XBP1 expression in pancreatic islet cells is associated with poor glycaemic control especially in young non-obese onset diabetes across ancestries

Abstract

Background: Individuals of South and East Asian ancestry have a higher risk of type 2 diabetes, often driven by insulin deficiency due to impaired beta-cell function. The transcription factor XBP1 supports beta-cell survival by reducing cellular stress, but its role in diabetes risk and glucose regulation remains unclear. This study aimed to evaluate the impact of XBP1 expression on diabetes risk, beta-cell function, glycaemic traits, and treatment response across ancestries.

Methods: We performed colocalisation analyses to test whether XBP1 expression in pancreatic islets and type 2 diabetes share causal variants. A lead variant regulating XBP1 expression was identified and analysed in two South Asian cohorts from India to assess associations with beta-cell function and glucose levels. We further assessed glycaemic control using HbA1c in cohorts of British South Asians and white Europeans. We examined the effect of the variant on drugs designed to improve insulin secretion.

Results: XBP1 expression colocalises with diabetes risk in East Asians but not in white Europeans, and lower expression is associated with higher risk of diabetes. The lead SNP of the eQTL (rs7287124) is more common in East (65%) and South Asians (50%) compared to white Europeans (25%). rs7287124 is associated with lower beta-cell function using HOMA-B (P = 5 × 10^-3, n = 470). In trans-ancestry meta-analyses rs7287124 is associated with 4.32 mmol/mol (95% CI: 2.60-6.04, P = 8 × 10^-7) higher HbA1c. In individuals with young, non-obese onset diabetes, the trans-ancestry effect is 6.41 mmol/mol (P = 2 × 10^-4). Variant carriers show impaired response to sulphonylureas.

Conclusions: XBP1 expression is associated with diabetes risk with particular value in under-represented populations at risk of young, non-obese onset diabetes.

Fig. 1. Colocalisation and Mendelian Randomisation methods for XBP1 expression and type 2 diabetes risk.

a Colocalisation probabilities using eCAVIAR and HyPrColoc between XBP1 expression in both pancreatic islets and GTEx pancreas samples and type 2 diabetes (T2D) risk in East Asians (EAS) from Biobank Japan and white Europeans (EUR) from DIAMANTE. Orange line indicates eCAVIAR recommended threshold for significant colocalisation. In both cases, East Asians are more likely to have overlapping causal variants. b Mendelian Randomisation (MR) ratio estimates of association between XBP1 expression in pancreatic islets and T2D risk in Biobank Japan. These plots have been made using 25 variants with LD < 0.4 in the XBP1 region +/−5KB. Weighted median estimate for the effect was −0.083 (SE: 0.025), P value = 9.2 × 10⁻³, and inverse variance weighted estimate was −0.065 (SE: 0.020), P value = 7.17  × 10 ⁻⁴.

Fig. 2. Effect of variant on beta-cell function and 2 hour stimulated glucose levels.

a Box plot showing additive models of XBP1 eQTL variant (rs7287124) and log-transformed Homeostasis Model of Assessment of Beta-Cell function (HOMA-B) calculated using fasted glucose and stimulated C-peptide in 470 individuals with newly diagnosed type 2 diabetes at DMDSC, Chennai, India. After adjustment for age, sex, BMI, and HOMA-B, the association remained significant (P = 5 × 10⁻³). b Boxplot showing effect of rs7287124 on 2 hr stimulated blood sugar in 484 rural-dwelling Indians from the TREND study. After adjustment for age, sex, and BMI, the association remained significant (P = 0.02). Per allele was associated with 0.7 mmol/L higher 2 h glucose (95% CI: 0.1–1.4).

Fig. 3. Trans-ancestry meta-analyses of the effect of rs7287124 on HbA1c levels overall and in young, lean T2DM cases.

Forest plots showing random effects of XBP1 eQTL variant on HbA1c (mmol/mol). Effects are pooled across summary statistics available from the MAGIC consortium for East Asians, Europeans, and South Asians without T2DM (n = 173,186) and for 6482 individuals with newly-diagnosed T2DM across 4 cohorts: Tayside Scotland (TDS), urban-dwelling South Asian Indians (DMDSC), rural-dwelling South Asian Indians (TREND), and British South Asian Pakistani and Bangladeshis (G&H). The random effect meta-analysis shows an increase of 4.32 mmol/mol of HbA1c per risk allele (95% CI: 2.60–6.04). Forest plots of sub-groups of people with newly diagnosed T2DM showed a stronger effect in 477 individuals diagnosed young with non-obese BMI (6.41 mmol/mol), compared to no effect in those diagnosed older with non-lean BMI (n = 3382).