The negative adipogenesis regulator Dlk1 is transcriptionally regulated by Ifrd1 (TIS7) and translationally by its orthologue Ifrd2 (SKMc15)

Delta-like homolog 1 (Dlk1), an inhibitor of adipogenesis, controls the cell fate of adipocyte progenitors. Experimental data presented here identify two independent regulatory mechanisms, transcriptional and translational, by which Ifrd1 (TIS7) and its orthologue Ifrd2 (SKMc15) regulate Dlk1 levels. Mice deficient in both Ifrd1 and Ifrd2 (dKO) had severely reduced adipose tissue and were resistant to high-fat diet-induced obesity. Wnt signaling, a negative regulator of adipocyte differentiation, was significantly upregulated in dKO mice. Elevated levels of the Wnt/β-catenin target protein Dlk1 inhibited the expression of adipogenesis regulators Pparg and Cebpa, and fatty acid transporter Cd36. Although both Ifrd1 and Ifrd2 contributed to this phenotype, they utilized two different mechanisms. Ifrd1 acted by controlling Wnt signaling and thereby transcriptional regulation of Dlk1. On the other hand, distinctive experimental evidence showed that Ifrd2 acts as a general translational inhibitor significantly affecting Dlk1 protein levels. Novel mechanisms of Dlk1 regulation in adipocyte differentiation involving Ifrd1 and Ifrd2 are based on experimental data presented here.


Introduction
Adipogenesis is a complex process in which multipotent stem cells are converted into preadipocytes before terminal differentiation into adipocytes (Sarantopoulos et al., 2018).These mechanisms involve protein factor regulators, epigenetic factors, and miRNAs.TPA-induced sequence 7 (TIS7) protein has been shown to be involved in the mainly transcriptional regulation of differentiation processes in various cell types, for example, neurons (Iacopetti et al., 1996), enterocytes (Wang et al., 2005), myocytes (Vadivelu et al., 2004), and also adipocytes (Nakamura et al., 2013).
Multiple lines of evidence link the regulation of Wnt/β-catenin signaling to the physiological function of Tis7, known in human as interferon-related developmental regulator 1 (Ifrd1) (Iezaki et al., 2016;Vietor et al., 2005).Experiments with Ifrd1 knockout mice generated in our laboratory show a negative effect of Ifrd1 on Wnt signaling and a positive effect on adipocyte differentiation (Vietor et al., 2005;Yu et al., 2010, #3955).Ifrd1 deficiency leads to a significant upregulation of Wnt/β-catenin transcriptional activity in both primary osteoblasts and in mouse embryonic fibroblasts (MEFs) derived from Ifrd1 knockout (KO) mice.It was shown that Ifrd1 is also involved in the control of adipocytes differentiation in mice and was upregulated in both visceral white adipose tissue (vWAT) and subcutaneous white adipose tissue (sWAT) of genetically obese ob/ob mice (Nakamura et al., 2013).Ifrd1 transgenic mice have increased total body adiposity and decreased lean mass compared with wild type (WT) littermates (Wang et al., 2005).On high-fat diet (HFD), Ifrd1 transgenic mice exhibit a more rapid and proportionately greater gain in body weight with persistently elevated total body adiposity.Enhanced triglyceride (TG) absorption in the gut of Ifrd1 transgenic mice (Wang et al., 2005) indicated that Ifrd1 expressed in the gut epithelium has direct effects on fat absorption in enterocytes.As a result of impaired intestinal lipid absorption, Ifrd1 KO mice displayed lower body adiposity (Yu et al., 2010).Compared with WT littermates, Ifrd1 KO mice do not gain weight when chronically fed an HFD and Ifrd1 deletion results in delayed lipid absorption and altered intestinal and hepatic lipid trafficking, with reduced intestinal TG, cholesterol, and free fatty acid mucosal levels in the jejunum (Garcia et al., 2014).Ifrd1 protein functions as a transcriptional co-regulator (Micheli et al., 2005) due to its interaction with protein complexes containing either histone deacetylases (HDAC) (Park et al., 2017;Vadivelu et al., 2004;Vietor et al., 2002;Wick et al., 2004) or protein methyl transferases, in particular PRMT5 (Lammirato et al., 2016).The analysis of adipocyte differentiation in preadipocytic 3T3-L1 cells suggested an involvement of Ifrd1 in the regulation of adipogenesis in the Wnt/β-catenin signaling context (Nakamura et al., 2013).
SKMc15, also known as interferon-related developmental regulator 2 (Ifrd2), a second member of the Ifrd gene family, is highly conserved in different species (Latif et al., 1997).Mouse Ifrd1 and Ifrd2 are homologous, with a remarkable identity at both the cDNA and amino acid levels (58 and 88%, respectively).However, there was so far no information about the physiological function and mechanisms of action of Ifrd2 protein and its possible involvement in differentiation of various tissues.Recently, cryo-electron microscopy (cryo-EM) analyses of inactive ribosomes identified Ifrd2 as a novel ribosome-binding protein inhibiting translation to regulate gene expression (Brown et al., 2018).The physiological function of Ifrd2 matching the mechanism based on the abovementioned cryo-EM data was so far not shown.Ifrd2 could be involved in adipogenesis since a significant reduction of whole protein synthesis was previously shown as a major regulatory event during early adipogenic differentiation (Marcon et al., 2017).
Adipogenesis occurs late in embryonic development and in postnatal periods.Adipogenic transcription factors CCAAT/enhancer binding protein α (Cebpa) and peroxisome proliferator-activated receptor γ (Pparg) play critical roles in adipogenesis and in the induction of adipocyte markers (Farmer, 2006).Pparg is the major downstream target of Delta-like protein 1 (Dlk1).It is inactivated by the induction of the MEK/ERK pathway, leading to its phosphorylation and proteolytic degradation (Wang and Sul, 2009).Dlk1, also known as Pref-1 (preadipocyte factor 1), activates the MEK/ERK pathway to inhibit adipocyte differentiation (Kim et al., 2007).Cebpa is highly expressed in mature adipocytes and can bind DNA together with Pparg to a variety of respective target genes (Lefterova et al., 2008).Besides, Pparg binding to Cebpa gene induces its transcription, thereby creating a positive feedback loop (Lowell, 1999).Both proteins have synergistic effects on the differentiation of adipocytes that requires a balanced expression of both Cebpa and Pparg.
Experimental data presented here show involvement of Ifrd1 and Ifrd2 in the process of adipocyte differentiation.dKO mice had strongly decreased amounts of the body fat when fed with even regular, chow diet and were resistant against the HFD-induced obesity.Two independent molecular mechanisms through which Ifrd1 and Ifrd2 fulfill this function were found.The fact that these two genes use independent mechanisms of action supported by the observation that whole-body deficiency of both genes led to a stronger phenotype when compared to single knockouts of Ifrd1 or Ifrd2.Ifrd1 regulates the Wnt signaling pathway activity and restricts Dlk1 protein levels, thereby allowing adipocyte differentiation.In contrast, Ifrd2 KO did not affect Wnt signaling, but as we show here, cells lacking Ifrd2 have significantly upregulated translational activity.In addition, strongly enriched Dlk1 mRNA concentrations were identified specifically in polyribosomes isolated from Ifrd2 knockout MEFs when compared to the WT MEFs.This was true also for dKO, but not for the single Ifrd1 knockout cells.The ablation of both Ifrd1 and Ifrd2 genes significantly affected the expression of genes essential for adipocyte differentiation and function.Since dKO mice render a substantially leaner phenotype on chow diet, even without any challenge by HFD induction, Ifrd1 and Ifrd2 represent novel players in the process of physiological adipocyte differentiation.

