Early B Cell Factor 1 Regulates Adipocyte Morphology and Lipolysis in White Adipose Tissue

Cell Metabolism

Article Early B Cell Factor 1 Regulates Adipocyte Morphology and Lipolysis in White Adipose Tissue Hui Gao,1,6 Niklas Mejhert,2,6 Jackie A. Fretz,3 Erik Arner,2,4 Silvia Lorente-Cebria´n,2 Anna Ehrlund,2 Karin Dahlman-Wright,1,5 Xiaowei Gong,1 Staffan Stro¨mblad,1 Iyadh Douagi,2 Jurga Laurencikiene,2 Ingrid Dahlman,2 Carsten O. Daub,1,4 Mikael Ryde´n,2 Mark C. Horowitz,3,* and Peter Arner2,* 1Department

of Biosciences and Nutrition, Karolinska Institutet, Stockholm, SE-141 86, Sweden of Medicine (H7), Karolinska Institutet, Stockholm, SE-141 86, Sweden 3Department of Orthopaedics and Rehabilitation, Yale School of Medicine, New Haven, CT 06520, USA 4RIKEN Center for Life Science Technologies (Division of Genomic Technologies), RIKEN Yokohama Institute, Yokohama, Kanagawa, 230-0045, Japan 5Science for Life Laboratory, Solna, SE-171 21, Sweden 6Co-first Authors *Correspondence: [email protected] (M.C.H.), [email protected] (P.A.) http://dx.doi.org/10.1016/j.cmet.2014.03.032 2Department

SUMMARY

White adipose tissue (WAT) morphology characterized by hypertrophy (i.e., fewer but larger adipocytes) associates with increased adipose inflammation, lipolysis, insulin resistance, and risk of diabetes. However, the causal relationships and the mechanisms controlling WAT morphology are unclear. Herein, we identified EBF1 as an adipocyteexpressed transcription factor with decreased expression/activity in WAT hypertrophy. In human adipocytes, the regulatory targets of EBF1 were enriched for genes controlling lipolysis and adipocyte morphology/differentiation, and in both humans and murine models, reduced EBF1 levels associated with increased lipolysis and adipose hypertrophy. Although EBF1 did not affect adipose inflammation, TNFa reduced EBF1 gene expression. High-fat diet intervention in Ebf1+/ mice resulted in more pronounced WAT hypertrophy and attenuated insulin sensitivity compared with wild-type littermate controls. We conclude that EBF1 is an important regulator of adipose morphology and fat cell lipolysis and may constitute a link between WAT inflammation, altered lipid metabolism, adipose hypertrophy, and insulin resistance.

INTRODUCTION Disturbances in white adipose tissue (WAT) function, including increased local inflammation and fat cell lipolysis, are linked to insulin resistance, dyslipidemia, and atherosclerosis (Rosen and Spiegelman, 2014; Sun et al., 2011). Results in recent years have also highlighted the clinical importance of adipose tissue morphology regardless of body fat mass. Thus, in lean as well as in obese subjects, a phenotype characterized by fewer but larger fat cells (adipose hypertrophy) correlates closely with

WAT dysfunction and insulin resistance, while a phenotype characterized by many small adipocytes (adipose hyperplasia) is protective (Arner et al., 2010b, 2011; Hoffstedt et al., 2010). Furthermore, adipose hypertrophy confers an increased risk for the development of type 2 diabetes (Lo¨nn et al., 2010; Weyer et al., 2000). Although human adipocyte turnover (i.e., adipocyte birth/death rate) is significantly reduced in adipose hypertrophy (Arner et al., 2010b), the mechanisms promoting differences in adipose morphology are still largely unknown. In addition, while adipose morphology covaries with in vitro and in vivo insulin resistance as well as changes in WAT inflammation and lipolysis, the causal relationship between these factors is not known. Using an unbiased approach, we set out to identify transcription factors (TFs) associated with altered adipose morphology, dissect their mechanism of action, and evaluate their clinical relevance. This resulted in the identification of early B cell factor 1 (EBF1), a TF previously implicated in adipogenesis (Akerblad et al., 2002; Fretz et al., 2010), which through combined studies in human and murine models was shown to also be an important regulator of adipose morphology, lipolysis, and the development of insulin resistance. RESULTS Adipose Morphology Is Characterized by Distinct Functional and Transcriptional Alterations The relationship between adipose morphology and metabolic function was investigated in obese and nonobese women (cohort 1, n = 322), who were further subdivided into either those having hyperplastic or hypertrophic subcutaneous WAT. In both weight groups, subjects were matched for age, percent body fat, and body mass index (BMI) (Table S1 available online). Irrespective of morphology, WAT of obese compared with nonobese subjects was characterized by increased basal (i.e., non-hormonestimulated) (Figures 1A and S1A) and attenuated isoprenalineinduced lipolysis (Figure 1B), insulin resistance at the adipocyte level (Figure 1C) and whole-body level (Table S1), as well as elevated WAT release of the cytokines/chemokines tumor necrosis factor alpha (TNFa) (Figure 1D), chemokine (C-C motif) ligand 2 (CCL2) (Figure 1E), and interleukin 6 (IL6) (Figure S1B). Adipose Cell Metabolism 19, 981–992, June 3, 2014 ª2014 Elsevier Inc. 981

Cell Metabolism EBF1 Regulates Adipocyte Morphology and Lipolysis

b 6 a 4 2

Non-obese

Obese

D

Insulin-stimulated lipogenesis (nmol glucose/2h/107 cells)

6

Figure 1. Relationship between Adipose Morphology, Metabolic Phenotype, and Gene Expression

15 12 9 a 6

b

4

0

4

Non-obese

Obese

a d

a,d

3

a b

3

b

3

5

2

b

(A–F) Effects of adipose morphology on in vitro (A) basal lipolysis, (B) isoprenaline-stimulated lipolysis (expressed as the quotient over basal lipolysis), (C) insulin-stimulated lipogenesis, (D) TNFa secretion, (E) CCL2 secretion, and (F) overall gene expression (shown as a PCA). For (A)–(E), results are shown as mean ±SEM and were evaluated using ANOVA followed by Fisher’s PLSD post hoc test. Statistically significant differences (p < 0.05) are denoted as follows: a = different within the weight group, b = different from both measures in the other weight group, c = different both within the weight group and with both measures in the other weight group, and d = different from one of the measures in the other weight group. In (F), groups are separated with 95% confidence interval if the circles are not overlapping each other. See also Figure S1 and Tables S1 and S2.

2 1 1 0

0 Non-obese

Obese

c

1.5 b CCL2 secretion (log10 ng/2h/107 cells)

Isoprenaline-stimulated lipolysis (vs. basal lipolysis)

In vitro lipolysis (µmol glycerol/2h/107 cells)

C

c

8

0

E

B

10

1.0

a

Non-obese

F

150 First component (17.6%)

A

50

Obese

hypertrophy belonged to proinflammatory pathways (Figure S1D, left panel), while genes that were decreased were primarily associated with carbohydrate/ lipid metabolism and fat cell morphology (Figure S1D, right panel).

