GRAF1 Regulates Brown and Beige Adipose Differentiation and Function

Adipose tissue, which is crucial for the regulation of energy within the body, contains both white and brown adipocytes. White adipose tissue (WAT) primarily stores energy, while brown adipose tissue (BAT) plays a critical role in energy dissipation as heat, offering potential for therapies aimed at enhancing metabolic health. Regulation of the RhoA/ROCK pathway is crucial for appropriate specification, differentiation and maturation of both white and brown adipocytes. However, our knowledge of how this pathway is controlled within specific adipose depots remains unclear, and to date a RhoA regulator that selectively controls adipocyte browning has not been identified. Our study shows that expression of GRAF1, a RhoGAP highly expressed in metabolically active tissues, closely correlates with brown adipocyte differentiation in culture and in vivo. Mice with either global or adipocyte-specific GRAF1 deficiency exhibit impaired BAT maturation, reduced capacity for WAT browning, and compromised cold-induced thermogenesis. Moreover, defects in differentiation of mouse or human GRAF1-deficient brown preadipocytes can be rescued by treatment with a Rho kinase inhibitor. Collectively, these studies indicate that GRAF1 can selectively induce brown and beige adipocyte differentiation and suggest that manipulating GRAF1 activity may hold promise for the future treatment of diseases related to metabolic dysfunction.


Introduction
Adipose tissue is comprised of two major cell types-white and brown adipocytes-that have opposing energetic functions.White adipocytes primarily function to store excess lipid and an over-abundance of white adipose tissue (WAT) leads to unfavorable outcomes due to disruption in many metabolic and hormonal activities.In contrast, brown adipocytes consume glucose and lipid and brown adipose tissue (BAT) dissipates energy through uncoupled oxidative phosphorylation-induced heat generation, a process necessary for thermoregulation in small animals and infants.Interestingly, in response to various environmental cues, including following cold exposure, a subset of adipocytes within WAT undergo browning and these so-called "beige" fat cells exhibit many of the bene cial metabolic features of classical brown adipocytes including dissipation of energy through thermogenesis and the secretion of endocrine factors that promote whole body metabolism.As obesity results from a disparity between energy intake and expenditure, the identi cation of new molecules and signals that can drive fat browning may lead to attractive targets for the treatment of obesity and related cardiometabolic disorders.
WAT and BAT are located in discrete depots that are distributed throughout the body and are derived from a diverse group of progenitor cells.While we do not yet know the precise origins of cells within each of these tissues, current lineage tracing studies indicate that Myf5-expressing progenitor cells that originate in the dermomyotome give rise to greater than 90% of inter-and sub-scapular BAT (iBAT and sBAT) which are the largest BAT depots in mice 1,2 .Other adipose depots likely arise from a more heterogeneous population of progenitors.For example, the contribution of Myf5 progenitors to WAT varies tremendously (from 5 to 60%) based on anatomic locale and sex 1 .The mural cell compartment of the vasculature also appears to be a major contributing source of adipogenic precursors in various WAT compartments.
Despite their heterogeneous origins, white and brown adipogenic precursors share some common transcriptional programs.For example, peroxisome proliferator-activated receptor gamma (PPARγ) is a master regulator of adipocyte differentiation and this transcription factor drives expression of a number of genes common to all adipocytes including adiponectin and fatty acid binding protein 4 (FABP4) 3,4 .
Moreover, classical brown adipocytes that originate from the dermomyotome and beige adipocytes that are intermingled in WAT depots share PGC1α-and PRDM16-dependent transcriptional programs that drive mitochondrial biogenesis and induce a common subset of BAT-selective genes.These include, among others, Cidea which promotes lipolysis and lipogenesis and uncoupling protein 1 (UCP-1) which is responsible for facilitating thermogenesis through uncoupled respiration 5 .
Importantly, both white and brown adipocytes undergo dynamic and reversible phenotypic conversions in response to environmental cues.For example, thermoneutrality or exposure to high fat diet can induce BAT to undergo reversible WAT-like remodeling.Conversely, long-term cold exposure, agonist-induced activation of the β3 adrenergic pathway, or physical activity can promote WAT to undergo browning.The bene cial physical activity-induced phenotypic changes are mediated, at least in part, by skeletal-muscle derived factors including broblast growth factors(FGFs) and bone morphogenic proteins (BMPs) 6, 7 .BMP 4, 6, and 7 are particularly strong inducers of brown and beige adipogenesis and, while we do not fully understand the underlying signaling mechanisms, downregulation of RhoA-mediated signaling has been shown to be important [8][9][10][11][12] .For example, down-regulation of RhoA in white adipocytes promotes beige cell development and heterozygous germline depletion of the RhoA-dependent kinase, ROCK2, or treatment with Rho kinase inhibitors promoted WAT browning and led to protection from diet-induced obesity and insulin resistance in mice 13 .Interestingly, the dynamic regulation of RhoA is also important to control the fate of mesenchymal stem cells and multi-potent progenitors wherein low RhoA activity favors pre-adipocyte speci cation while high RhoA activity promotes speci cation towards osteoblast, smooth-or skeletal muscle cell fates 14 .Mechanistically, RhoA/ROCK signaling promotes these cell fate conversions by modulating both actin cytoskeleton-dependent shape changes and altering gene transcription (RhoA activity inhibits the pro-adipogenic transcription factor, PPARγ-and activates myogenic SRF/MRTFA-mediated gene transcription) 15,16 .
RhoA, like all small GTPases, cycles between an inactive GDP-bound form and an active GTP-bound form and its activity is enhanced by guanine nucleotide exchange factors (GEFs) and inhibited by GTPaseactivating proteins (GAPs).To date, two RhoA-speci c GAPs have been identi ed as critical regulators of adipogenesis, p190B and DLC1.P190B limits RhoA activity in myogenic precursors and is necessary for these cells to adopt a pre-adipocyte fate [17][18][19] .On the other hand, DLC1 limits RhoA activity in preadipocytes and functions to promote white and brown adipogenesis 20 .However, a GAP that selectively controls adipocyte browning has not yet been identi ed.Herein, we show that the multidomain containing RhoA-GAP termed GRAF1 (guanosine triphosphatase (GTPase) regulator associated with FAK-1, also named Arhgap26) 21,22 that was previously reported to be strongly expressed in highly metabolic tissues (including heart, brain, and skeletal muscle) is also strongly expressed in BAT and functions to selectively promote the activation of BAT and the browning of WAT.Our ndings highlight the possibility that GRAF1 could be a new therapeutic target to combat obesity and associated morbidity.