Results
Mice lacking Ifrd1 and Ifrd2 genes have lower body mass, less fat, and are resistant against HFD-induced obesity In order to clarify whether both Ifrd1 and Ifrd2 are involved in the regulation of the adipocyte differentiation and whether they act through the same or different mechanisms, mice lacking both genes were generated by crossing Ifrd1 with Ifrd2 single KO mice.dKO pups were viable, and adult male and female mice were fertile.At birth, body weights of dKO and WT mice were similar.Nevertheless, already during weaning, both the male and female dKO mice failed to gain weight when compared to their WT littermates and this persisted in the following weeks when the mice were fed regular diet (RD; chow diet) containing 11% kcal of fat.At 10 wk of age, dKO mice displayed 30% and later up to 44.9% lower body weight compared with WT mice (Figure 1A).In all further presented experiments, we used only male mice.
Based on dual-energy X-ray absorptiometry (DEXA) measurement, 6-month-old WT mice had substantially higher amounts of fat than their dKO littermates (Figure 1B Ifrd1 and Ifrd2 dKO was even more pronounced when the total fat and lean mass values were normalized to the body weight since the dKO mice were smaller than their WT counterparts were.The percentage of fat was in the WT mice 37.7 ± 4% vs. 6 ± 3% of the total body mass in dKO animals.Furthermore, the percentage of lean tissue mass in WT animals was lower than those in dKO animals (60 ± 4.6% vs. 92 ± 3.14%; Figure 1B, right panel).The dKO mice were slightly, but statistically significantly, smaller since there was a difference in body length, including the tail between WT and dKO mice (Figure 1-figure supplement 1A).Next, we analyzed the contribution of Ifrd1 and Ifrd2 to the whole-body fat content of mice.Three-dimensional reconstruction of images based on microcomputed tomography (micro-CT) of sex-/age-matched adult mice disclosed that both Ifrd1 and Ifrd2 single KO mice already had less abdominal fat than WT controls and that the dKO of Ifrd1 and Ifrd2 genes caused the strongest decrease in abdominal fat content and size (Figure 1C).Quantitative analyses of micro-CT measurements showed that Ifrd1 deficiency caused a substantial lack of the abdominal fat tissues (p=0.002 when compared to WT mice, Figure 1D).Whereas Ifrd1 KO mice had less fat mass but were not significantly lighter than their WT littermates (Figure 1E), Ifrd2 KO were lighter (p=0.007) and leaner (p=0.0001) and dKO mice were both significantly lighter (p=0.0001) and had significantly less abdominal fat (p=0.006)than the WT mice (Figure 1D and E).
The indirect calorimetry trial with sex-and age-matched WT and dKO animals did not identify any significant difference in respiratory exchange ratios (RER = VCO 2 /VO 2 ) of WT and dKO mice (Figure 1-figure supplement 1B).However, dKO mice showed significantly reduced body weight mainly because of lacking fat despite identical food intake, activity, and no major differences in several metabolic parameters.There were no statistically significant differences between WT and dKO mice even if the measured parameters were normalized to the smaller body mass (Table 1).To investigate potential links between Ifrd1, Ifrd2, and obesity, the response of dKO mice to HFD was studied.At 2 mo of age, male mice were housed individually and fed with HFD for 21 d.Food intake was measured every second day, and body weight was measured every fourth day.Feces were collected every second day to analyze the composition of excreted lipids, and blood samples were collected after the third week of HFD feeding to measure the concentrations of hepatic and lipoprotein lipases, respectively.Already within the first week of HFD feeding WT mice gained more weight than the dKO mice (Figure 1F), although there were no obvious differences in food consumption (Figure 1-figure supplement 2A) or in levels of lipolytic enzymes (Figure 1-figure supplement 2B and C).These differences in body weight gain continued to increase during the second and third weeks, at which time the body weight of WT mice increased additionally for 30% (30.0 ± 2.3%) when compared with the beginning of the HFD feeding period.In contrast, the weight of dKO animals increased only slightly (7.6 ± 1.6%) (Figure 1F).Both genes, Ifrd1 and Ifrd2, contributed to this phenotype, and we could see a stronger effect following their deletion (Figure 1F).
abdominal fat in single and dKO mice.Three-dimensional reconstitution of images of the abdominal fat mass distribution in WT and KO mice.Yellow color represents subcutaneous and red visceral fat mass.Black compartments are shadows resulting as part of the lightning model for the 3D volume rendering.(D) Mass contribution (%) of the abdominal fat in correlation to the total body weight (g).A linear regression is overlaid for each group individually.The regression results for the WT group are y = 0.342x -11.13 with an R 2 of 0.74; for the Ifrd1 KO group y = 0.143x -3.91 with an R 2 of 0.82; for the Ifrd2 KO y = 0.005x + 0.48 with an R 2 of 0.04 and; for the dKO y = 0.011x + 0.01 with an R 2 of 0.1.ANCOVA for the fat mass as a percentage of the body weight as covariant was performed.(E) Effect of single Ifrd1 or Ifrd2 knockout and Ifrd1 Ifrd2 double knockout on total body weight in (g) and body fat amount normalized to the body weight in (%).(F) Ifrd1 and Ifrd2 knockout reduced the gain of body weight of mice fed with HFD.7-week-old male WT and dKO mice were caged individually and maintained up to 21 d on HFD.Data shown are mean ± STD, n ≥ 9 per genotype.Data were analyzed applying one-way ANOVA with Holm-Šidák's multiple-comparisons test.*p<0.05,**p<0.01,****p<0.0001.
The online version of this article includes the following source data and figure supplement(s) for figure 1: Source data 1.DEXA and microCT measurements of IFRD1 and IFRD2 double knockout mice.Adipocyte differentiation in dKO mice is inhibited due to upregulated DLK1 levels A possible explanation of the lean phenotype of dKO mice was that Ifrd1 and Ifrd2 regulate adipocyte differentiation.Primary MEFs derived from totipotent cells of early mouse mammalian embryos are capable of differentiating into adipocytes and are versatile models to study adipogenesis as well as mechanisms related to obesity such as genes, transcription factors, and signaling pathways implicated in the adipogenesis process (Ruiz-Ojeda et al., 2016).To test whether Ifrd1 and Ifrd2 are required for adipogenesis MEFs derived from WT, Ifrd2 and Ifrd1 single and dKO mice were treated according to an established adipocyte differentiation protocol (Wang et al., 2015).Expression levels of both Ifrd1 and Ifrd2 mRNA increased during the adipocyte differentiation of WT MEFs.Ifrd1 expression reached maximum levels representing 5.7-fold increase compared to proliferating WT MEFs on day 3 (Figure 1-figure supplement 2D) and Ifrd2 reached on day 5 the maximum of 2.5-fold expression levels of proliferating MEFs (Figure 1-figure supplement 2E).These data suggested that both proteins play a regulatory role in adipogenesis; however, they differ in their mechanisms and timing.Eight days after initiation of adipocyte differentiation, a remarkable reduction of adipocyte differentiation potential in Ifrd2, Ifrd1 KO MEFs, and dKO stromal vascular fraction (SVF) cells isolated from inguinal WAT was observed, as characterized by the formation of lipid droplets stained by oil red O (Figure 2A).Quantification of this staining revealed that fat vacuole formation in cells derived from Ifrd2, Ifrd1, and dKO mice represented 23, 48, and 12% of the WT cells, respectively (Figure 2B).Stable ectopic expression of Ifrd2 significantly increased adipocyte differentiation in both single and double Ifrd1 and Ifrd2 knockout MEF cell lines (Figure 1 supplement 1C).These data indicated that both Ifrd1 and Ifrd2 were critical for adipocyte differentiation and that the defect in adipogenesis could be responsible for the resistance of dKO mice to HFD-induced obesity.
Table 1.Double knockout (dKO) mice do not differ from wild type mice in any metabolic parameter besides the body weight.A 21-hr indirect calorimetry trial monitoring gas exchange (oxygen consumption and carbon dioxide production), activity (distance and rearing), and food intake.The genotype effects were statistically analyzed using one-way ANOVA.Food intake and energy expenditure were analyzed using a linear model including body mass as a co-variate.Wnt/β-catenin signaling is an important regulatory pathway for adipocyte differentiation (Prestwich and Macdougald, 2007).Ifrd1 deficiency causes upregulation of Wnt/β-catenin transcriptional activity.Therefore, Wnt signaling activity was assessed in MEFs by measuring transcriptional activity using TOPflash, TCF-binding luciferase reporter assays.As shown in Figure 2C, Wnt signaling activity, when compared to the WT MEFs, was not regulated in Ifrd2 KO, but highly significantly (p<0.0001)upregulated in Ifrd1 KO and also in dKO MEFs.Supporting these findings, western blot analyses also identified increased β-catenin protein levels in inguinal WAT of Ifrd1 KO and in dKO mice (p<0.0001),but not in Ifrd2 KO mice (Figure 2D, left top panel), suggesting the involvement of Ifrd1 but not Ifrd2 in the Wnt signaling pathway regulation.
Next, the expression levels of Dlk1, a negative regulator of adipogenesis and at the same time known target of Wnt signaling (Paul et al., 2015), were analyzed.Dlk1 protein was significantly upregulated in inguinal WAT isolated from Ifrd1 and Ifrd2 KO as well as from dKO mice (Figure 2D, right top panel).A significant (p<0.001)upregulation of Dlk1 mRNA and protein levels in dKO inguinal WAT samples revealed the qPCR and confirmed western blot analyses (Figure 2E).Upregulation of Dlk1 in dKO mice both on RNA and protein results was confirmed in undifferentiated SVF cells isolated from inguinal fat (Figure 2F and G).Dlk1 mRNA levels were even stronger upregulated following the 8-days differentiation protocol of SVF cells (Figure 2H).Moreover, whereas in WT MEFs the expression of Dlk1 was strongly upregulated only during the first day of adipocyte differentiation and then over the next days declined to basal levels, in dKO MEFs Dlk1 was upregulated (p<0.001)throughout the entire 8 d of differentiation (Figure 3A).This result complemented protein analyses of lysates from 8-day differentiated adipocytes (Figure 3-figure supplement 1, middle panel).A rescue experiment confirmed that Dlk1 expression was Ifrd1-and Ifrd2-dependent.Dlk1 mRNA levels were analyzed by RT-qPCR in dKO MEFs stably expressing Ifrd1 and/or Ifrd2.Ectopic expression of Ifrd1, Ifrd2 and mainly their combination significantly (p<0.001)downregulated Dlk1 mRNA and protein levels (Figure 3B).Accordingly, these experiments documented that Ifrd1 and/or Ifrd1 were involved in the regulation of Dlk1 expression, but the molecular mechanism remained unclear.
The adipocyte differentiation deficiency of dKO inguinal SVF cells suggested that Cebpa and Pparg might also be regulated through Ifrd1 and/or Ifrd2.It was previously shown that elevated levels of the cleaved ectodomain of Dlk1 have been correlated with reduced expression of Pparg (Lee et al., 2003).Therefore, the differences in Pparg expression between WT and dKO inguinal SVF cells were analyzed.While Pparg and Cebpa mRNA levels were strongly induced in undifferentiated WT, these were barely detectable in dKO SVF cells (Figure 3C and D).Similarly, significantly decreased Pparg and Cebpa mRNA levels in Ifrd1, Ifrd2, and dKO inguinal SVF cells following 8-day adipocyte differentiation protocol were identified (Figure 3E and F).
Earlier chromatin immunoprecipitation (ChIP) analysis revealed that Ifrd1 binds directly to DNA and via interaction with PRMT5 regulates gene expression (Lammirato et al., 2016).Therefore, ChIP experiments were performed to study binding of Ifrd1 and Ifrd2 proteins, transcription factor β-catenin, and symmetrically dimethylated histone H4 at arginine residue 3 (H4R3me2s) (Paul et al., 2015) to regulatory elements of the Dlk1 gene.dKO MEFs treated 8 d with the adipocyte differentiation cocktail were increased β-catenin binding to the β-catenin/TCF binding site 2 of the Dlk1 regulatory   element found when compared to WT MEFs (Figure 3G).On the other hand, binding of H4R3me2s to the same Dlk1 regulatory element was significantly reduced (p<0.001).No direct binding of Irfd1 or Ifrd2 proteins to two different Dlk1 regulatory elements (Dlk1 region A and β-cat/TCFbs2), neither in WT nor in dKO MEFs, could be identified, suggesting rather an epigenetic regulation than via their direct binding to Dlk1 regulatory elements.These results suggested that proteins Ifrd1 and Ifrd2 are required to restrain the Dlk1 levels through the Wnt/β-catenin signaling pathway and yet another so far unknown mechanism.
After finding significantly increased Dlk1 protein levels in inguinal WAT of SKMc15 single knockout mice (Figure 2D), Dlk1 expression in MEFs generated from these mice was measured.These were significantly increased (>70-fold) when compared to WT MEFs (Figure 3H).In a search for a Ifrd2specific regulatory mechanism of Dlk1 levels, primarily the translational regulation was studied.It was previously shown that general reduction of protein synthesis and downregulation of the expression and translational efficiency of ribosomal proteins are events crucial for the regulation of adipocyte differentiation (Marcon et al., 2017).Initially, specifically polyribosome-bound Dlk1 RNA in Ifrd2 knockout MEFs was measured.This analysis detected significantly (p<8.26104E-07) higher Dlk1 mRNA levels in polyribosome RNA fraction of Ifrd2 KO when compared to WT MEFs (Figure 3I).Interestingly, Ifrd2 was recently identified as a novel specific factor capable of translationally inactivating ribosomes (Brown et al., 2018).Because Ifrd2 may play a crucial role in the adipogenesis regulation in the following experiment, the effect of Ifrd2 knockout on the general translational efficiency of WT, Ifrd1, Ifrd2 single and double KO MEFs was tested.Cells were incubated 30 min in the absence of methionine and cysteine, followed by 1 hr in the presence of 35 S-methionine.As shown in Figure 4A, there was a significant increase in the general translational activity of MEFs lacking Ifrd2 alone or both Ifrd1 and Ifrd2, but not in TIS7 single knockout cells.These data suggested that Ifrd2 alone, but not Ifrd1, inhibits the general translational activity necessary for the induction of adipogenic differentiation, also via the translational regulation of Dlk1.This finding supported the result shown in Figure 2D where the Dlk1 protein levels were significantly induced in inguinal WAT samples isolated from Ifrd2 single knockout mice.
TIS7 and SKMc15 regulate adipocyte differentiation through Dlk1, MEK/ERK pathway, Pparg, and Cebpa Dlk1 protein carries a protease cleavage site in its extracellular domain (Lee et al., 1995) and is secreted.The extracellular domain of Dlk1 is cleaved by ADAM17, TNF-α converting enzyme to generate the biologically active soluble Dlk1 (Wang and Sul, 2009).DLK1 mRNA and protein levels are high in preadipocytes, but Dlk1 expression is absent in mature adipocytes.Hence, adding soluble Dlk1 to the        Previous studies showed that soluble Dlk1 protein activates MEK/ERK signaling, which is required for inhibition of adipogenesis (Kim et al., 2007).As DLK1 was strongly upregulated in dKO SVF cells during adipocyte differentiation, the possible activation of the MEK/ERK pathway in gonadal WAT samples of WT and dKO mice was analyzed subsequently (Figure 4G).The phosphorylation of p44 and p42 was upregulated 1.8-fold (p<0.001) in the dKO when compared to the WT G WAT samples (Figure 4G).Next, the question of expression levels of Pparg and Cebpa in WAT depots was addressed.Both adipocyte differentiation regulators Pparg and Cebpa's mRNA expression levels were in gonadal WAT samples isolated from dKO mice significantly (p<0.01)downregulated when compared to the values of WT control animals (Figure 4H).Furthermore, the possible effect of Ifrd1, Ifrd2 and their combined knockout on the Dlk1 levels in brown adipose tissue (BAT) was analyzed.Western blot analysis identified a significant (p<0.001)increase in Dlk1 protein levels in BAT samples in knockout mice of all three genotypes (Figure 4I).The MEK/ERK pathway was similarly as in gonadal WAT, upregulated also in MEFs generated from dKO mice (Figure 5A).The phosphorylation of p42 and of p44 was upregulated 3-fold and 4.3-fold, respectively (p<0.01), in the adipocytedifferentiated dKO when compared to the WT MEFs (Figure 5B).Previously, activation of MEK/ERK by Dlk1 was shown to upregulate the expression of the transcription factor Sox9, resulting in the inhibition of adipogenesis (Sul, 2009).Therefore, we measured Sox9 mRNA expression by RT qPCR in 8-day adipocyte-differentiated MEFs.Sox9 expression was significantly (p<0.001)upregulated in 8-day adipocyte-differentiated dKO when compared to WT MEFs (Figure 5C).Pparg and Cebpa mRNA levels were both continuously upregulated during the 8-day differentiation protocol of WT MEFs (Figure 5D and E).On the contrary, no significant increase in Pparg and Cebpa mRNA levels was found in dKO MEFs (Figure 5D and E (p<0.001), almost up to the levels of WT MEFs (Figure 5F).The expression of Cebpa was also strongly upregulated by the co-expression of Ifrd1 and Ifrd2 (p<0.001) (Figure 5G).The conclusion of these results was that Ifrd1 and Ifrd2 regulate the expression of both Pparg and Cebpa, crucial regulators of adipocyte differentiation.
Lipid absorption is reduced in Ifrd1 and Ifrd2 dKO mice dKO mice were (Figure 1-figure supplement 1B, Figure 1-figure supplement 2B) leaner than their WT littermates despite identical food intake and RER.Therefore, the possibility that dKO mice store energy ectopically was tested.In the feces of dKO mice fed with HFD were identified significantly higher (p<0.05)amounts of free fatty acids than in that of their WT siblings (Figure 6A).Secondly, the energy content of dried feces from dKO mice determined by bomb calorimetry was significantly higher (p<0.001)than that of WT mice (Figure 6B).Pparg induces the expression of Cd36, a very long chain fatty acids (VLCFA) transporter in heart, skeletal muscle, and adipose tissues (Coburn et al., 2000).The regulation of Cd36 by Pparg contributes to the control of blood lipids.Interestingly, Cd36 null mice exhibit elevated circulating LCFA and TG levels consistent with the phenotype of dKO mice and Cd36 deficiency partially protected from HFD-induced insulin resistance (Wilson et al., 2016).
Because of downregulated Pparg levels in adipose tissues of dKO mice, the regulation of Cd36 was studied more in detail as well.In inguinal WAT from dKO mice were found strongly reduced Cd36 mRNA expression levels (64% of the WT values) (Figure 6C).Next, we analyzed Cd36 in WT and dKO MEFs before onset and during adipocyte differentiation.As long as Cd36 mRNA expression substantially increased in WT MEFs, there were almost undetectable transcript levels of Cd36 in dKO cells treated with the adipocyte differentiation cocktail (Figure 6D).Diacylglycerol acyltransferase 1 (Dgat1), a protein associated with the enterocytic TG absorption and intracellular lipid processing (Nozaki et al., 1999), is besides Cd36 another target gene of adipogenesis master regulator Pparg (Koliwad et al., 2010).Dgat1 mRNA levels are strongly upregulated during adipocyte differentiation (Cases et al., 1998), its promoter region contains a Pparg binding site (Ludwig et al., 2002), and Dgat1 is also negatively regulated by the MEK/ERK pathway (Tsai et al., 2007).Dgat1 expression was shown to be increased in Ifrd1 transgenic mice (Wang et al., 2005), and its expression was decreased in the gut of HFD-fed Ifrd1 KO mice (Yu et al., 2010).Importantly, Dgat1 expression in adipocytes and inguinal WAT is upregulated by Pparg activation (Koliwad et al., 2010).Therefore, to analyze the role of Ifrd1 and Ifrd2 on the regulation of this protein involved in adipogenesis and TG processing, Dgat1 mRNA levels were measured during the differentiation of WT and dKO MEFs into adipocytes.
As long as Dgat1 expression substantially increased during the differentiation of WT MEFs, there was no difference in Dgat1 mRNA levels in dKO cells treated with the adipocyte differentiation cocktail (Figure 6E).Gene expression analyses showed that Ifrd1 and Ifrd2 regulated expression of multiple proteins involved both in adipocyte differentiation and in fat uptake, thereby contributing to the lean phenotype of dKO mice through multiple means.

