Identification of functionally distinct fibro-inflammatory and adipogenic stromal subpopulations in visceral adipose tissue of adult mice

White adipose tissue (WAT) remodeling is dictated by coordinated interactions between adipocytes and resident stromal-vascular cells; however, the functional heterogeneity of adipose stromal cells has remained unresolved. We combined single-cell RNA-sequencing and FACS to identify and isolate functionally distinct subpopulations of PDGFRβ+ stromal cells within visceral WAT of adult mice. LY6C- CD9- PDGFRβ+ cells represent highly adipogenic visceral adipocyte precursor cells (‘APCs’), whereas LY6C+ PDGFRβ+ cells represent fibro-inflammatory progenitors (‘FIPs’). FIPs lack adipogenic capacity, display pro-fibrogenic/pro-inflammatory phenotypes, and can exert an anti-adipogenic effect on APCs. The pro-inflammatory phenotype of PDGFRβ+ cells is regulated, at least in part, by NR4A nuclear receptors. These data highlight the functional heterogeneity of visceral WAT perivascular cells, and provide insight into potential cell-cell interactions impacting adipogenesis and inflammation. These improved strategies to isolate FIPs and APCs from visceral WAT will facilitate the study of physiological WAT remodeling and mechanisms leading to metabolic dysfunction. Editorial note: This article has been through an editorial process in which the authors decide how to respond to the issues raised during peer review. The Reviewing Editor's assessment is that all the issues have been addressed (see decision letter).


Introduction
White adipose tissue (WAT) represents the principle site for safe and efficient energy storage in mammals. WAT, as a whole, is considerably heterogeneous. WAT is composed of energy-storing adipocytes, various immune cell populations, vascular cells, adipocyte precursor cells (APCs), and largely uncharacterized stromal populations. The development and function of adipose tissue is highly dependent on critical interactions between adipocytes, APCs, immune cells, and endothelial cells (Han et al., 2011;Hong et al., 2015).
WAT has a unique and remarkable capacity to expand and contract in size in response to changes in demand for energy storage. In the context of positive energy balance (nutrient excess), WAT expands to meet the increased demand for energy storage, leading ultimately to the condition of obesity. The manner by which WAT expands is a critical determinant of metabolic health in obesity. It has long been appreciated that individuals who preferentially accumulate WAT in subcutaneous regions are at a relatively lower risk for developing insulin resistance when compared to equally obese individuals with central (visceral) adiposity (Kissebah et al., 1982;Krotkiewski et al., 1983). It is now widely believed that visceral and subcutaneous WAT depots represent fundamentally distinct types of WAT (Karastergiou et al., 2013;Lee et al., 2013;Macotela et al., 2012;Yamamoto et al., 2010). Indeed, visceral and subcutaneous WAT depots emanate from distinct developmental lineages (Chau et al., 2014).
Importantly, another clear determinant of metabolic health in obesity is manner in which individual WAT depots expand and 'remodel' Lee et al., 2010). WAT 'remodeling' associated with obesity can be described as both quantitative and qualitative changes in adipocyte numbers and stromal-vascular cell composition. Pathological WAT expansion is characterized by the presence of enlarged adipocytes, excessive macrophage accumulation, and fibrosis (Divoux et al., 2010;Gustafson et al., 2009;Hardy et al., 2011;Klö ting and Blüher, 2014;Sun et al., 2013). The prevailing hypothesis is that as 'overworked' fat cells reach their storage capacity, adipocyte death, inflammation, and fibrosis ensue Sun et al., 2011). This is often associated with the deleterious accumulation of lipids in the liver, skeletal muscle, pancreas, and heart (termed 'lipotoxicity') (Unger and Scherer, 2010). Healthy WAT expansion occurs when adipose tissue expands through adipocyte hyperplasia (increase in adipocyte number through de novo differentiation) (Denis and Obin, 2013;Kim et al., 2014;Klö ting et al., 2010). This is associated with a lower degree of chronic tissue inflammation and fibrosis. These adipose phenotypes of the 'metabolically healthy' obese tightly correlate with sustained insulin sensitivity in these patients. To date, the factors dictating a healthy vs. unhealthy WAT expansion in obesity remain poorly defined. In particular, the array of cell types within the adipose stromal-vascular compartment contributing to the remodeling of WAT in obesity has remained largely undefined. eLife digest Fat tissue, also known as white adipose tissue, specializes in storing excess calories.
Much of this storage happens under the skin, but fat tissue can also build up inside the abdomen and surround organs, where it is known as 'visceral' fat. When visceral fat tissue is unhealthy, it may help diseases such as diabetes and heart disease to develop.
Unhealthy fat tissue contains enlarged fat cells, which may die from overwork. The stress this places on the surrounding tissue activates the immune system, causing inflammation and the buildup of collagen fibers around the cells (a condition known as fibrosis). Not all people develop this type of unhealthy fat tissue, but we do not yet understand why.
In many tissues, blood vessels serve as a home for several types of adult stem cells that help to rejuvenate the tissue following damage. To identify these cells, Hepler et al. analyzed the genes used by more than 3,000 cells living around the blood vessels in the visceral fat of adult mice. Recent work had already revealed that stem cells called adipocyte precursor cells live in this region. Hepler et al. now reveal the presence of a second group of cells, termed fibro-inflammatory progenitor cells (or FIPs for short).
To investigate the roles of each cell type in more detail, Hepler et al. developed a new technique to isolate the adipocyte precursor cells from other cell types. When grown in the right conditions in petri dishes, the adipocyte precursor cells were able to form new fat cells. They could also make new fat cells when transplanted into mice that lacked fat tissue. By contrast, the FIPs can suppress the activity of adipocyte precursor cells and activate immune cells. They may also help fibrosis to develop.
It is not yet clear whether FIPs are present in human fat tissue. But, if they are, understanding them in greater detail may suggest new ways to treat diabetes and heart disease in obese people.
The growing appreciation for the casual link between adipose tissue distribution and remodeling with systemic metabolic health has sparked considerable interest in defining the adipocyte precursors giving rise to fat cells in adults and the mechanisms controlling their differentiation in vivo . In male C57BL/6 mice, adipose tissues expand in diet-induced obesity in a depot-selective manner. The epididymal WAT depot expands through both adipocyte hypertrophy and adipocyte hyperplasia (Jeffery et al., 2015;Kim et al., 2014;Wang et al., 2013b). The inguinal subcutaneous WAT depot expands almost exclusively by adipocyte hypertrophy. We recently reported that visceral adipocytes emerging in association with HFD feeding originate, at least in part, from perivascular precursors expressing Pdgfrb . Pdgfrb encodes the platelet-derived growth factor receptor b chain (PDGFRb protein) and is a widely used marker of perivascular cells (Armulik et al., 2011). We previously employed a pulse-chase lineage tracing mouse model to track the fate of Pdgfrb-expressing cells in adipose tissue. Following HFD feeding, Pdgfrb-expressing cells give rise to white adipocytes within visceral WAT depots . The ability of these precursors to undergo de novo adipogenesis in the setting of dietinduced obesity is critical for healthy visceral WAT expansion (Shao et al., 2018). Inducible genetic disruption of Pparg, the master regulatory gene of adipocyte differentiation, in Pdgfrb-expressing cells leads to a loss of de novo adipogenesis from Pdgfrb-expressing cells in the visceral WAT depot of diet-induced obese mice; this exacerbates the pathologic remodeling of this depot (i.e. increased inflammation and fibrosis). Driving de novo adipogenesis from Pdgfrb-expressing cells through transgenic Pparg expression leads to a healthy expansion of visceral WAT (lower inflammation and small adipocytes) (Shao et al., 2018). The highly adipogenic subpopulation of PDGFRb+ cells in gonadal WAT (gWAT) is quantitatively enriched in the expression of Pparg, as well as its upstream regulatory factor, Zfp423 (Gupta et al., 2012;Tang et al., 2008;Vishvanath et al., 2016). PDGFRb + cells enriched in these adipogenic factors express several mural cell (pericyte/smooth muscle) markers and reside directly adjacent to the endothelium in WAT blood vessels (Gupta et al., 2012;Tang et al., 2008;Vishvanath et al., 2016). Using Zfp423 reporter mice (Zfp423 GFP BAC transgenic mice), we revealed that PDGFRb+ cells expressing high levels of Zfp423 (GFP+ or Zfp423 High ) represent highly committed preadipocytes while Zfp423 Low cells (GFP-) lacked significant adipogenic capacity, and exhibited significantly different global patterns of gene expression . These observations suggested that the pool of PDGFRb+ cells in visceral WAT is functionally heterogeneous, with cells possessing distinct cellular phenotypes.
In this study, we set out to explore the functional heterogeneity within Pdgfrb-expressing cells of visceral WAT from adult mice. Furthermore, our objective was to identify improved strategies to purify adipocyte precursor populations from these depots. Through single-cell RNA-sequencing, we identified functionally distinct subpopulations of Pdgfrb-expressing progenitor cells. We identified a unique population of cells that display fibrogenic and functional pro-inflammatory phenotypes, and lack inherent adipogenic capacity. These fibro-inflammatory progenitors (termed here as 'FIPs') can be purified by the use of commercially available antibodies (LY6C + PDGFRb+). On the other hand, LY6C-CD9-PDGFRb+ cells represent a distinct pool of highly adipogenic visceral adipocyte precursor cells ('APCs') that robustly differentiate spontaneously in vitro in growth media containing insulin. The frequency of these PDGFRb+ subpopulations is highly regulated under physiological conditions. These data reveal the functional heterogeneity of perivascular progenitors within visceral WAT and provide insight into how the adipose stroma can control WAT remodeling. Moreover, the molecular profiles obtained for FIPs and APCs from visceral WAT, along with the strategies to isolate these cells, will facilitate the study of physiological WAT remodeling in vivo.