Results
GRAF1 expression correlates with BAT maturation in mouse and human cells.
We previously reported that GRAF1 was transiently upregulated during skeletal muscle development (E17-P4) and again following adult muscle injury and that, in these contexts, GRAF1 promoted myoblast differentiation by limiting RhoA activity 23 .Interestingly, while dissecting various muscles from adult GRAF1 hypomorphic (GRAF1 gt/gt ) mice, we noticed that their iBAT depots were relatively pale compared to WT controls.Since myocytes and BAT share common Myf5-expressing precursors 1,18,24,25 and both cell types require RhoA inactivation for differentiation and maturation 26 ,we further investigated a possible role for GRAF1 in BAT development.
In mice, iBAT and sBAT depots develop during late embryogenesis, and, upon expression pro ling in juvenile mice (3 weeks postnatal), we found that GRAF1 was highly expressed in these major BAT depots.GRAF1 was also expressed (albeit at a lower level) in subcutaneous WAT (scWAT) (Fig. 1a).Interestingly, GRAF1 mRNA levels in iBAT increased nearly 20-fold from 1 week to 3 months of age (Fig. 1b) and this increase paralleled the expression of BAT maturation genes including PPARγ, the thermogenic protein UCP1, and the mitochondrial marker ND5.In contrast, GRAF1 levels did not signi cantly change during WAT maturation.
We next used cultured cell models to con rm and extend these ndings.First, using WT-1 cells (a validated in vitro model of brown adipogenesis) 27 , we found that GRAF1 expression was dynamically and transiently increased at the onset of differentiation (Fig. 1c,d) and that its induction occurred prior to, or concomitant with, expression of BAT marker genes UCP1, PPARγ, and ND5 (Fig. 1d).This nding is consistent with prior reports indicating RhoA/ROCK signaling is down-regulated upon the induction of adipocyte differentiation in these cells.A similar dramatic increase in GRAF1 expression was observed in human preadipocytes (Simpson-Golabi-Behmel syndrome, SGBS) following exposure to brown adipocyte differentiation medium (Figure S1a).Note that adipogenic differentiation of these cells was accompanied by lipid droplet accumulation and characteristic brown phenotype observed using light microscopy (Fig S1a, bottom panel).Finally, in mouse 3T3L1 adipocytes, which can be induced to form WAT or BAT, treatment with BAT-induction cocktail led to a more robust increase in GRAF1 expression than did treatment with a WAT-induction cocktail (Figure S1b,c).Collectively, these studies indicate that GRAF1 might play an important and conserved role in promoting BAT development.

