RNA Binding Protein Ybx2 Regulates RNA Stability During Cold-Induced Brown Fat Activation

Obesity has reached an epidemic scale in many countries, resulting in a steep escalation in health care expenditures and a growing burden of chronic obesity-related morbidities (1). An attractive approach to improve metabolic health is to augment the mass and activity of brown adipose tissue (BAT) (2–7). There are at least two types of thermogenic adipocytes in mammals, namely, classic brown adipocytes and inducible/beige adipocytes. Classic BAT is located as a discernible depot in the interscapular region in small mammals and human infants. Beige/inducible adipocytes exist in defined anatomical white adipose tissue (WAT) depots, particularly in subcutaneous WAT, and express a gene program more like WAT at thermoneutrality. In response to prolonged cold exposure, chronic treatment of β-adrenergic receptor agonist, or intensive exercise, the number of beige adipocytes dramatically increases, accompanied by enhanced Ucp1 levels and mitochondria biogenesis, a process known as “browning” (2,5,6).

Understanding the detailed mechanisms underlying BAT differentiation and function is an area of immense research interest. A vast array of factors has been identified that regulate BAT development and activity by acting at the transcriptional level (6–16). How these processes are regulated at the posttranscriptional level, however, has yet to be fully understood. RNA binding proteins (RBPs) comprise a large and diverse group (17,18) that lie at the center of posttranscriptional regulation by governing the fate of mRNA transcripts from biogenesis, stabilization, and translation to RNA decay. Several RBPs have been reported to modulate adipocyte development and lipid metabolism. SFRS10 (splicing factor arginine/serine-rich10) inhibits lipogenesis by controlling the alternative splicing of LPIN1, a key regulator in lipid metabolism (19,20). Sam68 (the Src-associated substrate during mitosis of 68 kDa) is required for WAT adipogenesis by regulating mTOR alternative splicing (21). Knockout of KSRP (KH-type splicing regulatory protein) promotes browning of WAT by reducing miR-150 expression (22). IGF2 mRNA binding protein 2 (IGF2BP2) is a widely expressed RBP, and a single nucleotide polymorphism in its intron is associated with type 2 diabetes by genome-wide association studies (23). Knockout of IGF2BP2 results in resistance to diet-induced obesity, largely resulting from an enhanced translational efficiency of Ucp1 and other mitochondria mRNAs in the knockout BAT (24). Paraspeckle component 1 (PSPC1) was recently identified as an essential RBP for adipose differentiation in vitro and in vivo by regulating the export of adipogenic RNA from the nucleus to the cytosol (25). Despite these advances, our understanding of RBPs in adipocytes, particularly in brown adipocytes, is still at its early stage, and the functions of most RBPs remain unknown.

In this study, we systemically profiled 413 RBPs in different fat depots, during white fat browning and brown adipogenesis, and identified 5 BAT-enriched RBPs. We demonstrated the role of Y-box binding protein 2 (Ybx2) in the development and activation of BAT in vitro and in vivo, which could be, at least partially, explained by stabilizing mRNA.

The animal experimental protocols in this study were approved by the Singapore SingHealth Research Facilities Institutional Animal Care and Use Committee. Ybx2 heterozygous mice (NSA [CF-1] background) were originally imported from Dr. Paula Stein (University of Pennsylvania). C57BL6 mice were obtained from The Jackson Laboratory and subsequently bred in house. All mice were maintained at the animal vivarium at Duke-NUS Medical School. For cold challenge experiments, animals were housed individually in a 4°C chamber for 6 h. The rectal body temperature was recorded with a probe thermometer (Advance Technology) at a constant depth.

Glucose tolerance tests and insulin tolerance tests were performed as previously described (26), and EchoMRI was used to measure fat and lean mass. For the in vivo insulin-signaling study, Ybx2 knockout (KO) and wild-type (WT) mice were fasted for 6 h at room temperature or 4°C. Then the mice were injected with insulin (1 unit/kg body wt). Mice were sacrificed after 5 min, and BAT was collected. Lipolysis assay was performed as previously described (26).

