Comparative genomic analyses of multiple backcross mouse populations suggest SGCG as a novel potential obesity-modifier gene

Abstract

To nominate novel disease genes for obesity and type 2 diabetes (T2D), we recently generated two mouse backcross populations of the T2D-susceptible New Zealand Obese (NZO/HI) mouse strain and two genetically different, lean and T2D-resistant strains, 129P2/OlaHsd and C3HeB/FeJ. Comparative linkage analysis of our two female backcross populations identified seven novel body fat-associated quantitative trait loci (QTL). Only the locus Nbw14 (NZO body weight on chromosome 14) showed linkage to obesity-related traits in both backcross populations, indicating that the causal gene variant is likely specific for the NZO strain as NZO allele carriers in both crosses displayed elevated body weight and fat mass. To identify candidate genes for Nbw14, we used a combined approach of gene expression and haplotype analysis to filter for NZO-specific gene variants in gonadal white adipose tissue, defined as the main QTL-target tissue. Only two genes, Arl11 and Sgcg, fulfilled our candidate criteria. In addition, expression QTL analysis revealed cis-signals for both genes within the Nbw14 locus. Moreover, retroviral overexpression of Sgcg in 3T3-L1 adipocytes resulted in increased insulin-stimulated glucose uptake. In humans, mRNA levels of SGCG correlated with body mass index and body fat mass exclusively in diabetic subjects, suggesting that SGCG may present a novel marker for metabolically unhealthy obesity. In conclusion, our comparative-cross analysis could substantially improve the mapping resolution of the obesity locus Nbw14. Future studies will throw light on the mechanism by which Sgcg may protect from the development of obesity.

Figure 1

Metabolic characterization of females from parental mouse strains 129P2, C3H and NZO. Development of body weight (A) and blood glucose (C) was monitored at weeks 3, 6, 10, 15, 18 and 20 weeks of age, whereas fat mass (B) was measured at weeks 3, 6, 10 and 15. Six-hours-fasting plasma insulin levels (D) were measured at week 22 by ELISA. Dots represent single female animals (129P2: n = 11–12; C3H: n = 10–12; NZO: n = 12). Statistical differences between strains were calculated by 2-way (A-C) or one-way analysis of variance (D) followed by post hoc Bonferroni test, ^*P < 0.05, ^**P < 0.01, ^***P < 0.001 by comparison to NZO unless otherwise stated.

Figure 2

Genome-wide linkage map of QTL for obesity-related traits in female mice derived from N₂(NZOxC3H) and N₂(NZOx129P2) outcross populations. Female NZO and male 129P2 or C3H mice, respectively, were used to generate the F₁ generations, and male F₁ offspring were subsequently backcrossed with NZO females to generate two backcross populations (N₂(NZOxC3H) and N₂(NZOx129P2)). The backcross populations were metabolically phenotyped, and genotyped for 105 N₂(NZOxC3H) and 110 N₂(NZOx129P2) informative SNP markers. Linkage analysis of the female backcross mice (n = 307–309) was conducted using the R-package as described in methods. Further information for each QTL is shown in Table 1. BW, body weight; FM, fat mass.

Figure 3

Genetic linkage of body weight and fat mass to loci on Chr.14 in outcross populations of N₂(NZOxC3H) and N₂(NZOx129P2) female mice. Chr.14 revealed significant linkage with body weight and fat mass in both, N₂(NZOxC3H) (A) and N₂(NZOx129P2) (B) females. The SNP-markers that were used for genotyping with the corresponding Mb-position are plotted below. The SNP markers in closest proximity to the QTL peak position are marked in red. The genotypic effects are indicated with the arrows. LOD scores were calculated using R/qtl software as described in methods. LOD, logarithm of the odds; N, NZO; wk, week.

Figure 4

Quantitative effect of Nbw14 in N₂(NZOxC3H) and N₂(NZOx129P2) mice. Mean values for body weight (± SEM) (A), fat mass (B) and lean mass (C) for homozygous NZO and heterozygous allele carriers for Nbw14 the N₂(NZOxC3H) females. Mean values for body weight (± SEM) (D), fat mass (E) and lean mass (F) for homozygous NZO and heterozygous allele carriers for Nbw14 from N₂(NZOx129P2) females. Nbw14^NZO/NZO (N₂(NZOxC3H): n = 181, Nbw14^NZO/C3H (N₂(NZOxC3H): n = 126, Nbw14^NZO/NZO (N₂(NZOx129P2): n = 133, Nbw14^NZO/129P2 (N₂(NZOx129P2): n = 170; N, NZO.

