Extended Data Fig. 11: Visceral-specific adipocyte subpopulation hAd6 is associated with thermogenic traits.
From: A single-cell atlas of human and mouse white adipose tissue
a, Regional visualization of associations of common genetic variants near EBF2 with VATadj. b, Effect size of association of rs4872393 with VATadj, ASATadj, GFATadj, and BMI per minor allele A; n = 37,641. Error bars reflect a 95% confidence interval around the effect size estimate from regression. c, VATadj raw data plotted according to rs4872393 carrier status; n = 36,185. For box plots, horizontal line = median, lower and upper bounds of the box = 1st and 3rd quartile respectively, lower and upper whisker = 1st quartile – 1.5 x interquartile range (IQR) and 3rd quartile + 1.5 x IQR respectively, outliers are plotted as points. d, Scatterplot showing the relationship between estimated cell type proportion and the LipocyteProfiler calculated feature Mitochondrial Intensity in visceral samples. e, Expression of mitochondrial and thermogenic genes in visceral ex vivo differentiated adipocytes stratified by estimated hAd6 proportion and matched for amount of differentiation using PPARG expression, n = 7 mAd6 low and 5 mAd6 high. Error bars represent SEM, P values were calculated using two tailed t-tests with no adjustments for multiple comparison, *, P < .05, **, P < .01. Exact P values: EBF2 = 0.027, TFAM = 0.019, CKMT1A = 0.049, CKMT1B = 0.005. f, Representative images of hAd6 low and high visceral in vitro differentiated cultures. Green represents BODIPY staining, red represents MitoTracker staining, and blue represents Hoechst staining. Scale bars are 100 μm, in total 3 random images/sample were analyzed from 5 hAd6 low and 5 hAd6 high samples. g, Violin plot of sNuc-seq data showing axon guidance genes in adipocyte subclusters. h, Scatterplots showing the relationship between calculated proportion of visceral subpopulations hAd2 and hAd6 and expression of pan-neuronal markers on the ambient RNA of individual visceral sNuc-seq samples. For scatterplots, P values were calculated using an F-test with the null hypothesis that the slope = 0.
