FSP1-positive fibroblasts are adipogenic niche and regulate adipose homeostasis

Adipocyte progenitors reside in the stromal vascular fraction (SVF) of adipose tissues that are composed of fibroblasts, immune cells, and endothelial cells. It remains to be elucidated how the SVF regulates adipocyte progenitor fate determination and adipose homeostasis. Here, we report that fibroblast-specific protein-1 (FSP1)+ fibroblasts in the SVF are essential to adipose homeostasis. FSP1+ fibroblasts, devoid of adipogenic potential, are adjacent to the preadipocytes in the SVF. Ablation of FSP1+ fibroblasts in mice severely diminishes fat content of adipose depots. Activation of canonical Wnt signaling in the FSP1+ fibroblasts results in gradual loss of adipose tissues and resistance to diet-induced obesity. Alterations in the FSP1+ fibroblasts reduce platelet-derived growth factor (PDGF)-BB signaling and result in the loss of preadipocytes. Reduced PDGF-BB signaling, meanwhile, impairs the adipogenic differentiation capability of preadipocytes by regulating matrix metalloproteinase (MMP) expression, extracellular matrix remodeling, and the activation of Yes-associated protein (YAP) signaling. Thus, FSP1+ fibroblasts are an important niche essential to the maintenance of the preadipocyte pool and its adipogenic potential in adipose homeostasis.

Adipogenic potential of preadipocytes is determined not only by their intrinsic properties but also by the surrounding microenvironment. Diet-induced obesity is regulated by the adipose microenvironment, but not by cell-intrinsic mechanisms, highlighting the importance of microenvironmental regulation in adipose homeostasis [10]. Adipose tissues contain multiple types of stromal cells, including fibroblasts, endothelial cells, and immune cells. While macrophage-related chronic inflammation in the adipose tissues was reported to play a crucial role in the development of obesity and obesity-related insulin resistance [11,12], roles of other cell types in preadipocyte fate determination and adipose homeostasis are largely unknown. Fibroblasts are resident cell types in the adipose tissues with a mesenchymal lineage origin. Fibroblast-specific protein 1 (FSP1; also known as S100A4) is a reliable marker for detecting tissue-resident fibroblasts, in addition to other mesenchymal markers αSMA, PDGFR-β, and NG2 [13][14][15]. While the roles of fibroblasts in adipose homeostasis remain elusive, fibroblasts have multifaceted regulatory roles in tissue morphogenesis, wound healing, fibrosis, and cancer by modulating the behavior and functions of epithelial cells and stem cells [13,[15][16][17].
Here, we report that FSP1 + fibroblasts in the SVF are a niche for adipogenesis. Activation of canonical Wnt signaling in the FSP1 + fibroblasts or ablation of FSP1 + fibroblasts disturbs adipose homeostasis and results in loss of adiposity. Alterations in the FSP1 + fibroblasts resulted in decreased platelet-derived growth factor (PDGF) expression. On the one hand, PDGF maintains the preadipocyte number. On the other hand, PDGF regulates the adipogenic potential of preadipocytes by regulating extracellular matrix remodeling in the microenvironment and Yes-associated protein (YAP) activation. Thus, FSP1 + fibroblasts are a niche for preadipocytes orchestrating the functions of preadipocytes and their microenvironment.