Adipocyte EBF1 Expression and Activity Is Linked to Adipose 0.5 Morphology -150 In order to identify TFs potentially causing the gene expression alterations observed 0 -250 in Figure 1F, we performed a Motif Activity Non-obese Obese -80 -40 0 40 80 Response Analysis (MARA), based on the Second component (9.7%) global transcriptomic data mentioned Non-obese hyperplasia Non-obese hypertrophy Obese hyperplasia Obese hypertrophy above, and combined it with our previously published MARA results during human adipocyte differentiation (Arner morphology had a marked and statistically significant impact on et al., 2012). This identified seven TFs fulfilling the following several of these parameters, particularly in the nonobese sub- criteria: (1) present in in-vitro-differentiated adipocytes and (2) jects, where hypertrophy was associated with a more pernicious exhibiting binding site motif activity patterns similar to that phenotype. Although cohort 1 consisted only of women, similar observed in the PCA (i.e., altered in obesity and in nonobese subfindings were also observed in men (cohort 2, n = 176, jects with hyperplasia versus hypertrophy) (Table S3). Two out of the seven TFs (EBF1 and hepatic leukemia factor [HLF]) disTable S1 and values not shown). To determine if morphology-associated metabolic profiles played alterations in both binding site activity (Figure 2A and were linked to changes in the transcriptome, analyses of global Table S3, respectively) and mRNA levels (Table S3) in nonobese gene expression in WAT from a previously described subset of subjects with hyperplasia. However, only EBF1 could be cohort 1 (cohort 3, n = 56) (Arner et al., 2012) were performed. confirmed by qPCR (Figures 2B and S2A), and this was corroboA principal component analysis (PCA), based on overall gene rated by increased protein levels in nonobese individuals (Figures expression, separated the four groups of subjects into three 2C and 2D). Thus, EBF1 mRNA, protein, and activity levels were distinct clusters: obese (both with hyperplasia and hypertrophy), selectively increased in the nonobese subjects with hyperplasia, nonobese hypertrophy, and nonobese hyperplasia (Figure 1F). and as a result, further studies were concentrated on this TF. Further comparisons demonstrated that (1) 619 genes were Consistent with previously published data in murine WAT (Jimealtered by morphology in nonobese subjects, and (2) this gene nez et al., 2007), EBF1 mRNA expression was enriched in the set displayed a considerable overlap (88%) with genes perturbed adipocyte fraction of WAT compared with intact adipose tissue in obese individuals (Figure S1C; Table S2). Gene set enrichment or leucocytes and macrophages (Figures 2E and 2F). A detailed analysis (GSEA) revealed that genes increased in nonobese analysis of EBF1 mRNA levels during human adipocyte -50

982 Cell Metabolism 19, 981–992, June 3, 2014 ª2014 Elsevier Inc.

Cell Metabolism EBF1 Regulates Adipocyte Morphology and Lipolysis

A

Figure 2. Characterization of EBF1 Activity, mRNA, and Protein Levels in Human Subcutaneous Adipose Tissue

B

C

D

G

E

F

H

differentiation showed an increase 12 hr after induction of adipogenesis that plateaued at day 4 (Figure 2G). Furthermore, adipocyte volume correlated negatively with EBF1 gene expression in isolated adipocytes (Figure 2H) and mRNA/activity in intact tissue (Figures S2B and S2C). There was also a significant association between EBF1 mRNA levels/activity and the gene expression of leptin (LEP) and the Wnt1-inducible signaling pathway protein 2 (WISP2) (Figure S2D), two previously established markers of adipose hypertrophy (Gustafson et al., 2013; Jerna˚s et al., 2006; Skurk et al., 2007). These results prompted us to functionally evaluate the role of EBF1 in WAT and whether the altered activity/expression of this TF is a cause or a consequence of differences in adipose morphology and/or metabolic function. EBF1 Regulates Genes Important for Adipocyte Lipid Metabolism and Differentiation To map direct transcriptional regulatory targets of EBF1, genome-wide EBF1 chromatin immunoprecipitation (ChIP)

(A and B) (A) Activity and (B) mRNA expression in WAT from nonobese and obese subjects with adipose different morphologies. (C–G) (C) Western blot and (D) quantification of EBF1 protein levels in WAT from nonobese individuals with hyperplasia or hypertrophy. Relative EBF1 mRNA expression in (E) paired samples of isolated adipocytes (Adipo) and intact WAT (n = 43); (F) isolated macrophages (Macro), leukocytes (Leuko), and adipocytes (n = 10); as well as (G) during adipocyte differentiation in vitro (n = 3). (H) Correlation between adipocyte EBF1 mRNA expression and fat cell volume. Results were evaluated using (in [A], [B], and [F]) ANOVA followed by Fisher’s PLSD post hoc test, (in [D] and [E]) Student’s t test, and (in [H]) linear regression analysis. Statistically significant differences (p < 0.05) are denoted as follows: a = different in a two-group comparison or within the weight group, b = different from both other measures, and d = different from one of the measures in the other weight group. See also Figure S2 and Table S3.

followed by DNA sequencing (ChIPsequencing) analysis was performed in human adipocytes differentiated in vitro. Using a stringent approach that intersects two independent peak-calling methods, 10,428 unique EBF1-binding regions were identified (Table S4). A de novo search for enriched sequence motifs in these sites identified an EBF1like binding motif as the most significantly enriched motif (Figure 3A), which was present in 78% of the identified binding regions. There was a clear enrichment of EBF1-binding sites in proximity to the transcriptional start site (TSS) of the nearest genes (Figure 3B), a distribution pattern that is in line with previous findings in B cells (Treiber et al., 2010) and more recently in 3T3-L1 cells (Griffin et al., 2013). In total, 2,501 genes were identified as potential EBF1 target genes (Table S5) with the detected EBF1-binding region close to the respective TSS (±10 kb). Pathway enrichment analysis revealed these genes to be primarily involved in lipid metabolism and WAT morphology (Figure 3C; Table S6). Because TF/DNA interactions do not necessarily imply transcriptional regulation, we performed global gene expression profiling in human adipocytes transfected with siRNA targeting EBF1 (knockdown efficiency shown in Figure S3A). A subset of the identified potential EBF1 target genes was also regulated by EBF1 RNAi treatment (Figure 3D; Table S5). A GSEA of the upregulated genes displayed only three significantly enriched gene sets, none of which were related to adipocyte metabolism (Table S7). In contrast, among the downregulated genes, there was a significant enrichment of several gene sets implicated in WAT morphology and adipocyte metabolism (in particular lipid Cell Metabolism 19, 981–992, June 3, 2014 ª2014 Elsevier Inc. 983

Cell Metabolism EBF1 Regulates Adipocyte Morphology and Lipolysis

Figure 3. Identification of Genes and Pathways Regulated by EBF1 (A) Cis-regulatory sequence motif identified in EBF1-bound sequences identified by ChIP-seq data. (B) Histogram displaying the distance of EBF1-binding regions to the transcription start site. (C) Identification of pathways enriched among genes bound by EBF1. Orange and purple bars depict mouse phenotype and gene ontology databases, respectively. (D) Genes bound by EBF1 and/or altered by EBF1 knockdown. According to ChIP-seq data, 2,501 genes bound EBF1; however, only 1,989 of these passed the nonspecific filter for microarray analysis and were included in this Venn diagram. Details about the filters are in the Experimental Procedures. (E) Identification of genes bound by EBF1, altered by EBF1 knockdown, and present in several of the enriched pathways. Genes bound and regulated by EBF1 as well as altered by morphology are indicated by red arrows. See also Figure S3 and Tables S4, S5, S6, S7, and S8.

