FOXP4 differentially controls cold-induced beige adipocyte differentiation and thermogenesis

Beige adipocytes, a population of thermogenic adipocytes distinct from brown adipocytes, have recently attracted mainstream attention. Their relevance to adult humans holds promise as a new therapeutic target in combating obesity and other metabolic disorders. Beige adipocytes emerge within white adipose tissue (WAT) depots in response to various environmental cues, including chronic cold acclimatization, exercise, β3-adrenergic receptor (AR) agonists, cancer cachexia, and tissue injury (Barbatelli et al., 2010; Rosenwald et al., 2013). As with classical brown adipocytes, beige adipocytes present with multilocular morphology and produce heat mainly through UCP1-mediated thermogenesis. Lineage-tracing studies have demonstrated that beige adipocytes consist of heterogeneous subpopulations of distinct origins depending on the nature of external induction stimuli (Berry et al., 2016). For example, beige adipocytes originate either from PDGFRα⁺ stromal progenitor cells (Han et al., 2021; Lee et al., 2015, 2012), or from mural perivascular cells when targeted by SMA-CreERT (Long et al., 2014). These two beige progenitor populations are not mutually exclusive. A recent study noted that PDGFRα⁺ progenitor cells are required for developmental adipogenesis, but not for adult beige adipogenesis (Shin et al., 2020). Further evidence showed that beige precursor cells are a population distinct from brown or white fat precursors with selective cell-surface markers, including CD81 (Oguri et al., 2020) and CD137 (TNFRSF9) (Wu et al., 2012).

Following cell fate determination and differentiation, beige adipocytes possess inducible and reversible thermogenic capacity, as well as plasticity in cellular morphology in response to environmental stimuli (Paulo and Wang, 2019). The first wave of de novo beige adipocytes express UCP1 with multilocular morphology. They appear in WAT of mice independently of temperature conditions at the peri-weaning stage of development (Wu et al., 2020). At the adult stage, these beige adipocytes regress, lose UCP1 expression and are morphologically identical to white adipocytes at room temperature. The dormant beige adipocytes can reappear in response to cold challenge, and reverse back to white-like cellular morphology in conditions of re-warming or withdrawal of β3-AR agonist (Roh et al., 2018; Rosenwald et al., 2013). Previous studies have proposed that cold-induced beige adipocytes predominantly arise from transdifferentiation of pre-existing white adipocytes (Barbatelli et al., 2010; Cattaneo et al., 2020; Rosenwald et al., 2013). However, this view requires cautious assessment because of the current technical limitations to determine how many of those pre-existing ‘white’ adipocytes are de facto latent beige adipocytes with UCP1⁺ history. Of note, the conversion between thermogenically latent and active states in beige fat cells may be attributed to reversible processes of mitochondrial biogenesis and clearance (Altshuler-Keylin et al., 2016), as well as to chromatin reprogramming and function of specific transcriptional machinery (Roh et al., 2018).

Despite these differences in developmental origins and physiological functions of brown, beige and white adipocytes, these cell types share a similar transcriptional cascade that controls the process of fat cell differentiation. Several factors are commonly employed by both brown and beige adipocytes, including Cebpα/β-PPARγ cascades for early cell fate commitment and differentiation, as well as activators of EBF2, Prdm16 and PGC1α (PPARGC1α) for thermogenesis (Shapira and Seale, 2019; Wang and Seale, 2016). However, as mentioned above, beige adipocyte plasticity in cell differentiation and thermogenesis involves cell-specific changes in morphology, transcription and the chromatin landscape. In contrast to our significant understanding of brown adipocyte transcription, the beige-selective regulatory machinery remains largely unclear.

Forkhead box P4 (FOXP4) typically functions as a transcription factor that regulates islet α cell proliferation (Spaeth et al., 2015), breast cancer invasion (Ma and Zhang, 2019) and speech/language (Snijders Blok et al., 2021). Our previous studies have shown that FOXP4 also controls endochondral ossification in a complex with FOXP1 and FOXP2 (Zhao et al., 2015). In this study, we employed SMA-Cre^ERT and AdipoQ-Cre mice to knock out Foxp4 in beige precursors and differentiated beige cells, respectively. Inactivation of Foxp4 in progenitor cells impaired beige fat cell differentiation. In contrast, Foxp4 deficiency in differentiated beige adipocytes exacerbated their cold-induced thermogenesis response. Mechanistically, we found that FOXP4 directly repressed transcription of Pdgfra, Pgc1a and Cebpb, thereby acting as a transcriptional ‘brake’ on these crucial components of beige adipocyte regulation. Together, our results suggest that FOXP4 not only primes early cell fate commitment of beige adipocytes, but also attenuates their cold-induced thermogenesis response.

