Exploration of Conserved Human Adipose Subpopulations Using Targeted Single-Nuclei RNA Sequencing Data Sets

Abstract

Background: Smooth-muscle cells and pericytes are mural cells. Pericytes can differentiate into myofibroblasts, chondrocytes, vascular smooth-muscle cells, and adipocytes, marking them as a distinct progenitor population. Our goal was to molecularly define the progenitor cell populations in human adipose tissues and test the adipogenic potential of human mural cells.

Methods: We used informatic analysis of single-cell RNA sequencing data from human tissues to identify and define pericytes and adipose progenitor cells found in human adipose tissues, including perivascular, brown, and white adipose tissues.

Results: We established tissue-specific patterns of gene expression in pericytes and other putative human adipocyte progenitor cells. PPARG-expressing pericytes were present in multiple human adipose depots with consistent expression of COL25A1, MYO1B, and POSTN. We also found evidence of tissue-specific pericyte markers. Although there is some conservation between human and mouse adipose tissues, human pericyte populations have unique, depot-specific gene expression signatures. Immunofluorescence staining of human adipose tissue revealed the presence of pericytes both distant from and adjacent to vasculature in human adipose tissue. Additionally, we demonstrated the potential of human brain pericytes and aortic vascular smooth-muscle cells to differentiate into adipocytes in vitro on the basis of intracellular lipid accumulation and expression of adipocyte markers.

Conclusions: Human adipose cell populations are distinct from mice, and the pericyte subpopulation in human adipose tissues are present across adipose depots. Given that vascular mural cells, including pericytes and smooth-muscle cells, can undergo adipogenesis, we postulate that they are a novel source of adipocytes in the vascular microenvironment.

Keywords: adipogenic differentiation; adipose progenitor cells; pericyte; perivascular adipose tissue; scRNA‐seq; vascular smooth‐muscle cells.

Figure 1. Analysis of cell populations in human aortic perivascular adipose tissue.

A, UMAP of 3 harmonized human PVAT samples (n=3). Data were filtered on the basis of number of unique features, percent.mt, and doublets were removed using Scrublet. Indicated for each cell population is the proportion of total, reflecting all cells that passed quality control filtration. B, UMAP with protein tyrosine phosphatase receptor type C+ immune cells removed. C, UMAP segregated by tissue donor, visualizing individual donor (n=3) contributions to each cell type. Overall, cell types identified for all present in each donor. D, Results of differential gene expression analysis between white and brown adipocytes displayed in a DotPlot. Brown adipocytes expressed uncoupling protein 1, phosphotyrosine interaction domain containing 1, and peroxisome proliferator activated receptor γ coactivator 1 α as well as heightened expression of PLIN5 and EBF2 when compared with white adipocytes in human PVAT. E, Population markers used to identify clusters displayed in a DotPlot. Percentage of marker expressing cells proportional to size of dot. Average expression value of a marker in a population displayed as color intensity. PDGFRβ, COL25A1, RGS5, ABCC9, and POSTN were used to characterize a newly identified pericyte population. F, UMAP of the immune cell population, showing separation of macrophages and TNK cells. G, Comparison of gene markers distinguishing TNK cells from macrophages in D. No formal statistical analyses were conducted due to the absence of an experimental control group. Observations are based on qualitative assessments of gene expression patterns and marker presence across identified cell populations. ABCC9 indicates ATP‐binding cassette subfamily C member 9; COL25A1, collagen 25 α 1; PDGFRβ, platelet‐derived growth factor receptor β; POSTN, periostin; PVAT, perivascular adipose tissue; RGS5, regulator of G protein signaling 5; TNK, thymic natural killer; and UMAP, uniform manifold approximation and projection.

Figure 2. Comparison of human pericytes and smooth‐muscle cells from heart and PVAT.

Single‐nucleus RNA sequencing data from human PVAT (n=3) and heart tissue (n=38) were harmonized. Cardiac tissue was derived from samples of patients undergoing valve repair surgery. A, UMAP of harmonized data, and (B) contribution from the heart and PVAT. Circled are the pericyte and smooth‐muscle cell data that were further analyzed. Overall data points due to analyzed nuclei were lower in PVAT due to smaller tissue size. C, UMAP focusing on the pericyte and smooth‐muscle cell populations from heart (left) and PVAT (right). When segregated by tissue of origin, pericytes and smooth‐muscle cells still cluster on the basis of cell type. D, Defined pericyte markers such as PGDFRβ, RGS5, ABCC9, and MYO1B were high in pericytes regardless of tissue origin. Markers such as POSTN, COL25A1, VIPR1, and TRPC6 were highly expressed only in PVAT pericytes. Smooth‐muscle markers were low in pericytes from both tissues. Due to variability in technical preparation methods and the integrated nature of the data, formal statistical analyses were not performed. Interpretations are based on qualitative assessments of marker presence or absence across cell populations. ABCC9 indicates ATP‐binding cassette subfamily C member 9; COL25A1, collagen 25 α 1; MYO1B, myosin 1B; PDGFRβ, platelet‐derived growth factor receptor β; POSTN, periostin; PVAT, perivascular adipose tissue; RGS5, regulator of G protein signaling 5; TRPC6, transient receptor potential cation channel subfamily C member 6; UMAP, uniform manifold approximation and projection; and VIPR1, vasoactive intestinal peptide receptor 1.

