TNFα Mediates Inflammation-Induced Effects on PPARG Splicing in Adipose Tissue and Mesenchymal Precursor Cells

Low-grade chronic inflammation and reduced differentiation capacity are hallmarks of hypertrophic adipose tissue (AT) and key contributors of insulin resistance. We identified PPARGΔ5 as a dominant-negative splicing isoform overexpressed in the AT of obese/diabetic patients able to impair adipocyte differentiation and PPARγ activity in hypertrophic adipocytes. Herein, we investigate the impact of macrophage-secreted pro-inflammatory factors on PPARG splicing, focusing on PPARGΔ5. We report that the epididymal AT of LPS-treated mice displays increased PpargΔ5/cPparg ratio and reduced expression of Pparg-regulated genes. Interestingly, pro-inflammatory factors secreted from murine and human pro-inflammatory macrophages enhance the PPARGΔ5/cPPARG ratio in exposed adipogenic precursors. TNFα is identified herein as factor able to alter PPARG splicing—increasing PPARGΔ5/cPPARG ratio—through PI3K/Akt signaling and SRp40 splicing factor. In line with in vitro data, TNFA expression is higher in the SAT of obese (vs. lean) patients and positively correlates with PPARGΔ5 levels. In conclusion, our results indicate that inflammatory factors secreted by metabolically-activated macrophages are potent stimuli that modulate the expression and splicing of PPARG. The resulting imbalance between canonical and dominant negative isoforms may crucially contribute to impair PPARγ activity in hypertrophic AT, exacerbating the defective adipogenic capacity of precursor cells.

Hence, in this work, we investigated whether the balance between canonical and dominant negative PPARγ isoforms is affected by pro-inflammatory factors and if such alteration contributes to the impaired neo-adipogenesis and insulin-resistance in hypertrophic AT. Herein, we demonstrate that MΦ-secreted pro-inflammatory factors affect PPARG splicing, increasing the PPARG∆5/cPPARG ratio. Notably, we report that TNFα induces the activation of the serine/arginine-rich protein 40 (SRp40) and PI3K/Akt signaling, representing the most powerful inflammatory stimulus able to modulate PPARG expression and splicing. Moreover, TNFA and PPARG∆5 mRNA levels show a positive correlation in the SAT of obese patients, further supporting PPARG splicing alteration as a feature of inflamed AT. Our work reveals a previously unrecognized role of macrophage-secreted factors on PPARG regulation, indicating the inflammation-mediated alteration of PPARG splicing in adipogenic precursors as a new potential contributor to the defective neo-adipogenesis of hypertrophic adipose tissue.

Human Samples
Adipose tissue samples were obtained from donors to the Leipzig Obesity Bio Bank (n = 55; mean age = 57.6 ± 16.7 years; mean BMI = 32.3 ± 9.9 kg/m 2 ) undergoing bariatric surgery [31,32] and available through previously reported collaborations [29,30]. The study has been reviewed and approved by the ethics committee of Leipzig University, Germany (approval numbers: 159-12-21052012 and 017-12-23012012), and carried out in accordance with the Declaration of Helsinki, the Bioethics Convention (Oviedo) and the EU Directive on Clinical Trials (Directive 2001/20/EC). Patients were randomly selected from the human tissue biopsy bank held by Prof. Bluher, with more than 3700 subjects having a wide range of overweight and obesity (BMI range: 13-129 kg/m 2 ), body fat distribution, insulin sensitivity and glucose tolerance. The random selection of samples, as well as the exclusion criteria of individuals were applied as described in Aprile et al. (2018). Briefly, individuals with the following criteria were excluded: age <18 years, obesity secondary to endocrine diseases, gastrointestinal inflammatory diseases, risk of upper gastrointestinal bleeding, mental disorders, previous or current malignancies, habitual consumption of alcohol or drugs. Patients were categorized according to their BMI into three subgroups: lean (BMI ≤ 25 kg/m 2 ; n = 14), overweight (25 < BMI < 30 kg/m 2 ; n = 17) and obesity (BMI ≥ 30 kg/m 2 ; n = 24).