293T cells for retroviral packing were cultured in DMEM containing 10% FBS (HyClone). Primary brown and white preadipocytes were isolated from 3- to 4-week-old C57BL6 mice. The procedure for preadipocytes isolation, culture, and differentiation and Oil Red O staining was described previously (26). Human primary interscapular brown adipocytes were obtained from Zenbio Inc. and cultured and differentiated as previously described (27).

A murine stem cell virus (MSCV)–based retroviral vector (MSCV-pgkGFP-U3-U6P-Bbs vector) (28) was used to generate short hairpin (sh)RNAs to infect preadipocytes; XZ201 vector (29) was used to overexpress Ybx2 for gain-of-function studies. All of the retroviruses were packaged in 293T cells with the pCL-eco packaging vector and then used to transduce preadipocytes in the presence of 4 mg/mL Polybrene (Sigma-Aldrich), followed by induction of differentiation. FuGENE 6 Transfection Reagent (Promega) was used for plasmid transfection according to the manufacturer’s instructions.

Primary brown and white adipocytes were infected with retroviral Ybx2 and differentiated for 4 days. RNA immunoprecipitation (RIP) was performed using the Magna RIP kit (Merck Millipore) according to the manufacturer’s instructions. RNA samples retrieved from anti-Ybx2 (Abcam) and IgG control with the Magna RIP kit were used for RNA sequencing (RNA-seq).

RNA pull-down was performed according to our published protocol with a few modifications (30). In this study, we used tissue lysate from mouse BAT instead of primary cell culture prepared as described above. The tissue lysate was prepared as described in rna immunoprecipitation. The rest of the experiment followed our published protocol (30).

Primary brown preadipocytes were seeded in an X-24 cell culture plate, infected by retroviral constructs, as indicated in the text, followed by induction of differentiation. Differentiated cells were analyzed by Extracellular Flux Analyzer (Seahorse Bioscience) according to the manufacturer's instructions. Oxygen consumption rates were normalized by protein concentration.

Animals were kept at 4°C for 6 h before experiments. BAT and skeletal muscle (gastrocnemius) were harvested and minced with a micromincer (Glen Mills Inc.). The minced tissue was kept in ice-chilled mitochondrial respiration media (MiR05) (EGTA, 0.5 mmol/L; MgCl₂⋅6H₂O, 3 mmol/L; lactobionic acid, 60 mmol/L; taurine, 20 mmol/L; KH₂PO₄, 10 mmol/L; HEPES, 20 mmol/L; d-sucrose, 110 mmol/L; and BSA, 1 g/L). Tissue lysate, 2 mg and 10 mg, respectively, was immediately loaded into Oroboros Respirometry together with substrates, including glutamate (10 mmol/L), malate (2 mmol/L), pyruvate (5 mmol/L), and ADP (5 mmol/L). The oxygen consumption rate (OCR) was monitored at basal level and when the samples were treated with different drugs: oligomycin (5 mmol/L), carbonyl cyanide-p-trifluoromethoxy phenylhydrazone (FCCP; 1 μmol/L), and antimycin (5 mmol/L).

Western blotting was performed to detect target proteins using Ybx2 (Abcam), Gapdh (Abcam), Ucp-1(Abcam), Pgc1α (Santa Cruz Biotechnology), Cidea (Santa Cruz Biotechnology), Pparγ (Santa Cruz Biotechnology), phosphorylated (p)-AKT (Cell Signaling), AKT (Cell Signaling), Cpt1a (Proteintech), MCad (Santa Cruz Biotechnology), β-actin (Sigma-Aldrich), and tubulin (Cell Signaling) antibodies.

Gene lists were analyzed for enrichment of Gene Ontology (GO) terms by using Database for Annotation, Visualization and Integrated Discovery (DAVID) Functional Annotation Tools (31,32). Gene set enrichment analysis (33) was performed by using default parameters with the preranked gene sets.