Discussion
Experimental data presented here show that simultaneous depletion of Ifrd1 and of its orthologue Ifrd2, a protein recently identified as a translational inhibitor, caused severe reduction of adipose tissues and resistance against high fat-induced obesity in mice.Two parallel mechanisms were identified, leading to a deficiency in adipocyte differentiation.Firstly, Ifrd1-regulated Wnt signaling affected Dlk1 transcription, and secondly, experimental evidence proved that Ifrd2 acted as a translational inhibitor that controls Dlk1 protein levels, thereby contributing to the regulation of adipogenesis (Figure 6F).dKO mice were phenotypically similar to both Cd36-deficient and Dlk1 transgenic mice, namely in decreased amounts of WAT and resistance to HFD-induced obesity.Previous studies showed that overexpression of Ifrd1 caused increased intestinal lipid transport, resulting in elevated body weight gain during HFD feeding (Wang et al., 2015).Knockout of Ifrd1 and Ifrd2 impaired absorption of free fatty acids from the lumen into enterocytes and reduced rate of fat absorption from intestines into the circulation.Simultaneously, higher concentrations of free fatty acids in the feces of dKO mice (Figure 6A) suggested that mice lacking Ifrd1 and Ifrd2 suffer from an intestinal lipid uptake deficiency, yet another contributing reason for the lean phenotype of these mice.In dKO mice, Pparg and Cebpa levels were inhibited, induced MEK/ERK pathway, and decreased expression of Cd36 and Dgat1, all hallmarks of upregulated Dlk1, a known adipogenesis inhibitor.Ifrd1 and Ifrd2 acted after the commitment to the preadipocyte stage since the elevated Dlk1 and β-catenin levels found in WAT of dKO mice were characteristic of the preadipocyte stage (Gautam et al., 2017).
Disruption of Wnt signaling in embryonic fibroblasts results in spontaneous adipocyte differentiation (Bennett et al., 2003), and in contrast, stabilized β-catenin keeps cells in the preadipocyte stage (Ross et al., 2000).A subpopulation of the SVF of adipose tissue is adipogenic and at the same time has a weaker Wnt/β-catenin signal (Hu et al., 2015).Consistent with this knowledge, upregulated β-catenin protein levels and increased Wnt signaling activity in WAT of Ifrd1 single and dKO MEFs were found.It has to be mentioned that the data presented here differ from those published by Nakamura et al., who showed that following the Ifrd1 overexpression in adipocytes Wnt/β-catenin signaling was upregulated and inhibited oil red O staining (Nakamura et al., 2013).However, one has to take into consideration the difference in cell systems used in these two studies.In contrast to Nakamura's study where results were obtained in 3T3-L1 cells fibroblasts overexpressing Ifrd1, the data presented here document the role of endogenous Ifrd1 in cells derived from WT or knockout mice.Moreover, Ifrd1 was found to be ubiquitously expressed in all WT mouse organs without any pretreatment such as hypoxia.On the other hand, Nakamura et al. identified upregulated Ifrd1 expression levels in WAT of obesity model mice.This result, however, fully supports the findings of lean phenotype in Ifrd1 and Ifrd2 dKO mice.
In dKO MEFs, induced for adipocyte differentiation, the increase in Wnt signaling led to upregulated binding of β-catenin to the Dlk1 gene regulatory elements, resulting in sustained Dlk1 expression.In adipogenesis, Dlk1 expression is downregulated through histone methylation by Coprs and PRMT5 proteins that prevent β-catenin binding to the Dlk1 gene (Paul et al., 2015).Interestingly, in dKO MEFs weaker binding of dimethylated H4R3 to the Dlk1 gene was found, consistent with the previous findings on the negative role of Ifrd1 in epigenetic regulation of gene expression including the PRMT5 activity (Lammirato et al., 2016).
The difference in the effects of ectopic expression of Ifrd1, Ifrd2 and their co-expression on Dlk1 levels (Figure 3B) confirmed the hypothesis that proteins Ifrd1 and Ifrd2 regulate Dlk1 levels via two independent pathways/mechanisms.Despite the fact that Ifrd2 knockout had no effect on Wnt signaling, it nevertheless affected Dlk1 levels, suggesting a contribution of Ifrd2.Since protein Ifrd2 was identified as a novel factor translationally inactivating ribosomes (Brown et al., 2018), downregulation of the protein synthesis machinery is an essential regulatory event during early adipogenic differentiation (Marcon et al., 2017).Therefore, there was a possibility that Ifrd2 regulates Dlk1 protein levels through translational regulation.The experimental data presented here confirmed specific regulation of Dlk1 through this mechanism.Elevated Dlk1 levels in dKO MEFs activated the MEK/ERK pathway, thereby decreasing Pparg and Cebpa levels important for adipogenic differentiation.Besides, the expression of Sox9 and Hes1 was upregulated, suggesting that the dKO MEFs keep their proliferative state and cannot enter differentiation into mature adipocytes (Kim et al., 2007).
Downregulation of Pparg and Cebpa implied changes in the expression of downstream adipogenesis-related genes.Among them, decreased Cd36 and Dgat1 levels were a plausible explanation of the dKO mice lean phenotype since it is known that both proteins play a functional role in the differentiation of murine adipocytes and their deficiency impairs fat pad formation independent of lipid uptake (Christiaens et al., 2012).The results presented here implicate that Ifrd1 and Ifrd2 play a up to now unknown role in the regulation of Cd36 and therefore possibly contribute to Cd36related pathogenesis of human metabolic diseases, such as hypoglycemia (Nagasaka et al., 2011), hypertriglyceridemia (Kashiwagi et al., 2001), and disordered fatty acid metabolism (Glazier et al., 2002;Tanaka et al., 2001).
It is possible that due to the reduced Cd36 expression levels in intestines, Ifrd1 Ifrd2 dKO mice accumulated lower amounts of body fat and were resistant to diet-induced obesity also because of limited intestinal fat uptake.Furthermore, Ifrd1 Ifrd2 dKO mice displayed increased plasma TG levels due to impaired clearance of VLCFA transport by skeletal muscles.Co-expression of Ifrd1 and Ifrd2 in dKO myoblasts partially rescued the Cd36 expression on the transcriptional level as documented by the increase of the Cd36 promoter activity.Rescuing the free fatty acids transport by Cd36 overexpression in dKO myoblasts confirmed the hypothesis that the deficit in fatty acid transport was due to the diminished Cd36 expression.Thus, Ifrd1 Ifrd2 dKO mice accumulated lower amounts of adipose tissue also because of impaired lipid transport.
It was previously shown that Ifrd1 is involved in both chow and HFD conditions in intestinal TG absorption.The experimental data shown here document for the first time the role of Ifrd1 and its orthologue Ifrd2 in adipocyte differentiation.Moreover, they confirm the physiological function of Ifrd2 as a translational inhibitor.Surprisingly, although Ifrd1 and Ifrd2 share sequence homology and a functional role in the adipogenesis, they use two independent regulatory mechanisms.