GRAF1 is necessary for BAT maturation and function
We next compared BAT and WAT marker gene expression in tissue isolated from WT mice or from global GRAF1-de cient mice (GRAF1 gt/gt ) which harbors the gene trapping vector VICTR48 within the rst intron of Graf1.These mice lines are viable and fertile with no obvious abnormalities under baseline conditions 28 .As shown in Fig. 2a, we found no differences in expression of UCP1, PPARγ or ND5 in scWAT or iBAT isolated from 1 week old WT and GRAF1 gt/gt mice, suggesting that GRAF1 does not play an important role in adipose tissue specialization.However, since adipose tissue maturation happens gradually after birth through young adulthood [29][30][31] and GRAF1 expression was robustly increased during this timeframe, we reasoned that GRAF1 might impact adipose differentiation/maturation.Indeed, as shown in Fig. 2b, BAT depots isolated from 3 month-old GRAF1-de cient mice exhibited a dramatic reduction in BAT marker gene expression relative to littermate control WT mice.For example, iBAT from GRAF1 gt/gt mice demonstrated a signi cant reduction in BAT thermogenesis genes (UCP1, Elovl3), mitochondrial genes (ND5, COX7a), and in PPARγ, the master regulator gene of adipocyte differentiation, indicating a differentiation/maturation defect in the iBAT of GRAF1 hypomorphs.Accordingly, H&E staining of iBAT from 2-3 month old GRAF gt/gt mice revealed a "white like" appearance as demonstrated by a substantial increase in lipid deposition and enlarged lipid droplets compared to iBAT from littermate WT control mice(Fig.2c).scWAT from 2-3 month old GRAF1-de cient mice also exhibited a signi cant downregulation of brown marker genes including UCP1 and Elovl3 but upregulation of the general adipose marker adiponectin, suggesting that GRAF1 promotes WAT browning but limits WAT expansion (Fig. 2d).Despite these changes, and consistent with our prior report 28 , there was no difference in the body weights of GRAF1 gt/gt mice and littermate control mice fed ad libitum (data not shown).This nding is consistent with other mouse models that exhibit defects in brown adipose development, yet do not develop obesity, such as those with the loss of the uncoupling protein UCP-1 32 , or the fatty acid metabolism gene Adipose acyl-CoA synthetase-1 33 .
To determine the impact of GRAF1-dependent changes in BAT differentiation on BAT function, we next exposed mice housed at sub-thermoneutral temperatures to overnight fasting followed by an acute bout of cold stress (6°C) and measured their body temperatures over time.As shown in Fig. 2e, GRAF1de cient mice exhibited a remarkable reduction in thermogenic capacity under these conditions.
Interestingly, this failure to thermoregulate was associated by a signi cant reduction in serum lactate levels (Fig. 2f).Since circulating lactate is the main fuel that drives the tricarboxylic acid cycle (TCA cycle) in a fasted state 34 , the decreased lactate might suggest enhanced TCA cycling due to betaoxidation defects in GRAF1 gt/gt mice.In support of this possibility, there was a trend towards increased muscle triglyceride levels in the cold-exposed GRAF1 gt/gt mice relative to similarly treated WT mice (Fig. 2g).Nonetheless, given our knowledge regarding GRAF1's role in skeletal muscle maturation, we realized that the drop in temperature coupled with the reduced lactate production observed in GRAF1 gt/gt mice could be due (at least in part) to a reduced capacity for shivering thermogenesis in these germlinede cient mice.To begin to distinguish between these two possibilities, we rst sought to determine if GRAF1 can act in a cell autonomous fashion to promote BAT maturation.To this end, we turned to primary pre-brown adipocyte cultures isolated from the stromal vascular fraction (SVF) of iBAT as these cells have been reported to faithfully recapitulate brown adipocyte maturation when exposed to serumcontaining media supplemented with insulin, triiodothyronine (T 3 ), dexamethasone, IBMX and rosiglitazone 35 .Importantly, upon induction with differentiation media, SVF cells transfected with GRAF1 siRNA exhibited a signi cant reduction in browning capacity compared to control siRNA treated cells as assessed by UCP1 expression and oil red O staining (Fig. 3a-c).These ndings indicate that GRAF1 levels can directly impact the differentiation of brown adipocytes.
Next, to further explore an adipocyte-autonomous role for GRAF1 in BAT formation and function in vivo, we developed a new mouse model using targeting vectors from Eucomm to conditionally target the GRAF1 allele.In this model, Cre-mediated recombination causes a frame shift and early stop codon (Supplemental Fig. 2a) that results in nonsense mediated mRNA decay.Southern blot analysis con rmed successful targeting in embryonic stem cells (ES) and the germline transmission of resulting chimeras(Tm1a).The Tm1a mice were subsequently crossed with Flp recombinase mice to remove the LacZ reporter and a neomycin resistance cassette used for selection of targeted ES cells to generate GRAF1 / (GRAF1 Tm1c mice; Supplemental Fig. 2b).Finally, we established an adipose-speci c GRAF1 knockout mouse line (GRAF1 AKO ) by crossing GRAF1 Tm1c mice with the Adiponectin Cre line (010803, Jackson Lab) 36 which led to a signi cant depletion of GRAF1 in scBAT and WAT (Fig. 3d).Importantly, while not as dramatic as observed in the GRAF1 gt/gt mice, the thermogenic capacity of GRAF1 AKO mice was signi cantly blunted when compared to genetic control mice (Fig. 3e).Also, activation of a thermogenic gene expression pro le in mature BAT was signi cantly reduced in GRAF1 AKO mice when compared to similarly-treated genetic control mice (Fig. 3f).Consistent with the more modest impact on thermogenesis, while BAT from GRAF1 AKO mice exhibited a signi cant reduction in UCP1 expression, the decrease was not as robust as observed in GRAF1 gt/gt mice (possibly due to incomplete recombination in our model).Nonetheless, GRAF1 AKO BAT also exhibited signi cantly reduced expression of the mitochondrial genes Cpt1a, Cpt1b and CS.In addition, GRAF1 AKO BAT also exhibited signi cant reductions in the expression of growth factors known to promote glucose homeostasis including FGF1, which acts in an autocrine fashion to promote glucose uptake in activated BAT and FGF21, an endocrine factor that has been linked to the cardiometabolic bene ts of brown fat activation through its ability to promote glucose homeostasis in BAT, skeletal muscle, liver and brain.Collectively, these data con rm that GRAF1 plays an important, adipocyte-autonomous role in BAT formation and function.