984 Cell Metabolism 19, 981–992, June 3, 2014 ª2014 Elsevier Inc.

Cell Metabolism EBF1 Regulates Adipocyte Morphology and Lipolysis

a

1.5 1.0 0.5

C

1.2

Relative mRNA levels (vs. control)

0

1.0

8 a

6

2 0

a

a 0.6 a

a b b

0.2

2.5 2.0 1.5 1.0 0.5

PLIN1

LIPE

siPLIN1

0.4

EBF1

Ratio (LIPE/PLIN1)

Merged

4

0

D

LIPE

10

a 0.8

PLIN1

Control

2.0

E

12

a a

a Control siEBF1 siPLIN1 siEBF1/siPLIN1

siEBF1/siPLIN1

Relative glycerol release (vs. control)

b b

siEBF1

B

2.5

Isoprenaline-stimulated lipolysis (vs. basal lipolysis)

A

0

Figure 4. Functional Evaluation of EBF1 Knockdown in Human Adipocytes (A) Basal lipolysis determined in human adipocytes transfected with EBF1, PLIN1, EBF1/PLIN1, or control (nonsilencing) siRNAs. (B) Effects of EBF1 gene knockdown on isoprenaline-stimulated lipolysis (expressed as the quotient over basal lipolysis). (C) mRNA expression of the indicated genes following transfection of RNAi oligonucleotides targeting the same genes as in (A). (D) Representative immunofluorescence images of adipocytes transfected as in (A). Cells were incubated with antibodies directed against PLIN1 (left column) and LIPE (middle column), and the obtained images were merged (right column, with PLIN1 in red and LIPE in green). (E) The quotient between LIPE and PLIN1 levels at the lipid droplet surface was determined for each transfection condition. For (A)–(C), results are based on experiments performed in at least triplicate wells using cells obtained from a minimum of two different donors. Results in (D) and (E) were obtained as described in Experimental Procedures. Data are shown as mean ±SEM and were evaluated using (in [A], [C], and [D]) ANOVA followed by Fishers PLSD post hoc test or (in [B]) Student’s t test. Statistically significant differences (p < 0.05) are denoted as follows: a = different from control and b = different from both control and siEBF1. See also Figure S4.

metabolism) (Table S7); 24 genes were considered as key EBF1 target genes because they were included in many of the identified pathways (Figure 3E). Several of these genes are well-established regulators of adipogenesis (e.g., peroxisome proliferatoractivated receptor gamma, PPARG, and nuclear receptor corepressor 2, NCOR2) and lipolysis (e.g., hormone-sensitive lipase [LIPE], patatin-like phospholipase domain containing 2 [PNPLA2], perilipin 1 [PLIN1], and cell-death-inducing DFFAlike effector c [CIDEC]) (Girousse and Langin, 2012; Nofsinger et al., 2008). By overlapping results from our EBF1 ChIPsequencing (ChIP-seq) analysis with the global transcriptome data of human WAT with different morphologies, we found that out of the 24 key genes bound and regulated by EBF1 (Figure 3E), seven (PPARG, LIPE, PNPLA2, PLIN1, CIDEC, diacylglycerol O-acyltransferase 1 [DGAT1], and adiponectin [ADIPOQ]) were also expressed at significantly lower levels in nonobese subjects with adipose hypertrophy versus hyperplasia (Figure S3B;

Table S8). These genes were considered to be of particular interest and are highlighted by red arrows in Figures 3E and S3B. EBF1 Depletion Alters Adipocyte Lipolysis As evident from our clinical characterization of cohort 1, adipose hypertrophy is associated with altered basal, as well as stimulated, lipolysis, increased proinflammatory cytokine/chemokine secretion, and decreased insulin sensitivity both in vivo and in vitro (i.e., at the fat cell level). Although the functional role of EBF1 in controlling adipogenesis is well-established, the effects on other aspects of adipocyte function have not been characterized. To this end, human in-vitro-differentiated adipocytes were transfected with siRNA targeting EBF1. While no effects were observed on glucose transport (basal and insulin stimulated; Figure S4A) or secretion of TNFa, CCL2, IL6, and adiponectin (Figure S4B), basal (increased) and isoprenaline-induced (decreased) lipolysis were significantly altered (Figures 4A Cell Metabolism 19, 981–992, June 3, 2014 ª2014 Elsevier Inc. 985

Cell Metabolism EBF1 Regulates Adipocyte Morphology and Lipolysis

B

Total body weight (g)

60

a

25

a

d

40 d

30 20

a,d

20 a,d

15

d 10 d 5

10 0

a

a,d

50 Total fat mass (g)

A

Week 0 Week 3

Week 7

0

Week 10

C

Week 0 Week 3

Week 7

Week 10

D 150

gWAT adipocyte diameter (µM)

iWAT adipocyte diameter (µM)

150 c

120

c c

90 60 30 0

Week 0 Week 3

Week 7

c

120

d

a

d

a

a

Week 10

E

d 90

c a d

a

a

a d

d

a d

60 30 0

Week 0 Week 3

Ebf1+/+ Cd

Week 7

Week 10

Ebf1+/+ Hfd

mWAT adipocyte diameter (µM)

rpWAT adipocyte diameter (µM)

90

c

c a d

(A–F) WT and Ebf1+/ mice were fed chow (Cd) or high-fat diet (Hfd) for up to 10 weeks. Effects on (A) total body weight, (B) total fat mass, and adipocyte diameters in (C) inguinal (iWAT), (D) gonadal (gWAT), (E) retroperitoneal (rpWAT), and (F) mesenteric (mWAT) adipose tissue were determined. Results are based on measurements in all available mice (5–10 per group) and are shown as mean ±SEM. For Cd at baseline, differences were assessed using Student’s t test and statistically significant differences (p < 0.05) are marked with an ‘‘a.’’ To evaluate diet- and genotype-specific effects, data were analyzed using ANOVA followed by Fishers PLSD post hoc test. Statistically significant differences (p < 0.05) between groups at 3, 7, and 10 weeks are denoted as follows: a = different between diet within a genotype, b = different from both measures in the other genotype, c = different both between diet within the genotype and with both measures in the other genotype, and d = different from one of the measures in the other genotype. See also Figure S5.

30 0

Week 0 Week 3

150

120

a,d d

60

F 150

a

Figure 5. Diet and Genotype Effects on Body Composition and Fat Cell Diameter

120 90

b

b

a

60 30 0

Week 0 Week 3

Ebf1+/- Cd

and 4B). Quantitative PCR and western blot confirmed the EBF1 RNAi microarray data (i.e., a significant 30%–40% downregulation in the mRNA and protein expression of key lipolytic genes, including PLIN1 and LIPE) (Figures 4C and S4C). A combined decrease in the expression of PLIN1 and LIPE can result in increased lipolysis in human adipocytes provided that the ratio between LIPE and PLIN1 is increased at the lipid droplet surface (Stenson et al., 2011). In order to dissect mechanisms by which EBF1 controls basal lipolysis, human adipocytes were transfected with siRNAs targeting EBF1, PLIN1, or both (Figure 4C). While EBF1 and PLIN1 RNAi induced similar, albeit quantitatively slightly different, increases in glycerol release (Figure 4A) and the ratio of LIPE/PLIN1 protein at the lipid droplet surface (Figures 4D and 4E), there was no additive effect of combining the two treatments. The in vivo relevance of these findings was corroborated by the significant negative association between EBF1 mRNA levels in WAT and in vivo lipolysis (cohort 3, Figure S4D). 986 Cell Metabolism 19, 981–992, June 3, 2014 ª2014 Elsevier Inc.

