Diet-induced adipose tissue expansion is mitigated in mice with a targeted inactivation of mesoderm specific transcript (Mest)

Interindividual variation of white adipose tissue (WAT) expression of mesoderm specific transcript (Mest), a paternally-expressed imprinted gene belonging to the α/β-hydrolase fold protein family, becomes apparent among genetically inbred mice fed high fat diet (HFD) and is positively associated with adipose tissue expansion (ATE). To elucidate a role for MEST in ATE, mice were developed with global and adipose tissue inactivation of Mest. Mice with homozygous (MestgKO) and paternal allelic (MestpKO) inactivation of Mest were born at expected Mendelian frequencies, showed no behavioral or physical abnormalities, and did not perturb expression of the Mest locus-derived microRNA miR-335. MestpKO mice fed HFD showed reduced ATE and adipocyte hypertrophy, improved glucose tolerance, and reduced WAT expression of genes associated with hypoxia and inflammation compared to littermate controls. Remarkably, caloric intake and energy expenditure were unchanged between genotypes. Mice with adipose tissue inactivation of Mest were phenotypically similar to MestpKO, supporting a role for WAT MEST in ATE. Global profiling of WAT gene expression of HFD-fed control and MestpKO mice detected few differences between genotypes; nevertheless, genes with reduced expression in MestpKO mice were associated with immune processes and consistent with improved glucose homeostasis. Ear-derived mesenchymal stem cells (EMSC) from MestgKO mice showed no differences in adipogenic differentiation compared to control cells unless challenged by shRNA knockdown of Gpat4, an enzyme that mediates lipid accumulation in adipocytes. Reduced adipogenic capacity of EMSC from MestgKO after Gpat4 knockdown suggests that MEST facilitates lipid accumulation in adipocytes. Our data suggests that reduced diet-induced ATE in MEST-deficient mice diminishes hypoxia and inflammation in WAT leading to improved glucose tolerance and insulin sensitivity. Since inactivation of Mest in mice has minimal additional effects aside from reduction of ATE, an intervention that mitigates MEST function in adipocytes is a plausible strategy to obviate obesity and type-2-diabetes.


Introduction
The global rise in obesity rates are caused by interactions between an individual's genetic predisposition, the environment and epigenetic factors. The genetic component has been clearly demonstrated in studies of monozygotic twins with heredity predicted to account for~40-70% of obesity [1,2]. However, recent genome wide association studies (GWAS) of 97 body mass index (BMI)-associated loci in >300K individuals was shown to account for only <3% of the variance of BMI [3]. Although it is estimated that risk alleles may account for up to~20% of this variance [3], a large percentage of BMI is unexplained which underscores the complexity of interactions between genomic, environmental and epigenetic components [4][5][6]. Therefore, the use of murine models in which the genetic background and environment can be carefully controlled is a powerful tool for the identification of potential epigenetic determinants associated with the development of obesity and related disease.
Variable development of diet-induced obesity (DIO) and type 2-diabetes (T2D) within inbred populations of mice has been well described [7][8][9][10][11] and shown to be stable among individual animals which suggests that an epigenetic mechanism is likely involved in its etiology [10,12,13]. Analyses of gene expression in white adipose tissue (WAT) of inbred C57BL/6J mice that showed diversity in DIO identified a novel set of genes that correlate with fat mass accretion [10,[13][14][15][16]. Mesoderm specific transcript (Mest), a maternally imprinted gene, showed variations of mRNA up to~80-fold in WAT from individual mice fed an obesogenic diet which was positively associated with the rate of fat mass deposition [10,16]. In addition, Mest mRNA in WAT biopsies prior to feeding mice an obesogenic diet was predictive of future inter-individual development of DIO [10]. Compelling evidence for a role for Mest in facilitating fat accumulation was obtained by transgenic overexpression of Mest in cell culture and in mouse adipose tissue; and, in the analyses of Mest gene and protein in WAT during developmental growth and in adult mice fed dietary fat [7,10,13,14,[16][17][18][19][20]. Recent studies have determined that adipose tissue as well as blood levels of Mest mRNA is a rapid and early biological indicator of adipose tissue expansion (ATE) in mice [21]. ATE, or fat mass expansion, is a characteristic of obesity caused predominantly by triglyceride storage and expansion of adipocytes (i.e. adipocyte hypertrophy), and to a lesser extent, increased adipocyte hyperplasia [22,23].
The positive association of Mest with ATE has been well established; however, little is known in regards to the catalytic function of MEST and whether its activity directly contributes to adipocyte hypertrophy and/or adipose tissue remodeling. MEST belongs to the α/βhydrolase family of proteins [24,25] and contains the catalytic triad serine-histidine-aspartate (amino acids 145-147) which is commonly associated with serine proteases, lipases and acyltransferases [24,26]. Evidence showing that MEST is localized within the endoplasmic reticulum/Golgi apparatus of the adipocyte further supports its function in the facilitation of fat storage in adipocytes [16,20,27]. Alternatively, MEST contains a conserved epoxide-coordinating tyrosine in its sequence and shows sequence similarity to epoxide hydrolases from M. tuberculosis [28,29]. Epoxide hydrolases modulate levels of epoxy fatty acids and their diol derivatives which have been shown to act as endogenous mediators of PPARs alpha and gamma [30][31][32][33][34]. Studies have shown that epoxyeicosatrienoic acid (EET) analog treatment suppresses adipogenesis of 3T3-L1 preadipocytes and reduces ATE and improves glucose tolerance in mice fed a high fat diet [35]. It has been suggested that mitigation of adipogenic differentiation by EET may occur via Pgc1α-dependent activation of the heme oxygenase-1 (Hmox1)-pAKT signaling pathway [36,37]. Other studies suggest that Mest may act to attenuate Wnt signaling in 3T3-L1 cells by inhibiting the maturation and exit of LRP6 from the endoplasmic reticulum by controlling its glycosylated state [27].
In adult mice, MEST protein is highly expressed in white adipose tissue with very little or undetectable expression observed in brown adipose tissue, whole brain, liver, heart, kidney and spleen [16]. Recent studies suggest that human MEST could be an epigenetic target for periconceptual or intrauterine environmental exposures, such as parental obesity and gestational diabetes, leading to changes in its methylated state [38,39]. Although few studies have examined MEST with respect to human adipose tissue, it has been shown that MEST is induced during adipogenesis of isolated human adipocytes [40,41] and human abdominal subcutaneous white adipose tissue biopsies show increased MEST mRNA in obese compared to non-obese individuals [41].
In this study we generated isogenic mouse models with both global and conditional targeted inactivation of Mest to investigate its role in adipose tissue development and expansion. Our data suggests that MEST contributes to ATE by facilitating lipid accumulation in adipocytes in an obesogenic environment thereby enhancing hypoxia and inflammation which lead to impaired glucose tolerance and insulin sensitivity.