To investigate the expression pattern of FOXP4 in adipose tissues, two representative subpopulations of adipocytes, interscapular brown adipose tissues (BAT) and subcutaneous white adipose tissues (sWAT), were obtained from 8-week-old wild-type C57BL/6J mice that were housed at room temperature (23°C). High levels of FOXP4 expression were detected within brown and white adipocytes by immunofluorescence analyses (Fig. 1A). Next, stromal vascular fraction (SVF) cells isolated from sWAT of wild-type mice were induced to beige adipocytes in vitro by culturing in beige adipogenic media for 7 days. As expected, Pparg and Ucp1 expression levels were elevated during beige adipocyte differentiation, whereas Foxp4 expression peaked at day 2 of induction and declined swiftly thereafter (Fig. 1B). In addition, western blotting and qPCR analysis revealed that FOXP4 expression in BAT and sWAT was slightly increased in response to cold exposure (Fig. S1A,B). These dynamic alterations in expression indicate a phase-specific function of FOXP4 during beige adipocyte differentiation.

To investigate the function of FOXP4 in brown/beige adipocyte differentiation, we first transfected the SVF cells from Foxp4^fl/fl sWAT with either retroviral pMSCV-Cre or, as a control, retroviral pMSCV-GFP. This approach allowed us to achieve efficient and extensive inactivation of Foxp4 in adipocyte progenitors. Minor alterations in brown adipocyte differentiation and thermogenesis were detected in BAT-SVF cells with Foxp4 deficiency (designated Foxp4_pMSCV^Δ/Δ) (Fig. 1C,D). In contrast, beige adipocyte differentiation in Foxp4-deficient sWAT-SVF cells was impaired, as evidenced by Oil Red O staining and downregulation of a set of thermogenic (Pparg, Pgc1a, Adrb3, Ucp1, Cidea, Dio2, Elovl3) and beige-elective signature genes (CD137, Slc27a1) (Fig. 1E-G). Of note, deletion of Foxp4 by pMSCV-Cre had little effect on white adipogenesis, as evidenced by Oil Red O staining and qPCR based on sWAT-SVF adipogenic cultures without triiodothyronine (T3) and thiazolidinediones (Fig. S1C,D). Thus, FOXP4 expression is required for beige adipocyte early differentiation from stromal progenitor cells.

A major proportion of adult beige adipocytes is reported to be derived from mural progenitor cells within sWAT, which could be selectively and conditionally targeted by SMA-Cre^ERT (Long et al., 2014). Unfortunately, when Foxp4 was inactivated by SMA-Cre, the knockout mice showed defects in postnatal development and growth (data not shown). Therefore, tamoxifen (TM)-inducible Foxp4 conditional knockout mice at adult stage (hereafter designated Foxp4_Sma^ErtΔ/Δ) were generated by crossing Foxp4^fl/fl with SMA-Cre^ERT mice. SVF cells obtained from sWAT of Foxp4_Sma^ErtΔ/Δ were maintained for 2 days within cultures with 4-OH TM, followed by another 8 days in beige adipogenic differentiation (Fig. 2A). Beige cell differentiation was impaired following loss of Foxp4, as evidenced by Oil Red O staining (Fig. 2B). In addition, we observed decreased expression levels of Foxp4 (Fig. 2C), beige-specific markers (CD137, Klhl13, Slc27a1) (Fig. 2D) and thermogenic markers [Ucp1, Cidea, Elovl3, Cox2 (Ptgs2), Cox4i1, Cox5b, Cox7a] genes (Fig. 2E). The defects of beige cell differentiation may stem from perturbed de novo biogenesis. We next tracked PDGFRα⁺ stem cell antigen 1 (Sca1)⁺ adipocyte progenitor cells (APCs) within WAT-SVF 48 h post-TM administration by flow cytometry (FACS) as described previously (Berry and Rodeheffer, 2013; Lee et al., 2015). The frequency of APCs [defined as CD31 (Pecam1)⁻CD45 (Ptprc)⁻PDGFRα⁺Sca1⁺] was relatively lower in SVF cells of Foxp4_Sma^ErtΔ/Δ compared with those of Foxp4^fl/fl (Fig. 2F,G).

In addition, we administered three consecutive injections of TM to the Foxp4_Sma^ErtΔ/Δ mice, cultured them for 2 weeks at room temperature and exposed them to cold (4°C) for 3 days before the SVF cells were harvested for FACS analysis (Fig. 2H). The knockout efficiency of FOXP4 in SVF cells was validated by western blot (Fig. S1E). In vivo FACS analysis confirmed the reduction of adipocyte progenitor cells (Fig. 2I). qPCR analysis also showed the relative downregulation of Slc27a1 and Tbx1, two beige-selective marker genes (Fig. 2J). These observations indicate that FOXP4 inactivation decreases the APC population proportion, thus potentially impairing beige cell early differentiation.

Interestingly, qPCR analysis revealed that an array of collagen fibril-related transcripts was relatively enriched in Foxp4_pMSCV^Δ/Δ beige cells compared with controls. Elevated expression levels of several of these pro-fibrotic genes (Col1a1, Col3a1, Col4a1, Col6a3, Mmp2, Timp1, CD9) were detected in mutant beige cells (Fig. S2A). Accordingly, similar alterations of fibroblast marker genes were observed in beige cells obtained from sWAT-SVF of Foxp4_Sma^ErtΔ/Δ mice following Foxp4 inactivation induced by TM (Fig. S2B). These observations indicate that more Foxp4-deficient beige progenitor cells are arrested at fibroblast progenitor state.

