Candidate gene identification
We identified potential candidate genes in the QTLs associated with plasma blood glucose levels based on three criteria: first, the gene should have a strong correlation with plasma glucose levels in OGTT; second, the gene should have a significant cis-expression QTL (cis-eQTL), and third, the gene should harbor coding sequence variants between B6 and D2 mice. The genes that met any two of these criteria were considered further. Finally, we identified Tead1 as the candidate gene in QTL1; Llns1a, Aqp11, Alg8, Ndufc2, Myo7a, and Tsku as the candidates in QTL2; and Hmcn1 as the candidate gene in QTL3 (Table 2). No gene in QTL4, QTL5 and QTL6 qualified as a candidate based on the filtering criteria considered by us.
Table 2
Candidate genes identified in the QTL regions associated with plasma blood glucose levels at OGTT.
QTL | Symbol | Chr | Phenotypes | Correlation | Cis-eQTL | Coding Variants |
QTL1 | Tead1 | 7 | 15min, 30min, 45min | √ | √ | × |
QTL2 | Clns1a | 7 | 30min, 90min | √ | √ | √ |
QTL2 | Aqp11 | 7 | 30min, 90min | √ | √ | √ |
QTL2 | Alg8 | 7 | 30min, 90min | √ | √ | √ |
QTL2 | Ndufc2 | 7 | 30min, 90min | √ | √ | × |
QTL2 | Myo7a | 7 | 30min, 90min | √ | √ | √ |
QTL2 | Tsku | 7 | 30min, 90min | √ | √ | √ |
QTL2 | Pold3 | 7 | 30min, 90min | √ | √ | × |
QTL3 | Hmcn1 | 1 | 30min, 90min | √ | × | √ |
Tead1 is a candidate gene for OGTT-related QTL1
A significant QTL (peak -log10p = 4.6) was mapped to Chr 7 at 107.5–118 Mb for plasma glucose levels at OGTT-15, 45 and 60 min, with 63 genes located in the 1.5 LOD confidence interval (Fig. 3A). Among them, Tead1 was found to be cis-regulated (Fig. 3B). Moreover, Tead1 had significant correlations with plasma glucose levels at OGTT-15 (r = 0.4721, p = 0.001), 45 (r = 0.472, p = 0.001) and 60 min (r = 0.422, p = 0.006) (Fig. 3C-E). This was further confirmed with ePheWAS analysis (Fig. 3F). Notably, Tead1 knockout mice exhibited much lower glucose (Fig. 3G-I) and lipid metabolism (Supplementary Fig. 1) than their WT counterparts. Tead1 was also found to play a role in multiple physical systems including growth/size/body and mortality/aging, especially in hematopoietic/metabolic system; the mice with Tead1 mutations showed impaired glucose tolerance (Fig. 3J). To uncover the molecular mechanisms underlying Tead1 in glucose metabolism, we performed correlation analysis and identified top 2000 Tead1-co-expressed genes in the liver transcriptome of BXD strains. These genes were analyzed using WebGestalt for KEGG pathway enrichment. Our results highlighted strong enrichment of cholesterol metabolism signaling, arachidonic acid metabolism and metabolic signaling pathway by the correlated genes (Fig. 3K).
Myo7a and Ndufc2 are the candidate genes for OGTT-related QTL2
A significant QTL (peak -log10p = 4.2) was mapped to Chr7 at 89.2-104.3 Mb for plasma glucose levels at OGTT-30 and 90 min, with 145 genes located in the 1.5 LOD confidence interval (Fig. 4A-B). Among them, 11 genes were cis-regulated, 40 genes had variations in the DNA sequence between B6 and D2, and 21 genes were significantly correlated with plasma glucose levels at OGTT-30 and 90 min. Llns1a, Aqp11, Alg8, Ndufc2, Myo7a, and Tsku qualified our screening criteria. Furthermore, Myo7a and Ndufc2 had also evidence based on public genetic resources, ePheWAS and IMPC. Myo7a was cis-regulated (Fig. 4C), and significantly correlated with glucose levels at OGTT-30 (r = 0.45, p = 0.0036) and 45 min (r = 0. 55, p < 0.0001) both in CD and HFD cohorts (Fig. 4D-E). This was further confirmed with ePheWAS (Fig. 4G). Comparing the coding sequence variations between the B6 and D2 mice, Myo7a harbored two missense variants identified in two strains (Fig. 4F). Moreover, Myo7a knockout mice exhibited much lower glucose (Fig. 4H-L) and lipid metabolism-related phenotypes (Supplementary Fig. 2) than the WT mice.