Generation of animal models
All animal experiments were performed in accordance with Austrian legislation BGB1 Nr. 501/1988Nr. 501/ i.d.F. 162/2005. .Mice lacking Ifrd1 were described previously (Vadivelu et al., 2004).Ifrd2 KO mice generation: targeting construct contained Ifrd2 gene locus exons.A loxP site and a neomycin resistance gene were inserted at position 105515 (AY162905).The neomycin cassette was flanked by two frt sites.A second loxP site was inserted at position 102082 (AY162905).Further downstream, 15 additional nucleotides, part of intron 1, exon 2 (splice site: donor and acceptor), and the CDS of hrGFP from the Vitality hrGFP mammalian expression vector pIRES-hrGFP-2a (Stratagene) were added.The targeted construct was electroporated into Sv129 mouse ES cells.After the selection with G418, single-cell clones were screened by PCR and confirmed by Southern blot analysis.After Cre recombinase treatment, cell clones were screened by PCR.Clones with floxed gene deleted were used for blastocyst injection into C57BL/6J mice.Two male chimeras with ≥40% or more agouti coat color were mated to C57BL/6J females.The knockout mouse strain was derived from one male mouse carrying the allele of interest.Heterozygous mice were back-crossed to C57BL6/J mice for nine generations.dKO mice were generated by crossing the Ifrd2 KO mice with the Ifrd1 single KO mice.The resulting mice were screened by PCR and double heterozygous mice used for further breeding until homozygosity.In order to achieve maximal homogeneity of experimental groups, in all experiments presented here we used only male mice.Ifrd2 knockout PCR genotyping analysis hrGFP insert in the 3′ UTR was detected by primer GFP2 5′-AGCC ATAC CACA TTTG TAGA G-3′ and RK101 3′ UTR (5′-TGAT GATA GCTT CAAA GAGA A-3′; 100617-100591 of the Ifrd2 locus (AY162905).PCR product 1700 bp.Ifrd2 detection: RG1; 5′-TGTG GCCT TTAT CCTG AGTC -3′; 102286-102266) and RG2; 5′-TGGC TTCA TTTA CACT ACTC CTT-3′; 101860-101882 primers (Figure 1-figure supplement 2F).WT allele PCR product 426 bp and the targeted allele PCR product 1772 bp (Figure 1figure supplement 2G).Ifrd1 genotype was tested as explained previously (Vadivelu et al., 2004; Figure 1-figure supplement 2H).