Figure 3. Comparison of human PVAT with human BAT and WAT.

Human PVAT (n=3), BAT (n=9), and WAT consisting of both subcutaneous and visceral depots (n=13) were harmonized to investigate tissue heterogeneity, shared populations, and depot specific transcriptomic signals. A, UMAP of the harmonized data. Immune cells were removed from these analyses. B, Markers used to identify cell types displayed in a DotPlot. Pericytes were found in all human adipose tissue sources. Adipocytes from BAT and PVAT are more transcriptionally similar than adipocytes found in WAT. WAT features a unique small population of pericytes and smooth‐muscle cells in addition to the larger groupings that are transcriptionally similar to those found in BAT and PVAT. C, Split UMAP divided by tissue type. D, FeaturePlot visualizing the expression of XIST, a sex specific gene located on the X chromosome. Color intensity indicates unscaled normalized expression. Of particular interest is the lack of expression of XIST in the satellite smooth‐muscle cells and pericytes contributed from the WAT data. E, Results of differential gene expression analysis performed to query transcriptomic differences in pericytes from different adipose tissues displayed in DotPlot. Overall, pericyte markers showed very similar expression patterns across tissue types, with the most notable differences being the expression of FABP4 in pericytes from WAT and PVAT, while FABP4 signal was absent in pericytes from BAT. Additionally, a greater proportion of cells express RGS5 in WAT pericytes in comparison with those found in BAT or PVAT. F, Markers in human adipose‐derived fibroblast (putative progenitors) and GATA6+ adipose progenitor cells as determined by differential expression analysis visualized in a DotPlot. Formal statistical analyses were not conducted due to variability in technical preparation methods and the integrated nature of the data. Instead, interpretations are based on qualitative assessments of marker presence or absence across cell populations. BAT indicates brown adipose tissue; FABP4, fatty acid–binding protein 4; PVAT, perivascular adipose tissue; UMAP, uniform manifold approximation and projection; and WAT, white adipose tissue.

Figure 4. Comparing adipose stem and progenitor cells from multiple human adipose depots.

Adipose stem and progenitor cells as well as mesothelial cells from human PVAT (n=3), BAT (n=9), and WAT consisting of both subcutaneous and visceral depots (n=13) were isolated informatically to investigate transcriptional signals of adipose progenitor subtypes in these tissues. A, FeaturePlot of the harmonized data displaying levels of fibroblastic adipose progenitor cell marker PGDFRA. B, FeaturePlot of the harmonized data displaying levels of putative thermogenic adipogenesis marker, observed highly in the mesothelial cells. Arrays of adipose stem cell (C) and adipose regulator (D) gene expression levels displayed in a split FeaturePlots visualizing cells by depot of origin. E, DotPlot displaying expression levels of relevant genes documented to be involved in adipogenesis. Due to methodological variability and data integration, analyses relied on qualitative assessment of marker presence across cell populations rather than formal statistical testing. BAT indicates brown adipose tissue; PDGFRA, platelet‐derived growth factor receptor α; PVAT, perivascular adipose tissue; and WAT, white adipose tissue.

Figure 5. Human adipose tissue production of pericyte markers.

Immunofluorescence representative images depicting the spatial distribution and colocalization of selected pericyte biomarkers within the PVAT and subcutaneous microenvironments of coronary artery bypass grafting patients. A, Subcutaneous and PVAT sections, demonstrating specific costained fluorescence labeling for CD31 (n=3) and RGS5 (n=3), ABCC9 (n=3), or PDGFRβ (n=3). B, PVAT sections exhibit distinctive costained fluorescence labeling, encompassing CD31 and additional markers including NG2 (n=3), TRPC6 (n=3), ABCC9 (n=3), and COL25A1 (n=4). Patterns were investigated using paired tissue samples of PVAT and subcutaneous tissue from 3 individual patient representatives to ensure consistency and reproducibility. Nuclei stained with DAPI. Scale bars=20 μm. No formal statistical analyses were performed due to the lack of an experimental control group. ABCC9 indicates ATP‐binding cassette subfamily C member 9; CD31, cluster of differentiation 31; COL25A1, collagen 25 α 1; FITC, fluorescein isothiocyanate; IgG, tissue‐matched isotype immunoglobulin G negative control; NEG, secondary‐antibody tissue‐matched negative control, NG2, chondroitin sulfate proteoglycan 4; PDGFRβ, platelet‐derived growth factor receptor β; PVAT, perivascular adipose tissue; RRX, rhodamine X; and TRPC6, transient receptor potential cation channel subfamily C member 6.