Validation of gene targeted Mest knockout mice
A targeting vector containing a neomycin (NEO) selection cassette was used to insert loxP sites immediately 5' and 3' of Mest exon 3 using C57BL/6J (B6) ES cells (Fig 1A). Subsequent breeding of mice with germline transmission of the targeted Mest allele with Ella-cre transgenic mice resulted in the selection of progeny with deletions of only the NEO cassette or the floxed region (NEO cassette and Mest exon 3) resulting in an intact floxed or a global deletion of Mest exon 3. Deletion of exon 3 of Mest creates a stop codon in exon 4 and full inactivation of Mest as shown by the absence of Mest mRNA in white adipose tissue (WAT) of homozygous knockout (Mest gKO ) mice ( Fig 1B). Mest, an imprinted gene that is almost exclusively expressed from the paternal allele in adult mice [42][43][44][45] and highly expressed in WAT [10,16,46], is completely abrogated in mice with a paternal allelic inactivation of Mest (Mest pKO ), but shows fully intact mRNA and protein levels when the maternal allele of Mest is selectively inactivated (Mest mKO ; Fig 1B and 1C). Unlike previous observations showing abnormal maternal behavior and high incidence of postnatal mortality in mice with global or paternal inactivation of Mest on a mixed B6/129 background (Lefebvre et al 1998), our studies indicate normal litter sizes reared by dams with inactivated Mest (~8.2 pups/litter); and non-skewed gender (29 female/30 male) and genotype ratios with 14 wildtype (WT), 30 Mest mKO or Mest pKO and 15 Mest gKO (chi-square 0.119; P = 0.998) born from crosses between heterozygous females and males. The basis for phenotypic differences in maternal behavior and early post-natal mortality between our studies and those of Lefebvre et al 1998 [43] are not known.
Since the Mest locus harbors a microRNA (miR-335), which has also been implicated in adipogenesis and lipid metabolism [47,48], within an intronic sequence between exons 2 and 3 of Mest transcript variant 2 (TV2), an analyses was performed to determine whether the close proximity of a loxP site (Fig 1D) from the targeting vector aberrantly affects expression of miR-335. Analyses of inguinal (iWAT) WAT of 6-7 day old mice, a developmental time period when fat mass is rapidly expanding and WAT Mest is highly expressed [16], showed no differences in miR-335 expression in mice with intact (WT and Mest mKO ) versus inactivated (Mest gKO and Mest pKO ) Mest (Fig 1E). In addition, RT-PCR analyses of 3 known transcriptional variants of Mest in mice that vary in the use of alternative exons (Fig 1D) shows that transcript variant 2 (TV2; NM_008590.2) is predominantly expressed in both subcutaneous iWAT and visceral epididymal (eWAT) WAT depots (Fig 1F). Mest TV1, which uses an additional alternative exon in the 5' coding region compared to TV2, also shows a low level of expression in WAT, whereas TV3, a variant of Mest mRNA that uses an alternative 5'UTR 4.5 kb upstream from the 5'UTR containing exon of TV1 and TV2, was absent in WAT ( Fig 1F).

Bodyweight and body composition
Since Mest pKO mice do not express Mest mRNA or protein in WAT (Fig 1B and 1C), male progeny (WT and Mest pKO ) for the dietary obesity study were generated from a cross between female WT and male Mest mKO mice. From weaning (week 3) until 8 weeks of age, Mest pKO mice showed a small but significant reduction in body weight (Fig 2A) but no differences in adiposity measured as the ratio of fat mass/lean mass ( Fig 2B) compared to WT littermate controls. However, when fed an obesogenic diet (58 kcal% fat) from 8 until 16 weeks of age, dietary fat-induced increases in bodyweight, adiposity and fat mass deposition in Mest pKO mice was significantly attenuated compared to WT mice (Fig 2A, 2B and 2C) whereas fat free mass (FFM) was unaffected (Fig 2C; inset). Regression analyses of the 17 WT mice at 16 weeks of age showed a strong correlation between eWAT Mest mRNA and fat mass deposition (R = 0.70; p<0.002) as well as Mest mRNA expression among adipose depots (iWAT vs eWAT Mest; R = 0.59; P<0.02) within each mouse which is consistent with previous studies [10,16]. Analyses of subgroups of WT mice within the lowest (n = 6; WT-LM) and highest (n = 6; WT-HM) tertiles for eWAT Mest mRNA (Fig 2D) showed significant differences in BWT and adiposity between groups (Fig 2E and 2F). However, no differences were observed between WT-LM and Mest pKO mice (Fig 2E and 2F). These data suggest that the low level of MEST in adipose tissue of WT-LM mice does not significantly contribute to additional ATE beyond phenotypic analyses between WT and pKO (A-C) was determined by 2-way ANOVA using the Holm-Sidak multiple comparisons test. Time points annotated with 1, 2, 3 or 4 asterisks indicate significant differences of P 0.05, P 0.01, P 0.001 and P 0.0001 respectively. Significance in eWAT Mest RNA expression between the WT-LM and WT-HM groups (D) was measured by a two-tailed unpaired parametric t-test and is indicated on the bar graph. Longitudinal data for WT-LM, WT-HM and pKO (E and F) was performed using 2-way ANOVA using Tukey's multiple comparisons test. Annotation with 'a' and 'b' indicates significant differences between WT-HM vs pKO and WT-HM vs WT-LM mice respectively.

Adipose tissue inactivation of Mest
Mice with a floxed paternal allele for Mest (Mest pFL ; Fig 1A) were crossed with Adipoq-cre transgenic mice to generate animals with an Adipoq-cre mediated inactivation of the paternal allele (Mest ApKO ) of Mest. Analyses of WAT Mest mRNA in cohorts of 5-17 age-matched male mice fed an obesogenic diet for 16 weeks showed a >98% reduction in both iWAT and eWAT Mest mRNA expression (Fig 3A) in Mest ApKO mice compared with the 3 control groups (i.e. WT, WT-cre and Mest pFL ; Fig 3A) and supports previous studies showing that Mest is expressed in the mature adipocyte fraction of adipocytes [10,20]. In addition, the presence of loxP sites in the 5' and 3' introns surrounding Mest exon 3 of Mest pFL did not adversely affect Mest expression. Longitudinal analyses of bodyweight and body composition in mice fed a HFD shows significantly reduced bodyweight gain and fat mass accumulation in the Mest ApKO mice (Fig 3B and 3C), but no changes in fat-free mass (Fig 3D), compared to the 3 control groups. Reduced bodyweight and adiposity in Mest ApKO mice is consistent with the phenotypes observed for the Mest pKO mice (Fig 2). Both genetic models suggest that MEST facilitates ATE in an obesogenic environment. An additional cohort of mice with Fabp4-cre mediated inactivation of Mest show similar phenotypic profiles compared to the Mest ApKO mice (S1 Fig).

