Appearance and disappearance of the mRNA signature characteristic of T_reg cells in visceral adipose tissue: Age, diet, and PPARγ effects

Significance

A unique population of Foxp3⁺CD4⁺ regulatory T (T_reg) cells resides in visceral adipose tissue of lean mice. VAT T_regs are important regulators of local and systemic inflammation and metabolism. Here, we show that the VAT T_reg signature is imposed early in life, well before the typical age-dependent expansion of the adipose-tissue T_reg population. VAT T_regs in obese mice lose the signature typical of lean individuals but gain an additional set of over- and underrepresented transcripts. In striking parallel to a pathway recently elucidated in adipocytes, the obese mouse VAT T_reg signature depends on phosphorylation of a specific residue of PPARγ. These findings are important to consider in designing drugs to target type 2 diabetes and other features of the “metabolic syndrome.”

Abstract

A unique population of Foxp3⁺CD4⁺ regulatory T (T_reg) cells resides in visceral adipose tissue (VAT) of lean mice, especially in the epididymal fat depot. VAT T_regs are unusual in their very high representation within the CD4⁺ T-cell compartment, their transcriptome, and their repertoire of antigen-specific T-cell receptors. They are important regulators of local and systemic inflammation and metabolism. The overall goal of this study was to learn how the VAT T_reg transcriptome adapts to different stimuli; in particular, its response to aging in lean mice, to metabolic perturbations associated with obesity, and to certain signaling events routed through PPARγ, the “master-regulator” of adipocyte differentiation. We show that the VAT T_reg signature is imposed early in life, well before age-dependent expansion of the adipose-tissue T_reg population. VAT T_regs in obese mice lose the signature typical of lean individuals but gain an additional set of over- and underrepresented transcripts. This obese mouse VAT T_reg signature depends on phosphorylation of the serine residue at position 273 of PPARγ, in striking parallel to a pathway recently elucidated in adipocytes. These findings are important to consider in designing drugs to target type 2 diabetes and other features of the “metabolic syndrome.”

Inflammation is a major link between obesity and type 2 diabetes (T2D) (1–3). Energy intake in excess of expenditure induces chronic inflammation of visceral adipose tissue (VAT), which eventually provokes inflammation at distant sites and, thereby, metabolic abnormalities such as insulin resistance and dyslipidemia, culminating in T2D and cardiovascular disease. At the core of this “metabolic syndrome” is VAT: Through the production of a variety of adipokines and other mediators, adipocytes in visceral fat depots can positively or negatively influence insulin sensitivity, lipid levels and appetite. Although innate immunocytes, notably macrophages (MFs), have historically been considered to be the drivers of adipose-tissue inflammation and metabolic dysregulation, several recent reports argued for an important effector or regulatory role for adaptive immunocytes, i.e., T, B, or NKT cells.

In particular, a unique population of Foxp3⁺CD4⁺ regulatory T (T_reg) cells accumulates in VAT of lean mice as they age (4, 5). Vis-à-vis their lymphoid-tissue counterparts, VAT T_regs are remarkable by several criteria. They are highly overrepresented in lean individuals (40–80% vs. 5–15% of the CD4⁺ T-cell compartment). Their transcriptome is distinct, especially the profile of transcripts encoding transcription factors, chemokines/chemokine-receptors, and cytokines/cytokine-receptors, as well as atypical expression of a set of transcripts specifying molecules involved in lipid metabolism. Thirdly, they have an unusual, clonally expanded, repertoire of T-cell antigen receptors (4).

It came as a surprise that PPARγ, the “master regulator” of adipocyte differentiation (6), is also a crucial molecular orchestrator of VAT T_reg cell accumulation, phenotype and function (5). T_reg-specific ablation of Pparg greatly reduced the VAT T_reg population, while not affecting lymphoid-tissue T_regs, in mice maintained on a normal-chow diet (NCD). Conversely, injection of the PPARγ agonist, pioglitazone (pio), into mice kept on a high-fat diet (HFD) expanded the T_reg population in VAT but not lymphoid tissues. Cotransduction experiments revealed that PPARγ worked together with Foxp3 to impose the unique transcriptome of VAT T_regs: the sets of genes over- or underrepresented in CD4⁺ T cells transduced with Foxp3 plus Pparg in comparison with Foxp3 alone was enriched for the previously defined VAT T_reg up- and down-signatures.