Results
Single-cell RNA sequencing reveals molecularly distinct Pdgfrbexpressing subpopulations in visceral adipose tissue We previously derived a doxycycline-inducible (Tet-On) lineage-tracing model that allows for the indelible labeling of Pdgfrb-expressing perivascular cells in adipose tissue of adult mice (Pdgfrb rtTA ; TRE-Cre; Rosa26R mT/mG ; herein, 'MuralChaser mice') . Prior to exposing animals to doxycyline, all cells within the stromal-vascular fraction (SVF) of adult gonadal WAT (gWAT) express membrane tdTomato from the Rosa26 locus. Following 9 days of exposure to doxycycline-containing chow diet, Cre-mediated excision of the loxP-flanked tdTomato cassette occurs in Pdgfrb-expressing cells, and membrane-bound GFP (mGFP) expression is constitutively activated ( Figure 1A). As previously reported and confirmed here, FACS analysis indicated that nearly all mGFP+ cells are PDGFRb+ as expected, and are devoid of CD45 (hematopoietic), CD31 (endothelial), and CD11b (monocyte/macrophage) expression (Figure 1-figure supplement 1A) . Moreover, mGFP expression following transient doxycycline exposure is confined predominately to peri-endothelial cells in adult gonadal WAT (Figure 1-figure supplement 1B) .
We set out to test the hypothesis that Pdgfrb-expressing perivascular cells in gonadal visceral WAT of adult mice are heterogeneous, with subpopulations harboring functionally distinct phenotypes. To this end, we performed single cell RNA-sequencing (scRNA-seq) of mGFP+ cells isolated from gWAT of lean (chow fed) 8 week-old male MuralChaser mice following 9 days of doxycycline exposure. tSNE analysis of 1045 cell transcriptomes revealed distinct cell clusters exhibiting unique transcriptional profiles ( Figure 1B,C). Many of the top 20 most enriched transcripts in Cluster 1A and Cluster 1B correspond to notable genes related to adipogenesis and/or adipocyte gene expression ( Figure 1D). In particular, the majority of cells in Clusters 1A and 1B express high levels of Pparg, Fabp4, Hsd11b1, and Lpl, indicating these clusters may represent the PDGFRb+ APC population within visceral WAT ( Figure 1E). Interestingly, Cluster 1B further enriches in the expression of Pparg, Cebpa, and other markers of terminal adipocyte differentiation, including Plin1, Fabp5, Car3, and Cd36 ( Figure 1F). Notably, the expression of Adipoq, Retn, and Adrb3, genes typically characteristic of mature adipocytes, were detected within some cells within Cluster 1B ( Figure 1F). Unbiased gene set enrichment analysis (GSEA) revealed that cells of Cluster 1A/B enriched for gene sets related to 'adipogenesis' and cells of Cluster 1B enriched for gene signatures of 'oxidative phosphorylation,' 'adipogenesis,' and fatty acid metabolism (Tables 1 and 2). These data suggest Cluster 1A and 1B represent 'adipocyte precursor cells' (APCs), with Cluster 1B representing a subpopulation of APCs that are 'committed preadipocytes'.
The cells in Cluster 2 were highly enriched in the expression of genes associated with fibrosis and inflammation, including Fn1, Loxl2, Tgfb2, and Ccl2 ( Figure 1D and G). GSEA revealed the enrichment of numerous gene signatures characteristic of a fibrogenic and inflammatory phenotype, including gene sets corresponding to 'inflammatory response,' 'TGFb signaling,' 'TNFa signaling,' and 'hypoxia' ( Table 3). This fibro-inflammatory molecular signature of Pdgfrb-expressing cells suggested this subpopulation represents 'fibro-inflammatory progenitors' (herein, termed 'FIPs').
Cluster 3 was molecularly quite distinct from Clusters 1A/B and 2. Interestingly, Cluster 3 had a mesothelial-like cell (herein, 'MLCs') expression profile. Mesothelial cells are epithelial cells of mesodermal origin that form a monolayer (mesothelium) lining the visceral serosa. Mesothelial cells and mural cells share a common developmental lineage. Multiple genetic lineage tracing studies in mice indicate that various stromal cell populations within visceral tissues, including APCs, descend from embryonic mesothelial cells (Chau et al., 2014;Rinkevich et al., 2012). Mesothelial cells have been linked to multiple aspects of adipose tissue development and remodeling, including adipogenesis and inflammation (Darimont et al., 2008;Gupta and Gupta, 2015;Mutsaers et al., 2015). Cluster 3 was enriched for genes representing common mesothelial/epithelial markers, such as Msln, Upk3b, Krt8, and Krt14 ( Figure 1D and H). The presence of this cluster suggested that the Pdgfrb rtTA transgene targets at least a subset of visceral WAT associated mesothelial cells. Indeed, following transient doxycycline treatment of MuralChaser mice, a few mGFP+ cells can be observed within in the outermost epithelial layer of gonadal WAT (Figure 1-figure supplement 1C). Moreover, mGFP+ cells can be occasionally observed within cultures of isolated mesothelial cells obtained from gWAT of doxycycline-treated MuralChaser mice (Figure 1-figure supplement 1D).
We performed a second scRNA-seq analysis utilizing independently isolated mGFP+ cells from gonadal WAT MuralChaser mice (Figure 1-figure supplement 2). From the second scRNA-seq dataset, we again identified the same major subpopulations of Pdgfrb-expressing cells. All together, these scRNA-seq data reveal molecularly distinct Pdgfrb-expressing subpopulations in visceral adipose tissue.   Molecularly distinct visceral WAT PDGFRb+ subpopulations can be isolated by FACS Next, we developed a strategy to isolate these molecularly distinct cell populations by flow cytometry from wild type mice. For this purpose, we treated Cluster 1A and 1B as one broad 'APC' population ( Figure 2A). Candidate cell surface markers were selected on the basis of their corresponding gene expression in the three PDGFRb+ subpopulations and the availability of commercial antibodies suitable for FACS. Of note, Ly6c1 expression was abundant in FIPs but not APCs ( Figure 2B). The expression of Cd9, a recently described marker of fibrogenic cells (Marcelin et al., 2017), was abundantly expressed in both the FIPs and MLCs ( Figure 2B). Therefore, we isolated the three populations based on these markers using fluorescence-activated cell sorting. PDGFRb+ cells (CD31-and CD45-) were subdivided on the basis of LY6C and CD9 immunoreactivity ( Figure 2B,C). Three distinct subpopulations of PDGFRb+ cells were apparent: LY6C-CD9-(APCs), LY6C+ (FIPs), and LY6C-CD9+ (MLCs) cells ( Figure 2C). Flow cytometry analysis consistently revealed that LY6C+ PDGFRb+ cells were more abundant than LY6C-CD9-PDGFRb+ cells and Ly6C-CD9+ PDGFRb+ cells ( Figure 2D). Importantly, gene expression analysis by qPCR revealed that LY6C-CD9-PDGFRb+ Figure 1 continued (mGFP) reporter expression. The gating strategy shows prospective isolation of tdTomato-GFP+ cells from the stromal vascular fraction of gonadal WAT (gWAT). (B) t-distributed stochastic neighbor embedding (tSNE) plot of 1045 tdTomato-GFP+ cells isolated from pooled gWAT depots from five male MuralChaser mice. Equal numbers of cells were combined from five individual mice for single-cell RNA-sequencing. Clustering was generated using k-means = 4. See        Genes annotated by the GO term GO:0006954. The immediate defensive reaction to infection or injury caused by chemical or physical agents. The process is characterized by local vasodilation, extravasation of plasma into intercellular spaces and accumulation of white blood cells and macrophages.