Week 7

Week 10

EBF1 Reduction Promotes Adipose Hypertrophy in Mice In order to determine whether murine Ebf1 expression was regulated by alterations in fat mass in a similar manner as c c a observed in humans, publically available b transcriptional data of murine WAT depots were retrieved. The results from this analysis demonstrated that Ebf1 mRNA expression was significantly reduced upon high-fat diet (Hfd) (p = 0.0034 for overall difference in five indiWeek 7 Week 10 vidual studies; Table S9). To directly determine the role of EBF1 in the development of adipose morphology, we Ebf1+/- Hfd used a murine knockout model. Because Ebf1 / mice are lipodystrophic (Festa et al., 2011; Fretz et al., 2010) and therefore not suitable for studies of WAT function, Ebf1+/ animals were generated and subjected to Hfd for up to 10 weeks. Recent results have highlighted qualitative differences in the expandability between gonadal (gWAT) and inguinal (iWAT) depots (Wang et al., 2013). Assessments were therefore performed in four separate regions: iWAT, gWAT, retroperitoneal (rpWAT), and mesenteric (mWAT). At baseline, compared to wild-type (WT), Ebf1+/ mice displayed similar total body weight, total fat mass, lean body mass, brown adipose tissue (BAT), as well as iWAT, gWAT, rpWAT, and mWAT depot weights (Figures 5 and S5). During and at the end of the Hfd, both genotypes had, to a similar degree, increased their total body weight, total fat mass, lean body mass, BAT, and all but one WAT depot (the exception being gWAT, which was slightly less increased in the Ebf1+/ mice at week 10) (Figures 5 and S5). Analyses of fat cell diameter demonstrated that, compared to WT littermates, heterozygous animals on chow diet displayed consistently larger

Cell Metabolism EBF1 Regulates Adipocyte Morphology and Lipolysis

C

Time (minutes)

Time (minutes)

Week 10

Time (minutes)

c 150

c c

c

100 50 0

150

50

c

90

100

200

120

150

Week 10

a,d a,d

60

Serum glucose (mg/dl)

200

0

Week 7

250

0

0

Week 0 Week 3

15 30

50

50

Serum glucose (mg/dl)

100

100

150

150

90 120

60

0

15 30

50

150

150

Week 7

250

200

200

0

Week 10

90

100

Week 7

120

Serum glucose (mg/dl)

150

Week 0 Week 3

Week 3

250

200

0

0

Lipe

60

Week 0

250

100

0

Ebf1 Plin1

d

Serum glucose (mg/dl)

0

E Serum glucose (mg/dl)

0.3

200

150

0

a

300

90

1

a

0.6

a

a,d

400

120

2

0.9

250 a

60

3

1.2

500

Serum insulin (pmol/l)

4

600

0

a

D

1.5

15 30

Relative iWAT mRNA levels (vs. Ebf1+/+ Cd)

5

15 30

B

Relative iWAT glycerol release (vs. Ebf1+/+ Cd)

A

Time (minutes)

Figure 6. Diet and Genotype Effects on Lipolysis and Measures of Insulin Resistance (A and B) (A) Lipolysis and (B) mRNA expression of indicated genes in inguinal adipose tissue (iWAT) from WT and Ebf1+/ mice. (C–E) (C) Insulin and (D) glucose levels as well as (E) insulin tolerance tests (ITTs) at baseline and during Cd or Hfd at 3, 7, and 10 weeks in WT and Ebf1+/ mice. Results are based on measurements in all available mice (5–10 per group) and are shown as mean ±SEM. In two-group comparisons in (A)–(D) (i.e., for animals at baseline) differences were assessed using Student’s t test and statistically significant differences (p < 0.05) are marked with an ‘‘a.’’ To evaluate diet- and genotype-specific effects, data were analyzed using ANOVA followed by Fisher’s PLSD post hoc test. Statistically significant differences (p < 0.05) between groups were observed at 10 weeks and are denoted as follows: a = different between diet within a genotype, c = different both between diet within the genotype and with both measures in the other genotype, and d = different from one of the measures in the other genotype. In (E), note that there was no sign of glucose recovery at the 150 min time point. As detailed in the Supplemental Experimental Procedures, this could depend on the fact that animals were fasted overnight prior to the assessments.

adipocytes in the studied WAT depots over time (Figures 5C–5F). While Hfd in general increased fat cell size, the differences between genotypes persisted in all WAT depots up to 7 weeks of hypercaloric intake. At 10 weeks of Hfd, the observed differences in adipocyte diameters were no longer evident in gWAT, borderline significant in rpWAT (p = 0.061), and remained significant in iWAT and mWAT. Thus, while the fat cell diameter in haploinsufficent mice fed Hfd plateaued at 115–120 mm after 3–7 weeks, WT littermates reached this size only at 10 weeks. Effects of EBF1 Reduction on Lipolysis and Insulin Sensitivity in Mice To evaluate the metabolic phenotype of Ebf1 haploinsufficiency, we determined lipolysis in intact iWAT, measured fasting plasma insulin/glucose levels, and performed insulin tolerance tests (ITTs). Compared to WT mice, lipolysis was higher in heterozygous animals (Figure 6A), and this was accompanied by reduced mRNA levels of Ebf1, Plin1, and Lipe (Figure 6B). Although Hfd induced a significant increase in fasting insulin and a trend toward higher glucose levels (Figures 6C and 6D), there were no effects of genotype on these measures. However, when performing ITTs, Ebf1+/ mice at 10 (but not baseline), 3, or 7 weeks of Hfd displayed a significantly attenuated response to insulin

between 30 and 150 min compared to pair-fed WT animals (Figure 6E). This was independent of whether the comparisons were made at each time point by calculating the area under the curve (p = 0.028; data not shown) or expressing the results as percent of baseline glucose values (data not shown). In contrast, although WT animals on Hfd displayed slightly higher glucose values at each time points in the ITT curves, they were not statistically different from those of chow-fed mice. It should be noted that the mice were generated on a mixed genetic background (C57BL/6-129X1/SvJ) and fed Hfd for a relatively short time period. These aspects may explain why WT littermates did not develop pronounced insulin resistance at the end of the diet intervention. EBF1 Expression Is Regulated by TNFa Because adipose morphology and WAT inflammation are closely associated, we hypothesized that proinflammatory adipokines could affect EBF1 expression. This notion was corroborated by the observation that EBF1 activity correlated negatively with WAT secretion of TNFa, a cytokine with well-established effects on adipocyte lipolysis, insulin sensitivity, and adipogenesis (cohort 3, Figure 7A). This association was independent of BMI (partial r = 0.47; p = 0.0001). Furthermore, in vitro experiments Cell Metabolism 19, 981–992, June 3, 2014 ª2014 Elsevier Inc. 987

Cell Metabolism EBF1 Regulates Adipocyte Morphology and Lipolysis

A

B

Figure 7. Relationship between EBF1 and TNFa (A) Linear regression analysis between WAT TNFa secretion in vitro and EBF1 activity. After correction for BMI, the correlation remained significant (see main text). (B) EBF1 mRNA expression following incubations with recombinant TNFa in in-vitro-differentiated human adipocytes. Results are based on experiments performed in duplicate wells using cells obtained from two different donors. Data are shown as mean ±SEM and evaluated using Student’s t test. The statistically significant difference (p < 0.05) is marked with an ‘‘a.’’