GRAF1 promotes subcutaneous adipose beigeing
As noted above, scWAT from adult GRAF1 gt/gt mice also exhibited lower BAT marker gene expression relative to similarly housed WT mice, indicating the possibility that GRAF1 might also be necessary for physiological WAT browning.To further explore a role for GRAF1 in beige fat induction, we treated mice with the browning agent CL316243, a β 3-adrenergic receptor agonist.Treatment with CL316234 for 10 days in WT mice promoted expression of several beige adipose markers in scWAT including the mitochondrial enzyme CPT1A (Carnitine Palmitoyltransferase 1) that is essential for fatty acid betaoxidation 37 and the secreted factors FGF1 and FGF21 that improve adipose and systemic glucose and lipid metabolism 38 (Fig. 4a).Interestingly, CL316243 treatment also signi cantly increased scWAT GRAF1 expression, further suggesting a role for GRAF1 in regulating agonist-induced browning (Fig. 4a).Indeed, the induction of each of these marker genes was signi cantly reduced in scWAT isolated from CL316243treated GRAF1 AKO mice when compared to scWAT isolated from similarly treated WT mice (Fig. 4b).Thus, in addition to promoting classical brown fat maturation and function, GRAF1 also promotes the development of metabolically favorable beige adipocytes in scWAT.