LY6C-CD9-PDGFRb+ cells are functional visceral white adipocyte precursors
The global molecular signature of LY6C-CD9-PDGFRb+ cells (Cluster 1) suggests this population represents APCs. Indeed, freshly sorted LY6C-CD9-PDGFRb+ cells are enriched in Pparg expression when compared to LY6C+ PDGFRb+ cells ( Figure 3-figure supplement 1A). We explored this hypothesis by testing the ability of these subpopulations to undergo adipocyte differentiation in vitro. We isolated and cultured all three subpopulations in growth medium containing 2% FBS and 1% ITS (insulin, transferrin, selenium). These represent culture conditions that we previously established for growth and differentiation of gWAT-derived PDGFRb+ cells . Under these growth conditions, LY6C+ PDGFRb+ cells proliferate at a greater rate than LY6C-CD9-PDGFRb+ cells; however, the two subpopulations appear morphologically indistinguishable, with both populations appearing fibroblast-like until reaching confluence (

HALLMARK_ COMPLEMENT
Genes encoding components of the complement system, which is part of the innate immune system. 0.003505879 FN1, DPP4, CD55, TIMP2, ATOX1, S100A13,  GNGT2, FYN, KIF2A, IRF7, PLA2G4A, PLAT, CXCL1,  CALM1, EHD1, PFN1, ADAM9, IRF2, SERPINE1,  GNAI3, RHOG, PRCP, MAFF, GCA, DOCK4, PLAUR,  GNB2, IL6, CEBPB, TIMP1, GNAI2, Cluster # 1 2 3 Collectively, these data indicate that LY6C-CD9-PDGFRb+ cells are highly adipogenic functional gonadal white adipocyte precursors, while LY6C+ PDGFRb+ cells are largely refractory to adipogenic stimuli. Several studies have defined APCs from gonadal WAT as SCA-1+ CD34+ CD24± cells that also express PDGFRa (Berry and Rodeheffer, 2013;Jeffery et al., 2015;Lee et al., 2012;Rodeheffer et al., 2008). In fact, most studies of gonadal WAT APCs isolate these cells on the basis of these markers. Additionally, recent studies identified CD38 as a marker of committed preadipocytes (Carrière et al., 2017). The scRNA-seq analysis and follow-up qPCR analyses of isolated subpopulations revealed that all three PDGFRb+ subpopulations indeed expressed Pdgfra, Ly6a (SCA-1), and Cd34; however, the mRNA levels of Ly6a and Cd34 are actually lower in LY6C-CD9-PDGFRb+ APCs than in LY6C+ PDGFRb+ cells (FIPs) (Figure 3-figure supplement 1F,G). As expected, all three subpopulations expressed Pdgfrb; however, mRNA levels of Pdgfrb were quantitatively lower in FIPs than in the APCs and MLCs. qPCR analysis indicated that levels of Cd24a were low in all three PDGFRb+ subpopulations. Cd38 was present predominately in LY6C-CD9-PDGFRb + cells, consistent with the notion that CD38 identifies APCs from this depot (Carrière et al. Recently, Burl et al. reported scRNA-seq profiles of adipose SVF cells, creating a cellular atlas of potential adipocyte precursor populations (perivascular and non-perivascular) (Burl et al., 2018). Notably, the authors identified two prominent populations within the gonadal WAT depot, termed adipose stem cell (ASC) 1 and ASC 2. Moreover, they identified two additional smaller ASC subpopulations that were considered 'differentiating' ASCs and 'proliferating' ASCs. The identified populations were not isolated and explored functionally in their study; however, a comparison of the molecular profiles strongly suggests that ASC 1 defined by the authors bears close resemblance to APC population defined in our study, while the ASC 2 population bears close resemblance to the FIPs discovered here (Figure 3-figure supplement 1I). Markers of the differentiated/proliferative ASCs aligned closely to the committed PDGFRb+ preadipocyte depicted in Figure 1B. Taken together, our data here suggest a refined strategy to isolate functional white adipocyte precursors from visceral WAT of adult mice. Our prior studies of Zfp423 GFP reporter mice indicated that gonadal WAT PDGFRb+ cells expressing GFP are enriched in the expression of Pparg and are highly adipogenic in vitro (Gupta et al., 2012;Vishvanath et al., 2016). Additional studies by others indicated that this reporter captures committed preadipocytes within the skeletal bone marrow microenvironment (Ambrosi et al., 2017). Endogenous Zfp423 mRNA levels were found in all PDGFRb+ subpopulations, albeit at highest levels in APCs. (Figure 3-figure supplement 4A). We re-examined . Taken all together, these data indicate that endogenous Zfp423 mRNA expression is not confined exclusively to the APC subpopulation of PDGFRb+ cells in gWAT; however, Zfp423 GFP reporter mice represent a genetic tool to localize and enrich for committed preadipocytes from this depot.
Functionally distinct stromal populations from visceral, but not subcutaneous, WAT depots can be revealed on the basis of LY6C and CD9 expression Transcriptional programs of white adipocyte precursors are depot-and sex dependent (Macotela et al., 2012). Thus, we asked whether similar functional heterogeneity exists amongst PDGFRb+ cells within various WAT depots, and whether functionally distinct subpopulations could be selected for using the same FACS strategy described above. Indeed, the same three populations can be observed within the mesenteric and retroperitoneal depots of adult male mice, with LY6C-CD9-PDGFRb+ cells representing the highly adipogenic subpopulation ( Figure 4A-H). We also examined LY6C expression within PDGFRb+ SVF cells obtained from the inguinal and anterior subcutaneous WAT depots. We previously demonstrated that the total pool of PDGFRb+ cells from inguinal WAT is very highly adipogenic in vitro (Shao et al., 2018); however, remarkably, all PDGFRb + cells within the inguinal and anterior subcutaneous WAT depots expressed LY6C ( Figure 4I). These data suggest that if heterogeneity exists amongst PDGFRb+ cells in these subcutaneous depots, subpopulations could not be discriminated on the basis of LY6C expression. Therefore, functionally distinct perivascular cell subpopulations from visceral, but not subcutaneous, WAT depots can be revealed on the basis of LY6C and CD9 expression. We also asked whether visceral WAT in female mice contains APCs and FIPs, bearing similar molecular and functional properties. Within the SVF of peri-ovarian WAT, the same three distinct subpopulations of PDGFRb+ cells can be discriminated, with FIPs being the predominant population ( Figure 4-figure supplement 1A,B). Importantly, gene expression analysis by qPCR confirmed that LY6C-CD9-PDGFRb+ cells were enriched in the expression of genes that defined the epididymal WAT APC population (Cluster 1) (Figure 4-figure supplement 1C,F), including Pparg isoform 2. LY6C+ PDGFRb+ cells enriched for the mRNAs that initially defined the epididymal WAT FIPs (Cluster 2) (Figure 4-figure supplement 1D,G,H), and LY6C-CD9+ PDGFRb+ cells expressed mesothelial/epithelial markers (Figure 4-figure supplement 1E). Moreover, LY6C-CD9-PDGFRb+ cells from peri-ovarian WAT are functional adipocyte precursors; these cells, but neither FIPs nor MLCs, differentiate spontaneously upon reaching confluence in culture (Figure 4-figure supplement 1I-K). Collectively, these data provide evidence that functional APCs from both male and female visceral WAT can be isolated through this cell sorting strategy.