Activation of canonical Wnt signaling in FSP1 + fibroblasts results in loss of adiposity
Wnt regulates adipose homeostasis by activating β-catenin in preadipocytes to inhibit their differentiation [8,9]. In addition to its direct inhibitory effect on preadipocyte differentiation [8], Wnt may regulate adipose homeostasis through cell types in the microenvironment, e.g., fibroblasts. Expression of β-catenin target genes were altered in the FSP1 + /tdTomato + fibroblasts upon HFD-induced obesity (S2A Fig). To investigate whether activation of Wnt signaling in FSP1 + fibroblasts would affect adipose homeostasis, Ctnnb1 exon 3 fl/+ mice were crossed with Fsp1-Cre mice to generate the Fsp1-Cre;Ctnnb1 exon 3 fl/+ (F-BCA) compound mice. Fsp1-Credriven deletion of exon 3 in β-catenin produced a smaller-molecular-weight β-catenin protein in tail tip fibroblasts (Fig 2A and S2B Fig). Despite the fact that the activated form of β-catenin was present in the SVF (Fig 2A and S2B-S2D Fig), there was no detectable activated form of βcatenin in the WAT or adipocytes from the F-BCA mice (Fig 2A and S2B Fig). These data further suggested that FSP1 + fibroblasts were not directly involved in the adipocyte lineage ( Fig  1). In line with this notion, tdTomato + /FSP1 + SVF cells were negative for preadipocyte markers in F-BCA mice (S2E Fig).
Fsp1-Cre-driven activation of canonical Wnt signaling in fibroblasts did not affect embryonic development. F-BCA mice were born at expected Mendelian ratio. Mice with an Fsp1-Cre or Ctnnb1 exon 3 fl/+ genotype were phenotypically similar to the wild-type (WT) littermates throughout the lifespan. Mice with a WT, Fsp1-Cre, or Ctnnb1 exon 3 fl/+ genotype were therefore classified as controls in comparison to their F-BCA littermates. At weaning, F-BCA male mice were phenotypically normal, with similar amounts of adipose depots to the control mice (S3A- S3D Fig). Dissected I-WATs and E-WATs were similar in size to those from the control mice (S3C Fig). However, F-BCA male mice gain less weight after puberty compared with the control mice (Fig 2B). At 4 months of age, F-BCA male mice had less fat compared with the control mice (Fig 2C-2F). Dissected I-WATs and E-WATs were significantly smaller in size and weighed less in the F-BCA mice (Fig 2E and 2F). Histology inspection on adipose tissue sections showed that the F-BCA adipocyte diameter was significantly smaller (Fig 2G). F-BCA male mice at 8 months of age were virtually free of subcutaneous or visceral adipose depots   on ND at 4 months of age. n = 10 for male Ctrl mice, and n = 5 for male F-BCA mice. (K) Quantification of the AUC of the ITT relative to Ctrl group. Data are presented as mean ± SEM. Statistical analyses were performed with two-tailed unpaired student t test (panel D, F, I, and K) or two-way ANOVA followed by Bonferroni's multiple comparison test (panel B, H, and J). Ã p < 0.05; ÃÃ p < 0.01; ÃÃÃ p < 0.001. Underlying data can be found in S1 Data. AUC, area under the curve; Ctrl, control; F-BCA, Fsp1-Cre;Ctnnb1 exon 3 fl/+ ; FSP1, fibroblast-specific protein-1; GTT, glucose tolerance test; HE, hematoxylin-eosin; ITT, insulin tolerance test; ND, normal-chow diet; NMR, nuclear magnetic resonance; NS, not significant; SVF, stromal vascular fraction; TTF, tail tip fibroblast; WAT, white adipose tissue; WT, wild-type. WAT is an important metabolic regulator. Adipose defects observed in the F-BCA mice mimic lipodystrophy, a disorder accompanied by metabolic disturbances, including hyperglycemia, insulin resistance, and ectopic lipid deposition in secondary organs [8]. F-BCA male mice did not show the metabolic abnormality observed in the classical lipodystrophy mouse models. Rather activation of β-catenin in the FSP1 + fibroblasts offered metabolic benefits in the mice (Fig 2H-2K). F-BCA male mice were able to more efficiently clear glucose (Fig 2H  and 2I) but retained the same insulin responsiveness as the control mice (Fig 2J and 2K). Livers weighed less, and no ectopic lipid deposition was observed in the livers in the F-BCA male mice (S4I- S4K Fig).