Animals
Epididymal AT samples were isolated from C57BL/6J mice ip-injected with LPS (2 µg/g of body weight, InvivoGen, San Diego, USA) or with vehicle (NaCL 0.9%) for 5 h. Mice and adipose tissue samples were already described in a previous publication [33]. The study was conducted according to the Principles of Laboratory Animal Care (NIH publication no. 85-23, revised 1985) and the European Union guidelines on animal laboratory care. All procedures were approved by Animal Care Committee of the Faculty of Medicine of Nice-Sophia Antipolis University, Nice, France, and the French ministry of national education (#05116.02 and #201505&9143792_v2).
Human monocytes were isolated from the fresh buffy coats of healthy donors recruited at the Transfusional Center of the Hospital Policlinico at "Federico II" University of Naples (Italy) after informed consent according to the Declaration of Helsinki. PBMCs were obtained by density centrifugation over Ficoll-Paque plus (GE Healthcare, Bio-sciences AB, Uppsala, Sweden), and, subsequently, CD14+ monocytes were isolated from PBMCs using anti-CD14 antibody-bearing magnetic microbeads (Miltenyi Biotec, Bergisch-Gladbach, Germany), according to the manufacturer's instructions. Primary monocytes were resuspended in RPMI-1640 medium (GIBCO, Life Technologies, Carlsband, CA, USA) containing 5% heat-inactivated human AB serum (Lonza, Pearland, TX, USA) and 50 mg/mL gentamicin (Sigma-Aldrich Inc., St. Louis, MO, USA) and cultured at a density of 0.5 × 10 6 cells/mL/well in 24-well culture plates (Costar, Corning, NY, USA). The purity of isolated cells was determined microscopically after cytocentrifugation and differential staining with a modified Wright-Giemsa dye (Diff Quik, Medion Diagnostics AG, Düdingen, Switzerland), while viability was determined by trypan blue dye exclusion. Only preparations with >98% purity was used.
All above-mentioned cell cultures, except for 3T3-L1 cells, were maintained in a humidified atmosphere of 95% air and 5% CO 2 at 37 • C. 3T3-L1 cells were maintained in a humidified atmosphere of 95% air and 7% CO 2 at 37 • C.
All media, sera and antibiotics were purchased from Thermo Fisher Scientific (Waltham, MA, USA).

Human Adipocyte Differentiation
HMSCs were differentiated in mature adipocytes as previously described [29,30]. Briefly, cells were plated at 3 to 4 × 10 3 /cm 2 density and grown to maximum confluence (i.e., timepoint, T = 0 h). Adipocyte differentiation was induced by alternatively using two different mixes: an adipocyte differentiation induction mix and a maintaining mix (AIM and AMM, respectively).

Differentiation of Primary and InVitro Human Macrophages
THP-1 human monocytes were seeded in a complete culture medium supplemented with 50ng/mL of Phorbol 12-Myristate 13-Acetate (PMA, Sigma-Aldrich Inc, St. Louis, MO, USA, Cat#P8139) at 10 × 10 6 /10 mL density in a 10 cm plate. The medium, containing 50 ng/mL of PMA, was replaced every two days until a complete macrophage differentiation was reached (i.e., 6 days after induction).
In order to generate in vitro monocyte-derived macrophages, fresh primary monocytes-isolated as described above-were cultured for 7 days in complete medium supplemented with 50 ng/mL M-CSF (R&D Systems, Minneapolis, MN) and replaced at day 3. Terminally differentiated macrophages (i.e., 8 days after differentiation induction) were incubated for 24 h in fresh complete medium (i.e., control cells) or supplemented with LPS (10 ng/mL) of Escherichia coli O55:B5 (Sigma-Aldrich Inc., St. Louis, MO, USA) and IFNγ 20 ng/mL (R&D Systems, Minneapolis, MN, USA) for cell activation in pro-inflammatory macrophages or supplemented with IL-10 (20 ng/mL; R&D Systems, Minneapolis, MN, USA) to obtain anti-inflammatory macrophages.

Conditioned Medium Preparation
The conditioned medium of human macrophages differentiated from THP-1 was obtained stimulating-or not-the cells with LPS (20 ng/mL; Sigma-Aldrich Inc, St. Louis, MO, USA, Cat# L2630) in RPMI supplemented with 0.5% bovine serum albumin. After 24 h, the medium was collected and centrifuged at 1000× g rpm for 10 min, and cell-free supernatant was used for treating hMSCs, as described below. Similarly, the conditioned media of primary macrophages-both pro-inflammatory (i.e., LPS/IFNγ-induced macrophages) and anti-inflammatory (IL10-induced macrophages)-were obtained by collecting supernatants after 24 h of stimulation (as described above) and centrifugation at 1000× g rpm for 10 min at 4 • C. Finally, J774A.1 mouse macrophages were stimulated or not with 0.5 ng/mL of LPS in DMEM supplemented with 5% FBS for 24 h, and the medium was collected and centrifuged at 1300× g rpm for 5 min.

Oil Red O Staining
Lipid accumulation was measured by Oil red O staining, as previously described [20,29]. In detail, murine adipocytes (i.e., 8 days after adipogenesis induction) were fixed with 4% paraformaldehyde for 20 min, washed three times with PBS and then stained with Oil red O (Sigma-Aldrich Inc., St. Louis, MO, USA, Cat# O0625) for 15 min. Cells were washed, and Oil red O was eluted by isopropanol 100%. The quantification of accumulated lipids was performed by optical density measurement at 510 nm using the spectrophotometer. The background signal was estimated in undifferentiated cells and subtracted from optical density values. The significance of the differences between the samples was calculated by a two-tailed one sample Student's t test.