Visceral LY6C+ PDGFRb+ cells are anti-adipogenic and appear molecularly distinct from inguinal WAT Aregs
It is notable that very little spontaneous adipocyte differentiation occurs in cultures containing the total pool of visceral adipose PDGFRb+ cells ( Figure 3A,E,I,M,Q), despite the presence of numerous APCs within this population. This suggested that perhaps the presence of FIPs within these cultures influenced the differentiation capacity of neighboring APCs in vitro. Therefore, we also tested the impact of conditioned media from cultured FIPs on the differentiation capacity of APCs residing in parallel cultures. Remarkably, APCs exposed to conditioned media from FIPs, but not from parallel cultures of APCs, expressed lower levels of Pparg ( Figure 5A). Moreover, APCs exposed to conditioned media from FIPs lost a significant degree of adipogenic capacity ( Figure 5B,C,E). Conditioned media from cultures of MLCs had only a slight inhibitory effect on the terminal differentiation of APCs ( Figure 5D,E). Collectively, these data not only suggest that FIPs lack significant adipogenic capacity, but highlight the notion that these cells can actually be anti-adipogenic.
Recently, Schwalie et al. identified anti-adipogenic stromal cells within the inguinal WAT of mice (Schwalie et al., 2018). These cells, termed Aregs, are defined, in part, by the expression of CD142 and ABCG1 and exhibit perivascular localization. From our scRNA-seq dataset, we observed that F3 expression (encoding CD142) is detected in all PDGFRb+ clusters of gonadal WAT, albeit not enriched in FIPs ( Figure 5F). Abcg1 expression was not detected by the sequencing analysis in any population. We also examined the levels of mRNA for these two markers directly by quantitative PCR analysis. Consistent with the sequencing data, neither marker was enriched in FIPs ( Figure 5G). We also examined additional genes (23 in total) whose expression defines Aregs, as identified by Schwalie et al, by assessing their expression level within our scRNA-seq dataset ( Figure 5H). Levels of transcripts corresponding to a number of these genes were detectable, albeit at low levels. Notably, there was no selective enrichment of the broader set of Areg markers within FIPs or APCs. As such, despite some shared functional similarities, inguinal adipose Aregs and the gonadal adipose FIPs described here appear molecularly distinct.
Visceral LY6C+ PDGFRb+ cells are fibrogenic and exert a functional pro-inflammatory phenotype As described above, GSEA of scRNA-seq profiles also identified a gene expression profile suggestive of active TGFb signaling within Cluster 2 cells ( Table 3). Indeed the expression of major collagens (Col1a1 and Col3a1) and some of the assayed genes associated with extracellular matrix accumulation were enriched in freshly isolated LY6C+ PDGFRb+ cells compared to the other PDGFRb+ subpopulations ( Figure 6-figure supplement 1A). In vitro, cultured FIPs and APCs were both responsive to treatment with recombinant TGFb; however, the expression of collagens examined remained higher and/or was further induced in FIPs ( Figure 6-figure supplement 1B). These data indicate that LY6C+ PDGFRb+ FIPs exhibit a phenotype characteristic of fibrogenic cells. The most striking result from GSEA was the enrichment of pathways related to active 'Tnfa signaling' and 'inflammatory response' in FIPs (Table 3). Remarkably, FIPs exhibited a robust inflammatory gene expression signature following acute exposure to pro-inflammatory molecules. Lipopolysaccharide (LPS) treatment induced inflammatory cytokine gene expression in both APCs and FIPs; however, the response was more robust in the latter population ( Figure 6A). The differential response to TNFa treatment was the most striking; FIPs, but not APCs, activate the expression of several proinflammatory cytokines under these conditions ( Figure 6B). These fibro-inflammatory cells displayed increased gene expression of numerous cytokines involved in the recruitment of leukocytes and the activation of immune cells. This suggested that FIPs have the potential to activate macrophages through cytokine production. To test this, we treated cultured bone marrow derived macrophages with conditioned media from LPS-treated FIPs, APCs, and MLCs ( Figure 6C). Macrophage cultures exposed to conditioned media from LPS-treated FIPs had the most robust induction of pro-inflammatory genes, including Tnfa, Il1b, and Il6 ( Figure 6D). These data highlight the potential for FIPs to exert a functional pro-inflammatory phenotype.
The frequencies and gene expression profiles of APCs and FIPs are differentially regulated in association with high-fat diet feeding In the setting of caloric excess, adipose tissue undergoes a dramatic remodeling as it expands to meet increased demands for energy storage. Shortly after the onset of high-fat diet (HFD) feeding, adipose tissue inflammation occurs (Hill et al., 2014;Xu et al., 2003). After 4-5 weeks of HFD feeding (60% kcal from fat), newly formed visceral adipocytes emerging from the PDGFRb+ lineage begin to appear . We asked if the frequency of FIPs and APCs were altered during the course of HFD feeding. Four weeks of HFD feeding did not appear to dramatically alter the absolute number of PDGFRb+ cells present in gWAT; however, the ratio of FIPs to APCs begins to increase by as early as one week of HFD feeding ( Figure 7A). We also analyzed BrdU incorporation into the mural cell populations during one week of HFD feeding. FIPs and the MLCs displayed the greatest BrdU incorporation ( Figure 7B,C). BrdU incorporation into APCs was significantly lower than observed in the FIPs ( Figure 7B,C). These data indicate that frequencies of APCs and FIPs are differentially regulated in vivo in association with high-fat diet feeding, with FIPs exhibiting a relatively higher degree of cell proliferation under these conditions.
The change in frequency of FIPs and APCs during HFD feeding prompted us to examine if their defining gene expression programs were altered under these conditions. One month of HFD feeding lead to a significant elevation in mRNA levels of Pparg isoform two expression in APCs, with a smaller increase occurring in MLCs ( Figure 7D). Pparg isoform two expression was not elevated in FIPs, consistent with their apparent lack of adipogenic potential ( Figure 7D). mRNA levels of proinflammatory cytokines and extracellular matrix components were further induced and/or remained more abundant in FIPs than in APCs or MLCs ( Figure 7E-M). Interestingly, APCs activated the expression of some of these genes (e.g. Il6, Tnfa, Col1a1, Col3a1) during HFD feeding. These data are consistent with the in vitro analyses highlighting the potential of APCs to trigger some degree of an inflammatory response in pro-inflammatory stimuli (see Figure 6). These data reveal that PDGFRb + subpopulations exhibit unique transcriptional responses to HFD feeding; however, these data also suggest that APCs have some capacity to adopt characteristics of FIPs in vivo. NR4A nuclear receptors regulate the pro-inflammatory phenotype of PDGFRb+ cells We sought to gain insight into the potential transcriptional mechanisms regulating the pro-inflammatory and adipogenic phenotypes of PDGFRb+ perivascular cells. A number of transcription factors were differentially expressed between FIPs and APCs; however, it was notable that the expression of all three members of the Nr4a family of nuclear hormone receptors was significantly enriched in the FIPs cluster ( Figure 8A). Gene expression analysis by qPCR of the isolated populations confirmed the significant enrichment of Nr4a1, Nr4a2, and Nr4a3 in FIPs isolated from chow-fed mice, with relatively lower expression in the APCs and MLCs ( Figure 8B).
Members of the NR4A family, including NR4A1 (NUR77), NR4A2 (NURR1), and NR4A3 (NOR1), have been implicated in the regulation of inflammation; however, their exact impact on inflammatory signaling appears cell-type specific (Rodríguez-Calvo et al., 2017). Following 4 weeks of HFD-feeding, the expression of Nr4a family members remained significantly enriched in FIPs ( Figure 8C). In vitro, the expression of all Nr4a family members in FIPs was increased following exposure to recombinant TNFa ( Figure 8D). Therefore, we assessed the consequences of retroviral-mediated overexpression of individual Nr4a family members on the inflammatory response of FIPs ( Figure 8E-G).
Overexpression of Nr4a1, Nr4a2, or Nr4a3, attenuated the pro-inflammatory response to TNFa ( Figure 8H-K). Moreover, we assessed the impact of retroviral-mediated knockdown of Nr4a1 on the inflammatory response in FIPs. Knockdown of Nr4a1 using three independent shRNAs led to an exaggerated response of FIPs to TNFa treatment ( Figure 8L-P). These gain-and loss of function studies suggest that FIPs activate Nr4a family members in response to pro-inflammatory stimuli, perhaps as a means to counter-regulate a sustained cellular inflammatory response. These data provide proof of concept that FIPs may be utilized as a tool to identify additional regulators of inflammatory signaling pathways.