F-BCA mice are resistant to diet-induced obesity
Activation of preadipocytes drives adipocyte hyperplasia in diet-induced obesity [10]. Such a process is regulated by the adipose microenvironment but not by cell-intrinsic mechanisms [10]. To investigate whether activation of canonical Wnt signaling in the FSP1 + fibroblasts would affect diet-induced obesity, F-BCA male mice were fed with an HFD for 12 weeks. On HFD, F-BCA male mice consumed more food and had significantly enhanced respiratory exchange ratio and physical activity compared with the control mice (S4E- S4H Fig). Unlike the control mice, which accumulated a significant amount of fat upon HFD feeding (Fig 3A-3F), F-BCA male mice were resistant to HFD-induced obesity (Fig 3A-3F). Control mice fed with an HFD were glucose intolerant and insulin resistant (Fig 3G-3J). F-BCA mice on an HFD cleared glucose as efficiently as those on an ND (Fig 3G and 3H). F-BCA mice on an HFD were more sensitive to insulin (Fig 3I and 3J). Although control mice accumulated a large amount of fat in the liver upon HFD feeding, F-BCA mice were protected from HFDinduced steatosis (S4J and S4K Fig).

FSP1 + fibroblasts maintain preadipocyte pool via PDGF signaling
Activation of canonical Wnt signaling in the FSP1 + fibroblasts disturbed adipose homeostasis. Gene set enrichment analysis (GSEA) on gene expression of SVF cells indicated impaired adipogenic potential of the SVF cells isolated from the F-BCA mice ( Fig 4A). Indeed, F-BCA SVF cells less efficiently differentiated into adipocytes upon adipogenic induction compared with the control SVF cells (Fig 4B-4E). Inefficient differentiation of F-BCA SVF cells may result from a reduced number of preadipocytes or from impaired differentiation potential of the preadipocytes. F-BCA SVF cells had lower expression levels of preadipocyte markers than the control SVF cells (Fig 4F). FACS analyses revealed reduced numbers of CD34 + Sca1 + preadipocytes in the F-BCA SVF (Fig 4G and 4H).
Conditioned medium from control SVF cells promoted F-BCA SVF cell differentiation ( Fig 5A and 5B), indicating that secreted factors regulated preadipocyte differentiation. Expression of PDGFB was drastically decreased in the F-BCA SVF cells compared with that in the control SVF cells (Fig 5C). Immunohistochemical staining on the control and F-BCA WAT sections indicated reduced PDGFR-β phosphorylation in the F-BCA WATs ( Fig 5D). Preadipocytes express PDGF receptors [2,5,6]. We next studied whether reduced PDGF expression is responsible for the loss of preadipocytes in the F-BCA SVFs. Treatment of the F-BCA SVF cells with PDGF-BB significantly increased the number of preadipocytes and restored the adipogenic potential (Fig 5E-5H). To study whether PDGF-BB is responsible for preadipocyte pool maintenance in vivo, growth factor-reduced Matrigel supplemented with PDGF-BB was implanted into the I-WATs of the F-BCA mice. Matrigel alone minimally affected the number and the adipogenic potential of the F-BCA SVF cells (Fig 5I-5K), whereas PDGF-BB containing Matrigel in the contralateral I-WAT significantly increased the percentage of preadipocytes and restored the adipogenic potential of the F-BCA SVF cells (Fig 5I-5K). Data are presented as mean ± SEM. Statistical analyses were performed with two-tailed unpaired student t test (panel C, E, H, and J) or two-way ANOVA followed by Bonferroni's multiple comparison test (panel A, G, and I). Ã p < 0.05; ÃÃ p < 0.01; ÃÃÃ p < 0.001. Underlying data can be found in S1 Data. AUC, area under the curve; Ctrl, control; E-WAT, epididymal white adipose tissue; F-BCA, Fsp1-Cre;Ctnnb1 exon 3 fl/+ ; FSP1, fibroblast-specific protein-1; GTT, glucose tolerance test; HE, hematoxylin-eosin; HFD, high-fat diet; ITT, insulin tolerance test; I-WAT, inguinal white adipose tissue; NMR, nuclear magnetic resonance; NS, not significant; WAT, white adipose tissue.

