A plasma lipidomic endophenotype for bipolar disorder: evidence from unaffected siblings

Background: Bipolar disorder (BD) is highly heritable, yet the biology linking genetic vulnerability to clinical phenotype remains incompletely understood. Endophenotypes (heritable, state-independent biological traits also detectable in unaffected first-degree relatives) offer a useful framework for bridging this gap. Membrane sphingolipids and glycerophospholipids regulate myelin integrity, receptor organisation and intracellular signalling, and disturbances in these classes have been reported in BD. Whether such alterations are present not only in patients but also in unaffected siblings, who share genetic risk without disease expression, has not been systematically evaluated. Aim: To compare plasma lipidomic profiles among euthymic BD patients, their unaffected first-degree siblings and healthy controls, in order to identify candidate lipidomic endophenotypes. Methods: A total of 114 participants were enrolled from psychiatric outpatient clinics: 38 patients with DSM-5-TR BD euthymic for at least 8 weeks (HDRS ≤ 7, YMRS ≤ 5), 38 unaffected first-degree siblings and 38 healthy controls. Groups were comparable for age, sex, body mass index, education, smoking and medication use. Untargeted plasma lipidomic profiling used liquid chromatography–mass spectrometry. Following annotation and quality-control filtering, fold-change analysis between patients and controls identified 242 differentially altered species; the 30 with the highest between-group variance were taken forward for ANCOVA. Group differences were tested by one-way ANCOVA including these six variables (age, sex, body mass index, education, smoking, medication use) as covariates, with Bonferroni-corrected pairwise comparisons and Benjamini–Hochberg false-discovery-rate correction; effect sizes are reported as partial η². Significant species were classified as endophenotype candidates (gradient: patients > siblings > controls), familial-clustering (patients = siblings ≠ controls) or disease-specific (patients ≠ siblings = controls). Analyses used IBM SPSS 26.0. Results: After covariate adjustment, 11 of 30 species showed a significant group main effect. The largest effects were for glutaryl-CoA (p < .001, η² = .639), phosphatidylcholine PC(48:12) (p < .001, η² = .424), phosphatidylethanolamine PE(10:0/22:6) (p < .001, η² = .390) and phosphatidic acid PA(14:1/38:0) (p < .001, η² = .352), all elevated in patients. Two glycosylated sphingolipids bearing a 2-hydroxylation motif were selectively decreased in patients: dihexosylceramide Hex2Cer(d14:1/16:0(OH)) (p < .001, η² = .300) and galactosylceramide GalCer(d15:2/18:1(OH)) (p < .001, η² = .233). Excluding one signal attributed to a pharmaceutical artefact, the 10 interpretable species fell into three patterns. PC(48:12) was the only species showing a monotonic stepwise gradient (controls < siblings < patients; patient–sibling p = .001, sibling–control p < .001) and emerged as the single-lipid endophenotype candidate. Diglyceride DG(40:2) showed a familial-clustering pattern (p < .001, η² = .176; sibling–control p < .001) compatible with shared genetic risk. Seven further species (glutaryl-CoA, PE(10:0/22:6), PA(14:1/38:0), PS(20:3/24:4), Hex2Cer, GalCer, PA(16:1/22:2)) followed a disease-specific pattern, separating patients from siblings and controls; one species remained unclassified. Conclusions: PC(48:12) emerged as a candidate single-lipid endophenotype for BD, while DG(40:2) showed a familial-clustering pattern compatible with shared genetic risk. Selective reduction of 2-hydroxylated glycosphingolipids in patients points to fatty acid 2-hydroxylase (FA2H)-related myelin lipid biology as a candidate disease-specific mechanism. Findings should be interpreted as exploratory given the modest sample size and cross-sectional design. Validation in larger independent cohorts is warranted.

References

No references