Metabolic parameters and adipocyte morphology
To prevent aberrant phenotypic measurements due to stress from over-manipulation of the animals, two additional cohorts of Mest pKO mice and littermate WT controls were used to examine diabetic and metabolic parameters. A cohort used to measure glucose (GTT) and insulin tolerance (ITT) showed no significant differences in bodyweight (WT, N = 8, 22.51 ± 0.26 g vs Mest pKO , N = 7, 22.13 ± 0.30 g; p = 0.35) or glucose tolerance ( Fig 4A) between chow-fed WT and Mest pKO mice at 8 weeks of age. However, after feeding mice a HFD from 8 to 16 weeks of age, Mest pKO mice showed reduced bodyweight (WT, 38.46 ± 0.96 grams vs Mest pKO , 33.71 ± 1.45 grams; p<0.02) and improved glucose ( Fig 4B) and insulin ( Fig 4C) tolerance compared to WT littermate controls.
An independent cohort of HFD-fed WT (n = 8) and Mest pKO (n = 8) mice that were subjected to indirect calorimetry showed similar differences in both bodyweight and adiposity ( Fig 4D and 4E) as previously observed. Surprisingly, caloric intake normalized to fat-free mass (kJ/d/g FFM) measured in mice fed HFD at 8 and 16 weeks of age was similar between WT and Mest pKO mice ( Fig 4F) suggesting that differences in food consumption is not a causative factor for the divergence in adiposity between genotypes. Respiratory quotient (RQ) measured in 16 week old mice after feeding a HFD for 8 weeks were identical between genotypes ( Fig 4G) and RQ values levels were consistently below 0.8 indicating that fat, as expected due to feeding the mice HFD, is the primary fuel source for both genotypes. Although no significant differences in energy expenditure (EE; kcal/hr/kg FFM) were noted between genotypes ( Fig 4H) at 16 weeks of age, Mest pKO mice did show a~4-5% elevation in 24 hour EE compared to WT mice. Analyses of physical activity based on wheel running at 16 weeks of age shows that Mest pKO mice consistently ran~15% further (Fig 4I; p 0.05) and faster (Fig 4J; p 0.01) than WT mice at night which could at least partially explain the slight increase in EE observed for Mest pKO mice.
Adipocyte morphology for both eWAT and iWAT, determined for the cohort of mice used in the HFD study described in Although the size of adipocytes in iWAT ( Fig 5C) was not different between genotypes, the general trend of increased adipocyte hypertrophy in mice within the HM group was evident in both WAT depots. These results are consistent with previous studies using B6 mice fed a HFD Adipoq-cre (ApKO; n = 17) mice after being fed a HFD from 8-24 weeks of age. Mest mRNA expression measured by TaqMan QRT-PCR is represented as the mean ± SEM of arbitrary units (AU) normalized to TATA-box binding protein (Tbp). Significance in Mest RNA expression between groups was determined via one-way ANOVA using Tukey's multiple comparisons test. (B) Data shows the longitudinal measurements of bodyweight (BWT) for WT (n = 9), WT-cre (n = 5), pFL (n = 11) and ApKO (n = 17) mice fed a HFD from 8 to 24 weeks of age as indicated by the arrow along the X-axis. (C) Longitudinal measurements of fat mass (g) and (D) fat-free mass (g) measured by DEXA at the times indicated on the X-axis are shown for WT (n = 9), WT-cre (n = 5), pFL (n = 11) and ApKO (n = 17) mice fed a HFD from 8 to 24 weeks of age as indicated by the arrow along the X-axis. (B-D) All data in the longitudinal studies are presented as the mean ± SEM. Significance at each time point of the longitudinal phenotypic analyses was determined by 2-way ANOVA and Tukey's multiple comparisons test. Time points annotated with 'a' and 'b' indicates significant differences between 'ApKO vs pFL' and 'ApKO vs WT'.  that showed a positive correlation between adiposity and adipocyte size with WAT Mest mRNA and protein [16].