Brown preadipocytes were cultured and differentiated for 5 days. We treated cells with 5 μg/mL actinomycin D (Sigma-Aldrich) and harvested RNA at different times as indicated in the figures. We took the same proportion of RNA from each sample at a different time, conducted reverse transcription with random primers and real-time PCR. CTs from each sample were used to calculate the remaining percentage of mRNA at each point. We fit these data into a first-phase decay model to derive mRNAs’ half-life,

with Y_t indicating the remaining percentage at a given time, Y₀ indicating the initial amount of RNA, t indicating time after transcription inhibition, and k_decay indicating the rate constant.

Data are presented as the mean ± SEM. Statistical significance was assessed using the unpaired two-tailed Student t test. Statistical significance for samples with more than two groups was determined by one-way ANOVA. The distribution difference between different cumulative curves was determined by the Kolmogorov-Smirnov test. P values of <0.05 were considered to be significant.

The accession number for the RNA-seq data reported in this report is National Center for Biotechnology Information Gene Expression Omnibus (GEO): GEO SEries (GSE) 66686, GSE29899, GSE86590, and GSE86338

To identify RBPs functionally important for BAT, we profiled the gene expression of 413 RBPs annotated in the RBP database (18) in interscapular BAT, inguinal WAT (iWAT), and epididymal WAT (eWAT), which led to identification of 26 BAT-enriched RBPs. To further assess whether these RBPs are dynamically regulated during WAT browning and brown adipogenesis, we examined their expression alternation during inguinal WAT browning induced by a β₃-agonist (CL-316,243) and in primary brown preadipocytes versus mature adipocytes. By intersecting these gene sets, we discovered six BAT-enriched RBPs that were induced during browning and brown adipogenesis, including Pgc1β, Larp4, Rbpms2, Grsf1, Akap1, and Ybx2 F (Fig. 1A–D), for further investigation.

Pgc1β is not a typical RBP so we excluded it from our subsequent experiments. The tissue enrichment and dynamic regulation of the five other candidates during WAT browning were successfully validated by real-time PCR across 15 major mice organs (Fig. 1E) and in iWAT after the animals were housed for 7 days at 4°C (Fig. 1G). In BAT, only Ybx2 was significantly induced by cold treatment (Fig. 1F). To test whether these RBPs were repressed by BAT and beige fat inactivation, we housed mice at thermoneutrality (30°C) for 7 days to induce “whitening” of BAT and iWAT. All five RBPs were downregulated during BAT and iWAT “whitening” (Fig. 1H and I). We next examined their expression during an in vitro differentiation time course of primary brown and white adipocyte culture. All five RBPs were upregulated during brown and white adipogenesis, with a higher expression level in brown adipocytes (Fig. 1J).

Finally, to test the human relevance of these observations, we examined their expression across a differentiation time course of primary preadipocytes isolated from human fetal interscapular BAT and subcutaneous WAT (27). The expression of YBX2 and RBPMS2 increased throughout the human cell differentiation course with higher levels in BAT adipocytes. AKAP1 exhibited a significant induction from day 0 to day 7 and then decreased toward the end of differentiation, but its level was still higher in brown adipocytes than in white adipocytes (Fig. 1K).

To investigate the function of these five RBPs in brown adipocyte differentiation, we depleted them by infecting brown preadipocytes with retroviral shRNAs and then induced cells to differentiate for 5 days. Depletion of each of these RBPs resulted in distinct phenotypes. Knocking down Ybx2 expression by ∼90% (sh-3) severely blocked lipid accumulation (Fig. 2A) and reduced the expression of panadipogenic markers Fabp4 and Pparγ2, indicating a block of the panadipogenesis gene program, whereas inhibiting Ybx2 by ∼70% (sh-1) affected BAT markers but did not affect panmarker expression and lipid accumulation (Fig. 2B), suggesting that the expression of BAT-selective genes is more sensitive to Ybx2 depletion. To determine the role of Ybx2 in cellular respiration, we inhibited its expression by ∼70% (sh-1) in brown adipocytes and used the Seahorse XFp Extracellular Flux Analyzer to measure the OCR. A significant decrease of OCR for basal respiration and proton leakage was observed (Fig. 2E).