PDGFRα⁺ stromal progenitor cells also give rise to beige adipocytes during early development (Gao et al., 2018). Activation of the PDGFRα pathway can also perturb the differentiation of beige adipocytes (Shin et al., 2020; Sun et al., 2017). We then investigated the impact of FOXP4 on Pdgfra gene transcription. qPCR analysis revealed that Pdgfra transcripts were increased in beige progenitor cells from Foxp4_pMSCV^Δ/Δ or in Foxp4_Sma^ErtΔ/Δ mice (Fig. 3A), as was the PDGFRα protein level (Fig. 3B). In addition, chromatin immunoprecipitation and sequencing (ChIP-seq) analysis of SVF progenitor cells detected relatively high enrichment of FOXP4-binding sites within the promoter region of Pdgfra (Fig. 3C, arrow). A putative FOXP4-binding site was detected at −107 upstream of the Pdgfra transcription start site (Fig. 3D). Luciferase reporter assays employing a Pdgfra promoter-driven luciferase as substrate validated that FOXP4 repressed transcription of Pdgfra in HEK293T cells (Fig. 3D). On the other hand, overexpression of FOXP4 by pMSCV retrovirus repressed PDGFRa expression in SVF progenitor cells (Fig. 3E). Together, these data suggest that FOXP4 controls the early differentiation of beige adipocytes by directly modulating Pdgfra transcription.

The first wave of de novo beige adipocyte biogenesis and UCP1 activation occurs within sWAT at the peri-weaning stage independently of temperature conditions (Wang et al., 2017; Wu et al., 2020). To evaluate the potential impact of Foxp4 deficiency on beige adipocyte thermogenesis, we eliminated Foxp4 in differentiated adipocytes with AdipoQ-Cre (Eguchi et al., 2011), hereafter designated Foxp4_AdipQ^Δ/Δ. We confirmed that FOXP4 was efficiently reduced at the mRNA and protein levels in BAT and sWAT from Foxp4_AdipQ^Δ/Δ mice (Fig. S3A-C). Foxp4_AdipQ^Δ/Δ mice appeared relatively normal in size and in the appearance of fat depots of BAT and sWAT at 3 weeks of age (Fig. 4A,B). Hematoxylin and Eosin (HE) staining and immunohistochemistry (IHC) analysis revealed that higher numbers of UCP1⁺ beige adipocytes resided in sWAT from Foxp4_AdipQ^Δ/Δ knockout mice than in controls (Fig. 4C). Consistent with that observation, qPCR analysis confirmed the upregulation of a set of thermogenic and mitochondrial signature genes (Dio2, Elovl3, Pgc1a, Ucp1, Cox2, Cox4i1, Cox5b, Cox8b) (Fig. 4D,E). However, the expression of several beige-selective marker genes (CD137, CD40, Klhl13, Tbx1) showed no obvious increases (Fig. 4F). These observations suggest that FOXP4 boosts juvenile beige adipocyte activation, but has little effect on their early biogenesis.

The transcriptional pathway underlying adult beige cell differentiation and thermogenic activation is distinct from that of the juvenile at the peri-weaning stage (Wu et al., 2020). Foxp4_AdipQ^Δ/Δ knockout mice appeared normal in size, weight and adiposity compared with controls at adult stages (Fig. 5A). They were modestly smaller at 5 months (Fig. 5B). Accordantly, the thermogenic program in the BAT of Foxp4_AdipQ^Δ/Δ mice did not appear to be activated at ambient temperature (20-22°C), as indicated by HE staining and IHC, as well as by qPCR analysis of a set of thermogenic marker transcripts (Fig. S3D,E). Indirect calorimetry analysis by CLAMS (Comprehensive Lab Animal Monitoring System) revealed that Foxp4 loss had no impact on O₂ or CO₂ consumption, as well as on energy expenditure following identical diet and locomotor activity as controls at 20∼22°C (Fig. S4). In contrast, thermogenic activation of beige adipocytes was mildly exacerbated in sWAT of Foxp4_AdipQ^Δ/Δ mice, as evidenced by relatively enriched levels of UCP1⁺ adipocytes and by upregulation of thermogenic or mitochondrial signature genes (Pparg, Dio2, Cidea, Ucp1, Cox2, Cox4i1, Cox5b, Cox8b) (Fig. 5C,D). We conclude from these analyses that loss of FOXP4 results in modest augmentation of thermogenesis in beige adipocytes at ambient temperature.

Beige adipocyte thermogenesis could combat obesity in human (Lidell et al., 2013). To examine the impact of Foxp4 deficiency on long-term adipose tissue metabolism, mice were fed a high-fat diet (HFD) for 3 months. We observed that Foxp4_AdipQ^Δ/Δ mice were leaner in body, with fewer adipose depots (Fig. S5A,B), and gained less body weight and adiposity than littermates after 12 weeks of HFD feeding starting at the age of 2 months (Fig. S5C). Foxp4_AdipQ^Δ/Δ mutant mice also retained more efficient glucose tolerance following HFD feeding, as evidenced by glucose tolerance test scores (Fig. S5D). Of note, as previously mentioned, UCP1-mediated beige thermogenesis was mildly diminished in Foxp4_AdipQ^Δ/Δ mutant mice at room temperature (Figs 4C,D and 5C,D), which may undermine the protective effect against HFD-induced obesity provided by Foxp4 deficiency.