Cell Treatments
HMSCs (i.e., T = 0 h) or hMSCs-derived mature adipocytes (i.e., T = 21 d after adipocyte-differentiation induction) were treated for 24 h with conditioned medium collected by human macrophages (differentiated from THP-1 as above described) or with RPMI 0.5% BSA supplemented or not with 20 ng/mL LPS, used as control. Moreover, hM-SCs (i.e., T = 0 h) were treated for 24 h with culture supernatants of primary not-polarized macrophages or were activated by LPS/IFNγ or IL-10.
To investigate PPARG mRNA half-life, hMSCs were treated for 3 h with 2 or 5 mg/mL of actinomycin D (Sigma-Aldrich Inc., St. Louis, MO, USA, Cat# A1410) in complete culture medium. Cells treated with the vehicle (i.e., dimethyl sulfoxide, DMSO) were used as control.
Conditioned medium collected by THP-1-derived macrophages was incubated with 0.5 mg/mL of human neutralizing antibody against TNFα (R&D system, Minneapolis, USA, Cat# MAB610-SP) or anti-IgG1 Isotype (R&D system, Minneapolis, MN, USA, Cat# MAB002) for 2 h at 4 • C. Then, the medium supplemented with antibodies was added to hMSCs for 24 h for mRNA expression analysis and 30 min for protein assays. Cells treated with RPMI 0.5% BSA plus anti-IgG1 Isotype were used as control.
For all treatments, cells were starved for 18 h with culture medium without serum.

Cell Transfection
HMSCs were transfected with two different siRNAs designed against SRSF5 (IDT, Coralville, Johnson County, IA, USA, hs.Ri.SRSF5.13.1) or with scrambled siRNAs, by using Lipofectamine 3000 (Thermo Fisher Scientific, Waltham, Massachusetts, MN, USA, Cat# L3000001), according to the manufacturer's instructions. Cell transfection was carried out in culture medium without antibiotics and serum. After 6 h, transfected cells were fed with medium supplemented with 10% FBS. The efficiency of silencing was calculated by quantitative PCR assays (qPCR) after 24 and 48 h of cell transfection, using different siRNAs in single or in combination. The siRNA hs.Ri.SRSF5.13.1 (40 nM) induced the silencing of SRSF5 with estimated efficiency of about 75% and was selected for further analysis. After 18 h of transfection, SRSF5-knockdown cells were treated for 24 h with human recombinant TNFα cytokine (10 or 20 ng/mL) or with the conditioned medium of THP-1-derived macrophages.

RNA Isolation, RT-PCR and Quantitative PCR
Epididymal adipose tissue from C57BL/6J mice was homogenized by using Precellys tissue homogenizer. Total RNA was isolated from homogenized tissues and cell lines byTRIzol Reagent (Thermo Fisher Scientific, Waltham, Massachusetts, MN, USA, Cat# 15596026), according to the manufacturer's instructions. Isolated RNA was quantified with NanoDrop spectrophotometer and was reverse transcribed using "High-Capacity cDNA Reverse Transcription kit" (Thermo Fisher Scientific, Waltham, Massachusetts, MN, USA, Cat# 4368813). Gene expression analysis was performed by quantitative PCR assays using iTaq Universal Sybr Green Supermix (Bio-Rad, Hercules, CA, USA, Cat# 1725125), according to the manufacturer's instructions, on a CFX Connect Detection System (Bio-Rad, Hercules, CA, USA). The specific primer pairs were designed using the Oligo 4.0 program and are listed in Table  S1. The specificity of the amplification reaction was confirmed by melt curve analysis. PPIA, RPS23 and 36b4 were selected as reference genes for analyzing human and mouse samples, respectively. Relative expression analysis was performed using the 2 −∆∆Ct method, except for the analysis of TNFA expression levels assessedin the Leipzig cohort by the ∆Ct method. All reactions were performed in duplicate in at least three independent experiments.

ELISA
Levels of the inflammatory cytokines were determined by ELISA (R&D Systems, MN, USA) in cell-free supernatants according to the manufacturer's instructions. The absorbance of assay wavelength was measured at 450 nm using a Cytation 3 imaging reader (BioTek, Winooski, VT, USA).

Statistics
All data are expressed as means ± SEM, except for expression data in human biopsies that are expressed as median ± DEVST. All assays were performed at least in triplicate. A Shapiro-Wilk test ("shapiro.test function") implemented in R language was used for assessing the normal distribution of data. Statistical significance of differences between testing and control samples was evaluated by a two-tailed (one or two sample) Student's t test (GraphPad Software Inc., La Jolla, CA, USA). Gene expression differences in the German cohort were analyzed by a Mann-Whitney U test. A linear model implemented in R language (lmp function) was used for correlation analysis, as described elsewhere [29]. Boxplots showing TNFA mRNA expression analysis were generated in R using the ggplot2 library and custom scripts. Differences between testing and control samples were defined as significant as p value ≤ 0.05.