GRAF1 promotes brown phenotypes by limiting RhoA/ROCK signaling
Because previous studies have shown that the downregulation of RhoA-ROCK signaling is necessary and su cient to promote adipogenesis, we reasoned that GRAF1 might promote BAT differentiation by controlling this pathway.To test this possibility, we quanti ed RhoA activity in WT-1 brown adipocytes using a standard GST-rhotekin precipitation assay.As shown in Fig. 5a and b, GRAF1-depleted WT-1 cells exhibited signi cantly higher levels of RhoA activity in comparison with control siRNA-treated cells following treatment with the RhoA agonist, sphingosine-1-phosphate. Importantly, treatment with the ROCK inhibitor, Y27632 at the onset of differentiation completely restored UCP-1 expression in GRAF1depleted WT-1 cells (Fig. 5c) and in GRAF1-depleted SGBS human brown adipocytes (Fig. 5d).Collectively, these data indicate that GRAF1 promotes brown adipocyte differentiation by limiting RhoA/ROCK signaling.

Discussion
Adipose tissue in mammals is a vital component of the energy regulation system and consists of two primary types: white adipocytes and brown adipocytes.White adipocytes are characterized by their large central lipid droplets, serving as reservoirs for storing excess energy in the form of triglycerides.In contrast, brown adipocytes feature small lipid droplets and a high number of mitochondria, enabling them to dissipate energy as heat through a process known as thermogenesis 39,40 .The discovery of metabolically active brown adipose tissue (BAT) in healthy adult humans and beige adipocytes within subcutaneous white adipose tissues(scBAT) has generated signi cant interest in the biology of brown and beige adipocytes due to their unique ability to improve whole-body glucose and lipid metabolism by consuming substantial amounts of blood glucose and lipids.Consequently, identifying molecules and signals that can regulate brown fat differentiation and function has become an attractive target for potential treatments of obesity and diabetes.Our study demonstrates for the rst time that GRAF1 regulates brown adipogenesis.Depletion of GRAF1 reduces brown adipocyte differentiation and thermogenesis function in vivo, highlighting its unique role in mediating BAT cell fate.
The differentiation, activation, and maintenance of brown and beige adipocytes are governed by a complex interplay of multiple factors.These factors include endocrine signals such as broblast growth factors(FGFs) and bone morphogenic protein factors(BMPs), as well as critical transcription factors like PRDM16, PGC1α, PPARγ, and Foxp1 6, 7, 41-43 .Of particular interest, the RhoA/ROCK pathway has emerged as a key player in regulating adipogenesis.Previous studies have shown that disrupting this pathway, either through the expression of dominant-negative RhoA or the inhibition of ROCK, promotes differentiation in various adipogenic cell types, highlighting the signi cance of this pathway.Mechanistically, induction of adipocyte differentiation leads to downregulation of RhoA-ROCK signaling, which promotes disassembly of F-actin stress bers and results in marked changes in cell shape that are thought to be important for lipid droplet accumulation.Depolymerization of F-actin also leads to accumulation of monomeric G-actin, which binds and sequester MRTFs in the cytosol, preventing their nuclear translocation.As MRTFs and their co-factor, SRF repress PPARγ, this critical step allows for the expression of PPARγ and its target genes, facilitating the development and maintenance of adipocyte characteristics during adipogenic differentiation 15,16,44 .
