The family at the center of the study comprised 20 members spanning three generations. Five of them had type 2 diabetes: the proband’s mother in the first generation, the proband herself, and three of the proband’s sisters in the second generation. The proband was a 55-year-old woman with a ten-year history of diabetes who met the 1999 World Health Organization diagnostic criteria, including fasting blood glucose of at least 7.0 mmol/L or 2-hour postprandial glucose of at least 11.1 mmol/L during an oral glucose tolerance test. Across the five affected individuals, the mean age at diagnosis was 43.2 years, and all were overweight or obese, with a body mass index of at least 25 kg/m2. Three patients were managed with metformin-based regimens while two required insulin therapy. Diabetic peripheral neuropathy affected four of the five patients, and the eldest, the proband’s mother, had both diabetic nephropathy and retinopathy.

The remaining family members, including the proband’s brother, husband, brothers-in-law, sister-in-law, and all members of the third generation, showed no obvious abnormality in blood glucose. There was no history of inbreeding or other hereditary disease in the family. Because the third-generation relatives are still young and could carry latent risk, the researchers designed their analysis to account for this possibility, requiring that any candidate mutation appear in all five diabetic patients, in none of the normal controls, and in no more than half of the third-generation direct relatives.

To hunt for the family’s shared mutations, the team collected fasting peripheral venous blood from all 20 members, extracted genomic DNA, and sent the samples to BGI in Shenzhen for whole exome sequencing on the BGISEQ-500 platform, using the SureSelect Human All Exon V6 capture kit. Whole exome sequencing targets the protein-coding regions of the genome, where most known disease-causing diabetes mutations reside, and it is particularly sensitive to the low-frequency and rare variants that family-based studies depend on. The sequencing produced an average of roughly 14,890 Mb of raw bases per sample, and after quality filtering an average of 148,819,404 clean reads per sample remained, with an average GC content of 52.14 percent. Quality scores were high throughout, with Q20 and Q30 values both exceeding 80 percent, indicating reliable base calling across the dataset.

The scale of variation detected was substantial. On average, 96,114 single nucleotide polymorphism sites were identified per sample, including 11,002 synonymous and 10,381 missense mutations in coding regions, along with 32 sites that converted stop codons to non-stop codons, 94 that introduced premature stop codons, 19 that disrupted start codons, and 93 that altered splice donor or acceptor sites. Indel analysis added an average of 15,142 insertion-deletion sites per sample, including 308 coding frameshift mutations and 93 splice-site changes. To separate signal from this background noise, the researchers applied a probability formula that weighs how often a given mutation would be expected by chance given its frequency in reference populations from the 1000 Genomes Project and the Exome Aggregation Consortium, then visualized the results with Manhattan plots using the qqman package in R.

After filtering, 485 mutation sites met the family-specific criteria, comprising 429 SNPs and 56 indels. Annotation of biological function narrowed this to 15 significant SNP sites and one significant indel, distributed across 16 genes: EPHX3, SLC52A1, HCN4, ACSBG1, FHOD1, ENKD1, HP, ALPPL2, LOXHD1, PODNL1, LRIT2, CYP1A2, FUT6, MXI1, HYDIN, and RNLS. Five of these variants, in HCN4, ACSBG1, ENKD1, HP, and LOXHD1, had never been reported before. Cross-referencing with the KEGG pathway database and published literature led the team to focus on two genes with plausible links to glucose metabolism: EPHX3 and FHOD1.

The EPHX3 variant, rs760409070, is a missense mutation that changes the 260th amino acid of the encoded protein from glutamic acid to lysine. According to the KEGG pathway analysis, EPHX3 interacts genetically with PIK3CB, the gene encoding a catalytic subunit of phosphoinositide 3-kinase. The PI3K/AKT signaling pathway is a central conduit of insulin action: when activated, it promotes the glucose transporter GLUT4 in taking up and transporting glucose, enhances glycogen synthesis through glycogen synthase kinase-3, and suppresses gluconeogenesis by inhibiting the transcription factor FOXO1. The researchers propose that the EPHX3 mutation inhibits PIK3CB expression and thereby dampens this entire cascade, producing insulin resistance, reduced glucose uptake, impaired glycogen synthesis, and increased glucose production by the liver. In silico tools lent partial support: SIFT scored the variant as deleterious with a value of 0.03, PolyPhen-2 called it possibly damaging at 0.503, and the CADD PHRED score of 24.7 places it among the top 1 percent most deleterious substitutions in the human genome, although REVEL predicted a neutral effect at 0.178. The variant is absent from 44,890 East Asian alleles in the gnomAD database, has a global allele frequency of just 1.18 × 10−5, and is classified in ClinVar as a variant of uncertain significance.