The overall goal of the set of experiments reported herein was to further elucidate the VAT T_reg signature: its appearance in lean mice as they age; disappearance in obese mice; and its response to PPARγ-mediated signaling events. Our data document a number of age- and diet-dependent influences on the VAT T_reg transcriptome, and reveal that, as in adipocytes, the effect of obesity on transcription in VAT T_regs does not reflect a reduction in their expression of the Pparg gene, but rather a posttranslational modification of PPARγ proteins.

Results

Appearance of the VAT T_reg Signature in Lean Mice as they Age.

Our first goal was to learn how the transcriptome of VAT T_regs evolves over time in lean mice, given that insulin resistance is an age-dependent process. The epididymal fat depot was removed from cohorts of male C57BL/6 (B6) mice at increasing ages, and was digested with collagenase to separate floating mature adipocytes from the other cells, referred to as the stromal vascular fraction (SVF). Foxp3⁺CD4⁺ T_regs witin the SVF were enumerated and further characterized by flow cytometry. Control splenic and/or lymph node (LN) T_regs from the same individuals were assessed in parallel. At 5 wk of age, the fraction of T_regs in the VAT CD4⁺ T-cell compartment was similar to, even a bit lower than, that typical of lymphoid organs (Fig. 1A). The fraction in VAT gradually rose at 14 and 25 wk of age, and then descended quite precipitously at 40 wk (Fig. 1A), accompanied by a decline in insulin sensitivity (Fig. S1). Meanwhile, the percentage of splenic T_regs remained stable at 15–20% (Fig. 1A). A similar rise and fall of the VAT, but not splenic, T_reg population was evident when their number, rather than fractional representation, was quantified (Fig. 1B).

Fig. 1.

T_reg dynamics and phenotype with aging. T_regs were isolated from VAT and spleen of lean B6 males at the indicated ages. (A) Cytofluorometric analysis of Foxp3⁺CD4⁺CD3⁺ T cells. (Left) Representative dot plots. Numbers indicate the percentage Foxp3⁺ of CD4⁺ cells for that particular experiment. (Right) Summary data from at least three independent experiments. (B) T_reg numbers for the same mice. (C) As per A, except Gata3⁺Foxp3⁺ cells were examined. Numbers refer to the percentage Gata3⁺ of Foxp3⁺ cells for that particular experiment. Mean ± SD; **P < 0.01; ***P < 0.001; NS, not significant; by the Student’s t test.

We previously reported that the transcription factor, Gata3, is a useful marker for bona fide VAT T_regs [given the absence of an appropriate monocolonal antibody (mAb) for cytofluorimetric detection of PPARγ] (5). The fraction of Gata3⁺ T_regs in VAT also rose at 14 and 24 wk of age and fell at 40 wk, whereas the fraction in the spleen was constantly low (Fig. 1C).

To follow corresponding evolution of the VAT transcriptome, we performed microarray-based gene-expression profiling of Foxp3⁺CD4⁺ T cells from the LNs vs. the SVF of VAT from B6 mice at increasing ages. Although the transcriptional profiles of VAT and LN T_regs were already distinct at 5 wk of age, the total number of loci differentially expressed ≥twofold in VAT mounted at 14 wk and peaked at 25 wk of age (Fig. 2A). This increase reflected evolution of the VAT T_reg transcriptome as there were very few differences in the gene-expression profiles of LN T_regs from 5- and 25-wk-old mice (Fig. 2B). Analogous to the numerical decline in Tregs at 40 wk of age, the VAT:LN T_reg gene-expression differential also dropped at this age (Fig. 2A).