The Rho GTPase can exist in either an inactive GDP-bound or an active GTP-bound form and is modulated by guanine nucleotide exchange factors (GEFs) and GTPase-activating proteins (GAPs) 45 .The comprehensive understanding of the diverse regulators of Rho GTPase/ROCK in adipose tissues and their in vivo functional consequences remains relatively unexplored.Our previous studies have shown that GRAF1 is a bone de RhoGAP and plays important role in processes like myoblast fusion, which require GAP-dependent actin remodeling 28 .Given GRAF1's established function as a RhoGAP in various tissues, we hypothesized that it might also play a pivotal role in modulating RhoA activity during adipogenesis.Indeed, we observed that GRAF1 is highly and selectively expressed in metabolically active tissues, including brown adipose tissue (BAT), brain, and heart, and its expression pro le closely correlated with brown adipocyte differentiation.This association is evidenced by the upregulation of GRAF1 in response to brown adipocyte differentiation medium in various cultured adipocytes and the signi cant increase in GRAF1 mRNA levels during the maturation stage of BAT in mice, but not in WAT.GRAF1 de ciency, as observed in primary pre-brown adipocyte cultures and in both global and adiposespeci c GRAF1-de cient mice, signi cantly blunted brown and beige adipose differentiation.Furthermore, GRAF1-de cient mice exhibited an inability to e ciently respond to cold challenge-induced thermogenesis.Moreover, development of beige adipocytes in scWAT was compromised in response to browning agent CL316243 stimulation via the β3-adrenergic pathway.Collectively, these data indicate that GRAF1 plays a key cell autonomous role in the development and maturation of beige and brown adipose tissue.
Our mechanistic studies showed that GRAF1 depletion in brown preadipocyte cell line increased Rho activity, suggesting that GRAF1 mediates brown adipocyte differentiation by GAPdependent Rho GTPase inhibition.Interestingly, GRAF1 also possesses a BAR (Bin/amphiphysin/Rvs) and PH (pleckstrin homology) domain that are involved in sensing and inducing membrane curvature and determining membrane binding speci city.Interestingly, these domains in combination with an isoformspeci c hydrophobic segment (found in brain-selective GRAF1a) were previously reported to drive GRAF1a association with lipid droplets, an event that promoted lipid droplet clustering and reduced lipolysis 46,47 .While our current study focused on mouse GRAF1.2 (the ortholog of GRAF1b in humans) which does not contain this hydrophobic region, future studies are warranted to determine the extent to which GRAF1.2/GRAF1b may regulate adipocyte lipid droplet homeostasis.
Previous studies have demonstrated that p-190 B is a RhoGAP can that shift myogenesis towards adipogenesis by inhibiting Rho GTPase in adipogenic precursor cells [17][18][19] .Another RhoGAP, DLC1, has been reported to promote both white and brown adipocyte differentiation and to provide a molecular link between PPARγ and Rho signaling pathways 20 .However, it's important to note that while p-190 B and DLC1 play signi cant roles in regulating adipocyte differentiation, neither of these RhoGAPs exhibits selective induction of brown adipocyte differentiation.In our current study, we present the unique contribution of GRAF1, a RhoGAP, to the differentiation of brown and beige adipocytes.This distinctive function differentiates GRAF1 from other RhoGAPs.Combining in vitro investigations across various cell types with in vivo experiments in GRAF1 knockout mouse models, we have demonstrated that GRAF1 promotes classical brown fat maturation and the development of beige adipocytes in scWAT.This regulation, at least partly, occurs through GRAF1's ability to suppress RhoA activity, thereby limiting the RhoA-ROCK signal pathway and promoting brown adipocyte differentiation and maturation.Future research may elucidate the potential for manipulating GRAF1 expression and its GAP activity to improve metabolic pro les, holding promise for the treatment of metabolic diseases, including obesity and insulin resistance.

