Extended Data Fig. 3: Validation with negative controls of tissue-specific taxa different between participants who were normoglycemic or type 2 diabetic.

Tissue-specific bacterial targets found to discriminate between disease state were identified by LefSe analysis. The relative abundance of these taxa (at family and genus level) in liver (a, b), mesenteric (MAT – c, d), omental (OAT – e, f) and subcutaneous (SAT – g, h) adipose tissue and plasma (i, j) was analyzed, without accounting for disease state distribution, against negative controls (NCs) using Mann-Whitney U test. P values are indicated at the top of each graph. Left side panels show the relative abundance of taxa, whereas right side panels depict relative abundance normalized by 16S rRNA gene count (that is, relative abundance x 16S count). Box plots depict the first and the third quartile with the median represented by a vertical line within the box; the whiskers extend from the first and third quartiles to the highest and lowest observation, respectively, not exceeding 1.5 x IQR. Each circle (Non-diabetic, ND) and triangle (Type 2 Diabetic, T2D) represents a sample, and their tissue-specific dispersion is presented using a log10 scale. The number of independent biological replicates tested was: Liver (n=39), MAT (n=40), OAT (n=40), SAT (n=40), Plasma (n=39), NC (n=23). All statistical tests were two-sided, and differences were considered statistically significant at P<0.05.