Fig. 2.

T_reg transcriptome evolution with aging. (A) Normalized microarray-determined expression values for transcripts from VAT vs. LN T_regs of B6 mice. Values in the corners refer to numbers of transcripts up- (Upper) or down- (Lower) regulated in VAT T_regs. Lines indicate the twofold cutoff. (B) As in panel a except LN transcriptomes from 5- and 25-wk-old mice are compared. (C) Same data as in A except the VAT Treg up- (pink) and down- (green) signatures are overlain (from Dataset S1). (D) Principal components analysis of the same microarray data.

Using microarray data from 25-wk-old lean mice, we generated “VAT T_reg” signatures, comprised of 437 genes up-regulated and 71 genes down-regulated ≥twofold in VAT T_regs vs. VAT conventional T (T_conv) cells, LN T_regs and LN T_conv cells (Dataset S1; see its legend for details of signature generation). This signature is more robust than the one previously reported (4) because the various T-cell populations were double-sorted to extremely high purity and because the newer microarray platform encompasses a higher fraction of the genome. Superimposing the VAT T_reg signature on the VAT vs. LN T_reg comparison plots revealed that the bulk of the signature genes were divergently expressed (i.e., displaced from the diagonal) already at 5 wk of age—notably, paradigmatic VAT transcripts such as Pparg, Gata3, Klrg1, Ccr2, and Il1rl1 (Fig. 2C). Nonetheless, the differential gradually increased to 24 wk, but shrank noticeably at 40 wk of age (Fig. 2C). This age-dependent evolution of the VAT Treg signature, and its persistent distinction from the LN gene-expression profile, is readily apparent from the principal components analysis illustrated in Fig. 2D.

Disappearance of the VAT T_reg Signature in Obese Mice.

Next, we addressed how the transcriptome of VAT T_regs responds to metabolic dysregulation. Two insulin-resistant mouse models of obesity were compared with their lean counterparts: 14-wk-old leptin-deficient (B6.Lep^ob/ob; hereafter referred to as B6-ob/ob) mice versus B6 wild-type (B6-WT) littermates; and B6 animals maintained on HFD versus NCD between 8 and 16 wk of age. As previously reported (4), both genetically promoted and diet-induced obesity resulted in a clear reduction in the T_reg fraction and number in VAT, while not in the spleen (Fig. 3 A and B). In addition, a lower proportion of VAT, but not splenic, T_regs expressed detectable levels of Gata3 in the two obese models (Fig. 3C). Under the conditions we chose, the VAT T_reg changes were more pronounced in the B6-ob/ob than the HFD model, not surprising given that the former were distinctly more obese (6 g heavier on average).

Fig. 3.

T_reg numbers and phenotype in obese mice. (A–C) T_reg cell representation in the CD4⁺ compartment. Cells were isolated from the VAT and spleen of 14-wk-old B6-ob/ob vs. B6-WT mice or of B6 mice fed HFD vs. NCD between the ages of 8 and 16 wk. Percentage Foxp3⁺ of CD4⁺ T cells (A), number of such cells (B), and percentage Gata3⁺ of Foxp3⁺ cells were analyzed and displayed as per the corresponding panels in Fig. 1. For A–C, mean ± SD; *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant; by the Student’s t test. (D) Volcano plots comparing transcriptomes of obese and lean mice in the two models. Overlain are the VAT T_reg up- (pink) and down- (green) signatures (from Dataset S1). P values from the χ² test. (E) Representative dot plots of Klrg1 (Left) and CCR2 (Right) expression by VAT and spleen T_regs from obese mice and their lean controls. Percentages refer to the fraction of Foxp3⁺ cells constituted by the gated population. (F) Volcano plots comparing transcriptomes of obese and lean mice for the two models. Highlighted are the transcripts up- (orange) or down- (aqua) regulated ≥1.5-fold in obese mice (with a P value of ≤0.05 in three replicates). The omVAT T_reg up- and down-signatures sum the altered transcripts from the two models (Dataset S2).

