The Primacy of Adipose Tissue Gene Expression and Plasma Lipidome in Cardiometabolic Disease in Persons With HIV

Figures

• 

  Figure 1.. Study overview. A total of 93 persons with HIV virologically suppressed on contemporary antiretroviral therapy and without diabetes were included. Participants had several cardiometabolic assessments including waist circumference, hip circumference, body mass index, HOMA2-IR, fasting blood glucose, hemoglobin A1c, high-sensitivity C-reactive protein, and several morphometric measures by computed tomography including VAT volume and density, SAT volume and density, liver density (lower density correlates with higher hepatic fat content), and pericardial fat volume. Participants had whole-SAT transcriptomics and circulating lipidomics, metabolomics, and proteomics. Simple Spearman correlations were initially performed between cardiometabolic measures and individual assays. Then, we used MOFA to integrate across assays to identify a MOFA factor that was most closely associated with cardiometabolic features. We further analyzed this cardiometabolic factor including feature (assay) variance analysis, pathway and enrichment analysis, and single-cell RNA-sequencing deconvolution. Abbreviations: Circ, circumference; FBG, fasting blood glucose; HbA1c, hemoglobin A1c; HIV, human…

• 

  Figure 2.. Cardiometabolic Measures are associated with species across different omic assays. A–D, Spearman correlation between the variable of interest (x-axis) and the individual measured gene or species (y-axis) plotted using the Ward D2 hierarchical clustering algorithm for both rows and columns. Select genes or species associated with insulin resistance or visceral adipose tissue volume were annotated. Only species at the cutoff of a false-discovery rate of 5% with any of the cardiometabolic measurements were included in the heatmap for (A) subcutaneous adipose tissue transcriptome, (B) circulating proteome, (C) circulating lipidome, (D) and circulating metabolome. Abbreviations: AcCar, acylcarnitine; BMI, body mass index; Circ., circumference; DG, diacylglycerol; FBG, fasting blood glucose; HbA1c, hemoglobin A1c; HOMA2-IR, homeostatic model assessment of insulin resistance 2; hsCRP, high-sensitivity C-reactive protein; PAT, pericardial adipose tissue; SAT, subcutaneous adipose tissue; TG, triacylglycerol; VAT, visceral adipose tissue; Vol., volume.

• 

  Figure 3.. Integration of omics data reveals a factor that correlates with cardiometabolic assessments. A, Heatmap of assay type (x-axis) and MOFA factor (y-axis) showing the percent variance explained (shading). B, Heatmap of unadjusted Spearman correlation between cardiometabolic variable measures (x-axis) and MOFA factor (y-axis). Correlations with a P value < .05 are shaded in red or blue. Columns were plotted using the Ward D2 hierarchical clustering algorithm. C, Gene set enrichment analysis of SAT transcriptome with MOFA weights as the ranking using the Reactome database with gene ratio (x-axis) and pathway description (y-axis) split by positive (improved cardiometabolic risk factor measurements) and negative (worsened cardiometabolic risk factor measurements) enrichment. The dot size represents the number of genes in the pathway and the dot color represents the adjusted P value. D, Pathway activity inference using the PROGENy database shows changes in activity in positively weighted SAT genes compared with negatively weighted SAT genes. *P value < .05. Black border in (A) and (B) highlights factor 4, most closely associated with several cardiometabolic measures. Abbreviations: ART,…

• 

  Figure 4.. Diacylglycerol and triacylglycerol lipid species are heavily weighted in factor 4. A, Bee swarm plot with lipid class (x-axis) and MultiOmics Factor Analysis (MOFA) weighting from the cardiometabolic factor (y-axis). Negatively weighted factors are associated with worse cardiometabolic measurements. The asterisks denotes significant over- or underenrichment at a false-discovery rate of 5%. B, MOFA weighting for 15 highest and 15 lowest weighted circulating protein species from the cardiometabolic factor. The weighting is plotted on the x-axis while the protein species is plotted on the y-axis. The color corresponds to positive (red) and negative (blue) weighting. C, MOFA weighting for 15 highest and 15 lowest weighted circulating metabolite species from the cardiometabolic factor. The weighting is plotted on the x-axis while the metabolite species is plotted on the y-axis. The color corresponds to positive (red) and negative (blue) weighting. Abbreviations: AcCar, acylcarnitine; Cer, ceramide; DG, diacylglycerol; ether-PC, ether-linked phosphatidylcholine; ether-PE, ether-linked phosphatidylethanolamine; LPC, lysophosphatidylcholine; PC, phosphatidylcholine; PE,…

• 

  Figure 5.. Highest weighted subcutaneous adipose tissue genes map to myeloid (negative) and stromal (positive) cells in single-cell/nuclei RNA-sequencing data. A–C, Heatmap of mean expression of the top 400 positively and negatively weighted genes from the MOFA factor 4 by cell type. The mean normalized expression for each cell type was scaled (y-axis) across cell types (x-axis) and plotted using the Ward D2 hierarchical clustering algorithm. Higher expression is in red while lower expression is in blue. The corresponding directionality of the MOFA gene weight (positive = red, negative = blue) is plotted on the left for (A) coarse cell type classification in the HATIM single-cell RNA-sequencing data set (n = 59), (B) fine cell type classification in the HATIM single-cell RNA-sequencing data set (n = 59), and (C) coarse cell type classification from single-nuclei RNA-sequencing data of subcutaneous adipose tissue (n = 13). Abbreviations: ASPC, adipose stem and progenitor cell; cDC1, conventional dendritic cell type 1; cDC2B, conventional dendritic cell type 2B; cMo, classical monocyte; DC, dendritic cell; EC, endothelial cell; FIP, fibroinflammatory progenitor; HATIM, Human Adipose Tissue…