Discussion
Visceral adipose tissue dysfunction in obesity is driven, at least in part, by chronic tissue inflammation, collagen deposition, and a loss of adipocyte precursor activity. WAT remodeling involves substantial qualitative and quantitative changes to the composition of the stromal compartment of the tissue; however, the functional heterogeneity of WAT stromal-vascular fraction has remained poorly defined and tools to isolate and study distinct subpopulations have been lacking. Here, we unveil functionally distinct PDGFRb+ stromal cell subpopulations in visceral WAT (Figure 9). Importantly, we have developed strategies to prospectively isolate these distinct populations using commercially available antibodies.
Functional analyses indicate that a relatively large subpopulation of PDGFRb+ perivascular cells in visceral gonadal WAT exert fibrogenic and pro-inflammatory phenotypes. These cells, termed here as 'FIPs,' lack adipogenic capacity in vitro but instead exhibit a fibrogenic phenotype. FIPs are physiologically regulated; the frequency of these cells increases upon HFD feeding. Clement and colleagues previously reported that fibrogenic cells residing in WAT could be identified by the expression of CD9 and PDGFRa (Marcelin et al., 2017). Indeed, FIPs express CD9 and PDGFRa; however, both CD9 and PDGFRa are also expressed in at least a subpopulation of mesothelial cells isolated from visceral WAT. As such, the selection of FIPs on the basis of LY6c and PDGFRb expression (LY6C+ PDGFRb+) represents a strategy to prospectively and specifically isolate FIPs from mouse gonadal WAT. Importantly, our data reveal a number of previously unappreciated ways in which perivascular cells may impact WAT remodeling (Figure 9). The LY6C+ PDGFRb+ cells described here have the capacity to inhibit adipocyte differentiation from APCs through the release of secreted factors. The presence of highly anti-adipogenic stromal cells within the total PDGFRb+ population may explain the apparent lack of adipogenic capacity that crude/unpurified visceral PDGFRb+ cultures possess vitro, despite the presence of APCs. It is notable that visceral adipose FIPs appear distinct from the recently identified Aregs of inguinal WAT. Stromal cell heterogeneity may be depot-specific, with different depots utilizing distinct cell types to control their function and plasticity. The exact identities of the secreted factors and mechanisms that mediate the anti-adipogenic activity of FIPs and Aregs are still unknown. Importantly, whether FIPs and Aregs act to suppress/restrain adipocyte hyperplasia under physiological settings in vivo needs to be further explored. FIPs also exert a functional pro-inflammatory phenotype in response to pro-inflammatory stimuli. It is notable that the frequency of FIPs increases following the onset of HFD feeding. This raises an intriguing hypothesis for future studies that perivascular stromal cells can modulate local tissue inflammation. This notion is in line with recent studies indicating that vascular mural cells can serve as 'gatekeepers' of inflammation in the lung (Hung et al., 2017). On the other hand, one may expect that the increased frequency of FIPs would also completely blunt the differentiation of APCs in this depot. This is clearly not the case as gonadal WAT in mice is able to expand through adipocyte hyperplasia in the setting of diet-induced obesity. One possibility is that the anti-adipogenic activity of FIPs (rather than the frequency per se) is diminished by local signals in an attempt to facilitate adipogenesis within the depot. Another possibility lies in the spatial distribution of activated FIPs and APCs; their proximity to one another may influence their activity. The identification of these populations from the MuralChaser mice essentially places them within the perivascular compartment of adipose tissue; however, where APCs and FIPs are localized and become activated within the tissue is still unknown. Clearly, additional studies of these cells will be needed in order to determine their precise contribution to WAT inflammation and health in various settings in vivo. Furthermore, it is certainly plausible that FIPs contribute to WAT remodeling beyond fibrosis, inflammation, and adipogenesis.
The ability to isolate functionally distinct subpopulations of mural cells affords the possibility of identifying factors regulating these diverse mural cell phenotypes. Our gene expression analysis revealed the enrichment in mRNA levels of Nr4a family members in FIPs. Several studies have implicated NR4A nuclear receptors as modulators of inflammatory signaling; the precise impact of NR4A members on inflammation appears context/cell type specific. In some studies, NR4A family members are observed as being pro-inflammatory (Pei et al., 2006). In other settings, NR4A expression is induced in response to pro-inflammatory stimuli but acts as a molecular brake on inflammatory signaling. Our gain-and loss of function studies suggest that FIPs activate Nr4a family members in response to pro-inflammatory stimuli to serve as a transcriptional brake on inflammatory cytokine gene expression. This counter-regulatory response may be a mechanism to limit cellular oxidative stress and apoptosis driven by inflammatory signaling (Rodríguez-Calvo et al., 2017). Chao et al. previously demonstrated that NR4A family members are potent regulators of adipocyte differentiation (Chao et al., 2008). Thus, NR4A members may play multiple roles in controlling of the fate and function of adipose perivascular cells. Additional studies involving the genetic manipulation of Nr4a family members in PDGFRb+ cells in vivo will be needed to elucidate the exact requirements of NR4A members in WAT remodeling in vivo. Importantly, the ability to isolate FIPs and APCs affords the possibility of employing several different types of genomic approaches in an effort to reveal novel molecular mechanisms controlling adipose tissue inflammation, fibrosis, and adipogenesis.