FSP1 + fibroblasts regulate adipogenesis through extracellular matrix remodeling and YAP activation
Activation of Wnt signaling in fibroblasts promotes tissue fibrosis [18][19][20]. Sirius red staining revealed significantly more collagen deposition in the F-BCA adipose tissues, particularly around the SVF zones ( Fig 6A). More abundant type I collagen (Col I) expression was detected in the F-BCA SVF cells (Fig 6B), although the mRNA levels were unchanged upon activation of canonical Wnt signaling in the FSP1 + fibroblasts (Fig 6C). The matrix metalloproteinase (MMP) family is primarily responsible for the degradation and remodeling of extracellular matrix [6,21]. Proteolytic activity of MMPs is regulated by the tissue inhibitors of metalloproteinase (TIMPs) [6,21]. MMPs and TIMPs are differentially expressed in the adipose tissues during obesity [22] and modulate adipocyte differentiation [22][23][24][25][26]. Inhibition of MMP activity diminished the adipogenesis capability of preadipocytes [22]. F-BCA SVF cells had much reduced MMP expression and up-regulated TIMP-3 expression (Fig 6D and 6E). Gelatin zymography revealed significantly lower MMP expression in the F-BCA cells (Fig 6F). PDGF-BB treatment increased MMP expression and decreased TIMP3 expression (Fig 6G,   Despite the fact that the mRNA levels of collagen were not changed, Col I protein levels decreased upon PDGF-BB treatment (Fig 6H and S6C Fig). Excess collagen deposition and a stiff microenvironment inhibit preadipocyte differentiation [25][26][27][28]. Hippo pathway transcription factors YAP and TAZ are the major effectors sensing the mechanical signals exerted by extracellular matrix (ECM) physical property [28][29][30][31][32] that inhibits adipogenesis of mesenchymal stem cells [26,28,33]. Indeed, YAP protein accumulated in the F-BCA SVF cells (Fig 6B). The YAP signature was significantly enriched, and expression of YAP target genes CTGF and ANKRD1 was significantly up-regulated in the F-BCA SVF cells (Fig 6I and  6J). PDGF-BB treatment reduced the expression of YAP and its target genes (Fig 6H and 6K).
To study whether YAP activation is responsible for the impaired adipogenic potential of F-BCA SVF cells, F-BCA SVF cells were pretreated with YAP pharmacological inhibitor verteporfin (VP) [34]. Inhibition of YAP signaling with VP restored the adipogenic potential of the F-BCA SVF cells, without affected Col I and MMP expression (Fig 6L-6N

Ablation of FSP1 + fibroblasts disturbs adipose homeostasis
To investigate the physiological requirement of the FSP1 + fibroblasts in the adipose development, FSP1 + fibroblasts were ablated by generating the Fsp1-Cre;Rosa26-DTA (F-DTA) compound mice (Fig 7A). At 4 months of age, F-DTA mice were significantly lean, with markedly reduced fat compared with the control mice (WT, Fsp1-Cre, or Rosa26-DTA) (Fig 7B and 7C,  S7A and S7B Fig). Dissected WATs were smaller in size and weighed less in the F-DTA mice (Fig 7D and 7E). F-DTA male mice had significantly more food uptake, similar respiratory exchange ratio and physical activity, and slightly reduced energy expenditure compared with the control mice (S8A- S8D Fig), whereas F-DTA female mice had comparable food uptake, respiratory exchange ratio and physical activity, and slightly increased energy expenditure (S7C- S7F Fig). F-DTA mice more efficiently cleared glucose, but retained the same responsiveness to insulin, similar to that of the F-BCA mice (S8E- S8H Fig). No ectopic lipid deposition was observed in the livers of the F-DTA mice (S8I- S8K Fig). F-DTA SVF cells were defective in adipogenesis (Fig 7F-7H). F-DTA SVF cells contained much fewer preadipocytes (Fig 7I). PDGFB expression decreased in the F-DTA SVF cells (Fig 7J). PDGF-BB treatment increased the number of preadipocytes and restored the adipogenic potential of F-DTA SVF cells (Fig 7K-7M). Similar to the F-BCA SVF cells, F-DTA SVF cells had much reduced MMP expression, up-regulated TIMP-1 and TIMP-3 expression, and YAP activation (S9 Fig, Fig 7N  and 7O). Inhibition of YAP signaling restored the adipogenic potential of the F-DTA SVF cells (Fig 7P and 7Q).