In both obese models, the T_regs that persisted in the adipose tissue were no longer bona fide VAT T_regs. This finding is perhaps most evident from the Fold-Change (FC) versus P value volcano plots of Fig. 3D, which show a significant underrepresentation of the VAT T_reg up-signature and overrepresentation of the corresponding down-signature. Dampening of the characteristic VAT T_reg gene-expression profile could reflect alterations at the molecular level (i.e., genes turned up or down), the cellular level (e.g., VAT T_regs replaced by lymphoid-tissue T_regs) or a mixture of the two. The fact that the transcript “cross-overs” were never complete, some elements of the up-signature remaining overexpressed and some members of the down-signature still underexpressed in VAT T_regs of obese mice, argues that the transcriptome changes do not simply reflect replacement of VAT by lymphoid-tissue T_regs. We were able to confirm transcript changes at the protein level via cytofluorimetric analysis of gated T_regs (Fig. 3E).

To generate a transcriptional signature for the VAT T_regs characteristic of obese mice (which will be referred to as the omVAT T_reg signature), we combined the sets of transcripts differentially expressed ≥1.5-fold in either of the two models (Fig. 3F and Dataset S2). The substantially larger number of overrepresented transcripts in the genetically promoted than in the diet-induced obesity model may reflect leptin-dependent transcriptional effects or may simply result from the more advanced obesity in B6-ob/ob mice. Pathway analysis using the Ingenuity and DAVID programs highlighted lymphocyte activation (P = 10⁻⁷), cytokine/cytokine-receptor (P = 10⁻⁴) and chemokine/chemokine receptor (P = 10⁻³) pathways in the omVAT Treg up-signature. Sterol (10⁻⁷) and steroid (10⁻⁶) biosynthesis pathways emerged from pathway analysis of the corresponding down-signature.

Phosphorylation of PPARγ in Cultured VAT T_regs Recapitulates the Transcriptome Changes Provoked by Obesity in Vivo.

Interestingly, the omVAT T_reg up-signature was significantly enriched and the corresponding down-signature significantly impoverished in the transcriptome of T_regs isolated from mice lacking PPARγ only in T_regs (compared with their WT littermates) (Fig. 4A). These observations argue that obesity might exert its impact somewhere at the level of PPARγ. However, in neither model did VAT T_regs from obese mice express fewer Pparg transcripts than their lean counterparts did (Fig. 4B). Hence, we were prompted to explore other known aspects of PPARγ biology.

Fig. 4.

PPARγ and transcriptome changes with obesity. (A) Volcano plot comparing the VAT T_reg transcriptomes of mice lacking PPARγ only in T_regs (T_reg Pparg mut) vs. of their WT littermates (Pparg WT). Highlighted are the omVAT T_reg up- (orange) and down- (aqua) signatures (from Dataset S2). P values from the χ² test. (B) Pparg transcript levels. Extracted from the microarray data of Fig. 3. Mean ± SD. NS = P value not significant by the Student’s t test.

Spiegelman and colleagues recently discovered that many PPARγ ligands exert anti-diabetic activities through a previously unsuspected biochemical pathway independent of classical PPARγ agonism and its downstream transcriptional consequences (7–9). Diet-induced obesity activates cyclin-dependent kinase (Cdk)5 and ERKs in adipocytes, leading to phosphorylation of the serine residue at position 273 (Ser273) of PPARγ, which in turn results in dysregulation of a set of genes abnormally expressed in the obese state. Certain PPARγ ligands, such as rosiglitazone and MRL24, are antidiabetic because they block Cdk5-induced phosphorylation of PPARγ. These processes were detected in adipose tissue ex vivo, and could be mimicked in adipocyte cultures treated with tumor necrosis factor (TNF)α.