Animals
GRAF1 gene trap mice were generated and obtained from the Texas A&M Institute for Genomic Medicine (College Station, TX) and were described previously 28 .ES cells carrying Arhgap26 targeted knockout rst conditional-ready alleles were obtained from EuMMCR.Germline transmission of the allele was con rmed from F0 chimeric mice with C57BL/6 genetic background.Next, we crossed Flp recombinase transgenic mice with F1 to generate GRAF1 oxed mice.Finally, we established adipose speci c GRAF1 knockout mouse line (GRAF1 AKO ) by crossing Adiponectin Cre line (010803, Jackson Lab 36 ) with GRAF1 oxed mice.All mice were housed in pathogen-free facilities under a 12-hour light/dark cycle with unrestricted access to food and water.Animals were treated in accordance with the approved protocol of the University of North Carolina (Chapel Hill, NC) Institutional Animal Care and Use Committee, which is in compliance with the standards outlined in the guide for the Care and Use of Laboratory Animals.All methods are reported in accordance with ARRIVE (Animal Research: Reporting of In Vivo Experiments) guidelines.
To induce GRAF1 knockout in vivo, 80mg of Tamoxifen power (Sigma, T5648) was dissolved in 750ul 100% ethanol(molecular biology grade) (ThermoFisher Scienti c, T038181000) for 40 mins at 40°C, then was mixed with 3.25 ml ltered corn oil(Sigma, C8267) to make nal concentration of 20mg/ml.Tamoxifen was administered at 100mg/kg via oral gavage once every 24 hours for a total of 5 consecutive days. 1 week later, mice were subject to experiment treatment.To induce the browning program of white fat depots, control and Adcre-GRAF1ko (GRAF1 AKO ) mice received daily intraperitoneal injections of β3-adrenergic agonist CL316243 at 1mg/kg (Sigma, C5976)for 10 days as described previously 48 .

Cold exposure
To assess sensitivity to cold exposure, we monitored rectal temperature at hourly intervals following the placement of mice in a cold room maintained at 6-8 ℃.The mice underwent an overnight fast, which continued throughout the duration of the cold exposure.Subsequent to the cold exposure period, mice were humanely euthanized by CO2 inhalation.

Isolation and differentiation of Stromal Vascular Fraction (SVF) Cells
Subcutaneous WAT (inguinal WAT) or interscapular BAT tissues were used for Beige cell source, and followed a previously described isolation and differentiation method 49 .In brief, tissues were digested and centrifuged to get SVF cells.SVF cells were cultured to 100% con uence in complete medium (DMEM/F12 containing 10% FPS and P/S).SVF cells were cultured for 2 days in induction medium (complete medium plus 5µg/ml insulin, 1 nM T3, 125 µM Indomethacin, 2 µg/ml Dexamethasone, 0.5 mM IBMX, 0.5 µM Rosiglitazone).Then cell medium was changed to maintenance medium (complete medium plus 5µg/ml insulin, 1 nM T3) with 0.5 µM rosiglitazone for 2 days.At Day 4, cell medium was changed to maintenance medium with 1 µM rosiglitazone for 2 days.
WT-1 cells were provided by Dr. Yu-Hua Tseng (Joslin Diabetes Center, Harvard Medical School A liate, Boston, MA).Cells were cultured and induced differentiation with the same medium as SVF cells.

Real time PCR analysis
Total RNA was isolated from homogenized whole mouse tissues or cell cultures using RNeasy Mini Kit (Qiagen, 74106) according to manufacturer's instructions.After homogenizing, samples were placed on ice to remove lipid layer.Complimentary DNA (cDNA) was obtained from 1 µg of RNA isolated using the iScript cDNA Synthesis Kit (Bio-Rad, 1708897), and cDNA was used for qPCR with iTaq Universal SYBR Green Supermix kit (Bio-Rad, 1725124).The relative gene expression levels were calculated using deltadelta Ct method, also known as the 2-∆∆Ct method.Primer sequences were in Supplemental Table 1.