Discussion
Adipocytes continuously turn over in adults. Like other adult stem cells and progenitor cells, preadipocytes are resident in a highly specialized niche. In this report, we identified FSP1 + fibroblasts as the niche for preadipocytes. FSP1 + fibroblasts are crucial to the maintenance of adipose homeostasis. Alteration in FSP1 + fibroblasts disturbs adipose homeostasis and results in loss of adiposity.  student t test or two-way ANOVA followed by Bonferroni's multiple comparison test (panel B). Ã p < 0.05; ÃÃ p < 0.01; ÃÃÃ p < 0.001. Underlying data can be found in S1 Data. CD34, cluster of differentiation 34; Ctrl, Preadipocytes rapidly expand from the preexisting pool during the first postnatal month [3]. During adulthood, preadipocytes and adipocytes are constantly renewed [35,36]. Turnover of preadipocytes and adipocytes is observed in both humans and mice [35,36]. Obese mice have increased adipocyte formation [35,36]. Adipogenic niches, including macrophage-related chronic inflammation [11,12], may have predominant roles in regulating the turnover of preadipocytes and adipocytes during adulthood and obesity. Numbers of FSP1 + fibroblasts increase during obesity. FSP1 + fibroblasts regulate adipose homeostasis by maintaining the preadipocyte pool and its differentiation potential. Alterations in FSP1 + fibroblasts result in diminished adipose depots at adulthood but not at puberty. Therefore, FSP1 + fibroblasts may represent a class of microenvironmental cues in regulating the turnover of preadipocytes and adipocytes and adipose homeostasis during adulthood and obesity. FSP1 is broadly expressed in mesenchymal cells. It should be noted that fibroblasts are highly heterogeneous. The heterogeneity of fibroblasts are reflected not only by the marker expression but also their biological functions. Subtypes of fibroblasts, similar to those of macrophages, were proposed [13]. Obesity induces macrophage polarization in adipose tissues [12]. It warrants further investigation into whether FSP1 + fibroblasts undergo similar polarization in obesity and whether subpopulations of FSP1 + fibroblasts regulate preadipocyte renewal and maintenance of adipose homeostasis.
Conditioned medium from the control SVF cells can restore the differentiation potential of the F-BCA SVF cells, suggesting that FSP1 + fibroblasts regulate preadipocyte renewal and adipose homeostasis via soluble factors. Alterations in FSP1 + fibroblasts greatly reduce PDGF-BB expression. PDGF-BB can restore preadipocyte numbers both in vitro and in vivo. In addition to the maintenance of the preadipocyte population, PDGF-BB also regulates adipogenic differentiation capability of the preadipocytes. Preadipocytes express PDGF receptors [2,5,6]. PDGF receptor expression in preadipocytes not only reflects the mesenchymal origin of the preadipocytes but may also play a central role in adipose homeostasis by maintaining both the preadipocyte pool and its adipogenic potential. The regulation of maintenance of the preadipocyte population and adipogenic potential by PDGF-BB may be exerted through distinct signaling pathways. Alterations in FSP1 + fibroblasts disturb extracellular matrix remodeling and YAP signaling in the adipose tissue microenvironment. PDGF-BB treatment restores extracellular matrix remodeling and alleviates YAP activation. However, inhibition of YAP signaling restores the adipogenic capability of the SVF cells but does not affect the number of preadipocytes.