Growth monitoring and body composition measurement
Mice were weaned at 3 wk, regular chow diet, and weighed weekly.DEXA was measured with Norland scanner (Fisher Biomedical).Micro-CT experiments were performed using vivaCT 40 (Scanco Medical AG).The scans were performed using 250 projections with 1024 samples, resulting in a 38 µm isotropic resolution.Tube settings: 45 kV voltage, 177 µA current, integration time 300 ms per projection.Image matrix 1024 * 1024 voxels and a grayscale depth of 16 bit.The length of the image stack was individually dependent, starting from the cranial end of the fist lumbar vertebrae to the caudal end of the fifth lumbar vertebrae.The image reconstruction and postprocessing were performed using the Scanco Medical system software V6.6.For the adipose tissue evaluation, an IPL (image processing language) script by Judex et al., provided by Scanco Medical AG, was modified to the scanner individual parameters, leading to two values lower threshold than in the original script for the adipose tissue filters 76.The script calculated the total abdominal volume without potential air in the cavities.A separation of subcutaneous and visceral fat mass was used only for visualization.For the quantitative fat mass analysis, we used 56 male 5-12-month-old mice (mean age 9.35 ± 2.03 mo) reflecting an adult to mid-aged cohort defined by Flurkey et al., 2007;Neeland et al., 2013 #4207.The development of adipose tissue in healthy mice is stable between 4 and 12 mo (Lemonnier, 1972).Therefore, sex and age could have only very limited influence on the experimental results.For the quantitative comparison, the percent contribution of the abdominal fat to the body weight was calculated using a mean weight of 0.9196 g/ml for adipose tissue (Neeland et al., 2013).Statistics were performed using an ANCOVA with a Bonferroni corrected post hoc testing on the µCT fat data and the body mass data.