in human adipocytes confirmed that recombinant TNFa downregulated EBF1 expression (Figure 7B). DISCUSSION In vitro studies have conclusively demonstrated that EBF1 plays a central role in adipogenesis, and the in vivo relevance of these findings is evident from the fact that Ebf1 / mice are lipodystrophic (Fretz et al., 2010). Our unbiased approach identified, among a large number of TFs, adipocyte EBF1 expression and activity to be closely linked to adipose morphology in human WAT. The EBF1 regulatory cistrome included key genes involved in lipid metabolism and adipocyte differentiation/morphology, and EBF1 depletion led to altered lipolysis, increased insulin resistance, and WAT hypertrophy. Although we cannot exclude the possibility that reduced Ebf1 expression in other tissues may contribute to the observed metabolic phenotype, these data indicate that EBF1 is a central factor controlling the morphologic and metabolic phenotype of WAT as well as whole-body insulin sensitivity. The gene set bound and regulated by EBF1 in primary human adipocytes was determined and validated by combining transcriptional profiling, ChIP-seq, and RNAi. Overall, this approach identified genes involved in lipid metabolism and adipogenesis, two processes that are essential for adipocyte function. In support of these observations, EBF1 RNAi resulted in increased basal and attenuated stimulated lipolysis, most probably due to reduced mRNA and protein levels of PLIN1 and LIPE, as well as an altered ratio between these two proteins at the lipid droplet surface. This is consistent with the lipolytic phenotype observed in WAT hypertrophy and obesity. Current models posit that the relative expression and the lipid-droplet-associated interaction between coating proteins (e.g., PLIN1, CIDEA, and CIDEC) and lipases (LIPE and PNPLA2) control the rate of basal and stimulated lipolysis (reviewed in Girousse and Langin, 2012). It is therefore conceivable that other factors, such as CIDEA and PNPLA2, may also influence EBF1-mediated effects on lipolysis. In contrast, the specific genes involved in determining adipose 988 Cell Metabolism 19, 981–992, June 3, 2014 ª2014 Elsevier Inc.

morphology are less evident. The observation that WAT hypertrophy is associated with reduced adipocyte turnover (Arner et al., 2010b) suggests that genes involved in controlling adipocyte birth/death rates may be of importance. Moreover, it has been suggested that enlarged adipocyte size may represent a state of impaired WAT expandability due to altered adipogenesis (Tan and Vidal-Puig, 2008; Virtue and Vidal-Puig, 2010). Consistent with this, EBF1, PPARG, and CEBPA mRNA levels in WAT were lower in nonobese hypertrophy. As these three TFs cross-talk by regulating the expression of each other, as well as similar target genes (Griffin et al., 2013), it is tempting to speculate that EBF1 reduction, with concomitant alterations in PPARg and CEBPa levels, results in WAT hypertrophy via impaired fat cell differentiation. Because there are no established in vitro techniques to evaluate direct effects on morphology, the causal link between EBF1 downregulation and WAT hypertrophy was studied in Ebf1+/ mice. Compared with WT littermates, heterozygous animals displayed no significant changes in body composition or WAT/ BAT depot sizes. However, analogous with our findings in humans, the expression of lipolytic regulators was decreased, while spontaneous (basal) glycerol release was increased in iWAT. Moreover, fat cell size was significantly larger in all four studied WAT depots, but despite this, there was no difference in insulin sensitivity when animals were fed chow diet. Because both increased lipolysis and adipocyte size confer an increased risk for metabolic complications, these findings prompted us to explore the effects of a Hfd intervention. As expected, irrespective of genotype, 10 weeks of high-fat feeding induced a significant increase in body weight, WAT depot sizes, and circulating insulin concentrations. Heterozygous mice displayed significantly larger fat cell size throughout the observation period, as well as attenuated insulin sensitivity (determined by ITT) at the end of the study. Notably, while our assessments in humans were based on WAT from the abdominal subcutaneous region, our murine data demonstrate that Ebf1 haploinsufficiency results in similar alterations in adipose morphology in four different WAT depots, including two visceral regions (gWAT and mWAT). Taken together, these results suggest that reduced Ebf1 expression results in WAT hypertrophy and increased basal lipolysis and confers an increased risk of developing insulin resistance under hypercaloric conditions. To what extent the secondary reductions in Pparg/Cebpa contribute to this phenotype is presently unknown. It would therefore be of interest to study the effects of thiazolidinediones in Ebf1+/ animals as well as to determine the consequences of overexpressing Ebf1 in mice with genetically modified levels of Pparg/Cebpa. Although the current clinical and transcriptomic analyses confirm a tight association between adipose hypertrophy, reduced insulin sensitivity, elevated basal lipolysis, and increased local inflammation, the causal relationship between these changes is not known. Even if our data suggest that EBF1 affects adipose morphology and lipolysis, a combination of ChIP-seq, transcriptional profiling, RNAi, and ELISA consistently demonstrated that EBF1 reduction did not influence the activity of inflammatory pathways in human adipocytes. In contrast, TNFa downregulated adipocyte EBF1 expression, and the levels of TNFa secretion correlated negatively with EBF1 activity. Thus, adipose hypertrophy may not be a cause,

Cell Metabolism EBF1 Regulates Adipocyte Morphology and Lipolysis

but rather a consequence, of increased local inflammation acting via reduced EBF1 expression/activity. This hypothesis is supported by recent findings demonstrating that subcutaneous WAT from lean and healthy young women display a high correlation between adipose secretion of TNFa and adipose morphology (Arner et al., 2010a). Recent work explored the effects of Ebf1 depletion by adenoviral shRNA in the immortalized murine cell line 3T3-L1 (Griffin et al., 2013). In analogy with our data, a significant but selective decrease in adipogenic genes such as Pparg and Cebpa was observed without a global reduction of fat-cell-enriched transcripts. These results, together with the observation that EBF1 RNAi did not alter insulin sensitivity or adipokine secretion, suggest that EBF1 knockdown does not induce significant dedifferentiating effects in vitro. However, in clear contrast to our present data, Griffin and coworkers (Griffin et al., 2013) found a significant enrichment in inflammatory signaling (although this was probably regulated via indirect mechanisms), reduced glucose transport, and attenuated TLR signaling (including reduced cytokine release and LPS-stimulated lipolysis). These differences imply that the cellular context in which EBF1 is studied impacts strongly on the data, a notion which is supported and discussed by Griffin et al. (2013) after comparing their ChIP-seq with similar analyses performed in non-adipose cells. Several groups have aimed to elucidate the mechanisms underlying the development of adipose hypertrophy/hyperplasia under isocaloric or hypercaloric conditions in different murine WAT depots (reviewed in Berry et al., 2014). Using different techniques to determine adipocyte formation, these studies have come to somewhat divergent conclusions suggesting that upon Hfd, hyperplasia primarily occurs in either iWAT (Joe et al., 2009) or gWAT (Wang et al., 2013). In the latter work, the authors also proposed that large adipocytes, at least under some conditions, may in fact be new adipocytes formed through adipogenesis. Our data, demonstrating similar phenotypes in four different WAT depots of Ebf1+/ , both before and after Hfd, suggest that the effects of Ebf1 on adipocyte formation are region independent and persist at the onset of diet-induced obesity. Our combined clinical and transcriptomic analyses identified three distinct phenotypes—nonobese with hyperplasia, nonobese with hypertrophy, and obese subjects—implying that nonobese with WAT hypertrophy constitute a specific subgroup that, although anthropometrically nonobese, displays a metabolic profile that more closely resembles the obese state. In contrast, obesity appears to override the effects of adipocyte size and number, most probably due to several other associated disturbances. These observations, together with previous findings demonstrating that WAT hypertrophy confers an increased risk for future development of type 2 diabetes (Lo¨nn et al., 2010; Weyer et al., 2000), suggest that lifestyle interventions should be commenced early based on assessments of adipose morphology already in the nonobese state. Taken together, EBF1 is an important regulator of WAT morphology and fat cell lipolysis by controlling distinct gene sets and may therefore constitute a pathophysiological link between adipose hypertrophy and common metabolic disorders. Whether EBF1 is also a suitable therapeutic target remains to be demonstrated.