Although knockdown of Akap1 slightly reduced lipids accumulation (Fig. 2A), it did not affect panadipogenic marker expression, but the BAT-selective markers were downregulated (Fig. 2C, Supplementary Fig. 1A). Inhibiting Rbpms2 had a slight influence on lipid accumulation (Fig. 2A) and panadipogenic marker expression (Fig. 2D) but stronger effects on BAT-selective markers (Fig. 2D and Supplementary Fig. 1A). OCR analysis consistently showed a significant decrease of OCR attributed to proton leak in the Rbpms2- and Akap1-depleted cells (Supplementary Fig. 1B and C). In contrast, inhibiting Grsf1 and Larp4 did not affect lipid accumulation (not shown) or marker expression (Supplementary Fig. 2A–D).

Ybx2 harbors an ultraconserved cold shock RNA binding domain (CSD). Proteins bearing CSDs, known as cold shock proteins, have been reported to regulate cellular adaptation response, mainly at posttranscriptional levels, to cold stress in prokaryotes (34,35). BAT is a major organ for cold adaption in mammals; thus, the presence of CSD in Ybx2 suggests that Ybx2 may play a role in BAT thermogenesis. We validated the expression of Ybx2 at the protein level by Western blot in different adipose depots (Fig. 4A) and during brown and white adipogenesis (Fig. 2F). Consistent with its mRNA expression pattern, Ybx2 protein level is higher in BAT and induced during differentiation. To determine its function in beige adipocytes, we knocked it down in preadipocytes isolated from inguinal WAT, followed by induction of differentiation, and observed a clear reduction of BAT markers (Supplementary Fig. 2E–G). To ensure the phenotypes of Ybx2 knockdown are not the result of off-targeting effect, we further targeted different regions in its mRNA using a different shRNA retroviral vector. Inhibiting Ybx2 invariably impaired lipid accumulation and BAT marker expression in both BAT and iWAT adipocyte cultures (Supplementary Fig. 2H–L).

We next tested whether Ybx2 is sufficient to promote beige and brown adipogenesis by overexpressing Ybx2 in primary white and brown preadipocytes with retroviral vector (Fig. 3A and D), followed by induction of differentiation. Ectopic expression of Ybx2 in white adipocytes enhanced lipid accumulation assessed by boron-dipyrromethene staining (Fig. 3B) and increased the expression of key BAT markers such as Ucp1 and Pgc1α (Fig. 3C). Overexpression of Ybx2 in primary brown adipocyte culture also enhanced lipid accumulation (Fig. 3E) and BAT marker expression (Fig. 3F) in the early phase of differentiation (day 3), which was accompanied by a higher basal OCR and proton leakage OCR (Fig. 3G). Western blot showed elevated protein levels of Ucp1 and two fatty acid oxidization regulators, Mcad and Cpt1a, at day 3 of differentiation (Fig. 3H). After 6 days of differentiation, the expression of BAT markers in control cells caught up with that in the Ybx2-overexpressing cells, probably because the abundance of endogenous Ybx2 at this stage is sufficient to support full induction of the BAT-selective gene program. Taken together, these observations indicate that Ybx2 can promote brown adipogenesis in white adipocyte culture and accelerate brown adipogenesis in brown adipocyte culture.