Gene expression analyses
Gene expression markers for macrophage infiltration/inflammation were used to determine whether improved glucose tolerance in HFD-fed Mest pKO mice ( Although differences in the expression of these gene markers between groups were not as prominent in iWAT compared with eWAT, they were consistent with that observed for eWAT except for Itgax which showed no differences between groups. The absence of differences in gene expression between the WT-LM and performed via DEXA. Indirect calorimetric measurements of (G) respiratory quotient (RQ), (H) energy expenditure (EE), (I) wheel running distance (WhDst) and (J) wheel speed (WhSpd) were performed on mice for 1 week after being fed a HFD for 8 weeks. All data is presented as the mean ± SEM. Significance between groups was determined using two-tailed unpaired parametric t-tests and 1 or 2 asterisks indicate significant P-values of 0.05 and 0.01 respectively.
https://doi.org/10.1371/journal.pone.0179879.g004 Mest pKO mice suggest that changes in gene expression in Mest pKO mice are associated with the reduction of ATE and suppression of inflammation. Furthermore, the analyses of hepatic gene expression in WT vs Mest pKO mice show no differences in the expression of Fas (WT; 10.09 ± 0.75 vs Mest pKO ; 10.47 ± 1.53 AU/Tbp; p = 0.83) and Scd1 (WT; 11.55 ± 0.70 vs Mest pKO ; 11.93 ± 1.22 AU/Tbp; p = 0.79), genes involved in hepatic de novo lipogenesis that have been shown to be positively associated with hepatic steatosis [49][50][51][52]. These data indicate that reduced ATE in Mest pKO mice does not result in the development of a lipodystrophic state characterized by excessive hepatic accumulation of lipid. Gene expression was determined via TaqMan qRT-PCR in total RNA from eWAT and iWAT of wildtype (WT; n = 17), mice with a paternal inactivation of Mest (pKO; n = 10) and subgroups of WT mice with low (LM; n = 6) and high (HM; n = 6) WAT Mest expression after being fed a HFD from 8-16 weeks of age. Gene expression is represented as arbitrary units (AU) normalized to Tbp and presented as the mean ± SEM. Statistical analysis between genotypes within each WAT depot was performed using two-tailed unpaired parametric t-tests and datasets annotated with the same letter indicate no significant differences between groups. https://doi.org/10.1371/journal.pone.0179879.g006 Activation of both canonical and non-canonical Wnt signaling has been shown to inhibit adipogenesis, mediate adipose tissue development and expansion and promote adipose tissue inflammation in obesogenic conditions [53][54][55]. Because previous studies have shown that adipose tissue Mest mRNA expression is positively associated with a subset of genes that include Wnt signaling inhibitors (i.e. Sfrp5 and Nkd1), in addition to the Tgfb-pathway mediator Bmp3 and the serine protease inhibitor Serpine1 [10,13,16], an analyses of this subset of genes in eWAT and iWAT of WT and Mest pKO mice from the HFD study described in Fig 2 was performed to determine effects of Mest inactivation on their expression. Surprisingly, inactivation of Mest did not significantly alter WAT expression of Sfrp5, Bmp3, Nkd1 and Serpine1 (Fig 7B,  7C, 7D and 7E) with levels of expression comparable to WT-LM mice. However, consistent with previous studies [10,13,16], WT-HM mice did show elevated levels of Sfrp5, Bmp3, Nkd1 and Serpine1 expression compared with WT-LM mice for both WAT depots (Fig 7B, 7C  subgroups of WT mice with low (LM; n = 6) and high (HM; n = 6) WAT Mest expression after being fed a HFD from 8-16 weeks of age. Gene expression was determined via TaqMan qRT-PCR in total RNA from eWAT and iWAT and represented as arbitrary units (AU) normalized to cyclophilin b (Ppib). Gene expression is presented as the mean ± SEM and significance between groups was determined using two-tailed unpaired parametric t-tests. Datasets annotated with the same letter indicate no significant differences between groups. https://doi.org/10.1371/journal.pone.0179879.g007 A global analysis of gene expression via ABI SOLiD SAGE sequencing was used to identify genes and genetic pathways that are dysregulated in adipose tissue of mice with a targeted inactivation of Mest (Fig 8). Gene profiles for eWAT were analyzed for 6 WT and 5 Mest pKO mice selected from the cohort used in the HFD study in Fig 2. Adiposity indices (FM/LM ratio) of the WT mice were significantly higher than Mest pKO mice (WT, 0.724 ± 0.030; Mest pKO , 0.571 ± 0.019; p = 0.003) after the 8 weeks of HFD. A Venn diagram (Fig 8A) shows surprisingly few genes from the 15,453 gene targets detected in eWAT to have significantly higher (n = 27) or lower expression (n = 9) in WT vs Mest pKO mice when FDR-adjusted to P<0.1 (S1 Table). Mest, as expected, was significantly reduced >15-fold in eWAT of Mest pKO mice and TaqMan qRT-PCR validation showed undetectable (reduced >100-fold) levels of Mest in eWAT and iWAT of Mest pKO mice ( Table 1). Analyses of 14 additional selected genes with RNA from the individual eWAT samples used for ABI SOLiD SAGE sequencing via qRT-PCR showed significant and consistent differences between WT and Mest pKO mice in the expression of 10 genes whereas 2 genes (Negr1, Mrap) showed no significant differences between groups and 2 genes (Plagl2, Tph2) could not be detected (Table 1). Additional analyses of eWAT RNA from the entire cohort of 17-WT and 10-Mest pKO mice also showed comparable patterns of expression for the 10 differentially-expressed genes with the global and sample validation studies (Table 1). Importantly, 7 of the 10 genes showed similar expression patterns for both iWAT and eWAT whereas the expression of 3 genes (Gsta3, Ucp2 and Ces1f) only showed differences in eWAT.
Ontology and enrichment analyses of the 27 genes with reduced expression in eWAT of Mest pKO mice showed associations with several biological processes including cation transport (Tfr2, Atp6v0a1, Atp6v0d2, Slc5a7, Ccdc109b and Trdn; P<10 −4 ), regulation of growth of symbiont in host (Lbp, Irf8; P<10 −3 ), myeloid leukocyte activation (Lbp, Pirb, Ubd; P<10 −3 ) and acute phase response (Lbp, Hp; P<10 −3 ); however, none of the associations were significant after FDR correction. Genes more highly expressed in eWAT of Mest pKO showed modest association with mycotoxin/aflatoxin catabolism and metabolism (Gsta3; P<10 −3 ) and monoterpenoid metabolism (Cyp2e1; P<10 −3 ). Further enrichment analyses performed on non-FDRadjusted gene sets with a P-value cutoff of 0.05 and a minimum of 50 sequence tags for each mRNA in at least one of the groups identified 695 gene targets that were differentially expressed in eWAT of WT vs Mest pKO mice (S2 Table). The 395 genes with reduced expression in eWAT of Mest pKO mice showed highly significant associations with immune system processes (Padjusted <10 −10 ) including defense, immune, inflammatory and stress responses whereas genes with increased expression in eWAT of Mest pKO mice (n = 301) showed enrichment for single-organism metabolic processes (Padjusted<10 −3 ) including oxidation-reduction, organic acid metabolism; and, fatty acid, acyl-CoA, thioester and monocarboxylic acid metabolism (Fig 8B). An examination of the expression of the mature mouse macrophage gene marker F4/80 (Emr1; Adgre1), and gene markers that delineate classically (M1) versus alternatively (M2) macrophage subtypes [56] only showed significantly elevated expression for the M1 macrophage marker Gpr18 (p = 0.04) and M2 macrophage marker Egr2 (p = 0.02) in eWAT of WT mice. No differences in expression between genotypes were observed for F4/80, Cd38, Fpr2, Cd163, Cd68 and Myc.

Adipogenic differentiation of Ear-derived mesenchymal stem cells
In vivo inactivation of Mest both globally and in adipocytes was shown to reduce adiposity (Figs 2 and 3); therefore, studies were initiated to determine whether similar effects of Mest inactivation can be observed in an adipogenic population of primary ear-derived mesenchymal progenitor cells (EMSC) from mice [57]. Results showed that EMSC from Mest gKO mice  Table).
were comparable in their capacity for adipogenic differentiation and accumulation of lipids compared to EMSC from WT mice as indicated by the panels showing oil red O (ORO) staining of lipids and crystal violet (CV) staining of cell nuclei ( Fig 9A); and, in longitudinal profiles of lipogenesis measured by spectrophotometric determination of ORO in two independent experiments (Fig 9B and 9C). Although these results are somewhat contradictory to what is observed with in vivo studies, one possible explanation for the lack of effect of Mest inactivation in EMSC adipogenesis could be that cellular access to abundant nutrients and robust expression of lipid biosynthetic glycerol-and acyl-glycerol phosphate acyl transferases (GPATs and AGPATs) simply overwhelms the contribution MEST might have in facilitating lipid accumulation. By using an approach similar to one that used siRNA knockdown of GPAM in HEK293 cells to confirm GPAT-function for AGPAT6 [58], now renamed GPAT4, it was hypothesized that attenuation of GPAT/AGPAT function in EMSC might allow the contribution of MEST in lipid accumulation in EMSC undergoing adipogenesis to emerge. To test this possibility, EMSC were transduced with a lentiviral (LV) construct expressing shRNA against Gpat4, a glycerol-3-phosphate O-acyltransferase that catalyzes the acylation of glycerol-3-phosphate in the first and rate limiting step of de novo triacylglycerol and glycerophospholipid synthesis. Gpat4 was selected because, like Mest, it is localized within the endoplasmic reticulum (ER) [59] and is expressed during adipogenic differentiation of EMSC derived from both WT and Mest gKO mice (Fig 9D and 9E). Expression of Gpat4 in EMSC from WT mice transduced with LV shRNA against Gpat4 (GP4KD) was reduced~65-70% 5 days after initiation of adipogenic differentiation (Fig 9D). The expression of Agpat2 and Gpat3, which have 42% and 67% coding sequence identity with Gpat4 and are highly expressed in EMSC during adipogenic differentiation, are unaffected in the GP4KD EMSC (Fig 9D). Interestingly, Mest expression was  modestly increased~30% in GP4KD EMSC which suggests that its activation is a possible compensatory response to the attenuation of GPAT4 function. An independent longitudinal study using EMSC from WT and Mest gKO mice in the absence or presence of GP4KD shows comparable temporal patterns of Gpat4 expression during 7 days of adipogenic differentiation among genotypes which was significantly attenuated by the GP4KD (Fig 9E). Mest expression in WT GP4KD EMSC was initially lower than WT EMSC during the first 3 days after adipogenic differentiation was initiated (i.e. D0-D3), but was higher than WT EMSC by day 5 postdifferentiation (Fig 9F) which is consistent with previous results (Fig 9D). While GP4KD transduced WT and Mest gKO EMSC showed highly attenuated lipogenesis (Fig 9G) compared to non-transduced cells (Fig 9A), GP4KD was significantly more effective in mitigating lipogenesis in Mest gKO EMSC compared with WT EMSC as observed in the ORO/CV stained image ( Fig 9G) and in the spectrophotometric measurements of ORO stained lipids in two independent longitudinal studies (Fig 9H and 9I). These data suggest that intact MEST function in WT EMSC diminishes GP4KD-mediated reductions of lipogenesis and supports a role for MEST in facilitating lipid accumulation.