Western blotting
To examine protein levels, lysates from cells or tissues were prepared by lysing in a modi ed RIPA buffer with 1x HALT phosphatase & protease inhibitor cocktail(ThermoFisher Scienti c, 78438 and 78427).Protein concentration was determined by using a colorimetric BCA assay (Pierce, 23227).Lysates were electrophoresed on SDS-polyacrylamide gel, transferred to nitrocellulose and immunoblotted with speci c antibodies overnight at 4℃ as indicated using a 1:1000 dilution.The following primary antibodies were used in western blot: GAPDH(cell signaling technology, 5174S), b-actin(cell signaling technology, 3700S), α-Tubulin(Sigma, T6199), RhoA(Santa Cruz, SC-418).Rabbit anti-GRAF1 polyclonal antibody is homemade antibody in our lab.Blots were washed in TBST (TBS plus 0.1%Tween20) followed by incubation with horseradish peroxidase conjugated antibody at a 1/1,000 dilution.Blots were visualized after incubation with chemiluminescence reagents (ThermoFisher Scienti c, 32106).

Oil O staining and quanti cation
Cells were rinsed with PBS and then xed in 4% PFA for 30 minutes.Then cells were left in the air until completely dry.Oil Red-O working solution (0.3%) was freshly prepared and stained cells on the shaker for 10 minutes.Then cells were rinsed with PBS 3 times and used for imaging.After imaging, liquid was removed from cells completely.To elute the oil red O dye, 100% isopropanol was added to the plates.The plates were incubated for 10 min at room temperature on an orbital shaker.Absorption was measured at 518 nm on a plate reader.

Statistics
Unless stated otherwise, all data represented at least three individual experiments and presented as means ± standard error of the mean (SEM).Means of normally distributed data were compared by twotailed Student's t-test, one-way ANOVA (followed by Tukey's post-hoc correction) or linear regression where indicated and statistical signi cance was reported as p-values.A p-value < 0.05 was considered signi cant.Sample sizes were chosen based on an extensive literature search and standard exclusion criterion of two standard deviations from the mean were applied.Figure 2 GRAF1 does not alter adipocyte speci cation but is required for brown fat formation and function.a. iBAT and scWAT isolated from 1 week old WT and GRAF1 gt/gt mice exhibited similar levels in adipose marker gene expression as assessed by qRT-PCR (n=6-9/group), indicating that GRAF1 is not necessary for white or brown adipose tissue speci cation.b, iBAT isolated from 3 month old GRAF1 gt/gt mice exhibited signi cantly lower levels of brown fat marker genes that mediate thermogenesis compared to littermate control mice as assessed by qRT-PCR (n=7).c, Representative H&E stain of iBAT isolated from 3 month old GRAF1 gt/gt and WT mice.Note increased levels of lipid droplets (un-stained white vesicles) indicative of BAT whitening in GRAF1 gt/gt iBAT depot.(n=3/group).d, scWAT isolated from 3 month old GRAF1 gt/gt mice exhibited signi cantly higher levels of white fat marker genes and lower levels of beige-fat associated genes compared to littermate control mice as assessed by qRT-PCR.(n=6-7/group).e-g, GRAF1 gt/gt and WT mice were housed at sub-thermoneutral temperatures (22-25°C) and subjected to overnight fasting followed by an acute bout of cold stress (6°C) and their body temperatures were measured over time using a rectal thermometer(e) and serum lactate(f)and soleus triglyceride(g) level were measured .(n=4-5 WT, n=3 GRAF1 gt/gt ).Data(b,d,e,f,g) are represented as mean ± SEM, *P < 0.05; **P < 0.01; ***<0.001;****P< 0.0001 by two-tailed student's t-test.Figure 4 GRAF1 promotes subcutaneous adipose beigeing.a. WT mice were treated for 10 days with CL316243, and , qPCR analysis was performed on scWAT to assess mRNA levels of indicated genes; n=5/group.b. qPCR analysis of scWAT from WT and GRAF1 AKO mice revealed decreased Cpt1a, FGF1 and FGF21 in GRAF1 AKO mice.n=5-7/group.Data are represented as mean ± SEM, ns, not signi cant; *P < 0.05; **P < 0.01 ***<0.001;****P < 0.0001 by two-tailed student's t-test.

Declarations Figures
Figure 5