The Inflammatory Milieu Affects Pparg Expression and Splicing In Vitro and In Vivo
Pro-inflammatory stimuli suppress PPARγ transcriptional activity in murine primary adipocytes and pre-adipocyte cell lines linking the inflammation of AT to insulin resistance in obese patients [19]. We recently reported the increase of PPARG∆5 and of the PPARG∆5/cPPARG ratio in the SAT of obese patients [29], and a tight association with hypertrophic-rather than "metabolically-healthy" hyperplastic-obesity and its related metabolic alterations [30].
To assess in vivo whether inflammation-hallmark of hypertrophic obesity-affects PPARG splicing, we measured the relative abundance of canonical (cPparg) and dominant negative (Pparg∆5) Pparg transcripts in the epididymal AT of a well-established mouse model of systemic inflammation (previously described in Pastor et al., 2017) [33]. As expected, Pparg transcription is markedly repressed (about 70%; p = 0.0027) in the epididymal fat of LPS-injected lean mice ( Figure 1A, left panel) in line with previous findings [19]. Conversely, Pparg∆5 mRNA levels are not affected ( Figure 1A, left panel), determining in turn a 3.5-fold increase (p = 0.024) of Pparg∆5/cPparg ratio ( Figure 1A, right panel). This finding points out a previously unrecognized effect of inflammation on Pparg splicing in the AT. Accordingly, in line with the downregulation of cPparg levels and increased Pparg∆5/cPparg ratio, the expression of Adipoq, Slc2a4 and Cd36-validated Pparγ target genes-is significantly reduced ( Figure 1B), as we similarly observed in human lipid-engulfed hypertrophic adipocytes [30].
Moreover, our previous findings indicate that high PPARγ∆5 levels are associated with impaired adipogenic capacity of precursor cells [29], which represents a hallmark of hypertrophic obesity. Therefore, we explored Pparg splicing in murine precursor cells differentiating toward mature adipocytes in presence of a pro-inflammatory microenvironment. Hence, 3T3-L1 cells were induced to differentiate in the presence of conditioned media (CM) collected from murine J774.A1 (macrophage-like, MΦ cell line) activated and not with LPS. Interestingly, the analysis of the relative variation of all Pparg transcripts along the adipogenic process reveals that only the secretome of LPS-activated MΦ markedly affects Pparg splicing, inducing an increase of Pparg∆5/cPparg ratio in terminally differentiated adipocytes (1.5-fold, p = 0.03; Figure 1C and Figure S1A,B). As a consequence, pro-inflammatory microenvironment induces an impairment of the adipogenic differentiation ( Figure 1D) paralleled by the downregulation of Pparγ target genes ( Figure 1E), according to in vivo data ( Figure 1B).
These results indicate that the inflammatory microenvironment of the AT affects Pparg splicing, altering the relative amount between canonical and dominant negative isoforms. These results indicate that the inflammatory microenvironment of the AT affects Pparg splicing, altering the relative amount between canonical and dominant negative isoforms.