We hypothesized that analogous phosphorylation of PPARγ at the Ser273 position might underlie at least some of the transcriptome changes induced in VAT T_regs by obesity. Unfortunately, by-far-inadequate cell numbers precluded testing this hypothesis on ex vivo T_regs by biochemical means. Instead, we made use of a TNFα-supplemented culture system parallel to that which had been used for adipocytes by Spiegelman and colleagues. Conventional Foxp3⁻CD4⁺ T cells were isolated from lean B6 mice, activated in vitro, and cotransduced with retroviruses expressing Foxp3 and Pparg. We had previously demonstrated that the concomitant expression of these two genes in naïve CD4⁺ T_conv cells induced much of the VAT T_reg signature typical of lean mice (5). We first checked to what extent TNFα treatment of Foxp3/Pparg-transduced cells mimicked the T_reg transcriptome changes induced by obesity in vivo. As illustrated by a volcano plot comparing the effects of 24-h TNFα versus vehicle treatment, the omVAT T_reg up-signature was strongly induced by TNFα, whereas the corresponding down-signature was essentially unaltered (Fig. 5A). We then determined to what extent the TNFα-induced transcriptional changes reflected Ser273 phosphorylation. In the absence of TNF-α, the Foxp3 transductants cotransduced with wild-type Pparg (Pparg-WT) and those transduced with Pparg encoding an Ala substitution for the Ser at position 273 (Pparg-S273A) showed indistinguishable expression of the omVAT T_reg up- and down-signatures (Fig. 5B); however, in the presence of TNFα, the up-signature was greatly enriched in the double-transductants expressing WT compared with the mutant PPARγ (Fig. 5C). Furthermore, pretreatment of the Foxp3/Pparg double-transductants with SR1664, a Cdk5 inhibitor with no agonistic activity for PPARγ (7), abolished the TNFα-induced differential between WT and mutant PPARγ (Fig. 5D).

Fig. 5.

Effect of PPARγ phosphorylation on the VAT T_reg signatures. Naïve CD25⁻CD4⁺ T cells were stimulated ex vivo, and were transduced with retroviruses encoding Foxp3 and WT Pparg (Pparg WT) or Pparg bearing the S273A phosphorylation-site mutation. Infected cells were treated with TNFα or vehicle in the presence or absence of the Cdk5 inhibitor, SR1664, and then were sorted for GFP and Thy1.1 positivity before RNA processing. (A–D) Microarray analysis. Volcano plots comparing the effects of TNFα (A), of the S273A mutation in the absence (B) or presence (C) of TNFα, and of the Cdk5 inhibitor, SR 1664, in the presence of TNFα (D). Highlighted are the omVAT T_reg up- (orange) and down- (aqua) signatures. P values from the χ² test. (E) Western blotting. Same samples as in A–D, except an additional sample from cells treated with TNFα plus pio was included. (Left) An example blot using Abs that detect total PPARγ (clone E-8, Santa Cruz) or PPARγ phosphorylated at position 273 (7). (Right) Quantification from two experiments. For each experiment, the value for vehicle-treated cells was set at 1. Mean ± SD; *P = <0.05; NS, not significant; by the Student’s t test.

Western blotting with mAbs recognizing PPARγ or specifically the phosphorylated Ser273 residue of PPARγ confirmed the expected phosphorylation status of the Foxp3/Pparg double-tranductants (Fig. 5E). TNFα induced phosphorylation at position 273 in WT, but not mutant, PPARγ; and TNFα-induced Ser273 phosphorylation was dampened by both the PPARγ ligand, pio, and the Cdk5 inhibitor, SR1664.

Discussion

The transcriptome of VAT T_regs is strikingly different from that of classical T_regs circulating through lymphoid tissues (4, 5). Notably, transcripts encoding certain transcription factors (e.g., PPARγ, Gata-3), chemokines or their receptors (e.g., CCR1, CCR2), cytokines or their receptors (e.g., IL10, IL1rl1), and a set of proteins involved in lipid metabolism (e.g., Dgat1, Pcyt1a) are overrepresented in T_regs residing in VAT. The overall goal of this study was to further elucidate the VAT T_reg signature: how it responds to aging in lean mice, to metabolic perturbation in obese mice, and to certain PPARγ-mediated signaling events discovered in adipocytes.