To examine a potential cell-autonomous effect of FOXP4 on beige adipocyte energy metabolism, SVF cells were isolated from sWAT depots of Foxp4_AdipQ^Δ/Δ and control mice and then induced for beige adipocyte differentiation in vitro. We observed advanced beige adipocyte thermogenesis in Foxp4-deficient SVF progenitors, as evaluated by elevated mRNA levels of a set of thermogenic genes (Cidea, Pgc1a, Ucp1, Cox2, Cox7a1, Cox5b, Cox8b) (Fig. 5E,F). In addition, oxygen consumption rates (OCR) of beige adipocytes from Foxp4_AdipQ^Δ/Δ mutant mice exhibited higher basal and uncoupled OCR compared with controls (Fig. 5G,H). These results indicate that loss of FOXP4 led to elevated mitochondrial respiration. The thermogenic potency of SVF-derived brown adipocyte was also potentiated in Foxp4-deficient mice, as evidenced by qPCR and OCR analyses (Fig. S3F-I). This line of evidence indicates that FOXP4 suppresses beige adipocyte thermogenesis in a cell-autonomous manner.

We demonstrated that beige adipocytes could be mildly activated within sWAT at room temperature at the adult stage. We then examined beige adipocyte thermogenesis and activation in Foxp4_AdipQ^Δ/Δ mice under cold conditions. Compared with controls, Foxp4 knockout mice had relatively higher rectal temperature during a 6-h 4°C exposure (Fig. 6A). Foxp4_AdipQ^Δ/Δ mice also displayed a ‘browning’ feature of sWAT after 1 week at 4°C (Fig. 6B). This result was consistent with HE staining and IHC results demonstrating higher levels of UCP1⁺ beige adipocytes (Fig. 6C). Elevated beige adipocyte thermogenesis was validated by RT-qPCR analysis of signature genes (Cidea, Dio2, Elovl3, Pgc1a, Ucp1, Cox2, Cox5b, Cox8b) (Fig. 6D). However, long-term cold exposure had no significant effect on beige adipocyte de novo biogenesis in mutant mice, as evidenced by compromised levels of beige signature genes (Klhl13, Slc27a1, Tbx1, CD137) (Fig. 6E). Consistent with these observations, transmission electronic microscopy (TEM) and qPCR analysis revealed relative enrichment of mitochondria within beige adipocytes of Foxp4_AdipQ^Δ/Δ sWAT (Fig. 6F,G). We further observed that cold exposure had little effect on thermogenesis in the BAT of Foxp4_AdipQ^Δ/Δ mice (Fig. S6A,B). Together, these results suggest that FOXP4 acts as a repressor of the thermogenic gene program in beige adipocytes.

Thermogenesis of beige adipocytes could also be activated by adrenergic signaling (Jiang et al., 2017). However, when treated with 7 consecutive daily injections of the β3-AR agonist CL-316,243 Foxp4_AdipQ^Δ/Δ mice exhibited no evident activation of the thermogenic program and no increase in beige adipocyte biogenesis, as determined by HE staining, IHC and RT-qPCR analyses (Fig. S7). These findings suggest that FOXP4 controls cold-induced beige adipocyte activation through adrenergic-independent signaling.

To explore the molecular mechanism underlying the impact of FOXP4 on beige adipocyte thermogenesis and activation, RNA-sequencing (RNA-seq) analysis was performed on beige adipocytes derived from sWAT-SVF of Foxp4_AdipQ^Δ/Δ mice. As shown in Fig. 7A,B, expression of an array of thermogenic or mitochondrial gene markers was elevated in Foxp4-deficient beige adipocytes. Next, we conducted ChIP-seq analysis of SVF-induced beige adipocytes by employing anti-FOXP4 and anti-H3K27Ac antibodies for pulldowns. As shown in Fig. 7B, 22 common target genes were detected by both RNA-seq and ChIP-seq, including the classical thermogenic genes Pgc1a and Cebpb. Western blotting confirmed the upregulation of PGC1α, Cebpβ and UCP1 at the protein level under conditions of decreased FOXP4 (Fig. 7C). Promoter occupancy as determined by ChIP-seq validated the relative enrichment of FOXP4-binding sites within the chromatin of Pgc1a and Cebpb upstream promoter regions (Fig. 7D, arrows). In support of this, reporter assays employing a Pgc1a promoter-driven luciferase vector revealed that FOXP4 repressed the transactivation ability of PPARγ in 3T3-L1 cells (Fig. 7E). Moreover, an array of putative FOXP4-binding site-enriched genes, such as Pgk1, Arrdc3, Pim1, Etfbkmt, Ggnbp1, were altered in Foxp4-deficient beige adipocytes (Fig. 7B). Gene ontology enrichment analysis showed that they mostly function in energy metabolism, thermogenesis or mitochondrial metabolism (Fig. S8). These findings demonstrate that FOXP4 restrains beige adipocyte thermogenesis by directly repressing the thermogenic program, in particular the expression of Pgc1a and Cebpb.