Pro-Inflammatory Macrophage Secretome Perturbs PPARG Splicing in Human Mesenchymal Stem Cells
To determine whether inflammation also affects the PPARG splicing pattern in humans-similarly to that observed in mouse models-the relative abundance of canonical and dominant negative PPARG transcripts was measured in hMSCs and in vitrodifferentiated adipocytes exposed to the conditioned media of human MΦ differentiated from THP-1 monocytes activated and not with LPS ( Figure 2A, upper panel). As shown in Figure 2A (lower left panel), the exposure of undifferentiated hMSCs to conditioned media of LPS-activated MΦ (THP-1) causes a drop in the expression of all PPARG mRNAs, but with a more pronounced downregulation (about 80% reduction) of canonical transcripts (p = 0.0001). Accordingly, PPARG∆5/cPPARG ratio shows a 2.6-fold increase in exposed cells (Figure 2A, p = 0.0214; lower left panel). In mature adipocytes-differentiated in vitro from hMSCs (T = 21d)-exposed to LPS-activated MΦ (THP-1)-conditioned media, a significant drop in the levels of cPPARG (about 95% reduction; p = 0.0038) and PPARG∆5 (about 91% reduction; p = 0.0004) transcripts is observed (Figure 2A, lower right panel). However, likewise mature murine adipose cells ( Figure 1C), PPARG∆5/cPPARG ratio markedly increases (about 2-fold; p = 0.0115; Figure 2A, lower right panel).
Considering that PPARG∆5 inversely correlates with the adipogenic capacity of hM-SCs [29] and that inflammation perturbs PPARG splicing in these cells (Figure 2A), we investigated which inflammatory stimuli may account for the unbalanced PPARG∆5/cPPARG ratio. Hypertrophic AT is characterized by reduced adipogenesis and massive increase of pro-inflammatory macrophages by both infiltration and local proliferation [34]. Thus, we exposed hMSCs to the conditioned media of ex vivo-isolated human primary monocytes polarized toward two opposite distinct fates: pro-inflammatory (i.e., LPS/IFNγ-induced MΦ) or anti-inflammatory (IL10-induced MΦ) macrophages. As expected by our previous observation (Figure 2A), hMSCs exposed to the secretome of human primary pro-inflammatory MΦ display a strong transcriptional repression of the entire PPARG locus ( Figure 2B). Strikingly, a more pronounced reduction is observed for canonical PPARG transcripts (about 60% reduction; p = 0.0001; Figure 2B, left panel), resulting in a 1.5-fold increase of PPARG∆5/cPPARG ratio (p = 0.0208; Figure 2B, right panel). On the other hand, the secretome of primary anti-inflammatory MΦ-known to promote adipocyte differentiation [35]-induces a 2-fold increase of both canonical and dominant negative PPARG transcripts (p = 0.0061 and p = 0.076 respectively; Figure 2B, left panel), with no effects on their relative ratio (p = 0.9228; Figure 2B, right panel). different half-life, a mRNA stability assay was performed by culturing hMSCs in the pres-ence of growing amounts of the transcription inhibitor actinomycin D (ActD). Even at higher ActD doses, a similar degradation kinetic is observed for all PPARG transcripts ( Figure 2C, left panel), without any effect on their relative ratio ( Figure 2C, right panel), thus excluding differential mRNA decay as a possible source of bias. type. hMSCs at 0 h treated with the CM of non-polarized primary monocyte-derived macrophages were used as a reference sample (dotted lines) and PPIA as a reference gene. Data are reported as mean ± SEM of at least three independent experiments. * p ≤ 0.05, ** p ≤ 0.01 and *** p ≤ 0.001. (C) Relative mRNA quantification (qPCR) of cPPARG, PPARG∆5 (left panel) and the PPARG∆5/cPPARG ratio (right panel) in hMSCs at 0 h treated for 3 h with 2 or 5 mg/mL of actinomycin D. hMSCs at 0h treated with vehicle (i.e., DMSO) were used as a reference sample (dotted lines). PPIA was used as reference gene. Data are reported as mean ± SEM of at least three independent experiments.
To exclude that any variation observed in PPARG transcripts' abundance is due to a different half-life, a mRNA stability assay was performed by culturing hMSCs in the presence of growing amounts of the transcription inhibitor actinomycin D (ActD). Even at higher ActD doses, a similar degradation kinetic is observed for all PPARG transcripts ( Figure 2C, left panel), without any effect on their relative ratio ( Figure 2C, right panel), thus excluding differential mRNA decay as a possible source of bias.
Overall, our in vitro settings reveal that a pro-inflammatory microenvironment affects both PPARG expression and splicing, providing new insights into the mechanisms underlying the link between impaired adipogenesis and inflammation in AT.

