Transcriptomic profiling of human brown and white adipose tissue depots

Figures

• 

  Fig. 1. Multimodal characterization of human adipose depots: imaging, gene expression, and cellular composition. a 18FDG-PET images, frontal (left) and sagittal (right), of a healthy female subject (23 y), with each depot labeled: subcutaneous (SC), omental (OM), pericardial (PC), paravertebral (PV), adrenal (AD), and supraclavicular (SV). b UCP1 immunofluorescence in SC, PV, and AD depots. UCP1 immunofluorescence in SC, PV, and AD depots. UCP1 fluorescence intensity was normalized to DAPI for each sample. Data represent the average normalized UCP1 fluorescence from two experimental sessions (n = 2) for one subject, with the AD depot used as the no-primary-antibody control. Scale bar depicts 50 µm. c UpSet plot showing the number of genes expressed in each depot (PC, AD, OM, PV, SC, and SV) and their overlaps. Using log-transformed, normalized counts per million (CPM) across all depots, a gene was considered expressed in a specific depot if its trimmed mean of M values (TMM) normalized, log-transformed CPM exceeded the threshold of 0. d MA plots visualizing the differential expression: between PV vs. SC (left) and PV vs. SV (right). Significance was calculated using a two-sided Wald test…

• 

  Fig. 2. Gene ontology of human white and brown adipose tissues and mouse adipose tissue transcriptomics. a, b Dot plots showing representative pathways from gene ontology (GO) analysis of enriched genes in the (a) PV depot compared to the SV depot and the (b) PV depot compared to the SC depot. Significance was calculated via Gene Set Enrichment Analysis (GSEA) using 10,000 permutations to assess both positive and negative enrichment. Multiple testing adjustments were applied using the False Discovery Rate (FDR) method (FDR < 0.05). Two-sided comparisons were used. c, d Principal component analysis (PCA) of mouse microarray data based on differential gene expression (DGE) analysis shows clustering by (c) six adipose tissue depots—interscapular BAT, perigonadal WAT, inguinal WAT, mesenteric WAT, perirenal WAT, and subcutaneous WAT—compared to (d) clustering by individual mice under different dietary conditions. e Hierarchically clustered heatmap of (top) human genes and (left side) donors using average linkage clustering generated from the top 100 most variable genes across all depots after variance stabilizing transformation. The color scale represents Z-scores. Source data are provided…

• 

  Fig. 3. Comparison of intrasubject and intradepot transcriptomic variability in humans. a PCA was performed on variance-stabilized counts to explore sample variance across depots for each donor. Lines connecting the points represent Euclidean distances. Euclidean distances between points within the two-dimensional PCA space (PC1 and PC2) were calculated geometrically to quantify sample similarity; no statistical hypothesis testing was applied to these distances. Two-sided comparisons were made. b PCA of DGE results across donors for each depot. Lines connecting the points represent Euclidean distances. Source data are provided as a Source Data file.

• 

  Fig. 4. Predictive modeling of mouse and human donors and functional corroboration of human adipose tissue function. a Sankey diagram of a Random Forest analysis of microarray data using as the training set the inguinal WAT depot from 12 wk-old DBA/2 J mice fed 6 wk with either standard breeding diet (SBD) or high-fat diet (HFD). The model achieved 20% accuracy (pink = true; green = false) in predicting whether the sample originated from the same mouse. b Sankey diagram of a Random Forest analysis using (top) the human subcutaneous WAT depots as the training dataset and considering adipose samples only. The model achieved 83% accuracy in predicting the donor identity for autopsy tissues from donors 1, 4, 9, and 10. When the urinary bladder (bottom) was included in the model, the accuracy was 50%. c Axial images of six human adipose tissue depots in a 23 y woman after pharmacological stimulation of WAT and BAT metabolic activity as measured by 18F-FDG PET/CT. Source data are provided as a Source Data file.