Beige fat cells harbor distinctive molecular signatures from white and brown adipocytes during development. Once committed from progenitor cells, differentiated beige adipocytes demonstrate a plastic morphology that is tightly coupled with their reversible thermogenic potency. In this study, we dissected the phase-specific role of transcription factor FOXP4 in beige fat cell differentiation and thermogenesis. Foxp4 deletion impaired the early differentiation of beige adipocytes from progenitor cells, but exacerbated cold-induced thermogenesis in latent beige adipocytes. Our findings indicate that FOXP4 primes beige cell differentiation, but acts as a ‘brake’ for beige adipocyte thermogenesis under cold challenge (Fig. 7F).

To evaluate the impact of FOXP4 on beige cell differentiation, two independent knockout models were employed in our study. In our first model, we employed pMSCV-Cre to delete Foxp4 in all adipocyte precursor cells. Loss of Foxp4 led to defective beige adipogenesis, but not brown or white adipogenesis, during differentiation from progenitor cells. At the adult stage, the majority of cold-induced beige adipocytes arise from mural progenitor cells within vascular compartments of WAT (Long et al., 2014; Shamsi et al., 2021). As our second model, TM-inducible, Cre-mediated recombination was used to delete Foxp4 in SMA⁺ SVF cells prior to beige adipocyte differentiation. FACS analyses in vivo and in vitro for Foxp4-knockout SVF cells revealed a smaller cell proportion of adipocyte progenitor cells in the knockout mice compared with wild-type controls. Of note, the in vivo knockout of Foxp4 by TM injections in SMA-Cre^ERT mice did not exhibit an obvious deficiency in beige adipocyte biogenesis, unlike in vitro knockout in SVF cultures by 4-OH TM or by pMSCV-Cre. In addition, expression levels of general adipogenesis markers (Cebpa/b/g, Fabp4, Ppparg) were not evidently altered in SMA⁺ beige cells with Foxp4 deficiency (Fig. S2C,D). This may be due to the targeting efficacy by SMA-Cre^ERT. It was previously shown that SMA-Cre^ERT labeling could account for ∼60% of cold-induced beige adipocytes (Berry et al., 2016). Taken together, these experiments demonstrated that FOXP4 was required for early adipocyte differentiation, especially for beige cell biogenesis.

PDGFRα-positive progenitor cells within sWAT harbor multiple potency in cell differentiation. They are precursor cells for both white and beige adipocytes (Gao et al., 2018), as well as for fibroblasts (Cattaneo et al., 2020). PDGFRα-positive progenitors give rise to beige adipocytes in response to β3-adrenoceptor activation and high-fat diet (Lee et al., 2012), rather than to cold induction (Berry et al., 2016; Shin et al., 2020). In addition, PDGFRα expression levels seem to be precisely controlled during adipogenesis. PDGFRα expression precedes beige adipocyte differentiation (Gao et al., 2018). Its continuous expression or activation in sequential phases drive progenitor cells towards a fibroblastic cell fate (Iwayama et al., 2015; Marcelin et al., 2017; Sun et al., 2017). In line with those findings, Foxp4-deficient SVF cells display fibroblast signatures, as well as increased Pdgfra expression. When combined with our ChIP-seq and luciferase reporter data, the evidence suggests that FOXP4 primes beige adipocyte differentiation, partially by modulating Pdgfra expression levels. Of note, we cannot exclude the possibility that the higher Pdgfra expression may be partially due to more residual PDGFRα⁺ undifferentiated cells in cultures with FOXP4 deletion. The observations of PDGFRα expression may be a combined output from these two pathways.

Prior to the present report, only a few beige-selective transcription factors had been characterized, including MRTFA (McDonald et al., 2015) and Tbx1 (Wu et al., 2012). FOXP4 is expressed in both BAT and sWAT of mice. However, our genetic analysis showed that deletion of Foxp4 in adipocytes had little effect on in vivo BAT development and thermogenesis. Only oxygen consumption and expression of several thermogenic marker genes were slightly increased in in vitro cultures of SVF-derived brown adipocytes from Foxp4-deficient mice. Upon cold exposure and adrenoceptor agonist stimulation, we observed no significant differences in vivo in BAT thermogenesis. Its dynamic expression level in beige adipocytes during cell differentiation and cold exposure suggests that FOXP4 is a selective regulator of beige adipocyte development and cold-induced thermogenesis.