TNFα Alters PPARG Splicing in Human Mesenchymal Stem Cells
To identify which soluble factors released by pro-inflammatory macrophages are able to affect PPARG splicing, primary human monocytes were differentiated into macrophages and then polarized into pro-inflammatory and anti-inflammatory by LPS/IFNγ or IL-10 treatments, respectively. In line with previous studies [13,16,33,36], as shown in Figure  S2A, we assessed higher levels of IL-1β, IL-6, IL-8 and TNFα in the conditioned media of LPS/IFNγ-induced MΦ (both vs. not polarized and IL10-induced MΦ),. Thus, PPARG expression and splicing pattern were analyzed in human mesenchymal precursors treated with all the above-mentioned cytokines. HMSCs exposed to human recombinant IL-6 or IL-8 do not display any variation in the expression-nor in the splicing pattern-of PPARG ( Figure 3A, left panel). Differently, cPPARG mRNA is significantly reduced upon treatment with TNFα or IL-1β (about 40%, p = 0.001, and 30%, p = 0.0221, reduction respectively; Figure 3A, left panel). Thus, it is noteworthy that only TNFα significantly affects PPARG splicing, increasing the PPARG∆5/cPPARG ratio (1.4-fold, p = 0.0272; Figure 3A, right panel). Similar results have been obtained by treating 3T3-L1 cells with murine recombinant TNFα and IL-1β ( Figure S2B). Moreover, an high amount of TNFα was observed only in the conditioned media of mouse and human in vitro cultured macrophages ( Figure S2C,D), able to increase PPARG∆5/cPPARG ratio in adipocyte precursor cells and mature adipocytes ( Figures 1C and 2A). Interestingly, a suppressive dose-dependent effect is observed on cPPARG expression-but not on PPARG∆5-when hMSCs are exposed to increasing TNFα doses ( Figure 3B, left panel). Whereas the exposure to the highest TNFα amount (50ng/mL) causes massive cell death (data not shown), TNFα sublethal dose (20ng/mL) markedly increases the PPARG∆5/cPPARG ratio (2.5-fold, p = 0.0123; Figure 3B, right panel). Notably, the addition of recombinant neutralizing anti-TNFα antibody (IgG) to the conditioned media of LPS-activated MΦ (THP-1) does not restore the expression of cPPARG, but reduces PPARG∆5 levels ( Figure 3C, left panel) reverting the TNFα-induced alteration on PPARG splicing in hMSCs ( Figure 3C, right panel). correlation was observed between TNFA and PPARG canonical transcripts (r = 0.19, p = 0.37; Figure S3B). These data reveal a previously unrecognized effect of TNFα on PPARG splicing, suggesting it as a relevant pro-inflammatory molecule able to negatively affect the transcription and splicing at the PPARG locus. Thus, although the transcriptional repression induced by TNFα on the PPARG gene has already been reported [21,22], our data reveal that TNFα does not equally affect all PPARG transcripts, determining an altered balance of canonical and dominant negative isoforms. hMSCs treated with control medium supplemented with antibody against anti-IgG1 Isotype were used as reference samples (dotted lines). PPIA was used as the reference gene. Data are reported as mean ± SEM of at least three independent experiments. * p ≤ 0.05 and *** p ≤ 0.001. (D) Boxplot showing TNFA levels in three different subgroups of individuals from the German cohort, classified according to their BMI in lean (n = 14), overweight (n = 17) and obesity (n = 24). Data are reported as 40-∆Ct value ± DEVST. RPS23 was used as reference gene. *** p ≤ 0.001. Moreover, to investigate ex vivo the correlation between TNFα and PPARG splicing alteration, we evaluated TNFA expression in the subcutaneous adipose tissue (SAT) biopsies of a subset of lean, overweight and obese individuals (n = 56) from a German cohort, described in our previous studies [29,30]. As expected, TNFA mRNA expression was markedly higher in the obese group (vs. lean individuals; p = 0.0018; Figure 3D), in line with the increasing accumulation of pro-inflammatory macrophages in AT. Of note, we observed a mild positive correlation between TNFA expression and PPARG∆5 mRNA levels (r = 0.45, p = 0.027; Figure S3A) only in the SAT of obese individuals, whereas no correlation was observed between TNFA and PPARG canonical transcripts (r = 0.19, p = 0.37; Figure S3B). These data reveal a previously unrecognized effect of TNFα on PPARG splicing, suggesting it as a relevant pro-inflammatory molecule able to negatively affect the transcription and splicing at the PPARG locus. Thus, although the transcriptional repression induced by TNFα on the PPARG gene has already been reported [21,22], our data reveal that TNFα does not equally affect all PPARG transcripts, determining an altered balance of canonical and dominant negative isoforms.