Discussion
Since adipose tissue Mest was found to be positively associated with variable development of diet-induced obesity within a genetically homogeneous population of mice [10], a number of studies by our laboratory and others have utilized Mest as a sensitive biological indicator for adipocyte hypertrophy and fat expansion [7,14,16,21,60]. The initial characterization of mice with a targeted inactivation of Mest on a mixed genetic background (Mest tm1Lef ) did not describe adiposity-related phenotypes; however, these mice were smaller than control animals, had high rates of early post-natal lethality and showed abnormal maternal behavior [43]. Subsequent analyses of a subset of 9-month old Mest tm1Lef mice on a mixed B6/129 background showed modestly reduced bodyweights and fat pad weights suggesting a role for MEST in the storage of lipid in adipocytes [16]. To avoid problems associated with abnormal maternal behavior and post-natal lethality in Mest-deficient mice, which would confound the interpretation of MEST function in diet-induced obesity and adipose tissue expansion (ATE), we generated an isogenic (C57BL/6J) mouse model with loxP sites flanking the 3 rd exon of Mest to selectively inactivate Mest in a tissue-specific manner. Surprisingly, the creation of mice with Ella-cre-mediated global inactivation of Mest using our targeted model, unlike the Mest tm1Lef mice [43], did not result in increased risk of early post-natal lethality, reduced post-natal longitudinal growth, or abnormal maternal behavioral phenotypes. The reasons for differences between our targeted Mest mouse model and Mest tm1Lef are not well understood but could be due to differences in the genetic background of the mice, the gene targeting construct, or mutation(s) in the host ES cell used for targeting [61]. Recently it was reported that backcrossing targeted Mest tm1Lef mice 5 generations onto a129/SVJ background also eliminated the early postnatal lethality and maternal behavioral phenotypes of mice with a paternal inactivation of Mest; however, the explanation why this occurred was not clearly defined [62].
A microRNA (miR-335) suggested to have a role in the regulation of lipid metabolism, adipogenesis, adipose tissue inflammation and chondrogenesis [47,48,63] is located within the 2 nd intron of Mest TV2 (Fig 1D) and only~595 bp upstream (5') from the loxP/NEO cassette used to generate the targeted Mest allele. The regulation of miR-335 is complex with some studies suggesting that it is a paternally expressed imprinted microRNA that is co-regulated with Mest whereas others suggest that it can be regulated via its own promoter [48,64]. Since it is plausible that perturbed miR-335 function could modify ATE in our mouse model, it was necessary to determine whether miR-335 is dysregulated in mice with targeted Mest. Analyses of miRNA-335-5p expression in iWAT of 6-7 day old mice, at a time when fat accretion is rapid and Mest is highly expressed [16], showed comparable levels of miR-335 for all genotypes (Fig 1E) with no evidence for dysregulation of miR-335 in mice with inactivated Mest. Normal miR-335 expression in mice with a targeted allele of Mest suggests that phenotypic differences primarily result from loss of MEST function.
HFD-fed mice with inactivation of Mest on the paternal allele (Mest pKO ) showed significantly reduced bodyweight (Fig 2A), adiposity (Fig 2B and 2C), adipocyte hypertrophy (Fig 4) and improved glucose and insulin tolerance (Fig 4B and 4C); but, relatively few differences in metabolic parameters measured by indirect calorimetry (Fig 4G, 4H, 4I and 4J) compared to WT littermate controls. Although a reduction in bodyweight of just a little over 4 grams between male Mest pKO and WT mice after being fed a HFD for 8 weeks appears to be modest, this would be equivalent to a 177 vs 200 pound individual with~75% (~17.3 lbs) of the additional bodyweight being fat mass. Additional studies using Mest pFL mice with either Adipoqor Fabp4-cre mediated inactivation of the paternal Mest allele primarily in adipocytes (Fig 3;  S1 Fig) showed similar reductions in both bodyweight and adiposity compared to their respective littermate controls suggesting that inactivation of Mest in adipocytes is predominantly responsible for the observed differences in adiposity related phenotypes between genotypes.
A predicted complication in assessing phenotypic differences between WT mice and mice with global or adipose tissue specific inactivation of Mest was the expected variability in the development of obesity within the WT cohort and its positive association with WAT MEST [7,10,16]. This was in fact the case and was exemplified by the analyses of tertiles (n = 6) of WT mice with the lowest (WT-LM) and highest (WT-HM) eWAT Mest mRNA expression (Fig 2D  and 2E). Our analyses showed that differences in bodyweight and adiposity were more pronounced when comparing Mest pKO with WT-HM mice and were consistent with previous studies of interindividual variation of Mest in the absence of genetic heterogeneity. When tertiles for male WT mice based on eWAT Mest expression are extrapolated to human body weight, WT-HM and WT-LM would weigh 215 vs183 pounds respectively. The lack of difference observed for bodyweight, adiposity, and adipocyte hypertrophy between WT-LM and Mest pKO mice fed a HFD suggests that 'basal' levels of ATE do not require MEST. Additionally, studies showing increased adipocyte hypertrophy in mice with transgenic overexpression of Mest [20] coupled with high interindividual variability in WAT Mest expression in a genetically homogeneous population of mice [10,16,65] supports the premise that MEST could be an 'epigenetically' regulated rheostat for the control of ATE beyond defined basal levels established by the genetic background.
A concern with the ablation of Mest in a murine mouse model was that it could reduce the capacity for WAT to expand in an obesogenic environment leading to adipose dysfunction, ectopic accumulation of hepatic and intramuscular lipid, and impaired insulin sensitivity similar to that observed in lipodystrophy [66,67]. However, Mest pKO mice fed a HFD for 8 weeks showed markedly improved glucose homeostasis compared with WT controls (Fig 4B and 4C) which suggest that the suppression of ATE in Mest pKO mice does not adversely affect metabolic health of WAT. These observations are supported by reduced WAT expression of markers for inflammation, hypoxia and macrophage infiltration in Mest pKO and WT-LM mice compared to WT-HM mice (Fig 6) and lack of differences in hepatic expression of genes shown to be associated with the development of hepatic steatosis [49][50][51][52]. Data suggests that improved glucose tolerance is at least partially caused by reduced ATE in mice with inactivated or low WAT Mest expression. Improved glucose metabolism in association with reduced adipocyte lipid accumulation in mice with deficient Mest, caused by reduced adipocyte lipid accumulation shows similarity to the maintenance of normal glucose metabolism by caloric restriction; or, by mechanisms that enhance energy expenditure such as physical activity or cold-activated thermogenesis [68][69][70][71][72][73].
Although WT mice accrued significantly more bodyweight and adiposity than Mest pKO mice after being fed a HFD for 8 weeks (Fig 4D and 4E), there were no differences in caloric intake between genotypes when measured during the initial (wk 8) and final week (wk 16) of high fat feeding (Fig 4F). Since indirect calorimetric measurements showed a consistent but non-significant (~4-5%) increase in energy expenditure for Mest pKO mice during the day and night (Fig 4H) as well as significantly increased nighttime activity (Fig 4I and 4J), it is possible that these small differences in energy expenditure during the 8-week longitudinal study could contribute to some of the phenotypic differences in bodyweight and adiposity between genotypes.
Genes (i.e. Sfrp5, Bmp3, Nkd1 and Serpine1) expressed in WAT previously identified as being positively correlated with WAT Mest mRNA and variable development of adiposity in a genetically homogeneous population of mice fed obesogenic diet [7,10,16] were not significantly effected in WAT of Mest pKO mice (Fig 7B, 7C, 7D and 7E). However, consistent with previous studies, analyses of WAT gene expression in the WT mice used in the HFD study in Fig 2 showed significant positive associations between eWAT Mest vs Sfrp5, Bmp3 and Nkd1; and, between iWAT Mest vs Sfrp5, Bmp3, Nkd1 and Serpine1. Since these genes were expressed at similar levels in WAT of HFD-fed Mest pKO and WT-LM mice (Fig 7A,  7B, 7C, 7D and 7E), both of which show comparable weight gain and adiposity (Fig 2E and  2F), this suggests that differences in adiposity rather than WAT MEST affects their expression.
RNA-Seq analyses performed with RNA from eWAT of WT and Mest pKO mice to gain insight into gene pathways and mechanisms related to MEST function revealed relatively few differences in gene expression (~36 genes) between genotypes when corrected for FDR at a pvalue of <0.1 (Fig 8A) which suggests that ablation of Mest has relatively modest effects on gene expression in adipose tissue. This is an exciting feature of the association of reduced MEST and ATE with improved glucose/insulin metabolism because, unlike many other antiobesity therapies, it provides encouragement that targeting the MEST pathway to reduce ATE may have a minimal number of side effects. Pathway analyses of genes with reduced and increased expression in eWAT of Mest pKO mice (Fig 8A) showed modest associations with immune system-based pathways and metabolism/catabolism of toxins, respectively. The use of a broader set of genes with less stringent criteria for significance (S2 Table) in the gene ontology enrichment analyses also showed genes with reduced expression in eWAT of HFD-fed Mest pKO mice to be associated with immune system processes (Fig 8B), including inflammation, which is consistent with improved glucose tolerance and insulin sensitivity in these mice. Genes more highly expressed in eWAT of Mest pKO were categorized within multiple metabolic processes including those associated with metabolism of acyl-CoA, fatty acids and lipids ( Fig  8B) and included a subset of carboxylesterases (Ces1d, Ces1f), carnitine palmitoyltransferase 2 (Cpt2) and glycerol-3-phosphate acyltransferases (Gpat4, Gpat3; S2 Table). Since elevated Mest is associated with increased adiposity, the increased expression of lipid biosynthetic genes in the absence of MEST may reflect a compensatory response. GPAT4, an integral membrane protein is important for the first step in catalyzing triglyceride synthesis and has been suggested to diffuse from the endoplasmic reticulum (ER) to lipid droplets via membrane bridges [74]. Since evidence suggests that MEST is localized within the ER [16,20,27], it's plausible that interactions between MEST and GPAT4, or with other ER-localized GPATs or AGPATs, could facilitate triglyceride synthesis and storage in adipocytes. Studies are currently underway to delineate temporal patterns of MEST localization within the ER and its proximity to nascent lipid droplet formation.
Catalytic promiscuity and moonlighting are common among α/β-hydrolase fold proteins thus making it difficult to associate a specific catalytic function for MEST based on sequence homology [75][76][77]. Although the precise catalytic mechanism for MEST has not been defined, evidence showing a role for MEST in augmenting ATE and lipid accumulation in adipocytes [7,16,20] coupled with its intracellular localization within the ER [16,20,27] and the presence of a catalytic triad associated with proteases, lipases and acyltransferases within its peptide sequence [24,26], all suggest that MEST acts to facilitate the accumulation of lipid in adipocytes.
The observation that mesenchymal progenitor cells (EMSC) derived from mice with a targeted inactivation of Mest (Mest gKO ) showed no difference in adipogenic potential and lipid accumulation compared to WT EMSC (Fig 9A, 9B and 9C) is contradictory to results showing increased lipid accumulation in 3T3-L1 adipocytes with transgenic overexpression of Mest [20]; and, underscores the complexity involved in understanding MEST function using in vitro models. However, in lentiviral shRNA knockdown of GPAT4, a gene whose inactivation in mice causes subdermal lipodystrophy and reduced diet-induced obesity [78], in both WT and Mest gKO EMSC resulted in the emergence of differences in adipogenesis between the two genotypes with EMSC derived from Mest gKO mice showing significantly reduced lipid accumulation compared to WT EMSC (Fig 9G, 9H and 9I). These data provide strong evidence that MEST at physiological levels may augment lipid accumulation and storage during adipogenic differentiation by its own endogenous acyltransferase activity, interact with or stabilize proteins associated with fatty acid uptake or lipid droplet formation, or reduce adipogenesis-suppressing intracellular levels of EET [35][36][37] via endogenous epoxide hydrolase activity. Proteomic and lipodomic studies are needed to further delineate the precise catalytic mechanisms associated with MEST and its role in mediating ATE and glucose homeostasis in WAT.
In summary, our data shows that reduced diet-induced ATE in MEST-deficient mice diminishes expression of markers of hypoxia and inflammation in WAT which leads to improved glucose tolerance and insulin sensitivity. Since minimal additional physiological effects are observed in mice with inactivation of Mest, an intervention that represses the function of MEST, or its associated pathway, could be a feasible strategy to mitigate obesity and the inception of T2D.