To determine the function of Ybx2 in BAT in vivo, we imported Ybx2 KO mice. The KO animals were infertile (36) but viable and born at expected Mendelian ratios. We confirmed their lack of Ybx2 by Western blot (Fig. 4A). The KO animals did not exhibit significant alteration in their body weight (Fig. 4B), fat, and lean mass (Supplementary Fig. 3A and B). The iWAT and eWAT of KO animals also did not change significantly in size (Supplementary Fig. 3C and D), whereas their iBATs were moderately but significantly smaller (Fig. 4B), coincident with slightly smaller lipid droplets under microscopy (Fig. 4C and D). To study the effect of Ybx2 KO at the molecular level, we quantified the expression of panadipogenic and BAT-selective marker genes by real-time PCR and observed no change in panadipogenic markers (Supplementary Fig. 3E) but a detectable decrease in Ucp1, Prdm16, and Dio2 (Supplementary Fig. 3F). RNA-seq was performed to examine the global effects of Ybx2 KO on gene expression, but very few genes showed a significant difference (Supplementary Table 2), indicating that Ybx2 is dispensable for BAT to maintain its gene-expression program at room temperature. In iWAT, we did not observe significant change of BAT-selective markers or of HoxC10, a WAT marker (Supplementary Fig. 3G). The glucose tolerance test revealed a glucose intolerance (Supplementary Fig. 3H), and the insulin tolerance test detected a trend of insulin intolerance, but the difference was not statistically significant (Supplementary Fig. 3H). Nevertheless, to what extent the impaired glucose tolerance can be accounted for by a smaller BAT or by systemic effects from other organs needs to be investigated in the future.

Because whole-body Ybx2 deficiency may have indirect effects on BAT phenotypes, to confirm whether Ybx2 KO may exert a cell-autonomous effect, we isolated brown preadipocytes from KO and WT mice for differentiation. Real-time PCR revealed decreased expression of panadipogenic markers and a more significant reduction of BAT-selective markers in the KO cells (Supplementary Fig. 4A–C), consistent with the shRNA knockdown phenotypes. RNA-seq was then performed to profile the genome-wide effect of Ybx2 KO, and gene set enrichment analysis revealed that the pathways of adipogenesis, fatty acid oxidation, oxidative phosphorylation, and cellular respiration were significantly downregulated (Supplementary Fig. 4D). Thus, Ybx2 should have cell-autonomous effects on brown adipocyte differentiation in vitro, but such an effect was much ameliorated in vivo.

To determine the role of Ybx2 in BAT activation, we exposed WT and KO animals to 4°C for 6 h. The WT BAT mass upon cold activation became smaller than that at room temperature, but the KO BAT mass did not decrease after cold exposure (Fig. 4E). Hematoxylin and eosin staining consistently revealed that lipids in WT BAT but not the KO BAT were largely depleted (Fig. 4F and G), indicating that the KO BAT failed to combust lipids upon cold activation. To directly assess the effects of Ybx2 KO’s function, we measured the OCRs for cold-activated WT and KO BAT with Oroboros respirometry. We observed a decreased OCR in KO BAT before but not after FCCP treatment, which suggested that loss of Ybx2 did not change the maximal OCR capacity but reduced the cold-provoked mitochondria activity (Supplementary Fig. 5A). Consistently the core body temperature of KO mice dropped faster than that of WT animals at cold temperature (Fig. 4H). Although the BAT defect is a certain culprit of the cold intolerance, we cannot preclude the possibility that the effect of Ybx2 KO on other organs can also contribute to this phenotype.

We examined the lipolysis rates in the WT and KO BAT and did not observe any significant change (Supplementary Fig. 5B), indicating that the larger BAT mass in the KO BAT is unlikely caused by any change in lipolysis. In addition, to test whether loss of Ybx2 affects insulin signaling in BAT, we performed Western blot to detect p-AKT in BAT and found that cold challenge could enhance the insulin sensitivity in WT but not in KO BAT (Supplementary Fig. 5C). To test whether Ybx2 KO may affect BAT-selective gene expression in beige adipocytes, we performed real-time PCR for iWATs and found a significant downregulation of Ucp1 but not other detected markers (Supplementary Fig. 5D).