EXPERIMENTAL PROCEDURES Subjects Cohorts 1–3, described in detail in Table S1 and Supplemental Experimental Procedures, comprised healthy women and men subdivided into obese/ nonobese and hyperplastic/hypertrophic WAT. The study was approved by the regional board of ethics, and written informed consent was obtained from all participants. Transcriptional Profiling and MARA Gene expression profiling was performed using Affymetrix Human Gene 1.0 (cohort 3, previously presented in Arner et al., 2012) and Affymetrix Human Gene 1.1 (EBF1 knockdown versus control) ST arrays (Affymetrix, Inc., Santa Clara). Data were analyzed with packages available from Bioconductor (http:// www.bioconductor.org). Normalization and calculation of gene expression was performed with the Robust Multichip Average expression measure using oligo package (Carvalho and Irizarry, 2010). Prior to further analysis, a nonspecific filter was applied to include genes with expression signal >30 in at least 20% of all samples. Limma package (Smyth, 2004) was used to identify the differentially expressed genes that were ranked in the gene set enrichment analyses, and PCA was performed using the FactomineR package. MARA was performed exactly as described (Arner et al., 2012; Suzuki et al., 2009) and detailed in Supplemental Experimental Procedures. ChIP Adipocytes differentiated in vitro were fixed in 1% formaldehyde for 15 min at room temperature and quenched for 5 min by adding glycine to a final concentration of 0.125 M. Cells were harvested, and nuclear fractions were enriched by a brief sonication in lysis buffer A (10 mM Tris/Hcl [pH7.5], 10 mM NaCl, 3 mM MgCl2, 10mM KCl2, 0.05% Nonidet P-40, and 13 protease inhibitors [Roche Complete EDTA-free cocktail tablets]) followed by centrifugation at 5000 3 g for 5 min. The collected pellets were suspended in 1,000 ml cell lysis buffer B (50 mM Tris [pH 8.0], 1 mM EDTA, 0.5 mM EGTA, 1% Triton X-100, 0.1% Na-deoxycholate, 150 mM NaCl, and protease inhibitor). ChIP assays were then performed following the protocol described in Gao et al., (2008). Samples were immunoprecipitated with 5 mg custom-made EBF1 antibody (Abmart, P.R. China) or rabbit IgG (Santa Cruz Biotechnology) at 4 C overnight. ChIP-Seq ChIP-seq libraries were prepared using NEBnext ChIP-seq Library Prep Master Mix Set (New England Biolabs, Ipswich) from 5 ng of anti-EBF1 and anti-IgG ChIP DNA, respectively. Sequence data were generated with Illumina HiSeq 2000 single-read sequencing and aligned against the human genome (hg19, NCBI) using Burrows-Wheeler Aligner (Li and Durbin, 2009) with default parameters. The uniquely mapped reads were analyzed to identify EBF1-binding regions (with anti-IgG DNA as a control) by two different approaches— MACS (Zhang et al., 2008) and Homer (Heinz et al., 2010)—using default settings. Regions that were detected by both methods were considered as EBF1-binding regions. An independent EBF1 ChIP-seq experiment demonstrated a high reproducibility of the identified regions (Pearson correlation r = 0.89) when comparing the normalized tag counts between replicates. Gene Annotation Analysis Genomic Regions Enrichment of Annotations Tool (GREAT) (McLean et al., 2010) was employed to perform the enrichment analysis of gene annotation in the proximity of EBF1-binding regions against the whole genome background. In the GREAT analysis, association of EBF1 genomic regions with nearby genes was determined by the ‘‘basal plus extension’’ rule with the parameter of 5 + 1 kb basal promoter and a more limited 10 kb extension. Significantly enriched gene sets were selected by FDR q value < 0.05 for both binomial and hypergeometric tests. Significantly enriched gene sets between 15 and 600 genes were included. GSEA GSEA (Subramanian et al., 2005) was employed to analyze the altered gene expression profiles following EBF1 knockdown in adipocytes and between the human WAT with different morphologies. In order to identify core EBF1 target genes/pathways affected by altered EBF1 mRNA/protein levels in

Cell Metabolism 19, 981–992, June 3, 2014 ª2014 Elsevier Inc. 989

Cell Metabolism EBF1 Regulates Adipocyte Morphology and Lipolysis

human adipocytes, 207 gene sets were compiled. These gene sets were derived from the significantly enriched terms identified in the GREAT analysis described above and contained only the genes associated with EBF1-binding regions identified by ChIP-seq (Table S5). Mice Ebf1 / mice were originally generated and supplied by Dr. Rudy Grosschedel (Lin and Grosschedl, 1995). C57BL/6 and 129X1/SvJ mice were obtained from the National Cancer Institute and The Jackson Laboratory (Bar Harbor), respectively, and housed at Yale University School of Medicine. Ebf1+/ and age-matched WT littermate control mice were generated and characterized as described in the Supplemental Experimental Procedures. The Yale Medical School Institutional Animal Care and Use Committee approved the purchase and use of all animals. Cell Culture Cell culture experiments were performed on human WAT adipocytes and mesenchymal stem cells differentiated in vitro. The isolation procedure and culture/differentiation conditions have been described previously (Ehrlund et al., 2013; van Harmelen et al., 2005). In-vitro-differentiated adipocytes were stimulated with human recombinant TNFa (50 ng/ml, Sigma-Aldrich) for 6 hr, upon which effects on EBF1 mRNA levels were measured. RNAi Human adipocytes differentiated in vitro were transfected with ONTARGETplus SMARTpool (Thermo Fisher Scientific, Lafayette) targeting EBF1 (L-011848-01) or Non-targeting Control Pool (D-001810-10) siRNA and HiPerfect Transfection Reagent (QIAGEN, Hilden) as described (Kulyte´ et al., 2014). The cells were incubated for 48–72 hr, upon which functional assays were performed and/or RNA/protein and culture media were collected. RNA Isolation, cDNA Synthesis, and qPCR Total RNA was extracted using RNeasy Lipid Tissue Mini Kit/miRNeasy mini kit (QIAGEN, Hilden) or NucleoSpin RNA II kit (Macherey-Nagel, Du¨ren) according to the manufacturers’ instructions. Subsequently, RNA concentration and quality were measured using Nanodrop 2000 spectrophotometer (Thermo Fisher Scientific, Lafayette) and Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto), respectively. mRNA was reverse transcribed to cDNA using Omniscript First-strand cDNA synthesis kit (QIAGEN) and random hexamer primers (Invitrogen, Carlsbad). Assessment of mRNA expression was performed as described (Pettersson et al., 2011)m and relative levels were calculated using a comparative Ct-method (i.e., 2DCt-target gene/2DCt-reference gene). 18S rRNA and LRP10 (human samples) as well as ACTB (murine samples) were used to normalize the expression of all analyzed genes. TaqMan and SYBR Green assays used in this study are listed in the Supplemental Experimental Procedures. Immunofluorescence and Imaging All analyses were performed in human in-vitro-differentiated adipocytes as described in detail in the Supplemental Experimental Procedures. Fractionation of WAT The stroma-vascular fraction of subcutaneous WAT from ten subjects was washed in PBS/0.5% BSA/2 mM EDTA buffer and stained with anti-CD45 Pacific blue clone T29/33 (DakoCytomation, Glostrup) and anti-CD14 PE (BD Biosciences, New Jersey) antibodies. After washing, cells were resuspended in PBS/0.1% BSA/2 mM/1 mM MgCl2/25 mM DNase buffer, passed through a 70 mm cell mesh (BD Biosciences), and sorted using a FACSAria III cell sorter (BD Biosciences). Leucocytes were defined as CD45+/CD14 while CD45+/CD14+ cells corresponded to the macrophage/monocyte fraction. Cell purity was investigated after each sorting and constituted 95.1% ± 2.9% for the monocyte/macrophage and 98.8% ± 0.87% for the leucocyte fraction. Glycerol Release and Glucose Transport The effects of EBF1 knockdown on basal/insulin-stimulated glucose uptake and basal or isoprenaline-stimulated lipolysis in human adipocytes differenti-