Figure 6. Adipogenic differentiation and lipogenesis of human smooth‐muscle cells in vitro.

Human primary VSMC populations were screened for pericyte markers. A, Normal human aortic VSMC populations (n=4) number 70008916 (lane A) and number 70045161 (lane B) from American Type Culture Collection, #10 and #53 primary VSMC isolated from aortic biopsy adjacent to abdominal aneurysm, dVSMC (n=6) isolated from abdominal aortic aneurysm biopsy, and HBVPs, n=1 were cultured in 6‐well plates in growth medium until 70% to 80% confluency. Cells were lysed and cell lysates were subjected for immunoblotting analysis as indicated target proteins and followed by appropriate housekeeping protein HSP90 or tubulin. Target proteins were normalized to the housekeeping protein on the same membrane and graphed in (B). C through F, Primary human VSMC (American Type Culture Collection; no. 70008916), at passages 6 to 9 were subjected to adipogenic/lipogenic induction medium (smooth‐muscle cell growth medium 2 containing 250 μmol/L IBMX, 2 nmol/L T3, 100 nmol/L dexamethasone, 10 nmol/L hydrocortisone, 5 μmol/L indomethacin, 1.25 μmol/L insulin, 33 μmol/L biotin with either 1 μmol/L rosiglitazone (Adipo‐R) or 5 μmol/L T0901317 (Lipo‐T) for 6 days, and then maintained in smooth‐muscle cell growth medium 2 containing 1.25 μmol/L insulin, 33 μmol/L biotin with either 1 μmol/L rosiglitazone or 5 μmol/L T0901317, respectively, for another 6 days. Cells were fixed for oil red O staining, data representing 1 biological replicate, with methodologies performed in technical triplicate (C) and quantification of lipid (D). Cell lysates were subjected to immunoblotting to detect proteins indicated (E), and changes in relative protein levels normalized by appropriate housekeeping protein on the same blot quantified (F). G through K, The same experiment was performed with primary dVSMC isolated from abdominal aortic aneurysm (ID no. 64, biological [n=1]). G, Oil red O staining and quantification (H) after differentiation. I, Immunoblot with proteins indicated, with quantification in (J). K, Total RNA was collected for reverse transcriptase quantitative polymerase chain reaction as indicated. Relative mRNA levels were calculated using δδCt method (details in Methods section). Scale bars=200 μm. Statistical analysis was performed using a Brown‐Forsythe test prior to using 1‐way ANOVA followed by Tukey's post hoc multiple comparisons test. When evaluating sterol regulatory element‐binding protein‐1 protein abundance, a 2‐way ANOVA followed by Sidak's post hoc multiple comparisons test was conducted. Data are presented as mean±SD, with error bars representing the SD, analyses conducted in Prism 7 (GraphPad Software). Statistical significance was determined at *P<0.05; **P<0.01; ***P<0.001. Ct indicates number of amplification cycles reaching the threshold fluorescence; dVSMC, dilated vascular smooth‐muscle cell; HBVP, primary human brain pericyte; IBMX, 3‐isobutyl‐1‐methylxanthine; and VSMC, vascular smooth‐muscle cell.

Figure 7. Adipogenic differentiation and lipogenesis of human brain pericytes in vitro.

Primary HBVPs were grown in 6‐well plates in growth medium until 70% to 80% confluency. At passages 6 to 9, cells were subjected to adipogenic/lipogenic induction medium (medium containing 250 μmol/L IBMX, 2 nmol/L T3, 100 nmol/L dexamethasone, 10 nmol/L hydrocortisone, 5 μmol/L indomethacin, 1.25 μmol/L insulin, 33 μmol/L biotin with either 1 μmol/L rosiglitazone (Adipo‐R) or 5 μmol/L T0901317 (Lipo‐T) for 6 days, and then maintained in smooth‐muscle cell growth medium 2 containing 1.25 μmol/L insulin, 33 μmol/L biotin with either 1 μmol/L rosiglitazone or 5 μmol/L T0901317, respectively, for another 6 days. Cells were fixed for oil red O staining after differentiation (A), and quantification of lipid (B). C, Immunoblot to assess proteins indicated was performed, and quantification shown (D). Data represent 1 biological replicate, with methodologies performed in technical triplicate. E, Reverse transcriptase quantitative polymerase chain reaction was performed to assess gene expression in pericytes after differentiation, and the relative mRNA levels calculated using δδCt method were represented. Scale bars=200 μm. Statistical analysis was performed using a Brown–Forsythe test before using 1‐way ANOVA followed by Tukey's post hoc multiple comparisons test. Data are presented as mean±SD, with error bars representing the SD; analyses conducted in Prism 7 (GraphPad Software). Statistical significance was determined at *P<0.05; **P<0.01; ***P<0.001. Ct indicates number of amplification cycles reaching the threshold fluorescence; HBVP, primary human brain pericyte; and IBMX, 3‐isobutyl‐1‐methylxanthine.