The thermogenic program of beige adipocytes can be activated through various pathways (Barbatelli et al., 2010; Rosenwald et al., 2013). Most beige adipocytes are activated through the β3-adrenergic signaling pathway (Lee et al., 2015). However, it has also been reported that cold-induced activation of beige adipocytes requires the β1 adrenergic receptor (Adrb1), but not the β3 adrenergic receptor (Adrb3) (Jiang et al., 2017). Recently, a glycolytic beige population was identified that could be induced by chronic cold adaptation in the absence of β-adrenergic receptor signaling (Chen et al., 2019). Excess calorie intake also can trigger the activation of CHRNA2-dependent beige adipocytes (Jun et al., 2020), which also are glycolytic and β-adrenergic signaling independent. Interestingly, Foxp4-inactivated beige adipocytes appeared to only react to cold exposure, not to an adrenoceptor agonist (Fig. S7C,D), and had elevated expressions of glycolytic marker genes compared with controls in response to cold exposure (Fig. S6C,D). This suggests that FOXP4 controls beige cell thermogenic activation through an adrenergic signaling-independent pathway, perhaps partially through a glycolytic pathway.

A recent report pointed out that different transcriptional machinery governs beige adipocyte development and activation between peri-weaning and adult stages (Wu et al., 2020). However, we observed that inactivation of Foxp4 only slightly exacerbated UCP1-mediated thermogenesis at both stages. Thus, we suggest that FOXP4 is shared as a regulator of arresting thermogenesis in both juvenile and mature beige adipocytes. This view is consistent with previous studies from our laboratory that showed that FOXP1, a highly conserved paralog of FOXP4, repressed both beige adipocyte differentiation and UCP1-mediated thermogenesis (Liu et al., 2019). Given that FOXP1 and FOXP4 form dimers in various tissues (Li et al., 2012, 2004; Sin et al., 2015), we cannot exclude the possibility of a cooperative function of FOXP1 and FOXP4 in controlling beige cell thermogenesis.

Collectively, the data presented here reveal a role of FOXP4 in beige adipocytes differentiation and cold-induced thermogenesis. We suggest that a more thorough understanding of the underlying causes of FOXP4-selective functions will allow us to manipulate beige cells specifically to improve systemic energy metabolism and to combat obesity.

The Foxp4^fl/fl constructed by our lab has been described in previous studies (Zhao et al., 2015). Adiponectin-Cre was obtained from The Jackson Laboratory (Stock no. 028020). SMA-CreERT mice was kindly provided by Prof. Gang Ma (Shanghai Jiaotong University, China). The genetic backgrounds of all knockout mice were C57Bl/6J. Mice were bred with standard rodent chow food or HFD. For cold treatment, mice were bred in a 4°C environment for a week. For continuous β-adrenergic stimulation, mice were injected intraperitoneally with CL-316,243 (0.75 mg/kg) each day for up to 7 days. Male mice were used in the experiments unless otherwise indicated. The experiments were not randomized, and the investigators were not blinded to allocation during experiments or outcome assessments. Mice are maintained under a constant environmental temperature (22°C) and a 12-h light/12-h dark cycle. All animal experiments were performed according to the ethical guidelines of Shanghai Jiao Tong University (SYXK 2011-0112).

Minispec TD-NMR Analyzers (Bruker Instruments) were used to evaluate adiposity composition on anesthetized animals. Food intake, energy expenditure, O₂ consumption, CO₂ production and physical activity were measured using an indirect calorimetry system (CLAMS; Oxymax, Columbus Instruments), installed under a constant environmental temperature (22°C) and a 12-h light (07:00-19:00 h), 12-h dark cycle (19:00-07:00 h). Mice in each chamber had free access to food and water. The raw data were normalized by body weight and the histograms of day (07:00-19:00 h) and night (19:00-07:00 h) show the mean value of all points measured during the 12-h period.

Adipose tissues were fixed in 4% paraformaldehyde for 16 h at 4°C, embedded in paraffin or tissue freezing medium (Leica) and sectioned at 4 μm. HE staining was conducted according to standard protocols. For immunofluorescence, heat-induced antigen retrieval with sodium citrate buffer (10 mM sodium citrate, 0.05% Tween 20, pH 6.0) was performed before sections were blocked with 5% bovine serum albumin in Tris-buffered saline with 0.1% Tween 20 detergent (TBST) (pH 7.6) for 30 min at 37°C, then incubated overnight at 4°C with primary antibodies to mouse Foxp4 (Millipore, ABE74, 1:100) or UCP1 (Abcam, ab10893, 1:100). Subsequently, sections were incubated with secondary fluorescence-conjugated or horseradish peroxidase (HRP)-conjugated antibodies at 37°C for 30 min in the dark. Samples were imaged using a Leica TCS SP5 confocal microscope, Leica DM2500 microscope or Leica 3000B microscope. TEM of beige adipose tissue was carried out in accordance with a previous study (Liu et al., 2019).