TNFα Modifies PPARG Splicing through PI3K/Akt Signaling
Since TNFα does not equally repress all PPARG transcripts-affecting only cPPARG levels- (Figure 3A,B) and canonical and alternative PPARG mRNAs have similar stability ( Figure 2C), we hypothesized a direct effect of TNFα on members of the splicing machinery. In line with this hypothesis, TNFα is known to induce the splicing of different genes by phosphorylating multiple SR proteins [37][38][39]. Moreover, we previously reported the contribution of ASF2/SF2 to PPARG exon 5 skipping [29]. Therefore, we evaluated SR proteins phosphorylation in hMSCs exposed to TNFα, observing a pronounced phosphorylation of SRp40 (alias SRSF5; Figure 4A). Otherwise, we did not detect marked ASF/SF2 phosphorylation levels. Moreover, a dose-dependent increase of pSRp40 was observed ( Figure 4A). Similarly, the conditioned media collected from LPS-activated MΦ (THP-1) are able to induce SRp40 phosphorylation ( Figure 4B), which is reduced-but not completely abolished-by the addition of neutralizing anti-TNFα IgG ( Figure 4B). This finding suggests that other MΦ-secreted pro-inflammatory factors may sustain SRp40 phosphorylation in the AT.
Furthermore, we attempted to verify how TNFα promotes SRp40 phosphorylation in hMSCs. Interestingly, PI3K/Akt signaling-a known target of TNFα [40,41]-has been reported to induce alternative splicing of PKCbII gene via SRp40 phosphorylation [42]. Hence, we hypothesized that the stimulation of the PI3K/Akt pathway may mediate the effects of TNFα on SRp40. According to this hypothesis, pAkt levels (Ser473) significantly increase upon TNFα exposure ( Figure 4C). Moreover, the addition of neutralizing TNFα antibody to the conditioned media of LPS-activated MΦ (THP-1) is able to reduce Akt phosphorylation ( Figure 4D). Accordingly, PI3K inhibition -induced by cell treatment with wortmannin-blocks Akt phosphorylation and in turn TNFα-mediated phosphorylation of SRp40 (Figures 4E and S4A).  In light of these data, we first evaluated the impact of SRp40 silencing on TNFα-mediated PPARG splicing in hMSCs. Surprisingly, SRSF5 knockdown ( Figures 5A and S4B,C), in cells treated with TNFα or the conditioned media of LPS-activated MΦ (THP-1), significantly increases PPARG exon 5 skipping and, in turn, the PPARG∆5/cPPARG ratio. This unexpected result suggests that SRp40 is required to balance the inclusion/exclusion rate of PPARG exon 5 in the presence of inflammatory stimuli.
Thus, these results further support the role of PI3K/Akt signaling in linking inflammation to PPARG splicing and suggest that the inhibition of SR proteins facilitates the activity of other competing splicing factors-such as heterogeneous nuclear ribonucleoproteins (hnRNPs) [46][47][48]-in the modulation of PPARG exon 5 inclusion/exclusion.  and the PPARG∆5/cPPARG ratio (right panel) in hMSCs at the undifferentiated stage treated with 10 or 20 ng/mL of human recombinant TNFα cytokine in combination or not with KHCB19 inhibitor. hMSCs at 0h treated for 24 h with 10 or 20 ng/mL of human recombinant TNFα cytokine and with vehicle (i.e., DMSO) were used as reference samples (dotted lines). PPIA was used as the reference gene. Data are reported as mean ±SEM of at least three independent experiments. * p ≤ 0.05. (D) Relative mRNA quantification (qPCR) of cPPARG, PPARG∆5 (left panel) and the PPARG∆5/cPPARG ratio (right panel) in hMSCs at the undifferentiated stage treated with 10 or 20 ng/mL of human recombinant TNFα cytokine and with or without Wortmannin inhibitor. hMSCs at 0 h treated for 24 h with 10 or 20 ng/mL of human recombinant TNFα cytokine and with vehicle (i.e., DMSO) were used as reference samples (dotted lines). PPIA was used as the reference gene. Data are reported as mean ± SEM of at least three independent experiments. * p ≤ 0.05 and ** p ≤ 0.01.