To examine the effect of Ybx2 on BAT activation at the molecular level, we conducted RNA-seq of BAT isolated from WT and KO mice at both room temperature and after cold activation. One of the most striking observations was that the cold-induced thermogenic program in BAT was severely hindered in KO animals. Ucp1, Dio2, Pgc1α, and Elvol3 were among the most significantly depleted genes in KO upon cold exposure (Fig. 5A and Supplementary Fig. 6A), which we validated by real-time PCR and Western blot (Fig. 5C and D). Consistent with the individual markers, pathways analysis revealed that one of the most enriched pathways associated with the downregulated genes is oxidation reduction (Fig. 5B).

To integrate the gene expression profiles at room temperature and after cold exposure, we calculated the fold change of each gene after cold exposure in WT and KO BAT (Supplementary Fig. 6B) and looked for enriched pathways among the most differentially regulated genes. The mitochondrion and fatty acid metabolic process pathways were among the top downregulated pathways (Supplementary Fig. 6C). Importantly, the BAT mass and gene expression changes in KO animals were not sex-dependent and were also observed among female animals (Supplementary Fig. 7A–D). Thus, although BAT can still form in the absence of Ybx2, its thermogenic response to cold temperature is impaired.

As a part of BAT adaptation to cold exposure, glucose uptake, lipogenesis, and combustion of long-chain fatty acids are increased in coordination with stimulation of β-oxidation and thermogenesis (37,38). In Ybx2 KO BAT, besides thermogenic genes, those involved in glucose uptake (Glut4), lipogenesis (Scd1, Fasn, Dgat1, Dgat2, Acaca), and long-chain fatty acid generation (Elvol3, Elvol6) were also reduced (Fig. 5E and Supplementary Fig. 7E), which was further supported by pathway analysis (Fig. 5B and Supplementary Fig. 6C). Thus, Ybx2 is a regulator orchestrating glucose metabolism, lipid metabolism, and thermogenesis during BAT activation.

To identify the mRNA targets of Ybx2, we performed RNA immunoprecipitation, followed by RNA-seq (RIP-seq), by using an antibody against Ybx2 in both brown and white adipocyte culture (Supplementary Fig. 8A). First, we confirmed the successful Ybx2 precipitation by Western blot (Fig. 6A). We then selected candidates with at least eightfold enrichment in the Ybx2 IP sample compared with the IgG control, which revealed 800 and 1,822 potential mRNA targets in BAT and WAT, respectively. Targets in BAT and WAT significantly overlapped, leading to identification of 414 common targets (Fig. 6B). As expected, Ybx2 can target many mRNAs encoding proteins involved in posttranscriptional RNA processing, a general feature of RBPs (39–42). Functional terms enriched among Ybx2 targets include ribosome, ribonucleoprotein complex, and translation (Fig. 6C). Interestingly, these targets were also enriched for mitochondria term (Fig. 6C). To further confirm this observation, we calculated the relative abundance of each mRNA in Ybx2 versus the IgG sample and plotted the cumulative distributions for mitochondrion-related genes (206 genes) as well as for all genes detectable in the RIP-seq assay (4,095 genes). The cumulative curve of mitochondrion significantly shifts toward the right (Fig. 6D), confirming that the targets of Ybx2 are enriched for mitochondrial functions.

Next, we asked whether Ybx2 exerts a functional effect on its mRNA targets. We used RNA-seq data to calculate the fold change of each mRNA between KO and WT BAT and plotted the cumulative distributions for Ybx2 target and nontarget mRNAs. At room temperature, distributions of targets and nontargets did not show a significant difference (Fig. 6E), but upon cold activation, targets of Ybx2 were markedly repressed in KO BAT (Fig. 6F). These data support a role of Ybx2 in stabilizing its target mRNAs, which was suggested by earlier work in oocytes (43).