For SVF cell isolation, primary BAT and sWAT were digested with 1 mg ml⁻¹ collagenase type I (Sigma-Aldrich) in DMEM (Invitrogen) supplemented with 1% bovine serum albumin for 25 min at 37°C, followed by filtration and density separation by centrifugation (10 min at 250 g). The freshly isolated SVF cells were seeded and cultured in growth medium containing DMEM, 20% fetal bovine serum (FBS), 1% penicillin/streptomycin at 37°C with 5% CO₂ for 3 days, followed by feeding with fresh medium every 2 days to reach confluence. For brown/beige adipocyte differentiation, the cells were induced with induction medium containing DMEM, 10% FBS, 5 μg ml⁻¹ insulin, 0.5 mM isobutylmethylxanthine (Sigma-Aldrich), 1 μM dexamethasone (Sigma-Aldrich), 50 nM T3 (Sigma-Aldrich) and 5 μM troglitazone (Sigma-Aldrich) for 48 h, and further in growth medium supplemented with insulin, T3 and troglitazone for 6 days, followed by 0.5 mM cyclic AMP (Sigma-Aldrich) treatment for another 4 h. For inducible knockout in beige cells, SVF from Foxp4_Sma^ErtΔ/Δ and Foxp4^fl/fl mice were expanded for 2 days with growth medium, then treated with 4-OHTM at a final concentration of 1 µM in cultures for 48 h, followed by 6-day differentiation culture without 4-OHTM. HEK293T cells (ATCC) were cultured in DMEM with 10% FBS. For Oil Red O staining, cultured cells were washed with PBS and fixed with 10% formaldehyde for 15 min at room temperature. The cells were then stained using 1 ml Oil Red O working solution (5 g/l in isopropanol) and 4 ml H₂O for 30 min. After staining, the cells were washed with 60% isopropanol and imaged.

Two-month-old Foxp4 knockout mice by SMA-CreERT were injected once intraperitoneally with TM dissolved in sunflower oil (Sigma-Aldrich, 100 mg/kg). After 48 h, SVF cells were isolated from mice euthanized with CO₂ according to protocols described above. Progenitor cells from SVF were expanded for 2 days in growth cultures and treated with 4-OHTM (a final concentration of 1 µM) for 24 h. Those cells were then cultured with fresh medium without TM for another 24 h before being collected for FACS analysis. 4-OHTM-administrated SVF cells were suspended in FACS buffer to a final concentration of 10⁵-10⁶ cells/100 μl, incubated on ice for ∼30-45 min with combinations of PerCP-CD45 (BioLegend, 103131), PerCP-CD31 (BioLegend, 102419), APC-Sca-1 (BioLegend, 122511), PE-CD140α (BioLegend, 135905), CD24-FIFC (BioLegend, 101815) antibodies. The cells then washed twice with PBS before analysis on a BD FACSCalibur. The proportion of adipocyte progenitors was analyzed with FlowJo v10 software.

Primary SVF cells from BAT and sWAT were isolated and cultured for 4 days before being plated in XF cell culture microplates (Seahorse Bioscience). SVF cells (10,000 cells) were seeded in each well, and each treatment included cells from three BAT or sWAT replicates. After 6 days of differentiation, cultured adipocytes were washed twice and pre-incubated in XF medium (supplemented with 25 mM glucose, 2 mM glutamine and 1 mM pyruvate) for 1-2 h at 37°C without CO₂. The OCR was measured using an XF Extracellular Flux Analyzer (Seahorse Biosciences). Oligomycin (2 mM), FCCP (2 mM), and antimycin A (0.5 mM) were preloaded into cartridges and injected into XF wells in succession. OCR was calculated as a function of time (pmoles per minute per µg protein).

Luciferase assays were performed in HEK293T (National Collection of Authenticated Cell Cultures, SCSP-502) or 3T3-L1 (#SCSP-5038) cells. The reporter plasmid Pgc1α-Luc, containing a 2.6 kb fragment of the promotor region of the Pgc1a gene, was obtained from Dr JiQiu Wang of the Ruijin hospital (Shanghai, China). Pdgfrα-Luc plasmid, containing 1.3 kb of the 5′ flanking region of the Pdgfra gene was constructed by our lab. The primers used for amplification are shown in Table S1. The expression plasmids of Foxp4 and Pparg were constructed into the pcDNA3.0 vector. Cells were transfected using FuGENE HD (Promega) in 24-well plates. The transfection amount of each plasmid was 200 ng, and the total amount of transfected DNA across each transfection was balanced by pcDNA3.0 plasmids when necessary. After 36 h, a dual luciferase assay was performed according to the manufacturer's protocols (Promega).

We used TRIzol (Vazyme, R401) for total RNA extraction according to the manufacturer’s instructions. Extracted RNA (1 µg) was converted into cDNA using the HiScript III SuperMix for qPCR (Vazyme, R323-01). Quantitative RT-PCR (qRT-PCR) was performed using a LightCycler 480 II (Roche) and SYBR Green PCR Master Mix (Vazyme, Q711-02). Fold change was determined by comparing target gene expression with the reference gene β-actin. The primers used for qRT-PCR are shown in Table S1. For RNA-seq, total RNA extracted from sWAT with TRIzol was used for library construction and RNA sequencing (Personal Biotechnology Co., Shanghai, China).

For western blotting, adipose tissue was homogenized and lysed with RIPA (Beyotime, P0013B). Protein samples were incubated with primary antibodies against Foxp4 (Millipore, ABE74, 1:1000), Ucp1 (Abcam, ab10893, 1:1000), Cebpβ (Santa Cruz Biotechnology, sc-150, 1:500), PGC1α (Abways, CY6630, 1:1000), β-actin (Selleck Chemicals, A1016, 1:2000) at 4°C overnight. Proteins were visualized using HRP-conjugated secondary antibody and chemiluminescent HRP substrate (Millipore).