Gene targeting
A cre/loxP system was used to create mice with global and conditional inactivation of mesoderm specific transcript (Mest). DNA amplified from a C57BL/6J genomic BAC clone (RP24-211g11) was used to generate a targeting vector with loxP sites flanking exon 3 of Mest transcript variant/protein isoform 2 (NM_008590.2). A third loxP site flanking a NEO cassette (G418 resistance) 5' of exon 3 was included as a selectable marker. Targeted C57BL/6J embryonic stem cells carrying the conditional floxed allele were introduced into C57BL/6J-Tyrc-2J host blastocysts and chimeric progeny crossed with C57BL/6J-Tyrc-2J mice to generate mice heterozygous for the targeted allele. The presence and correct orientation of the 3 loxP sites with respect to Mest exon 3 and the NEO cassette were confirmed via PCR. Generation of mice with global inactivation of Mest and with a floxed Mest exon 3 in the absence of the NEO cassette was achieved by crossing mice carrying the targeted allele to Ella-cre mice which has restricted expression of cre recombinase under the control of the adenoviral Ella promoter in oocytes and in preimplantation embryos [79]. Progeny that show positive PCR amplification patterns that correspond to a complete deletion of the floxed region and/or deletion of the NEO cassette were backcrossed to C57BL/6J mice to ensure germline transmission of the mutation and removal of the Ella-cre transgene.