The FHOD1 variant, rs142023201, substitutes serine for asparagine at position 498 of the formin homology domain-containing 1 protein, a cytoskeletal protein found in adipocytes and skeletal muscle cells. FHOD1 is not directly annotated in glucose metabolism pathways, but prior studies suggest it interacts with insulin-regulated aminopeptidase, or IRAP, a major component of GLUT4 storage vesicles that helps regulate their trafficking to the cell membrane. Earlier work also found that FHOD1 protein interacts with actin-binding fibrin to increase insulin-stimulated glucose uptake, and that FHOD1 lacking the IRAP-binding domain fails to do so. The researchers speculate that the p.Asn498Ser change reduces FHOD1 function, downregulates IRAP activity, and thereby impairs GLUT4-mediated glucose transport. Notably, all four prediction tools scored this variant as tolerated or benign, with SIFT at 0.5, PolyPhen-2 at 0, CADD at 9.086, and REVEL at 0.061, and ClinVar classifies it as likely benign. The authors argue that computational predictors, which rely heavily on sequence conservation, may perform poorly for structural and cytoskeletal proteins whose function depends on protein-protein interactions, so the negative predictions do not override the co-segregation evidence within the family. Like the EPHX3 variant, it is absent from East Asian populations in gnomAD and has no recorded associations in the GWAS Catalog.

Both heterozygous mutations were carried by all five diabetic family members, were absent from the unaffected second-generation controls, and appeared in fewer than half of the third-generation relatives. The authors suggest that third-generation carriers, who are currently young and undiagnosed, may face elevated risk and could benefit from early adoption of healthy lifestyle measures to reduce or delay onset, though this recommendation rests on the assumption that the variants are genuinely pathogenic. Whether the two mutations act independently, synergistically, or as primary driver and secondary modifier remains unknown, and functional studies such as cell-based assays of PI3K/AKT activity or GLUT4 translocation are needed to resolve this.

The study carries important limitations. It analyzes a single pedigree, so the findings require validation in larger cohorts and in sporadic diabetic populations before any general claims can be made. Environmental modifiers such as diet and physical activity were not assessed, body mass index data for the non-diabetic relatives were unavailable, and detailed insulin resistance or secretion measures such as HOMA-IR and C-peptide were missing for most patients because data collection was retrospective, limiting genotype-phenotype correlations. Yunnan Province is notable for its ethnic diversity and distinct genetic architecture, and previous regional studies have implicated genes such as MTHFR, ARL15, CDH13, RANTES, and CCR5 in diabetes risk, but large-scale population genomic data for Yunnan remain scarce. The researchers therefore cannot determine whether the EPHX3 and FHOD1 variants are specific to this one family or represent broader regional risk factors, and they frame the two variants as novel candidate genes whose contribution to familial type 2 diabetes now awaits experimental confirmation.

Subject of Research: Rare EPHX3 and FHOD1 variants identified by whole exome sequencing in a Chinese family with type 2 diabetes

Article Title: Analysis of Whole Exome Sequencing Results of Type 2 Diabetic Pedigree

Article References: Chen, J., Li, J.-L., Yin, H.-J., Guo, W.-C., & Li, H.-F. (2026). Analysis of Whole Exome Sequencing Results of Type 2 Diabetic Pedigree. Molecular Genetics & Genomic Medicine, 14(10), Article e70296. https://doi.org/10.1002/mgg3.70296

Image Credits: AI Generated

DOI: 10.1002/mgg3.70296

Keywords: type 2 diabetes, whole exome sequencing, EPHX3, FHOD1, genetics, pedigree study, insulin resistance, GLUT4, PI3K/AKT pathway, rare variants, Yunnan China, bioinformatics