Metabolic measurements
The indirect calorimetry trial monitoring gas exchange, activity, and food intake was conducted over 21 hr (PhenoMaster TSE Systems).Body mass and rectal body temperature before and after the trial were measured.The genotype effects were statistically analyzed using one-way ANOVA.Food intake and energy expenditure were also analyzed using a linear model including body mass as a co-variate.

HFD feeding
Age-matched (7-week-old) male Ifrd1-/-Ifrd2-/-and WT mice were caged individually and maintained up to 3 wk on a synthetic, HFD diet (TD.88137;Ssniff).Animals were weighted every fourth day between 08:00 and 10:00.Intestines, liver, muscles, and adipose tissue were collected for total RNA and protein isolation.Unfixed intestines were flushed with PBS using a syringe, embedded in Tissue-Tek (Sakura, 4583) and frozen in liquid nitrogen for immunohistochemical analysis.

Quantitative food consumption and fecal fat determination
Adult mice were acclimatized to individual caging and to the HFD for a week, and monitored for weight gain and their food intake daily.The daily food intake data were pooled for the following 7 d, and the food intake was estimated (g/day).Feces were collected daily and weighted for 7 d after the second week of the HFD consumption.Lipids were extracted and methylated according to Lepage et al., 1989.After freeze-drying and mechanical homogenization, aliquots of feces were subjected to the same procedure as described by Lepage et al., 1989.The resulting fatty acid methyl esters were analyzed by gas chromatography to measure the total and individual amounts of major fatty acids (Minich et al., 2000).The energy content of dried egested feces samples (~1 g per sample) was determined by bomb calorimetry (IKA C 7000, IKA, Staufen, Germany) (Pfluger et al., 2015).

Plasma cholesterol and triglyceride analyses
Serum cholesterol and TG were measured using cholesterol/TG reagent (Cobas, Roche) according to the manufacturer's instructions.Lipoprotein profiles were analyzed by fractionation of pooled serum using two Superose 6-columns (Cytiva) in series (FPLC), followed by cholesterol measurement (Demetz et al., 2020).