Animals and study design
Since Mest is almost exclusively expressed from the paternally-derived allele in adult mice, experimental cohorts of mice with a paternal inactivation of Mest (Mest pKO ) were generated by crossing wildtype (WT) C57BL/6J female mice with male mice heterozygous for the inactivated (KO) Mest allele on the maternal allele (Mest mKO ). This cross yields equal ratios of WT and Mest pKO mice. Complete knockout of Mest in tissues of Mest pKO mice have been verified by both gene expression and Western blot analysis (Fig 1B and 1C). Longitudinal dietary obesity studies were performed using male WT and Mest pKO mice because male C57BL/6J has greater susceptibility for the development of diet-induced obesity than females. Experimental cohorts of mice to study the conditional inactivation of Mest were generated via crosses between female mice that are hemizygous for either the Fabp4-cre (B6.Cg-Tg (Fabp4-cre) 1Rev/J; JAX Stock# 005069) or Adipoq-cre (B6;FVB-Tg(Adipoq-cre)1Evdr/J; JAX Stock# 010803) transgenes with male mice containing either a maternal (Mest mFL ) or paternal (Mest pFL ) floxed allele for Mest. The resulting genotypes will include mice with Fabp4-or Adipoq-cre inactivation of the paternal allele for Mest (Mest FpKO or Mest ApKO respectively) and control groups consisting of WT, WT-cre (Fabp4-or Adipoq-cre) and Mest pFL mice. Prior to use in generating experimental mice, the Adipoq-cre transgenic mice were crossed to C57BL/ 6J for several subsequent generations to eliminate potential genomic contribution from the FVB strain. Genomic analyses of the Adipoq-cre using a 1449 SNP marker panel (Taconic, Germantown, NY) confirms 99.79% C57BL/6J with the only residual genome of FVB located between~83.5 and~92.1 mb on mouse Chr 9, the putative genomic insertion site of the Adipoq-cre transgene [80]. The

Glucose and insulin tolerance tests
Glucose tolerance test (GTT) was performed in 7 week old male mice prior to initiation of high fat diet and repeated after 8 weeks on high fat diet. Mice were fasted overnight and administered 2 g/kg BW glucose intraperitoneally (IP). Blood glucose levels were determined using Accu-check Aviva plus system (Roche Diagnostics). Insulin tolerance test was conducted in mice fed high fat diet for 8 weeks. Mice were fasted for 6 hours and injected with human recombinant insulin (Humulin-R, Eli Lilly and Company) at a dose of 1.5 kg/BW IP.

Indirect calorimetry
The Promethion Metabolic Cage System (Sable Systems) was used for indirect calorimetry. The mice were acclimated to the system for 12-hrs followed by 72-hr data collection.

Determination of body composition
Nuclear magnetic resonance. Body composition (body fat, lean mass and free fluid) was analyzed by Minispec NMR (Bruker) which allowed for non-invasive, serial body composition measurements in the mouse. The Minispec uses the contrasting hydrogen density and/or hydrogen spin properties from adipose tissue and muscle for estimating body composition. A quality control check of NMR parameters using a standard provided by the manufacturer was performed at the beginning of each day of testing.
Dual-energy X-ray absorptiometry. Mice were scanned using the Lunar PIXImus densitometer (GE Medical Lunar; DEXA). Whole-body and femoral areal bone mineral density, and body composition excluding the head were determined in each mouse under isoflurane anesthesia (IsoFlo, Abbot Labs). The Piximus was calibrated at the beginning of each day of testing using a phantom according to the manufacturer's instructions.

Histology and morphometrics of white adipose tissue
Inguinal (iWAT) and epididymal (eWAT) adipose tissue were fixed in Bouin's solution (Sigma-Aldrich), and paraffin-embedded sections were stained with H & E. Adipocyte area and size distribution were analyzed using Image J software at specific parameters to measure area of each adipocyte based on size and exclusion limits.

Quantitative reverse transcription PCR
Expression of mRNA. Quantitative reverse transcription-PCR was performed using total RNA with specific primers and probes (S3 Table) designed using Primer Express software v3.0.1 (Life Technologies) essentially as previously described [16] except that a BioRad CFX Real-Time System was also used for analyses. TaqMan probes (Biosearch Technologies) were used for gene quantification using TaqMan 1 RNA-to-CT 1-Step Kit (Life Technologies). Gene expression data were normalized to level of cyclophilin b (Ppib), or TATA box binding protein (Tbp). In addition, the total nanograms of input RNA/reaction was quantified using Qubit HS RNA Assay Kit (Life Technologies) and can be used to normalize gene expression by RNA input if necessary.
Expression of miRNA. Total RNA including miRNAs was isolated from white adipose tissue using miRNeasy mini kit (Qiagen). cDNA synthesis was performed using the TaqMan 1 MicroRNA Reverse Transcription Kit and TaqMan miRNA specific primers (Life Technologies) for mmu-miR-335-5p, mmu-miR-335-3p and snoRNA202 (NCBI Accession #AF357327). The cDNA was amplified using the TaqMan Universal Mastermix II, no UNG and TaqMan Small RNA Assay (Life Technologies). Quantitive PCR was performed using the CFX384 Real Time PCR detection system (BioRad). Relative abundance of miR-335 was quantified using the 2 -ΔΔCT method using snoRNA202 for normalization of expression.

Western blot analyses
Total tissue lysate from eWAT and iWAT were prepared in RIPA buffer containing 1% protease, phosphatase I and phosphatase II inhibitors (Sigma). Protein content was determined using Sigma BCA protein assay kit. Proteins were resolved on 10% SDS-PAGE gel and after separation electro-transferred to PVDF (Immobilon-P, EMD Millipore) or nitrocellulose (Hybond-ECL, GE Healthcare) membranes. Blots were incubated in antibodies against MEST (1:5000) and GAPDH (1:1000, ABCAM ab9484). Bands were visualized and quantified using the Odyssey imaging system (Li-Cor Bioscience).