990 Cell Metabolism 19, 981–992, June 3, 2014 ª2014 Elsevier Inc.

ated in vitro were determined as described previously (Dicker et al., 2009; Mejhert et al., 2010). Protein Secretion and Western Blot Protein levels of human IL6, CCL2, TNFa (D6050, DCP00, and QTA00B; R&D Systems, Minneapolis), and adiponectin (10-1193-01; Mercodia, Uppsala) in conditioned media were measured using ELISA according to the manufacturer’s instructions. Western blot was performed as described previously (Mejhert et al., 2013). Primary antibodies were PLIN1 (Progen, Heidelberg), LIPE (Cell Signaling Technologies, Beverly), and b-actin (Sigma-Aldrich). Statistical Analyses Analyses were performed using standard software packages, and relevant statistical methods are detailed in the figure legends and under the respective subheadings in the Experimental Procedures. When performing ANOVA, post hoc tests were only performed if the overall ANOVA p value was <0.05. To facilitate the overview of statistical differences in multigroup comparisons, a uniform lettering system was adopted and is explained in detail under each figure legend. For data sets displaying nonnormal distribution, nonparametric statistical tests or log-transformed values were used. Error bars in Figures 1 and 4–7 represent SEM. Materials and Data Transcriptional profiles are accessible using Gene Expression Omnibus (GEO) accession numbers GSE25402 and GSE42680. For ChIP-seq results, both raw and mapped data are available using GEO accession number GSE54889. SUPPLEMENTAL INFORMATION Supplemental Information includes five figures, nine tables, and Supplemental Experimental Procedures and can be found with this article online at http://dx. doi.org/10.1016/j.cmet.2014.03.032. AUTHOR CONTRIBUTIONS H.G., N.M., M.R., and P.A. designed the study and analyzed the data. N.M. and M.R. wrote the first version of the manuscript. All authors read, gave input, and approved the final version of the manuscript. I.D., M.R., and P.A. generated the clinical data (cohort 1–3). H.G., N.M., X.G., S.S., S.L-C., A.E., Iy.D., and J.L. performed the human adipocyte experiments. J.A.F. M.C.H, P.A., M.R., and N.M. characterized the Ebf1+/ and WT mice. H.G., N.M., E.A., K.D-W., and C.O.D. performed the bioinformatic analyses. None of the authors declare any competing financial interests. ACKNOWLEDGMENTS We thank Dr. Rudy Grosschedel (Department of Cellular and Molecular Immunology, Max Planck Institute of Immunobiology and Epigenetics, Germany) for the Ebf1-deficient mice and Elisabeth Dungner, Eva Sjo¨lin, Kerstin Wa˚hle´n, Gaby A˚stro¨m, Britt-Marie Leijonhufvud, Katarina Hertel, and Yvonne Widlund (Department of Medicine [H7], Karolinska Institutet, Sweden) for excellent technical assistance. We also thank Dr. Sylvie Le Guyader (Department of Biosciences and Nutrition, Karolinska Institutet, Sweden) for assistance in microscopic imaging analyses. This work was supported by grants from the Swedish Research Council; the NovoNordisk Foundation; and the European Association on the Study of Diabetes, together with Eli-Lilly; the Swedish Diabetes Foundation; Diabetes Wellness Fund; the Strategic Research Program in Diabetes at Karolinska Institutet; the National Institute of Diabetes and Digestive and Kidney Diseases; the National Institute of Arthritis and Musculoskeletal and Skin Diseases/National Institutes of Health grants R24DK092759, RO1AR052690, and 1K99DK093711; the Yale Core Center for Musculoskeletal Disorders P30AR046032; the Yale Diabetes Endocrinology Research Center P30DK045735; the Department of Orthopaedics and Rehabilitation at Yale University School of Medicine; the Ministry of Education, Culture, Sports, Science & Technology in Japan; and the EU-FP7-Systems microscopy Network of Excellence. Microscopic imaging was performed at the Live Cell Imaging unit at the Department of Biosciences and Nutrition at Karolinska Institutet,

Cell Metabolism EBF1 Regulates Adipocyte Morphology and Lipolysis

supported by grants from The Knut & Alice Wallenberg Foundation, the Center for Biosciences at Karolinska Institutet, and the Jonasson donation. Received: June 10, 2013 Revised: February 11, 2014 Accepted: March 26, 2014 Published: May 22, 2014 REFERENCES Akerblad, P., Lind, U., Liberg, D., Bamberg, K., and Sigvardsson, M. (2002). Early B-cell factor (O/E-1) is a promoter of adipogenesis and involved in control of genes important for terminal adipocyte differentiation. Mol. Cell. Biol. 22, 8015–8025. Arner, E., Ryde´n, M., and Arner, P. (2010a). Tumor necrosis factor alpha and regulation of adipose tissue. N. Engl. J. Med. 362, 1151–1153. Arner, E., Westermark, P.O., Spalding, K.L., Britton, T., Ryde´n, M., Frise´n, J., Bernard, S., and Arner, P. (2010b). Adipocyte turnover: relevance to human adipose tissue morphology. Diabetes 59, 105–109. Arner, P., Arner, E., Hammarstedt, A., and Smith, U. (2011). Genetic predisposition for Type 2 diabetes, but not for overweight/obesity, is associated with a restricted adipogenesis. PLoS ONE 6, e18284. Arner, E., Mejhert, N., Kulyte´, A., Balwierz, P.J., Pachkov, M., Cormont, M., Lorente-Cebria´n, S., Ehrlund, A., Laurencikiene, J., Hede´n, P., et al. (2012). Adipose tissue microRNAs as regulators of CCL2 production in human obesity. Diabetes 61, 1986–1993. Berry, R., Jeffery, E., and Rodeheffer, M.S. (2014). Weighing in on adipocyte precursors. Cell Metab. 19, 8–20.