Although the fraction of T_regs in the VAT CD4⁺ T-cell compartment was routinely much higher than that in lymphoid organs in 25-wk-old mice, the differential was not apparent until about 10–15 wk of age, and 5-wk-old mice harbored a T_reg population in VAT fractionally similar to, or even smaller than, that in the spleen and LNs (ref. 4 and Fig. 1 A and B). One possibility was that VAT is populated by T_regs of lymphoid-type early in life, which are gradually replaced by their VAT-type counterparts, optimally adapted to thrive and proliferate. However, we found that the VAT T_reg signature was already evident at 5 wk of age. Paradigmatic VAT T_reg transcripts were already overrepresented, and for most of them there was little additional increase in expression at 25 wk: e.g., Pparg (1.3-fold), Gata3 (1.4-fold), Ccr2 (2.6-fold), Klrg1 (1.7-fold), Il1rl1 (1.8-fold). There was, however, a score of transcripts enriched substantially more (>4 times) in the VAT T_reg population of 25- than of 5-wk-old mice; their continued increase might reflect additional local adaptation to the lipophilic, hypoxic adipose-tissue environment.

The first suggestion that VAT T_regs might have an important function in guarding against adipose-tissue inflammation and insulin resistance was their conspicuous reduction in insulin-resistant models of obesity, a notion ultimately substantiated by gain- and loss-of-function experiments (4, 5). In the present study, we showed this numerical decrease to be accompanied by loss of a large swathe of the VAT T_reg signature in the remaining Foxp3⁺CD4⁺ T-cell population, no matter whether obesity was provoked by a genetic or dietary alteration. Interestingly, about 1/3 of the VAT T_reg up-signature was not down-regulated in each of the two models.

In this study, we also defined an omVAT T_reg signature, consisting of transcripts up- or down-regulated in obese mice vis-à-vis their lean counterparts. Surprisingly, although this signature showed a strong correlation with transcriptional changes provoked by ablation of PPARγ (Fig. 4A), the level of transcripts encoding PPARγ, itself, was not altered in the obese state. Instead, experiments on a previously and currently validated (ref. 8 and Fig. 5A) TNFα-supplemented cell-culture model led us to conclude that PPARγ must be phosphorylated at the Ser273 position by Cdk5. This posttranslational modification led to dysregulated transcription akin to what was previously reported for adipocytes (ref. 8 and Fig. 5).

These findings point to a remarkable parallelism in PPARγ’s modus operandi in adipocytes and VAT T_regs. It is thus easier to rationalize our previous observation that the PPARγ displayed on T_regs is a major conduit for the insulin-sensitizing effects of the PPARγ agonists, such as pio and rosiglitazone, that were for many years front-line drugs for treatment of T2D (5). And one might anticipate that new PPARγ agonists designed to have potency in the absence of weight gain, fluid retention, and cardiac abnormalities (7) will effectively target VAT T_regs as well as adipocytes. Lastly, this remarkable parallelism between the T_reg population and its parenchymal cohabitants highlights the malleability of T_regs in adapting to a particular tissue context, operating in concert with its defining cellular participant.

Methods

Mice.

B6, B6.Lep^ob/ob, and B6.Pparg-flox (10) mice were bred in our specific-pathogen-free facilities at Harvard Medical School, or were purchased from Jackson Laboratory. T_reg-Pparg mut mice were generated as described (5). HFD-fed animals were maintained on a diet of 60 kcal% fat, purchased from Research Diets (catalog no. D12451). NCD-fed control animals were kept on a diet containing 10 kcal% fat, obtained from the same vendor (catalog no. 12450B). In designated experiments, pio (Actos, Takeda) was introduced into the diet at a concentration of 100 mg per kg of food. Experimental and control animals were generally littermate-matched males.