In addition, our data also found Ndufc2 expression to be significantly correlated with plasma glucose levels at OGTT-30 (r = 0.578, p < 0.0001) and 45 min (r = 0. 432, p = 0.0005), both in CD and HFD cohorts (Fig. 5B-C). This gene was cis-regulated (Fig. 5D) and had a significant association with OGTT-related phenotypes in ePheWAS (Fig. 5E). Ndufc2 knockout mice exhibited much lower glucose (Fig. 5F-I) and lipid metabolism (Supplementary Fig. 3) than their WT counterparts. To uncover the molecular mechanisms associated with Ndufc2 in glucose metabolism, we performed pathway enrichment analysis of Ndufc2 co-expressed genes. Our results highlighted significant enrichment of metabolic, glycolysis/gluconeogensis pathway and non-alcoholic fatty liver disease (NAFLD) pathways (Fig. 5J).
Functional validation of TEAD1, MYO7A and NDUFC2 in vitro
To investigate whether TEAD1, MYO7A and NDUFC2 are essential for glucose metabolism in the liver, we used siRNAs to knockdown the expression of TEAD1, MYO7A and NDUFC2 in THLE-3 and HepG2 cells. The qPCR results found that the transcripts of all three genes were significantly decreased after siRNA transfection in THLE-3 and HepG2 cells (p < 0.05) (Fig. 6A-C). Furthermore, we assessed the expression of the genes related to insulin signaling pathway, including insulin receptor (ISR) [37], insulin like growth factor 2 binding protein 2 (IGF2BP2) [38], PPARG (peroxisome proliferator activated receptor gamma) [39], pancreatic and duodenal homeobox 1 (PDX1) [40], as well as hepatic glucose metabolism pathway, including glucose-6-phosphate dehydrogenase (G6PD) [41], hexose-6-phosphate dehydrogenase (H6PD) [42], glucose transporter 4 (GLUT4) [43], glucose transporter 1 (GLUT1) [44] and glucokinase (GCK) [45] in TEAD1, MYO7A or NDUFC2 knockdown cells and controls, respectively (Fig. 6A-C). Our results demonstrated that the expression of the genes associated with insulin signaling, including INSR, PPARG, PDX1, and glucose metabolism pathway, including G6PD, GCK and H6PD was significantly decreased with TEAD1 knockdown (Fig. 6A). Similar results were obtained in MYO7A or NDUFC2 knockdown cell models, except for PPARG and GLUT1 (Fig. 6B-C). Further, we observed that knockdown of the candidate genes resulted in a significant inhibition of insulin-induced glucose uptake by HepG2 cells (Fig. 6D-F). Overall, these results indicate that TEAD1, MYO7A and NDUFC2 influence glucose uptake and metabolism in hepatocytes by modulating the expression of genes related to insulin and glucose metabolism pathways.
TEAD1, MYO7A and NDUFC2 are cis-regulated and influence liver metabolism in humans
To assess the potential causative effects of TEAD1, MYO7A and NDUFC2 in genetic regulation and liver metabolism in humans, we explored these genes using GTEx and GWAS databases. The results showed that TEAD1, MYO7A and NDUFC2 were cis-regulated in human liver tissue (Table 3). Furthermore, TEAD1 was found to be associated with body weight, high-density lipoprotein (HDL) and model assessment of beta-cell function (HMOA-B) [46], while MYO7A was associated with alanine aminotransferase (ALT) [47] and aspartate transaminase (AST) [48], and NDUFC2 was associated with body mass index (Table 4).
Table 3
List of cis-eQTLs for TEAD1, MYO7A and NDUFC2 in human liver.
Symbol | Variant ID | Chr | Start | End | Ref | Alt | P-value |
TEAD1 | rs4756756 | chr11 | 12674591 | 12944483 | T | C | 3.00E-05 |
MYO7A | rs73499629 | chr11 | 77128264 | 77215239 | C | G | 3.86E-05 |
NDUFC2 | rs11236546 | chr11 | 78068352 | 78080219 | T | G | 5.15E-05 |
Table 4
List of TEAD1, MYO7A and NDUFC2 associated traits in human GWAS.
Variant ID | P-value | Mapped gene | Trait(s) |
rs371148512 | 4.00E-09 | TEAD1 | Body weight |
rs4237723 | 5.00E-08 | TEAD1 | HOMA-B |
rs7924536 | 6.00E-08 | TEAD1 | High density lipoprotein cholesterol measurement |
rs7927472 | 3.00E-09 | MYO7A | Aspartate aminotransferase measurement |
rs3819170 | 2.00E-14 | MYO7A | Aspartate aminotransferase measurement |
rs606164-G | 5.00E-06 | NDUFC2 | Body mass index |