Among the top Ybx2 targets was the Pgc1α mRNA that is significantly decreased in KO BAT. We used Pgc1α as an example to illustrate how Ybx2 recognizes and affects its targets. To confirm binding between Pgc1α mRNA and Ybx2 in vivo, we performed the RIP-PCR in tissue lysate from KO and WT BAT to detect Pgc1α mRNA precipitated by Ybx2. A clear reduction of Pgc1α signal was detected in RIP from KO BAT, but such a reduction was not observed for Fabp4 mRNA, which bears a short 3′ untranslated region (UTR; 180 bp) and is used as a control (Fig. 7A). To dissect Ybx2-binding sites within Pgc1α mRNA, we performed an RNA pull-down assay using four in vitro transcribed sequential RNA fragments from Pgc1α 3′UTR and found that a 1,101 nucleotide RNA fragment (fragment 3) can readily retrieve Ybx2 protein (Fig. 7B). We then generated eight small RNA fragments (200–300 bp) from this segment for a second round of pull-down assays and found that fragments 3.1, 3.3, 3.5, and 3.7 can retrieve Ybx2 (Fig. 7C). Intersecting segments 3.1 vs. 3.5 and 3.3 vs. 3.7 locate two regions that harbor Ybx2 binding sites. Further truncation of these two fragments abolished their interactions with Ybx2 (data not shown), suggesting that a secondary or tertiary nucleotide structure may be necessary for Ybx2 binding. To test whether this identified RNA fragment can define the Ybx2 binding site in humans, we blasted the human Pgc1α 3′UTR and mouse Pgc1α 3′UTR and identified a ∼1-kb segment that was >90% homologous to the fragment 3 in Fig. 7B (Fig. 7D). We cloned this fragment for pull-down assay. As expected, this fragment could retrieve Ybx2 in BAT lysate (Fig. 7D), indicating that the Ybx2-Pgc1α interaction is conserved.

To test whether the interactions between Ybx2 and Pgc1α is enhanced at cold exposure, we performed RIP-PCR in BAT before and after cold exposure and found that Ybx2 could retrieve more Pgc1α mRNAs upon cold exposure (Fig. 7E). To further study whether this apparent increase is attributable to an enhanced binding affinity or an elevated Pgc1α mRNA abundance upon cold exposure, we inhibited transcription with actinomycin D in cultured brown adipocytes and then performed RIP-PCR to detect the Ybx2-Pgc1α mRNA interaction in the presence or absence of norepinephrine treatment. Interestingly, in the actinomycin D treatment cells, the Pgc1α mRNA retrieved by Ybx2 antibody was similar before and after norepinephrine treatment (Supplementary Fig. 8B). Therefore, BAT activation likely increases the Ybx2-retrived Pgc1α mRNA by stimulating Pgc1α mRNA expression but not by changing their binding affinity.

To examine the influence of Ybx2 knockout on Pgc1α mRNA stability, we used actinomycin D to stop mRNA transcription in WT and KO brown adipocyte culture and measured the decay rates for Pgc1α and Fabp4 mRNA. The half-life of Pgc1α mRNA decreased from 2.39 to 1.29 h in the absence of Ybx2 (Fig. 7F), supporting a role of Ybx2 in stabilizing Pgc1α mRNA. To investigate whether the above-identified Ybx2-binding sites in Pgc1α 3′UTR can mediate the mRNA-stabilizing effect from Ybx2, we constructed two reporter plasmids: one containing an ∼2kb Pgc1α 3′UTR after the Renilla luciferase (WT) and another containing a truncated 3′UTR without the Ybx2-binding fragment (mutant). We measured the decay rates of the Renilla luciferase mRNA in 293 cells in the presence and absence of a vector expressing Ybx2. In the absence of Ybx2, both reporter constructs manifested similar decay rates (Supplementary Fig. 8C); in the presence of Ybx2, the mRNA decay rate of the WT reporter is approximately twofold slower than that of the mutant reporter (∼8.4 h vs. ∼4.1 h), indicating our identified Ybx2-binding sites (Fig. 7B and C) in the Pgc1α 3′UTR are required for Ybx2’s mRNA stabilization function.