Construction of expression plasmids and generation of stable cell lines
The pRRL CMV GFP Sin-18 plasmid (Zufferey et al., 1998) was used to generate Ifrd1, Ifrd2, and Dlk1 expression of lentiviral constructs.For this, the corresponding cDNAs were cloned into the BamHI and SalI sites of the pRRL CMV GFP Sin-18 plasmid.A cap-independent translation enhancer (CITE) fused to the puromycin resistance pac gene and the woodchuck hepatitis virus post-transcriptional regulatory element (WPRE) were introduced downstream of the Ifrd1, Ifrd2, and Dlk1 coding sequences.All DNA constructs were verified by sequencing.

Analysis of translation by metabolic labeling
Pulse labeling of proteins was performed as described before (Popow et al., 2015), with the following changes: 1 × 10 6 wild-type and dKO MEFs cells per plate were seeded and cultivated for 24 hr.Cells were washed twice with PBS followed by one wash with labeling medium (ESC medium without cysteine and methionine) and a 30 min incubation in labeling medium.To label newly translated products, 200 µCi of [ 35 S]-methionine (10 mCi/ml; Hartman Analytic) were added and the cells were incubated for 1 hr.Cells were washed and incubated another 10 min in standard medium at 37°C.Cells were then harvested by trypsinizing, washed once with ice-cold PBS, extracted with ice-cold RIPA buffer containing protease-inhibitors, and briefly sonicated.Proteins were fractionated by gel electrophoresis in 16% Tricine gels (Thermo Fisher, EC66955BOX) and stained with Coomassie brilliant blue, and radioactive signals were visualized by phosphorimaging.Signal intensities were quantified using the Image Studio Lite (v5.2) software.

RT-PCR
Tissues from animals fed for 3 wk with HFD were snap-frozen and stored at -80°C.Total RNA was isolated using the TRIzol reagent (Invitrogen, 15596026).RNA was then chloroform-extracted and precipitated with isopropanol.The yield and purity of RNA were determined by spectroscopic analysis; RNA was stored at -80°C until use.

Quantitative RT-PCR and statistics
Total RNA were treated with DNAse1 and reverse transcribed to cDNA by Revert Aid First Strand cDNA Synthesis Kit (Thermo Scientific, K1622) with oligo dT primers.Quantitative RT-PCR was performed using TaqMan probes and primer sets (Applied Biosystems) specific for CD36 (assay ID Mm00432398_m1), Dgat1 (Mm00515643_m1), Pparg (Mm00440940_m1), Cebpa (Mm00514283_s1), Dlk1 (Mm00494477_m1), and Sox9 (Mm00448840_m1).Ribosomal protein 20 (assay ID Mm02342828_ g1) was used as normalization control for quantification by the ddCt method.PCR reactions were performed using 10 µl cDNA in PikoReal 96 real-time PCR system (Thermo Scientific).Quantification data were analyzed by two-tailed, homoscedastic t-tests based on the assumption that variances between the two sample data ranges are equal to type 2 Student's t-test.
Transient transfections and luciferase assay pGL2-Basic (Promega, E1641) or pGLCD36 (Shore et al., 2002) were used as reporter constructs.Expression constructs or empty vector DNA as a control were co-transfected.pCMV-β-Gal plasmid was used to normalize for transfection efficiency.For luciferase reporter assays, 1.5 × 10 5 cells were seeded into 24-well plates and transfected after 24 hr with the indicated plasmid combinations using Lipofectamine Plus Reagent (Invitrogen, 15338030).The total amount of transfected DNA (2 μg DNA per well) was equalized by addition of empty vector DNA.Cells were harvested 48 hr posttransfection in 0.25 M Tris, pH 7.5, 1% Triton X-100 buffer and assayed for both luciferase and β-galactosidase activities.Luciferase activity and β-galactosidase activity were assayed in parallel using the Lucy 2 detection system (Anthos).Transfections were performed in triplicates, and all experiments were repeated several times.

Chromatin immunoprecipitation (ChIP)
Chromatin was isolated from Ifrd1 WT and dKO formaldehyde-treated, 8-day adipocyte-differentiated MEFs using the EpiSeeker Chromatin Extraction Kit (Abcam, ab117152).ChIP analyses were carried out as described previously (Reintjes et al., 2016).The sequence of the oligonucleotides for two regions of the Dlk1 promoter, encompassing TCF and β-catenin binding sites, were as defined in Paul et al., 2015.Sonicated chromatin was centrifuged at 15.000 × g for 10 min at 4°C, and the supernatant (65 µg of sheared DNA per each IP) was diluted tenfold with cold ChIP dilution buffer containing 16.7 mM Tris-HCl pH 8.1, 167 mM NaCl, 0.01% (w/v) SDS, 1.1% (w/v) Triton X-100, and 1.2 mM EDTA with protease inhibitors.Samples were pre-cleared for 1 hr with protein A Sepharose CL-4B (Sigma-Aldrich, 17-0780-01) beads blocked with 0.2 µg/µl sonicated herring sperm DNA (Thermo Fisher, 15634017) and 0.5 µg/µl BSA (NEB, B9000S).Immunoprecipitations were performed at 4°C overnight.Immune complexes were collected with protein A Sepharose for 1 hr at 4°C followed by centrifugation at 1000 rpm and 4°C for 5 min.Beads were washed with 1 ml low salt wash buffer (20 mM Tris-HCl pH 8.1, 150 mM NaCl, 0.1% [w/v] SDS, 1% [w/v] Triton X-100 [Merck], 2 mM EDTA), high salt wash buffer (20 mM Tris-HCl pH 8.1, 500 mM NaCl, 0.1% [w/v] SDS, 1% [w/v] Triton X-100, 2 mM EDTA), LiCl wash buffer (10 mM Tris-HCl pH 8.1, 250 mM LiCl, 1% [w/v] sodium deoxycholate, 1% [w/v] IGEPAL-CA630, 1 mM EDTA) for 5 min at 4°C on a rotating wheel, and twice with 1 ml TE buffer (10 mM Tris-HCl pH 8.0, 1 mM EDTA).Protein-DNA complexes were eluted from antibodies by adding a freshly prepared elution buffer containing 1% SDS and 0.1 M NaHCO 3 .The eluate was reverse cross-linked by adding NaCl to a final concentration of 0.2 M and incubating at 65°C for 4 hr.Afterward the eluate was treated with proteinase K at 45°C for 1 hr.The immunoprecipitated DNA was then isolated by phenol/ chloroform precipitation and used as a template for real-time quantitative PCR.The primer pairs specific for regulatory regions of the Dlk1 gene were selected as described before (Paul et al., 2015).Reactions with rabbit IgG or with 1.23% of total chromatin (input) were used as controls.For real-time quantitative PCR, a PikoReal System was used.Signals were normalized to input chromatin and shown as % input.The raw cycle threshold (Ct) values of the input were adjusted to 100% by calculating raw Ct -log2(100/input).To calculate the % input of the immunoprecipitations, the equation 100 × 2[Ct (adjusted input to 100%) -Ct (IP)] was applied.