FSP1-positive cells as adipogenic niche
canonical Wnt signaling in the FSP1 + fibroblasts results in loss of adiposity and beneficial metabolism, similar to that observed in the mice with canonical Wnt signaling activation in the preadipocytes. Thus, Wnt ligands secreted by the preadipocytes may regulate adipose homeostasis via both autocrine and paracrine mechanisms orchestrating the functions of preadipocytes and their microenvironment.
In summary, FSP1 + fibroblasts are essential to the maintenance of the pool and differentiation potential of preadipocytes via PDGF signaling, extracellular matrix remodeling, and YAP activation. FSP1 + fibroblasts are a niche for preadipocytes orchestrating the functions of preadipocytes and their microenvironment.

Ethics statement
All mice were housed in a specific pathogen-free environment at the Shanghai Institute of Biochemistry and Cell Biology and treated in strict accordance with protocols approved by the Institutional Animal Care and Use Committee of Shanghai Institute of Biochemistry and Cell Biology (approval number: NAF-15-003-S325-006).

Mice
Fsp1-Cre mice [15] were from the Jackson Laboratory. β-cat exon3 flox/+ mice were a generous gift from Professor Lijian Hui (Shanghai Institute of Biochemistry and Cell Biology). Fsp1-Cre and β-cat exon3 flox/+ mice were backcrossed to the FVB background for >9 generations. Rosa26-tdTomato, Rosa26-mTmG, and Rosa26-DTA mice at C57Bl/6 and 129 mixed background were generous gifts from Professor Yi Zeng (Shanghai Institute of Biochemistry and Cell Biology). For diet-induced obesity studies, starting at 4 weeks of age, mice were fed with HFDs containing 60% kcal from fat for 12 weeks (Research Diets, New Brunswick, New Jersey).

Glucose tolerance tests and insulin tolerance tests
Glucose tolerance tests (GTTs) and insulin tolerance tests (ITTs) were performed as described [37]. For GTTs, mice were fasted overnight and received an intraperitoneal injection of 2 g/kg body weight glucose (Sigma-Aldrich, St. Louis, Missouri). For ITTs, mice were injected intraperitoneally with 0.5 U/kg body weight insulin (Biosynthetic Human Insulin, 100 U/mL; Novo Nordisk, Bagsvaerd, Denmark) after a 4-hour fast. Tail blood glucose levels were measured at 0, 15, 30, 60, 90, and 120 minutes after challenge using the Onetouch Ultra blood glucose monitoring system (LifeScan, Shanghai, China). AUC was calculated using GraphPad Prism.

Metabolic measurements
Metabolic measurements with indirect calorimetry were performed on mice as described [38]. Animals were maintained on a ND or an HFD at ambient temperature under 12-hour light and dark cycles. Mice were acclimated in a comprehensive lab animal monitoring system (Columbus Instruments, Columbus, Ohio) for 1 day before recording over 2 days following the manufacturer's instruction.

SVF cell isolation and culture
I-WAT and E-WAT were washed with PBS, cut into small pieces, and digested with 1 mg/mL Collagenase (Worthington, Lakewood, New Jersey) at 37˚C for 60 minutes. SVF cell culture, FACS analyses, and adipogenic induction were performed as described [40,41]. For PDGF and VP treatment experiments, SVF cells were treated with PDGF or VP for 4 days before FACS analyses and adipogenic induction. To isolate floated adipocytes, collagenase-treated adipose depot mixture was centrifuged at 200 g for 1 minute. Floating cell layer was collected as adipocytes.

Western blot analysis
Western blot analysis was performed as previously described [39].