We further examined the functional interactions between Ybx2 and Pgc1α by overexpressing a full-length open reading frame Pgc1α in the Ybx2-inhibited brown adipocytes. As described above (Fig. 2B), knockdown of Ybx2 reduced BAT marker expression, but Pgc1α overexpression could significantly rescue the phenotype (Fig. 7G). Therefore, although stabilizing Pgc1α alone is unlikely to account for all the phenotypes of Ybx2 KO, Ybx2 may be a key target of Ybx2, and the function of Ybx2 to some extent relies on Pgc1α expression.

Early studies have suggested a role of Ybx2 in global mRNA stabilization. Schultz and colleagues (44) knocked down Ybx2 in oocytes by expressing a transgenic Ybx2 hairpin double-stranded RNA. They observed a 60% reduction of Ybx2 protein and a 75–80% reduction of poly-(A) mRNAs. In another study, they generated a knockout strain and observed severe defects in spermatogenesis and oocyte development (36) accompanied by an ∼25% decrease of mRNAs in the mutant oocytes (43). Exogenous mRNAs injected into mutant oocytes were lower than that in WT cells, consistent with a decreased mRNA stability in the absence of Ybx2 (43). This is consistent with our conclusion, which demonstrated a role of Ybx2 in enhancing mRNA stability in a more systemic manner (Fig. 6F).

As an RBP, Ybx2 may affect multiple RNA processing steps, including but not limited to RNA stability. Given its cytosol localization (Supplementary Fig. 9C), the ability of Ybx2 to influence translational control for certain mRNAs is not surprising. This may explain why the changes in mRNA and protein levels for some genes, to some extent, may display discordance. The potential influence of Ybx2 on translation will be further studied in the future.

Skeletal muscle is known to contribute to nonshivering thermogenesis (NST) mainly through sarcolipin-mediated ATP hydrolysis by SERCA (45,46). To investigate whether Ybx2 KO can alter this pathway, we examined the NST genes in muscles. Despite an increase of sarcolipin expression observed in KO muscle, Ybx2 KO did not affect Serca1-3 expression (Supplementary Fig. 9A) and OCRs (Supplementary Fig. 9B) directly measured by Oroboros respirometry. Therefore, although whether the increase of sarcolipin expression in muscle is the result of a tissue autonomous effect or a cross-organ response to the compromised KO BAT is unclear, the NST function of muscle was not altered.

In sum, we profiled the expression of >400 RBPs across different fat depots, during adipogenesis and WAT browning, and identified Ybx2, a CSD-containing protein that orchestrates BAT activation. CSD-containing proteins are among the most phylogenetically conserved families and are known for their role in cold adaptation in prokaryotes (34,35). Because BAT activation is a part of cold adaptation in mammals, we speculated that CSD proteins, exemplified by Ybx2, may be evolutionarily conserved to mediate cold adaptation at the whole organismal level via roles in BAT activation.

Acknowledgments. The authors thank Dr. Paula Stein, University of Pennsylvania, for the KO mice as a generous gift and Dr. Manvendra Singh, Duke-NUS Medical School, for the coordination of mice transportation.

Funding. This work was supported by Singapore National Research Foundation (NRF) fellowship (NRFF) (NRF-2011NRF-NRFF 001-025) to L.S. This research is also supported by the Singapore NRF under its Cooperative Basic Research Grant (CBRG) grant administrated by the Singapore Ministry of Health's National Medical Research Council (NMRC) (NMRC/CBRG/0070/2014 and NMRC/CBRG/0101/2016).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. D.X., S.X., A.M.M.K., Y.C.L., S.Y.C., J.R.A.-D., and D.T.C.S. performed experiments. D.X. and L.S. designed experiments and wrote the manuscript. P.C. and M.K.-S.L. discussed the experiment design and critically reviewed the manuscript. L.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.