As described above, there has been tremendous interest in elucidating the identity of adipocyte precursors in adult adipose tissue. Pioneering studies from Friedman and colleagues led to a now widely-used strategy to prospectively isolate adipocyte progenitor cells from freshly isolated WAT (Rodeheffer et al., 2008). APCs have been isolated on the basis of CD29, SCA-1, and CD34 expression (CD29+ CD34+ SCA-1+ CD31-CD45-). These markers have proven to be quite useful for the selection of APCs from inguinal WAT and other WAT depots; however, a notable observation made here is that SCA-1 expression is in fact enriched in FIPs rather than APCs of the gonadal WAT depot. As such, the selection of cells on the basis of SCA-1 expression from this particular WAT depot yields a functionally heterogeneous population that likely includes FIPs, and perhaps even enriches for these cells. This may explain, at least in part, the notable lack of adipogenic potential that isolated gonadal CD34+ SCA-1+ cells possess in vitro (Church et al., 2014). Our prior work pointed to Zfp423 as a marker of committed preadipocytes; however, our scRNA-seq data reveal that Zfp423 expression is not confined exclusively to the APC subpopulation of PDGFRb+ cells; Pdgfrb-expressing MLCs and FIPs also express Zfp423. Nevertheless, the selection of GFP High PDGFRb+ cells from gWAT of Zfp423 GFP reporter mice enriches for committed APCs, perhaps reflecting increased promoter/transgene activity in these cells.
Recent scRNA-seq analyses from Granneman and colleagues provide a cellular atlas of putative adipocyte precursor populations in adipose tissue (Burl et al., 2018). Their analyses included all non-hematopoetic, non-endothelial cells of the isolated adipose stromal-vascular fraction. The strength of their approach is that it allows for one to capture both perivascular (PDGFRb+) and nonperivascular precursor populations. Our approach will identify precursor populations that express Pdgfrb/rtTA at the time of the pulse labeling. Pdgfrb expression declines as cell undergo differentiation into adipocytes. This means that cells even further committed to the adipocyte lineage (i.e. no longer express Pdgfrb) may not be captured through our analysis. Moreover, putative stem cell populations not yet expressing Pdgfrb may also be present and not captured (e.g. Pref1 rtTA targeted cells [Hudak et al., 2014]). Nevertheless, it is notable that most of the adipocyte precursor populations represented in the study by Burl et al. were indeed captured in our analysis. One cannot rule out the existence of additional adipocyte progenitor populations in any particular adipose depot; however, the congruency of the two independent studies suggests the MuralChaser model can identify and target the major APC populations residing within visceral WAT of mice. Moreover, the selection of LY6C-PDGFRb+ cells from gWAT using commercially available antibodies represents a refined and convenient strategy to isolate visceral adipocyte precursors from wild type mice or genetic mouse models of interest.
Single-cell transcriptomics has become very useful in revealing molecular heterogeneity amongst seemingly homogenous populations of cells. The challenge, however, is to determine whether molecularly distinct populations of cells represent distinct 'cell types,' or rather 'cell states' which are influenced by their local microenvironment. Here, we reveal molecular heterogeneity amongst PDGFRb+ cells within visceral white adipose tissue and begin to define the functional differences between the identified subpopulations. Our functional analyses suggest that the properties of visceral FIPs are at least somewhat stable; FIPs are quite limited in adipogenic potential in vitro and upon transplantation. They do not readily activate Pparg expression under the conditions examined. The phenotype of visceral APCs may be less stable. Visceral APCs have the potential to adopt characteristics of FIPs. In vitro and in vivo following HFD, APCs can activate the expression of pro-inflammatory cytokines. A caveat to most of our functional studies is that the cells are studied outside their native microenvironment. Under some physiological conditions, it is certainly plausible that multiple PDGFRb+ subpopulations give rise in vivo to adipocytes; such adipocytes might even possess unique functional characteristics. Our prior lineage-tracing studies using the MuralChaser model clearly established that adipocytes emerge from Pdgfrb-expressing cells; efforts to define the relative contribution of individual mural cell subpopulations will require more precise lineage-tracing approaches with more specific Cre drivers. Moreover, our current studies cannot exclude the possibility that even further heterogeneity exists amongst PDGFRb+ cells or within the identified subpopulations. Deeper sequencing, refined analyses, and further functional studies may unveil even more heterogeneity than appreciated.
It is noteworthy that Pdgfrb is expressed in a subset of WAT associated mesothelial cells and that Pdgfrb-expressing mesothelial cells express Pparg and some level of Zfp423. As described above, several lines of evidence point to a lineage relationship between embryonic mesothelial cells, APCs, and perivascular stromal cells within the visceral compartment (Chau et al., 2014;Rinkevich et al., 2012). PDGFRb+ MLCs did not exhibit robust adipogenic potential under the culture conditions utilized here, despite their expression of Pparg isoform two and Zfp423. Additional signals may be needed in order to drive adipocyte differentiation from these cells. Alternatively, this subpopulation of mesothelial cells may represent developmental intermediates between mesothelial cells and perivascular progenitors. Further insight into the functional significance of these various stromal subpopulations, their developmental origins, and their cellular plasticity will require additional genetic tools to manipulate individual populations selectively in vivo. Nevertheless, the molecular profiles obtained for FIPs and APCs from visceral WAT, along with the strategies to isolate these cells, will facilitate the study of physiological visceral WAT remodeling in vivo. Ultimately, unraveling the cellular and molecular determinants of WAT expansion and remodeling may lead to strategies to improve adipose tissue function and defend against metabolic disease.