Figure 5

Combined approach of haplotype- and gene expression analysis in gWAT of the parental strains for the identification of NZO-specific gene variants within Nbw14. (A) Haplotype analysis using Genome Reference Consortium Mouse Build 38 (GRC38) provided by the Wellcome Trust Sanger Institute. The red line represents the number of polymorphic SNPs according to NZO ≠ 129P2 and C3H, whereas the grey line shows the total number of SNPs (all SNPs annotated for C57B6/J reference genome with calls for C3H, 129 and NZO). Both lines overlap in the regions between 56.5–56.75 Mb, 60.25–61.5 Mb and 68.0–68.5 Mb. Genomic regions containing less than 100 polymorphic SNPs (NZO ≠ 129P2 and C3H) per window (250 kb) are represented with the white boxes. In contrast, regions with more than 100 polymorphic SNPs (NZO ≠ 129P2 and C3H) per window are highlighted by red boxes. Genes located in regions that were defined as polymorphic between NZO and the other strains are listed above. For a better overview, gene models are not included. (B) Significantly (P < 0.05) differential gene expression in NZO vs. C3H and 129P2 at the age of 6 weeks in gWAT detected by microarray analysis. Higher expression in NZO is indicated with an expression ratio (NZO/C3H and NZO/129P2) > 1, whereas downregulation in NZO is shown with an expression ratio (NZO/C3H and NZO/129P2) < 1. Differences between the strains were calculated by one-sided Wilcoxon signed rank test. (C) Venn diagram showing the total number of annotated genes (top), the number of genes revealed from the haplotype analysis (left), the number of genes revealed from the microarray analysis (right) and the overlap of genes (centre).

Figure 6

Sgcg and Arl11 mRNA expressions across different tissues from the parental strains and validation in gWAT by qRT-PCR. All tissues were collected from male NZO, C3H and 129P2 mice at 6-weeks of age. The expressions of Sgcg and Arl11 in liver, skeletal muscle, gWAT, BAT and pancreatic islets was measured by microarray analysis and the bars show the relative fluorescence units (A and C). Differential expression in gWAT tissue was confirmed by qRT-PCR (B and D). Tbp was used as endogenous control. Data represent mean values ± SEM from 5 (microarray) or 7 (qRT-PCR) mice each strain, respectively. Differences between the strains were calculated by two-sided Wilcoxon test (microarray) or One-way analysis of variance followed by post hoc Bonferroni test (qRT-PCR), ^*P < 0.05, ^***P < 0.001. gWAT, gonadal white adipose tissue; qPCR, qRT-PCR.

Figure 7

Expression QTL analysis for Sgcg and Arl11 expressions in gWAT from all N₂(NZOxC3H) mice. Genome-wide LOD scores distribution for the mRNA expressions of Sgcg (A) and Arl11 (B) in N₂(NZOxC3H) males, Chr.14 is highlighted in grey and the significance threshold (P < 0.05) is indicated with the horizontal line. (C) Expression QTL curves and metabolic QTL curves (body weight and fat mass) on Chr.14 were superimposed. The SNP-markers that were used for genotyping with the corresponding Mb-positions are plotted below. The genotypic effects are indicated with the arrows. All LOD scores were calculated with R/qtl package as described in methods. LOD, logarithm of the odds; Expr, expression; N, NZO; wk, week.

Figure 8

Measurement of 2-Deoxy-D-glucose uptake in Sgcg overexpressing 3T3-L1 vs. control adipocytes. Undifferentiated 3T3-L1 cells were infected with either a retrovirus carrying cDNA from Sgcg (Sgcg OE) or the empty pMSCV-puro vector as control and differentiated to adipocytes. The mRNA overexpression was confirmed by quantitative Realtime PCR (A). Hprt was used as endogenous control. 2-Deoxy-D-glucose uptake was determined in fully differentiated cells by measuring luminescence intensity at basal as well as after 60 min stimulation with 100 nM insulin (B). Bars represent mean values ± SEM from 3 (A) or 7 (B) different experiments. Statistical differences were calculated by two-way analysis of variance (B) followed by post hoc Bonferroni test, ^*P < 0.05.

Figure 9

SGCG expression analysis in human adipose tissue. Correlation of SGCG mRNA levels with BMI in subcutaneous (A) and visceral adipose tissue (B) from healthy (blue circles) vs. patients with type 2 diabetes (T2D, red squares). Correlation of SGCG mRNA levels with body fat in subcutaneous (C) and visceral adipose tissue (D) from healthy controls (blue circles) vs. T2D (red squares) subjects. Correlations were calculated using Pearson correlation test. n = 16–36 for BMI and 14–22 for body fat.