas the INSR-B isoform is expressed in adult differentiated cells. These two protein isoforms have different functions: INSR-A promotes growth through its ability to bind IGF-II and proinsulin, whereas INSR-B is a highly specific receptor for insulin. It is expressed predominantly in insulin-sensitive tissues and thus regulates glucose homeostasis (reviewed in Belfiore et al [1]). INSR splicing has mostly been studied in skeletal muscle in relation to type 2 diabetes. In most, but not all, of the studies high skeletal muscle expression of the INSR-B variant has been associated with type 2 diabetes and insulin resistance (reviewed in Belfiore et al [1]). An increased level of INSR-B was detected in isolated adipocytes from ten patients with type 2 diabetes compared with 11 normoglycaemic patients [2]. The effect of weight loss on adipose tissue INSR splicing has not been studied.

We have previously shown that expression of several splicing factors is reduced in liver and skeletal muscle of obese individuals [3]. In addition, we have shown that the splicing of TCF7L2 is regulated by weight loss in adipose tissue and liver [4]. Thus, we hypothesised that adipose tissue INSR splicing is modified by weight loss through changes in splicing factor gene expression levels. To test this hypothesis we determined the association of INSR splicing with metabolic variables and the expression of splicing factors in three independent studies (n = 189 combined): Kuopio Obesity Surgery study (KOBS, n = 108) [4], a very low calorie diet (VLCD) intervention (n = 32) [5] and the population-based Metabolic Syndrome in Men study (METSIM, n = 49) [4].

Methods

Participants and clinical studies

Subcutaneous adipose tissue samples were collected at the time of Roux-en-Y gastric bypass surgery from a total of 108 morbidly obese individuals participating in the ongoing KOBS study (45 with type 2 diabetes, 63 non-diabetic individuals). In addition, subcutaneous fat biopsies were taken 1 year after the surgery. Visceral biopsies (n = 81) were collected at the baseline [4]. Two independent study groups were used for the replication of the results. First, subcutaneous adipose tissue samples were taken from a study of 32 non-diabetic individuals recruited into a dietary intervention study consisting of a 7 week long VLCD followed by a 24 week weight-maintenance period. During the weight loss period the energy intake was 2,510 kJ/day (600 kcal/day) [5]. Subcutaneous tissue biopsies were collected at all visits (baseline, 7 weeks and 24 weeks). Second, subcutaneous adipose tissue samples from a total of 49 men (21 with type 2 diabetes, 28 with normal glucose tolerance) were included from the population-based METSIM study [4] (Table 1). Diabetic status was determined using the ADA 2003 criteria. The study protocols were approved by the Ethics Committee of Northern Savo Hospital District and carried out in the accordance with the Helsinki Declaration.

Table 1 Characteristics of the study groups

Gene expression and splicing analysis

A PCR-capillary electrophoresis method was used to determine the relative ratio of INSR splice variants (ABI Prism 3100 DNA Genetic Analyzer, Applied Biosystems, Foster City, CA, USA). Quantification of peak area was performed with Peak Scanner Software v1.0 ( Applied Biosystems, Foster City, CA, USA). Total gene expression of HNRNPA1, SF3A1, SFRS7, SFRS10 and INSR normalised to RPLP0 was analysed by quantitative (q)PCR, using SYBR Green chemistry (KAPA SYBR FAST qPCR Kit, Kapa Biosystems, Woburn, MA, USA).

Statistical analysis

Data are presented as mean ± SD. Data from the same individuals at different time points were compared using the two-tailed paired samples t test. Correlations were assessed using non-parametric Spearman’s correlations and with partial correlations. Insulin levels were logarithmically transformed (log10) to obtain a normal distribution. The main effects of BMI, age, sex, fasting insulin, fasting glucose and study group on INSR splicing were evaluated by general linear model. A p value <0.05 was considered significant.

Results

The proportion of INSR-B mRNA differed in the study groups at baseline (Fig. 1a). The relative proportion of INSR-B mRNA in the subcutaneous fat increased by 9.1% in response to surgery (p  = 1 × 10−5; Fig. 1b) and by 8.1% during the VLCD (p  = 1 × 10−4; Fig. 1c). The relative proportion remained increased by 6.7% after the weight-maintenance period in the VLCD study (p  = 2 × 10−4, Fig. 1c). No change in total gene expression of INSR was detected. Thus, the effects on splicing were independent of transcriptional regulation (electronic supplementary material [ESM] Fig. 1a). No difference in expression of INSR-B was observed between visceral and subcutaneous fat depots at baseline (ESM Fig. 1b).