Animals and diets
All animal experimens were performed according to procedures approved by the UTSW Animal Care and Use Committee. MuralChaser mice were derived from breeding Pdgfrb rtTA (JAX 028570), TRE-Cre (JAX 006234), and Rosa26R mT/mG (JAX 007676) mice, as previously described . Wildtype C57BL/6 mice mice were purchased from Charles River Laboratories and Zfp423 GFPB6 mice were described previously . Mice were maintained on a 12 hr light/dark cycle in a temperature-controlled environment (22˚C) and given free access to food and water. Mice were fed a standard rodent chow diet, doxycyline-containing chow diet (600 mg/kg doxycycline, Bio-Serv, S4107), or high-fat diet (60% kcal% fat, Research Diets, D12492i). For all experiments involving MuralChaser mice, 6 weeks-old mice were fed doxycyclinecontaining chow for 9 days, and then standard chow for additional 5 days before analysis (doxcycyline washout). For the high fat diet feeding experiments, mice were placed on the high fat diet beginning at 8 weeks of age.

Gene expression analysis by quantitative PCR
RNA was isolated from freshly sorted cells using RNAqueous-Micro Total RNA Isolation Kit (Thermo Fisher Scientific) or from cell cultures using Trizol, according to manufacturer's instructions. cDNA was synthesized using M-MLV Reverse Transcriptase (Invitrogen) and Random Primers (Invitrogen). Relative mRNA levels were determined by quantitative PCR using SYBR Green PCR Master Mix (Applied Biosystems). Values were normalized to Rps18 levels using the DD-Ct method. Unpaired Student's t-test was used to evaluate statistical significance. All primer sequences are listed within Table 4.

Isolation of adipose stromal vascular fraction (SVF) and flow cytometry
Adipose tissue was minced with scissors in a 1.5 mL tube containing 200 mL of digestion buffer (1X HBSS, 1.5% BSA, and 1 mg/mL collagenase D) and then transferred to a 50 mL Falcon tube containing 10 mL digestion buffer. The mixture was incubated in a 37˚C shaking water bath for 1 hr. The solution of digested tissue was passed through a 100 mm cell strainer, diluted to 30 mL with 2% FBS in PBS, and centrifuged at 500 x g for 5 min. The supernatant was aspirated and red blood cells in the SVF pellet were lysed by brief incubation in 1 mL RBC lysis buffer (Sigma). Next, the mixture was diluted to 10 mL with 2% FBS in PBS, passed through a 40 mm cell strainer, and then centrifuged at 500 x g for 5 min. The supernatant was aspirated, and cells were resuspended in blocking buffer (2% FBS/PBS containing anti-mouse CD16/CD32 Fc Block (1:200)). Primary antibodies were added to the cells in blocking buffer for 15 min at 4˚C in the dark. After incubation, the cells were washed once with 2% FBS/PBS and then resuspended in 2% FBS/PBS for sorting. Cells were sorted for collection using a BD Biosciences FACSAria cytometer or analyzed using a BD Biosciences LSR II cytometer (UTSW Flow Cytometry Core Facility). Flow cytometry plots were generated with FlowJo (V10).

BrdU assays
Eight week-old mice were administered 0.8 mg/mL BrdU in drinking water (replaced fresh every 2 days) and placed on chow or high fat diet for 1 week. At the end of the treatment period, adipose tissue SVF was isolated as described above and stained with the following antibodies: CD31, CD45, PDGFRb, LY6C, and CD9. Anti-BrdU staining of fixed cells was then conducted using the BrdU Flow Kit (BD Biosciences 559619), according to the manufacturer's protocol.