Quantitative RT-PCR
Total RNA was prepared from SVF cells using Trizol reagents (Invitrogen) as previously described [39]. Equal amounts of RNA were subjected to quantitative RT-PCR using SYBR green with the BIO-RAD Q-PCR Systems according to the manufacturer's protocol. Relative expression levels were calculated using the comparative CT method. Gene expression levels were normalized to Actin. The primers used are listed in S1 Table. RNA-Seq, GO analysis, and GSEA SVF cells were isolated from pooled WATs from control, F-BCA, and F-DTA mice. Total RNA was extracted and purified with TRIzol. Two biological replicates were subjected to complementary DNA library preparation and sequencing according to the Illumina standard protocol. RNA-seq reads were mapped to mm9 reference genome using TopHat2. Mapped reads of the 2 replicates in each group were merged together for further analysis. Expression for each known gene from RefSeq was determined by covered reads and normalized with RPKM. Normalized gene expression of the SVF cells are listed in S2 Table. Differentially expressed genes (DEGs) were identified by gene expression comparison between control and F-BCA or F-DTA SVF cells and were defined with parameters including p < 0.01 (Wald test), fold change ! 1.5 or 0.667, and expression level ! 5 RPKM in at least one sample. Gene ontology (GO) analysis was performed with the DAVID online tool. Top GO categories were selected according to the p-value after Benjamini-Hochberg correction. GSEA was performed on gene signatures obtained from the MSigDB database version 5.0 (March 2015 release) [42]. Statistical significance was assessed by comparing the enrichment score to enrichment results generated from 1,000 random permutations of the gene set to obtain p-values (nominal p-value).

Statistics
Sample sizes for each figure are denoted in the figure legends. Data are representative of at least 3 biologically independent experiments. For animal experiments, the sample size is determined on the basis of our prior knowledge of the variability of experimental output. Age of animals was matched. The experiments were not randomized, and the investigators were not blinded to allocation during experiments and outcome assessment. Statistical significance between conditions was assessed by unpaired or paired two-tailed student t tests or ANOVAs followed by Bonferroni's multiple comparison test. All error bars represent SEM, and significance between conditions is denoted Ã p < 0.05; ÃÃ p < 0.01; and Ã p < 0.001. "NS" denotes not significant. The numerical data used in all figures are included in S1 Data. , and XTOT (panel F) were recorded. n = 7 for female control mice, and n = 5 for female F-BCA mice. Data are presented as mean ± SEM. Statistical analyses were performed with two-tailed unpaired student t test or two-way ANOVA followed by Bonferroni's multiple comparison test (panel A). Ã p < 0.05; ÃÃ p < 0.01. Underlying data can be found in S1 Data. Ctrl, control; EE, energy expenditure; (K) Hepatic triglyceride levels in 4-month-old male control and F-DTA mice. n = 5 for each group. Data are presented as mean ± SEM. Statistical analyses were performed with two-tailed unpaired student t test or two-way ANOVA followed by Bonferroni's multiple comparison test (panel E and G). Ã p < 0.05; ÃÃ p < 0.01; ÃÃÃ p < 0.001. Underlying data can be found in S1 Data. AUC, area under the curve; EE, energy expenditure; F-DTA, Fsp1-Cre;Rosa26-DTA; FSP1, fibroblast-specific protein-1; GTT, glucose tolerance test; HE, hematoxylin-eosin; ITT, insulin tolerance test; NS, not significant; RER, respiratory exchange ratio; XTOT, physical activity. (E) Gelatin zymography of conditioned medium of F-DTA treated with or without 10 ng/mL PDGF-BB or 0.5 μg/mL VP. Data are presented as mean ± SEM. Statistical analyses were performed with two-tailed unpaired student t test. Ã p < 0.05; ÃÃ p < 0.01; ÃÃÃ p < 0.001. Underlying data can be found in S1 Data. F-DTA, Fsp1-Cre;Rosa26-DTA; FSP1, fibroblast-specific protein-1; MMP, matrix metalloproteinase; NS, not significant; PDGF, platelet-derived growth factor; RT-PCR, reverse transcription PCR; SVF, stromal vascular fraction; TIMP, tissue inhibitor of metalloproteinase; YAP, Yes-associated protein.