Magna ChIP Protein A+G Magnetic Beads (20 µl; Millipore, 16-663) were washed twice with 1 ml FA buffer (10 mM HEPES, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1%S sodium deoxycholate, 0.1% SDS and protease inhibitors). Beads were suspended with 1 ml FA buffer, and 4 μg antibody was added to the beads, which were then rotated for at least 2 h. Differentiated SVF cells were cross-linked using 1% formaldehyde in PBS at room temperature with rotation. Cells were then incubated with 125 mM glycine (Sangon, A610235-0500; 62.5 μl 2 M glycine/ml PBS) at room temperature for 10 min to stop cross-linking. After two washes with ice-cold PBS, cells were collected and diluted in 0.5 ml FA buffer. Cells were sonicated using a Sonics VCX130 with the following parameters: 5 s on; 10 s off; 6 min to make the DNA fragment at 300-500 bp. The sonicated solution was centrifuged at 13,000 rpm (16,200 g) for 5 min and the supernatant collected. The beads bound to antibody were then washed twice with FA buffer. Antibody-coated beads were added into the sonicated supernatant for 12-16 h at 4°C with rotation. The beads were then washed with FA buffer once, then with the following buffers (twice each): high salt buffer (10 mM HEPES, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, 0.1% SDS); LiCl buffer (10 mM Tris-HCl, pH 8.0, 0.25 M LiCl, 1% NP-40, 1 mM EDTA, 0.1% sodium deoxycholate); TE buffer (10 mM Tris-HCl, pH 7.5, 1 mM EDTA). The beads were then suspended with 270 elution buffer (50 mM Tris-HCl, pH 7.5, 1 mM EDTA, 1% SDS) and eluted by votexing at 900 rpm for 30 min on an Eppendorf Thermo Mixer at 68°C. After elution, 130 µl TE buffer containing 3 µl RNase A (Thermo Fisher, EN0531) was added into the eluate, and incubated at 37°C for 30 min. Then, 5 μl Proteinase K (Thermo Fisher, AM2546) was added and incubated at 65°C overnight to reverse cross-links. DNA was isolated and purified with a ZYMO DNA Clean & Concentrator (Zymo Research, D4013).

Libraries were constructed using a VAHTS Universal DNA Library Prep Kit for Illumina V3 (Vazyme, ND607) according to the manufacturer's protocol. Isolated Chip DNA was sequentially subjected to end repair/phosphorylation/A-tailing adding, and index adaptor ligation. AMPure XP beads (Beckman Coulter, A63880) were used for a post-ligation cleanup, DNA was eluted from beads and amplified by PCR for ten cycles. DNA sized between 200 and 700 bp was selected by a double-sided size selection strategy with AMPure XP beads. After elution with 10 mM Tris (pH 8.0), libraries were analyzed using the Qubit and Agilent 2100 Bioanalyzer, pooled at a final concentration of 12 pM and sequenced on a HiSeq2500. For ChIP-seq analysis, demultiplexed ChIP-seq reads were aligned to the mm10 mouse genome using Bowtie2 (Langmead and Salzberg, 2012) with the parameter ‘--no-discordant --no-unal --no-mixed’. PCR duplicates and low-quality reads were removed by Picard. Reads were processed using Samtools (Li et al., 2009) and subjected to peak-calling with MACS2 (Zhang et al., 2008) with a parameter ‘except -f BAMPE -p 0.01’. We converted the bam file to a bw file using deeptools (v3.3.0). Integrative Genomics Viewer (IGV, v2.7.2) was used for peak visualization. Overlaps between ChIP-seq and RNA-seq were performed and Venn diagrams were drawn in R.

For the glucose tolerance test, mice were given an intraperitoneal injection of 100 mg/ml D-glucose (2 g/kg body weight) after overnight fasting, and tail blood glucose concentrations were measured using a glucometer (AccuCheck Active, Roche).

For RNA-seq analysis, sequencing reads were filtered using Cutadapt and aligned to the mm10 mouse genome using HISAT2 (Kim et al., 2015). Filtered reads were assigned to the annotated transcriptome and quantified using HTSeq (Anders et al., 2015); FPKM was used as normalization method. The analysis below was all performed in R. We used DESeq for differential expression analysis (Anders and Huber, 2010). Genes were considered significant if they passed a fold change (FC) cutoff of |log2FC|>0.585 and a false discovery rate (FDR) cutoff of FDR<0.05%. Heatmap in R was used for gene expression cluster analysis and heatmap visualization. We used topGO for gene ontology analysis (Alexa et al., 2006). clusterProfiler was used for KEGG and GSEA analysis with default parameters (Yu et al., 2012).

For statistical analysis, all data are presented as mean±s.d. Two-tailed Student's t-tests were used for comparisons between two groups and two-way ANOVA for more than two groups.

We thank Prof. Haley O. Tucker from the University of Texas for editing the manuscript.