HOMA-IR, calculated as described (11), was determined for 25- and 40-wk-old B6 individuals. Mice were fasted overnight, weighed, then tested for fasting blood-glucose and blood-insulin concentrations by ELISA, performed by the Joslin Diabetes Center’s Specialized Assay Core.

Immunocyte Isolation and Analysis.

Epididymal adipose tissue (VAT) and splenic or LN immunocytes were isolated as described (4, 5). They were routinely stained with anti-CD45 (30-F11), -CD3 (145-2C11), -CD4 (GK1.5), -CD8 (5H10), and -CD25 (PC61) mAbs, (all eBiosciences) and, for some experiments, were fixed and permeabilized according to the manufacturer’s instructions, followed by intracellular staining of Foxp3 (FJK-16s) and/or Gata3 (TWAJ) (both from eBiosciences). Cells were double-sorted using the Moflo, or analyzed using an LSRII instrument (BD Bioscience) and FlowJo software.

Retroviral Transduction Experiments.

Immunocytes were harvested from pooled spleens and LNs of 6- to 8-wk-old B6 mice, and CD25⁻CD4⁺ T cells were prepared for retroviral transduction as described (5). Viruses were made by transfecting platE cells (12) with (i) retroviral expression plasmids [(MSCV IRES-GFP (pMIG2) and IRES-Thy1.1 (pMIT2)] (13), encoding Foxp3 (GFP), PPARγ WT (Thy1.1) or PPARγ S273A (Thy1.1); and (ii) the packaging construct pcl-ECO (14); using TransIT-293 (Mirus) according to the manufacturer's instructions. Naive CD4⁺ T cells were infected with retroviral supernatants 48 h after activation with anti-CD3/CD28 mAb-coated beads, and were cultured for an additional 72 h. Singly (Foxp3) or doubly (Foxp3 plus Pparg WT or Pparg S273A) transduced cells were double-sorted as (CD11b⁻ CD11c⁻B220⁻CD8⁻)CD3⁺CD4⁺GFP⁺(and/or Thy1.1⁺) by Moflo for RNA processing and microarray analysis. In designated experiments, 48 h after cell infection, single- and double-transductants were treated with TNFα (50 ng/mL) with or without SR1664 (2 μM) (7) or vehicle (DMSO) alone for 24 h.

Microarray Analysis.

LN or VAT T_regs (CD25^hiCD4⁺CD3⁺) or T_conv cells (CD25⁻CD4⁺CD3⁺) were double-sorted from B6-WT, B6-ob/ob, Pparg WT or T_reg-Pparg mut mice. GFP⁺ (i.e., Foxp3⁺)CD4⁺CD3⁺ double-transduced cells, treated with TNFα plus SR1664 or vehicle, were also double-sorted to ensure high purity. RNA was extracted, amplified for two rounds, biotin-labeled, and purified as described (5). The resulting cRNAs (three independent datasets for each sample type) were hybridized to Mouse Genome M1.0 ST arrays (Affymetrix) according to the manufacturer's protocol. Microarray data were background-corrected and normalized, and replicates were averaged as described (5). The “VAT T_reg” signatures included transcripts ≥twofold over- or underrepresented in VAT T_regs vs. LN T_regs, VAT T_conv cells and LN T_conv cells of 25-wk-old mice. The omVAT T_reg signatures included all transcripts ≥1.5-fold over- or underrepresented in VAT T_regs of obese mice vis-à-vis their lean comparators, summing those from the B6-ob/ob and HFD-fed models.

Immunoblotting.

Foxp3⁻CD4⁺ T cells were cotransduced with Foxp3 and Pparg(WT or encoding the S273A mutant), pretreated with pio or SR1664 in DMSO or just DMSO, before incubation with DMSO or TNFα in DMSO. Protein lysates were made, run on SDS/PAGE, and probed with an anti-PPARγ or an anti-phosphoPPARγ (pPPARg) antibody, all as per ref. 7.