Discussion
Insulin resistance in individuals with hypertrophic obesity is causally associated with low-grade chronic inflammation, characterized by the infiltration of T cells, macrophages and other immune cells [4][5][6][7][8][9][10]. The release of cytokines and chemokines by pro-inflammatory and metabolically activated macrophages [12,14,15] has a negative impact on the expression and activity of PPARγ [19][20][21][22]. This nuclear receptor has a well-recognized central role in differentiating new insulin-sensitive adipocytes, in keeping the metabolic homeostasis of adipose tissue, in repressing inflammatory genes (e.g., TNFα, iNOS, MMP9) [23][24][25][26][27][28]49] and in dictating monocyte polarization toward an anti-inflammatory phenotype [50][51][52]. Hence, the repressing effect induced by the inflammation on PPARγ generates a selfsustained vicious cycle, contributing to the impairment of neo-adipogenesis, which in turn promotes adipocyte hypertrophy, as well as local and global inflammation [12,18]. We previously described a marked overexpression of the dominant-negative isoform of PPARγ, PPARγ∆5, in the AT of obese/diabetic patients and its ability to impair adipocyte differentiation and PPARγ activity in hypertrophic adipocytes [29,30]. Moreover, the increased PPARG∆5/cPPARG ratio in SAT enriched of large adipocytes, as well as in our recently generated in vitro model of human hypertrophic-like adipocytes [30], further encouraged us to investigate the putative contribution of pro-inflammatory factors-highly secreted in hypertrophic AT-to the unbalance of PPARG isoforms.
Herein, we aimed to dissect the impact of inflammatory stimuli on PPARG expression and splicing, considering that the relevance of the latter has been often underestimated. The identification, for the first time, that inflammatory factors, other than repressing PPARG expression can modulate-in vitro and in vivo-its splicing pattern, adds an interesting piece to the puzzling regulation of PPARG. In this regard, our results provide evidence of a marked increase of the Pparg∆5/cPparg ratio-paralleled by the repression of Ppargregulated adipocyte markers-in the epididymal fat of LPS-treated mice, as well as in murine and human adipocytes exposed to macrophages-secreted pro-inflammatory factors. A similar effect on PPARG splicing is also observed in human mesenchymal stem cells, in which we recently described a role for PPARG∆5 in reducing their adipogenic capacity [29]. Taken together, our data provide an intriguing insight in the molecular mechanisms through which the proinflammatory microenvironment contributes to the impaired neo-adipogenesis in hypertrophic AT, suggesting for PPARG splicing a role as a hub. The increased PPARG∆5/cPPARG ratio in the SAT of overweight/obese individuals [29] and in lipid-engulfed adipocytes [30] further support this hypothesis. However, we cannot discern the relative contributions of functional isoforms, whose levels are reduced, and of increased dominant negative isoforms to the impairment of PPARγ activity leading to a diminished adipogenic differentiation.
Moreover, in line with the suggested role of TNFα as a key mediator of the effects induced by inflammation on preadipocytes and adipocyte differentiation [53,54], herein we reported TNFα-known to inhibit PPARγ activity through distinct mechanisms [21,22]-as a major pro-inflammatory cytokine able to alter PPARG splicing pattern. Ex vivo analyses on SAT biopsies revealed higher TNFA expression and a positive correlation with PPARG∆5 levels only in obese (but not in lean) individuals, further supporting also in human hypertrophic adipocytes the link between inflammation and PPARG splicing alteration. Interestingly, the finding that canonical and alternative PPARG transcripts have the same half-life but respond differently to TNFα, as well as to the conditioned media of proinflammatory macrophages, supports a direct effect of inflammation on the splicing machinery. Of note, the alteration of spliceosome machinery due to altered levels of phosphorylated proteins involved in mRNA splicing has been recently reported even in the context of obesity and T2D [55][56][57]. Moreover, TNFα is known to dynamically modulate the inclusion/exclusion rate of alternative exons by enhancing the phosphorylation of multiple SR proteins [37][38][39]. In our experimental settings-i.e., in hMSCs exposed to MΦ-secreted pro-inflammatory factors-we found a TNFα-mediated increase of SRp40 phosphorylation, suggesting a still unexplored role for this splicing factor in PPARG exon 5 inclusion/exclusion. However, as we did not definitely establish the exact role of SRp40 in PPARG splicing, we cannot exclude that multiple SR proteins may cooperate in the modulation of PPARG exon 5 inclusion/exclusion balance and in turn to the generation of the truncated PPARγ isoform.
The finding that MΦ-secreted TNFα impairs PPARγ functionality by promoting the generation of dominant negative isoforms over canonical ones has relevance also from a therapeutic perspective. Indeed, the recently developed anti-inflammatory drugs, which specifically targeting pro-inflammatory MΦ, revealed to be very promising to ameliorate obesity-related inflammation and metabolic dysfunctions [58]. Moreover, the mechanisms underlying polarization of MΦ into pro-inflammatory and metabolically activated cellsrather than in anti-inflammatory MΦ-in the AT of obese patients [12,14,15] remain to be elucidated. Of note, human monocytes/macrophages express PPARγ dominant negative isoforms [29,59]. Hence, given the key role of PPARγ in monocyte polarization, we might speculate that autocrine (or paracrine) signaling in a hypertrophic AT microenvironment may affect PPARG transcripts abundance in AT-resident MΦ, contributing to disease etiology.
In conclusion, our results demonstrate a novel mechanism through which inflammation impairs PPARγ activity in the AT. The imbalance among functional and dominant negative PPARG transcripts is likely to contribute to establishing healthy and unhealthy AT under positive energy intake. These data further indicate the relevance to target inflammation in adipose tissue of patients with hypertrophic obesity to minimize the detrimental effects that pro-inflammatory molecules have on PPARγ functionality in this tissue.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/cells11010042/s1, Figure S1: Conditioned media from mouse macrophages modulate the expression of cPparg and Pparg∆5 transcripts. Figure S2: Cytokines secretions by ex vivo and in vitro macrophages and their effects on Pparg splicing in 3T3-L1 cells. Figure S3: TNFA mRNA levels positive correlate with PPARG∆5 levels in obese individuals. Figure S4: Inflammatory stimuli modulate PPARG splicing through SRp40 phosphorylation. Table S1: List of primers sequence of gene analyzed by qPCR.

Institutional Review Board Statement:
The study was conducted according to the Principles of Laboratory Animal Care (NIH publication no. 85-23, revised 1985) and the European Union guidelines on animal laboratory care. All procedures were approved by Animal Care Committee of the Faculty of Medicine of the Nice-Sophia Antipolis University, Nice, France, and the French ministry of national education (#05116.02 and #201505&9143792_v2). The study has been reviewed and approved by the ethics committee of Leipzig University, Germany (approval numbers: 159-12-21052012 and 017-12-23012012) and carried out in accordance with the Declaration of Helsinki, the Bioethics Convention (Oviedo) and the EU Directive on Clinical Trials (Directive 2001/20/EC).
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.