Figure 1 .
Figure 1.Double knockout (dKO) mice display postnatal growth retardation, less body fat, and are resistant against high-fat diet (HFD)-induced obesity.(A) Growth curves of wild type (WT) and dKO mice on chow diet (n fluctuate depending on the time point of measurement).(B) Dual-energy X-ray absorptiometry (DEXA) measurements of WT and dKO mice.Left panel depicts the absolute values of fat and lean mass per animal, and the right panel represents values normalized to the total body weight of animals (n = 15).(C) Micro-computed tomography measurements identified a lack of Figure 1 continued on next page

Figure supplement 1 .
Figure supplement 1. Body length and respiratory exchange of double knockout (dKO) and wild type (WT) mice.

Figure supplement 1
Figure supplement 1-source data 1.Further characterization of dKO mice.

Figure supplement 2 .
Figure supplement 2. No difference between wild type (WT) and double knockout (dKO) mice in food consumption.

Figure supplement 2
Figure supplement 2-source data 1.Supplementary information for the Figure 1.
Figure1 continued -figure supplement 1C and D and Figure 2-figure supplement 1).Ectopic expression of Ifrd1 significantly induced the adipocyte differentiation in Ifrd1 single knockout MEFs (Figure 1-figure

Figure 2 .
Figure 2. Adipocyte differentiation, Dlk1 levels, and Wnt signaling pathway are significantly affected in double knockout (dKO) mice.(A) Representative oil red O staining of inguinal stromal vascular fraction (SVF) cells following 8 d of the adipocyte differentiation protocol.Scale bar is 20 µm.(B) Quantification of oil red O staining from three independent adipocyte differentiation experiments shown in panel (A).Ordinary one-way ANOVA ****p<0.0001.(C) Wnt signaling activity was measured in mouse embryonic fibroblasts (MEFs) transiently co-transfected with the Tcf reporter plasmid pTOPflash and β-galactosidase expression vectors.Transfection efficiency was normalized using the β-galactosidase values.The experiment was repeated three times.Data shown are mean ± STD.Ordinary one-way ANOVA Holm-Šídák's multiple-comparisons test ****p<0.0001.(D) Left top panel: western blot analysis of β-catenin amounts in inguinal WAT samples.Representative western blot from three biological repetitions is shown.Right top Figure 2 continued on next page panel: Dlk1 is significantly upregulated both in Ifrd1, Ifrd2 single KO and in dKO mice.Bottom panels: actin western blots were used for normalization of sample loading.(E) Dlk1 mRNA expression measured by qPCR in inguinal white adipose tissue (WAT) samples (top).Values were normalized on GAPDH, n = 3. Error bars indicate standard deviations.A representative western blot image detecting Dlk1 and GAPDH proteins (bottom).(F) Dlk1 mRNA expression detected in undifferentiated cells isolated from inguinal WAT SVF cells.Normalized on GAPDH.Error bars indicate standard deviations, ***p<0.001.(G) Dlk1 protein western blot analysis in undifferentiated cells isolated from inguinal WAT SVF cells.Mean of three biological repeats; inset is one representative western blot.***p<0.001.(H) Dlk1 mRNA expression detected in 8 d adipocyte-differentiated cells isolated from inguinal WAT SVF cells.Normalized on GAPDH.Error bars indicate standard deviations, ***p<0.001.The online version of this article includes the following source data and figure supplement(s) for figure 2: Source data 1.b-catenin and Dlk1 protein levels in inguinal WAT.Source data 2. Dlk1 mRNA levels in inguinal WAT.Source data 3. Dlk1 protein levels in inguinal WAT cells.

Figure 3 .
Figure2 continued MEF values were set as 1, n = 3; error bars indicate standard deviations.***p<0.001(C) Pparg mRNA expression detected in undifferentiated cells isolated from inguinal white adipose tissue (WAT) stromal vascular fraction (SVF) cells.Normalized on GAPDH.Error bars indicate standard deviations, ***p<0.001.(D) Cebpa mRNA expression detected in undifferentiated cells isolated from inguinal WAT SVF cells.Normalized on GAPDH.Error bars indicate standard deviations, ***p<0.001.(E) Pparg mRNA expression detected in 8-day adipocyte-differentiated cells isolated from inguinal WAT SVF cells.Normalized on GAPDH.Error bars indicate standard deviations, ***p<0.001.(F) Cebpa mRNA expression detected in 8-day adipocytedifferentiated cells isolated from inguinal WAT SVF cells.Normalized on GAPDH.Error bars indicate standard deviations, ***p<0.001.(G) Recruitment of indicated proteins to regulatory regions of the Dlk1 promoter in WT and KO MEFs was analyzed by chromatin immunoprecipitation (ChIP) at day 8 of the adipocyte differentiation.Values are expressed as the percentage of immunoprecipitated chromatin relative to input and are the mean of triplicates.ChIP analysis identified increased specific β-catenin binding to its Dlk1 regulatory element in dKO samples.Pre-immune serum and IgG were used as background controls.n = 3, data shown are mean ± STD, Student's t-test ***p<0.001.(H) Dlk1 mRNA expression detected in undifferentiated MEF cells.Normalized on GAPDH.Error bars indicate standard deviations, ***p<0.001.(I) Real-time qPCR detection of Dlk1 RNA in polyribosomes.Normalized on GAPDH.Error bars indicate standard deviations, ***p<0.001.The online version of this article includes the following source data and figure supplement(s) for figure 3: Source data 1.Supplementary information for the Figure 3 -WB images.Source data 2. qPCR measurements for Figure 3G.Source data 3. Supplementary information for the Figure 3.

Figure
Figure supplement 1-source data 1.WB images for the Figure 3-figure supplement 1.

Figure supplement 1
Figure supplement 1-source data 1.Original images of the Figure 4-figure supplements 1 and 2.
Figure 4 continued

Figure 6 .
Figure 6.Free fatty acid (FFA) uptake is inhibited in double knockout (dKO) mice.(A) FFA concentrations in feces of wild type (WT) and dKO mice.Feces were collected every second day, and the composition of excreted lipids was determined by capillary gas chromatography.Data shown are mean ± STD, *p<0.05.(B) Energy content of dried egested feces samples was determined by bomb calorimetry.Data shown are mean ± STD, **p<0.01.(C) qPCR analysis of Cd36 mRNA levels in inguinal white adipose tissue (WAT).Data shown are mean ± STD, *p<0.05.(D) qPCR analysis of Cd36 in mouse embryonic fibroblasts (MEFs) differentiating into adipocytes.Relative expression levels were normalized on GAPDH expression.WT MEF values were set as 1.Data shown are mean ± STD, ***p<0.001.(E) Dgat1 mRNA levels were downregulated in dKO adipocyte-differentiated MEFs.Data shown are mean ± STD, ***p<0.001.(F) Proposed model of Ifrd1 and Ifrd2 molecular mechanisms of action during adipocyte differentiation.Two parallel mechanisms leading to a deficiency in adipocyte differentiation: Ifrd1-regulated Wnt signaling affected Dlk1 transcription and Ifrd2 acting as a translational inhibitor also contributed to the regulation of adipogenesis.