Fig. 1
figure 1

The relative proportion of INSR-B in subcutaneous fat (a) in the METSIM, VLCD and KOBS studies at baseline and in response to (b) obesity surgery in the KOBS study (white bar, baseline; black bar, 1 year post surgery) and (c) to a 7 week VLCD followed by a 24 week weight-maintenance period (white bar, baseline; black bar, 7 weeks VLCD; grey bar, 24 weeks weight maintenance). (d) Relative proportion of the INSR-B variant in individuals with and without type 2 diabetes in the METSIM and KOBS studies. White bars, non-diabetic participants; black bars, participants with type 2 diabetes. (e) Scatter plot demonstrating the correlation of INSR-B with logarithmically transformed fasting insulin levels in pooled samples from the KOBS, VLCD and METSIM studies (r  =  −0.649, p  = 3 × 1022). Black diamonds, non-diabetic participants (n = 123, r  =  −0.676, p  =  4 × 1016); white circles, participants with type 2 diabetes (n  =  66, r  =  −0.601, 3 × 107). Mean ± SD shown. *** p < 0.001 (b,c) vs baseline of same study (d) and vs non-diabetic participants of same study

Next, we investigated clinical variables that associate with INSR splicing. Fasting glucose correlated negatively with INSR-B splicing in the KOBS study (r  =  −0.261, p = 0.010, ESM Table 1). INSR-B expression was 7.1% higher in normoglycaemic individuals compared with patients with type 2 diabetes in the METSIM study (p = 5 × 10−4; Fig. 1d). However, the most consistent finding was a negative correlation between fasting insulin levels and INSR splicing in subcutaneous fat in all three studies, KOBS (r = −0.348, p = 5 × 10−4), VLCD (r = -0.417, p = 0.020) and METSIM (r = −0.522, p = 1 × 10−4). Significant correlation was also observed between insulin levels and INSR splicing in visceral fat (KOBS n = 81, r = −0.338, p = 8 × 10−4; ESM Table 1). Additionally, negative correlation was observed between HOMA-IR and INSR-B at baseline in all three studies (ESM Table 2).

In pooled data from the KOBS, VLCD and METSIM studies (n = 189) INSR-B variant correlated strongly with fasting insulin levels (r  =  −0.649, p  = 3 × 10−22; Fig. 1e). Insulin was the strongest determinant of INSR-B splicing (p  = 9 × 10−9) followed by BMI (5 × 10−7), age (p =  0.009) and sex (p = 0.036) in a multivariate general linear model. After controlling for study group, insulin remained the strongest determinant of INSR-B splicing (p  = 7 × 10−7) (ESM Table 3). Fasting glucose level did not associate with INSR-B splicing.

Expression of HNRNPA1, SF3A1, SFRS7 and SFRS10 has been shown to be lower in the skeletal muscle and liver of obese individuals compared with lean individuals [3]. Therefore, we analysed the effect of weight loss on the adipose tissue expression of these genes and the association with INSR splicing. The expression of HNRNPA1, a known regulator of INSR splicing [6], was increased in response to surgery-induced weight loss (p = 0.001, data not shown) and its expression correlated negatively with INSR-B expression in the KOBS (r  = −0.427, p  = 1 × 10−5), but not significantly in the VLCD study (r  = −0.222, p  = 0.238). However, after pooling adipose tissue samples from all time points of the KOBS and VLCD studies and after controlling for insulin levels the correlation between INSR-B and HNRPNA1 was significant in both KOBS and VLCD studies (r = −0.226 , p = 0.005 and r  =  −0.330, p  =  0.012 respectively) (ESM Table 4). In addition, we observed negative correlations of INSR-B expression with SF3A1 and SFRS7 expression in the KOBS study (ESM Table 4). No correlation between examined splicing factors and fasting insulin levels was detected (ESM Table 5).

Discussion

In this study we report, for the first time, that weight loss regulates alternative splicing of INSR in adipose tissue. The proportion of INSR-B mRNA variant was increased in response to weight loss induced by both obesity surgery and VLCD (Fig. 1b, c). Interestingly, insulin levels correlated strongly with INSR splicing (Fig. 1e) in both subcutaneous and visceral fat (ESM Table 1), suggesting common regulatory mechanisms related to insulin action. Finally, we observed a correlation between alternatively spliced INSR variants and expression of HNRNPA1, a known regulator of INSR splicing [6].

Weight loss resulted in higher expression of INSR-B, the more active isoform in insulin signalling [7]. Accordingly, we detected a strong negative correlation between INSR-B and fasting insulin levels, suggesting an association between INSR-B splicing and insulin action. Multivariate analysis of the pooled samples from all three studies (n = 189) revealed that the main determinant of the expression of INSR splice variants in subcutaneous adipose tissue was fasting insulin. This is consistent with a previous study in monkeys, including an analysis of human data, suggesting a link between hyperinsulinaemia and INSR splicing [8, 9].

There are two potential mechanisms for an association of INSR splicing with fasting insulin levels. First, lower insulin levels in relation to INSR-B may reflect better peripheral insulin sensitivity as expected for this isoform. However, an increase in INSR-B in response to weight loss could also be secondary to lower insulin levels, as insulin is a known regulator of splicing factor activity through phosphorylation [10] and because hyperinsulinaemia associates with lower expression of splicing factors [3]. We found a correlation between INSR splicing and expression of HNRNPA1, SF3A1 and SFRS7 in the KOBS study (ESM Table 4). The strongest negative correlation was with HNRNPA1, previously reported to inhibit exon 11 inclusion in HepG2 and HEK293 cells [6]. We acknowledge that other regulators of INSR splicing exist [6], and they may also be modified by weight loss. One limitation of this study is that INSR protein isoforms created by alternative splicing could not be detected because the difference between the protein isoforms is only 1 kDa (data not shown).

In conclusion, we demonstrated that INSR splicing correlates strongly with fasting insulin and is regulated by weight loss. These changes may associate with changes in splicing factor activity.