In vitro cell culture
Experimental cohorts for in vitro experiments were generated by intercrossing Mest heterozygous mice to generate WT (+/+), Mest pKO (+/KO) and Mest gKO (KO/KO) mice. Ear mesenchymal stem cells (EMSCs) were isolated from WT and Mest gKO mice as previously described [57]. Cells were cultured in 5% CO 2 and maintained in DMEM/F-12 (Life Technologies) supplemented with 10% FBS (Life Technologies) supplemented with Primocin as an antibiotic (InVivoGen), and 10 ng/ml recombinant bFGF (PeproTech). To induce adipocyte differentiation, recombinant bFGF was removed and replaced with 10% FBS with 0.5 mM methylisobutylxanthine, 1 uM dexamethasone, 5 ug/ml insulin, and 5 uM troglitazone. On day 2, cells were fed 5 ug/ml insulin plus 5 uM troglitazone. On day 4 and every 2 days thereafter, cells were fed with 15% FBS. Lipid droplets were visualized using Oil Red O staining. Cells were washed with PBS followed by fixation with 10% neutral buffered formalin for 10 min. The cells were then washed once in dH 2 O and incubated in 60% isopropanol for 5 min. The isopropanol was removed and followed by an incubation in Oil Red O staining solution (0.35% Oil Red O in 60% isopropanol) for 10 min, washed once in 60% isopropanol and 3X for 10 min with dH 2 O. Oil Red O dye was eluted in 100% isopropanol for 15 minutes. The degree of adipogenic differentiation was quantified spectrophotometrically by measuring absorbance at 490 nm. A standard curve with known amounts of Oil Red O dye (ug/ml) was used to quantify lipid as total ug of Oil Red O eluted. After elution of Oil Red O dye, cells were then washed 1X with 100% isopropanol followed by dH 2 O and subsequently used for quantifying relative cell number by nuclear staining with crystal violet. Relative cell number was measured by staining cells with 0.05% crystal violet in deionized H 2 O for 5 minutes. The cells were then washed 3X for 10 min with dH 2 O then lysed in 1% SDS. Crystal violet in the 1% SDS lysate was measured at an absorbance of 560 nm a standard curve with known amounts of crystal violet (ug/ml) was used to quantify relative cell number and to normalize Oil Red O eluted from cells.

Gene knockdown
Five short-hairpin RNA (shRNA) constructs against Gpat4 (Agpat6) in lentiviral vector pLKO.1 were obtained from Sigma-Aldrich. For knockdown experiments, the construct chosen was TRCN0000099245. According to the manufacturer's protocol, EMSC were transduced in maintenance medium in the presence of polybrene (Sigma-Aldrich). Transduction was performed at MOI = 2. After overnight incubation, virus-containing media was removed and standard culture media was added. Cells were allowed to expand for 48 hours and then selection was initiated using puromycin (Sigma-Aldrich) at 4 ug/ml. Puromycin-resistant cells were expanded until P6 and then seeded in 12-well plates at a density of 100,000 cells per well. Prior to lentiviral transduction, a titration was performed to determine the minimum concentration of puromycin required to cause complete cell death after 3-5 days. Cells were harvested for gene expression, Oil Red O and crystal violet staining.

SOLiD SAGE analyses
SAGE gene expression profiling was performed with next-generation sequencing using the Life Technologies 5500XL SOLiD system. RNA was isolated from tissues using TRI Reagent; quantified and quality was determined using an Agilent 2100 Bioanalyzer. RNA was transcribed into cDNA via priming at the poly A+ tail. Double-stranded cDNA was digested with NlaIII restriction enzyme and ligated to adaptors containing an EcoP15L restriction site. Digestion was performed with EcoP15L and ligation of "bar code" oligonucleotides that specifically label each sample, and subsequent PCR amplification yielded a library for sequencing where each gene was represented with a 3'-expression tag. The alignment of sequence tags back to a reference genome was performed using a modified version of the Applied Biosystems SOLiD™ SAGE™ Analysis Software v1.10 that was installed, modified to meet stringent mapping criteria, and extensively tested in-house. RNA isolated from eWAT of 6 independent wildtype control mice and 5 Mest pKO mice were used for preparation of libraries and sequencing analyses. Raw count files were analyzed by the R/Bioconductor program DE-Seq31; significance and predictive analyses were performed using SAM (Significance Analysis of Microarrays; http://wwwstat.stanford.edu/~tibs/SAM/index.html) and PAM (Prediction Analysis for Microarrays; http://statweb.stanford.edu/~tibs/PAM/index.html) respectively [81,82]. P-values were adjusted for false discovery rate using the Benjamini-Hochberg method [83]. Pairwise analysis and hierarchical clustering of samples was performed using Bioconductor edgeR Version 3.2.3 and JMP Genomics respectively. Raw sequencing files from ABI SOLiD SAGE analyses have been deposited in NCBI's Gene Expression Omnibus (GEO; https://www.ncbi.nlm.nih.gov/ geo/) with the accession number GSE96968.

Statistics
Statistical calculations were performed using GraphPad Prism software V.6 and Microsoft Excel 2010. Differences between 2 groups were calculated using two-tailed unpaired parametric t-test with confidence level at 95%. Significance among multiple groups was calculated with ordinary one-way ANOVA followed by Tukey's multiple comparisons test with an alpha of 0.5 or by the Fisher's LSD test as indicated. Significance at each time point of longitudinal phenotypic datasets was determined using 2-way ANOVA with Holm-Sidak multiple comparisons test. Data is presented as the mean ± SEM. P<0.05 was considered to be significant.

Ethics statement
The use of murine models for the studies presented in the submitted manuscript was approved by the Institutional Animal Care and Use Committees (IACUC) at the Pennington Biomedical Research Center and the Maine Medical Center Research Institutes. Fabp4-cre (FpKO; n = 12) mice after being fed a high fat diet (HFD) from 8 to 20 weeks of age. Mest mRNA expression measured by TaqMan QRT-PCR is represented as the mean ± SEM arbitrary units (AU) normalized to cyclophilin b (Ppib). Significance in Mest RNA expression between groups was determined via one-way ANOVA using Tukey's multiple comparisons test. (B) Data shows the longitudinal measurements of bodyweight (BWT) for WT (n = 8), WT-cre (n = 11), pFL (n = 6) and FpKO (n = 12) mice fed a HFD from 8 to 20 weeks of age as indicated by the arrow along the X-axis. (C) Longitudinal measurements of adiposity index; (D) fat-free mass (g) and (E) fat mass (g) measured by NMR at the times indicated on the Xaxis are shown for WT (n = 8), WT-cre (n = 11), pFL (n = 6) and FpKO (n = 12) mice fed a HFD from 8 to 20 weeks of age as indicated by the arrow along the X-axis. (B-E) All data in the longitudinal studies are presented as the mean ± SEM. Significance at each time point of the longitudinal phenotypic analyses was determined by 2-way ANOVA and Tukey's multiple comparisons test. Time points annotated with 'a', 'b', 'c', and 'd' indicate significant differences between 'FpKO vs pFL'; 'FpKO vs WT'; 'FpKO vs WT-cre' and 'FpKO vs all genotypes respectively. (F) Data shows morphometric analyses of adipocyte size in eWAT in WT (n = 8), WTcre (n = 11), pFL (n = 6) and FpKO (n = 12) mice. Two-tailed unpaired parametric t-tests were used to determine significant differences in adipocyte size (F) between genotypes. Datasets annotated with the same letter indicate no significant differences between groups. (TIF) S1