Single-cell RNA-sequencing and analysis
Six-week-old male MuralChaser mice were fed doxycycline-containing chow diet for 9 days, followed by standard chow diet for 5 days. Following the 5 day washout period, gonadal WAT was isolated and digested as described above. tdTomato-mGFP+ cells were collected by FACS. Single cell library preparation was performed using the 10X Genomics Single Cell 3' v2 according to the manufacturer's instructions. After FACS isolation of gonadal WAT tdTomato-mGFP+ cells from MuralChaser mice, 10,000 cells were partitioned into droplets containing a barcoded bead, a single cell, and reverse transcription enzyme mix using the GemCode instrument. This was followed by cDNA amplification, fragmentation, end repair and A-tailing, adaptor ligation, and index PCR. Cleanup and size selection were performed using Dynabeads MyOne Silane beads (Thermo Fisher Scientific) and SPRIselect Reagent beads (Beckman Coulter). Libraries were sequenced on an Illumina NextSeq 500 High Output (400M) by the UT Southwestern McDermott Center Next Generation Sequencing Core. 75 pairedend reads were obtained using one flow cell with the following length input: 26 bp Read 1, 66 bp Read 2, 0 bp Index 1, and 0 bp Index 2.
Cell Ranger software (v2.1.0) was used to perform demultiplexing, aligning reads, filtering, clustering, and gene expression analyses, using default parameters. We recovered 1378 cells with a median UMI count of 10,879 per cell, a mean reads per cell of 277,212, and a median genes per cell of 3278. In order to ensure that our analysis was restricted to genetically marked Pdgfrb-expressing cells, we filtered the cells based on expression of tdTomato (<0) and GFP (>0) to only include cells in the final analysis that were devoid of tdTomato mRNA and expressed GFP transcript. After this screening, we obtained a total of 1,045 cells for the analysis shown in Figure 1. The Cell Ranger data was imported into Loupe Cell Browser Software (v1.0.5) for t-distributed stochastic neighbor embedding (tSNE) based clustering, heatmap generation, and gene expression distribution plots. The Cell Ranger files were imported into R Studio (v3.3.2) and the Seurat (v2.1.0) and Readr (v1.1.0) packages were used to generate gene cluster text (GCT) and categorical class (CLS) files, using the clustering generated from Cell Ranger (k-means = 4 for the analysis in Figure 1 and k-means = 3 for the analysis in Figure 2). The GCT and CLS files were input into Gene Set Enrichment Analysis (GSEA) (v3.0) using the Java GSEA implementation with default parameters. The single cell RNAsequencing experiment was repeated using cells isolated from pooled gonadal WAT from five additional MuralChaser mice to validate the identification of APCs, FIPs, and MLCs (Figure 1-figure supplement 1). The raw sequencing data from Figures 1 and 2 has been deposited to Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE111588).

Oil red O staining
Cells were fixed in 4% PFA for 15 min at room temperature and washed twice with water. Cells were incubated in Oil Red O working solution (2 g Oil red O in 60% isopropanol) for 10 min to stain accumulated lipids. Cells were then washed three times with water and bright field images were acquired using a document scanner or with a Keyence BZ-X710 Microscope.

Mesothelial cell isolation
Adipose associated mesothelial cells were isolated as described previously (Darimont et al., 2008). Epididymal adipose depots were harvested from 6 weeks-old male MuralChaser mice treated with doxycyline as described above. Intact whole adipose depots were placed in 10 mL of PBS containing 0.25% trypsin for 20 min at 37˚C with continuous end over end rotation. Next adipose tissues were removed, and remaining solution containing isolated cells was centrifuged at 600 x g for 5 min. The media was aspirated, and the cell pellet was resuspended in growth media (10% FBS in DMEM/F12 (Invitrogen)) and plated in a 12-well collagen-coated plate. The cells were incubated at 37˚C in 10 CO 2 and the media was replaced daily. Images were obtained using a Leica DMIL LED microscope and a Leica DFC3000g camera.

Cellular assays
To assess in vitro proliferation, sorted cells (APCs and FIPs) were plated at a density of 5 x 10 3 cells/ well in a 48-well plate containing 2% FBS in ITS Media. Cell numbers were assessed in parallel wells every 2 days by cell counting with a hemocytometer. To study the impact of conditioned media on adipogenesis, media from equally confluent cultures of APCs, FIPs, and MLCs was harvested and placed onto APCs beginning 48 hr after culture in a 1:1 ratio with 2% FBS in ITS media. Cells were harvested at the indicated time points for RNA expression analysis. Images were obtained using a Leica DMIL LED microscope and a Leica DFC3000g camera. Bone marrow derived macrophages (BMDMs) were derived from bone-marrow stem cells (BMSCs) isolated from the femurs and tibias of male mice, as previously described (Shan et al., 2017). BMSCs were maintained in differentiation medium derived from L929 cells for 7 days to allow for macrophage differentiation. In parallel, adipose tissue SVF was isolated and PDGFRb+ subpopulations were sorted as described above. Sorted cells (APCs, FIPs, and MLCs) were plated in a 48-well plate at 4 Â 10 4 cells/well in 2% FBS in ITS media. 48 hr later, the adipose-derived cells were treated with vehicle (PBS) or LPS (100 ng/ml) for 3 hr. Next the adipose-derived cells were washed with PBS and fresh media was added. 24 hr later, the conditioned media was harvested and placed on the BMDMs (at day 7) in a 1:1 ratio with 2% FBS in ITS media. After a 3 hr incubation, the BMDMs were harvested for RNA analysis.
For TGFb treatments, sorted cells were plated in 48-well plates at 2 Â 10 4 cells/well in 2% FBS in ITS media. 24 hr later, vehicle (PBS) or 1 ng/mL recombinant TGFb was added to the media for 3 days prior to harvest. The media was replaced daily under this period. For the LPS and TNFa treatments, cells were plated in 48-well plates at 4 Â 10 4 cells/well in 2% FBS in ITS media. 48 hr later, the cells were treated with vehicle (PBS), LPS (100 ng/ml), or TNFa (20 ng/ml). After 3 hr of treatment, cells were harvested for RNA isolation.
Cell transplantation assays 80,000 cells (APCs and FIPs) collected by FACS were suspended in 100 mL transplantation media (50% Matrigel in PBS, supplemented with 2 ng/mL FGF) and injected subcutaneously into the remnant inguinal WAT region of 3 month old lipodystrophic mice (Adiponectin-Cre; Pparg loxP/loxP ). Three weeks later, the remnant inguinal WAT depots were harvested for histology.

Retroviral production and transduction of primary cells
The pMSCV-Nr4a1, pMSCV-Nr4a2, and pMSCV-Nr4a3 plasmids were previously reported (kind gift from Dr. P. Tontonoz) (Chao et al., 2008). Retroviral production and packaging in phoenix cells was performed as previously described . Briefly, 10 mg of the pMSCV overexpression plasmids were co-transfected with 5 mg gag-pol and 5 mg VSV-g plasmids into phoenix packaging cells using Lipofectamine LTX (Thermo Fisher Scientific), according to the manufacturer's protocol. APCs and FIPs were transduced with diluted virus (1:10 ratio) in 2% FBS/ITS media containing 8 mg/ ml polybrene (Sigma). Following 16 hr of incubation with indicated viruses, cells were returned to 2% FBS/ITS media and assayed for TNFa responsiveness as indicated.

Statistical analysis
All data were expressed as the mean +SEM. We used GraphPad Prism 7.0 (GraphPad Software, Inc., La Jolla, CA, USA) to perform the statistical analyses. For comparisons between two independent groups, a Student's t-test was used and p<0.05 was considered statistically significant. For in vitro studies, we estimated the approximate effect size based on independent preliminary studies. Studies designed to characterize an in vitro difference in gene expression were estimated to have a slightly larger effect size of 30% with assumed 15% standard deviation of group means. To detect this difference at a power of 80% and an alpha of 0.05, we predicted we would need four independent replicates per group. We estimated this effect size based on independent preliminary studies. Statistical information, including p values, samples sizes, and repetitions, for all datasets are provided in Supplementary file 1.