Jerna˚s, M., Palming, J., Sjo¨holm, K., Jennische, E., Svensson, P.A., Gabrielsson, B.G., Levin, M., Sjo¨gren, A., Rudemo, M., Lystig, T.C., et al. (2006). Separation of human adipocytes by size: hypertrophic fat cells display distinct gene expression. FASEB J. 20, 1540–1542. Jimenez, M.A., Akerblad, P., Sigvardsson, M., and Rosen, E.D. (2007). Critical role for Ebf1 and Ebf2 in the adipogenic transcriptional cascade. Mol. Cell. Biol. 27, 743–757. Joe, A.W., Yi, L., Even, Y., Vogl, A.W., and Rossi, F.M. (2009). Depot-specific differences in adipogenic progenitor abundance and proliferative response to high-fat diet. Stem Cells 27, 2563–2570. Kulyte´, A., Belarbi, Y., Lorente-Cebria´n, S., Bambace, C., Arner, E., Daub, C.O., Hede´n, P., Ryde´n, M., Mejhert, N., and Arner, P. (2014). Additive effects of microRNAs and transcription factors on CCL2 production in human white adipose tissue. Diabetes 63, 1248–1258. Li, H., and Durbin, R. (2009). Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25, 1754–1760. Lin, H., and Grosschedl, R. (1995). Failure of B-cell differentiation in mice lacking the transcription factor EBF. Nature 376, 263–267. Lo¨nn, M., Mehlig, K., Bengtsson, C., and Lissner, L. (2010). Adipocyte size predicts incidence of type 2 diabetes in women. FASEB J. 24, 326–331. McLean, C.Y., Bristor, D., Hiller, M., Clarke, S.L., Schaar, B.T., Lowe, C.B., Wenger, A.M., and Bejerano, G. (2010). GREAT improves functional interpretation of cis-regulatory regions. Nat. Biotechnol. 28, 495–501. Mejhert, N., Laurencikiene, J., Pettersson, A.T., Kaaman, M., Stenson, B.M., Ryde´n, M., and Dahlman, I. (2010). Role of Receptor-Interacting Protein 140 in human fat cells. BMC Endocr. Disord. 10, 1.

Carvalho, B.S., and Irizarry, R.A. (2010). A framework for oligonucleotide microarray preprocessing. Bioinformatics 26, 2363–2367.

Mejhert, N., Wilfling, F., Esteve, D., Galitzky, J., Pellegrinelli, V., Kolditz, C.I., Viguerie, N., Tordjman, J., Na¨slund, E., Trayhurn, P., et al. (2013). Semaphorin 3C is a novel adipokine linked to extracellular matrix composition. Diabetologia 56, 1792–1801.

Dicker, A., Astro¨m, G., Wa˚hle´n, K., Hoffstedt, J., Na¨slund, E., Wire´n, M., Ryde´n, M., Arner, P., and van Harmelen, V. (2009). Primary differences in lipolysis between human omental and subcutaneous adipose tissue observed using in vitro differentiated adipocytes. Horm. Metab. Res. 41, 350–355.

Nofsinger, R.R., Li, P., Hong, S.H., Jonker, J.W., Barish, G.D., Ying, H., Cheng, S.Y., Leblanc, M., Xu, W., Pei, L., et al. (2008). SMRT repression of nuclear receptors controls the adipogenic set point and metabolic homeostasis. Proc. Natl. Acad. Sci. USA 105, 20021–20026.

Ehrlund, A., Mejhert, N., Lorente-Cebria´n, S., Astro¨m, G., Dahlman, I., Laurencikiene, J., and Ryde´n, M. (2013). Characterization of the Wnt inhibitors secreted frizzled-related proteins (SFRPs) in human adipose tissue. J. Clin. Endocrinol. Metab. 98, E503–E508.

Pettersson, A.T., Mejhert, N., Jerna˚s, M., Carlsson, L.M., Dahlman, I., Laurencikiene, J., Arner, P., and Ryde´n, M. (2011). Twist1 in human white adipose tissue and obesity. J. Clin. Endocrinol. Metab. 96, 133–141.

Festa, E., Fretz, J., Berry, R., Schmidt, B., Rodeheffer, M., Horowitz, M., and Horsley, V. (2011). Adipocyte lineage cells contribute to the skin stem cell niche to drive hair cycling. Cell 146, 761–771. Fretz, J.A., Nelson, T., Xi, Y., Adams, D.J., Rosen, C.J., and Horowitz, M.C. (2010). Altered metabolism and lipodystrophy in the early B-cell factor 1-deficient mouse. Endocrinology 151, 1611–1621. Gao, H., Fa¨lt, S., Sandelin, A., Gustafsson, J.A., and Dahlman-Wright, K. (2008). Genome-wide identification of estrogen receptor alpha-binding sites in mouse liver. Mol. Endocrinol. 22, 10–22. Girousse, A., and Langin, D. (2012). Adipocyte lipases and lipid droplet-associated proteins: insight from transgenic mouse models. Int. J. Obes. (Lond) 36, 581–594. Griffin, M.J., Zhou, Y., Kang, S., Zhang, X., Mikkelsen, T.S., and Rosen, E.D. (2013). Early B-cell factor-1 (EBF1) is a key regulator of metabolic and inflammatory signaling pathways in mature adipocytes. J. Biol. Chem. 288, 35925–35939. Gustafson, B., Hammarstedt, A., Hedjazifar, S., and Smith, U. (2013). Restricted adipogenesis in hypertrophic obesity: the role of WISP2, WNT, and BMP4. Diabetes 62, 2997–3004. Heinz, S., Benner, C., Spann, N., Bertolino, E., Lin, Y.C., Laslo, P., Cheng, J.X., Murre, C., Singh, H., and Glass, C.K. (2010). Simple combinations of lineagedetermining transcription factors prime cis-regulatory elements required for macrophage and B cell identities. Mol. Cell 38, 576–589. Hoffstedt, J., Arner, E., Wahrenberg, H., Andersson, D.P., Qvisth, V., Lo¨fgren, P., Ryde´n, M., Tho¨rne, A., Wire´n, M., Palme´r, M., et al. (2010). Regional impact of adipose tissue morphology on the metabolic profile in morbid obesity. Diabetologia 53, 2496–2503.

Rosen, E.D., and Spiegelman, B.M. (2014). What we talk about when we talk about fat. Cell 156, 20–44. Skurk, T., Alberti-Huber, C., Herder, C., and Hauner, H. (2007). Relationship between adipocyte size and adipokine expression and secretion. J. Clin. Endocrinol. Metab. 92, 1023–1033. Smyth, G.K. (2004). Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat. Appl. Genet. Mol. Biol. 3, http://dx.doi.org/10.2202/1544-6115.1027. Stenson, B.M., Ryde´n, M., Venteclef, N., Dahlman, I., Pettersson, A.M., Mairal, A., Astro¨m, G., Blomqvist, L., Wang, V., Jocken, J.W., et al. (2011). Liver X receptor (LXR) regulates human adipocyte lipolysis. J. Biol. Chem. 286, 370–379. Subramanian, A., Tamayo, P., Mootha, V.K., Mukherjee, S., Ebert, B.L., Gillette, M.A., Paulovich, A., Pomeroy, S.L., Golub, T.R., Lander, E.S., and Mesirov, J.P. (2005). Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc. Natl. Acad. Sci. USA 102, 15545–15550. Sun, K., Kusminski, C.M., and Scherer, P.E. (2011). Adipose tissue remodeling and obesity. J. Clin. Invest. 121, 2094–2101. Suzuki, H., Forrest, A.R., van Nimwegen, E., Daub, C.O., Balwierz, P.J., Irvine, K.M., Lassmann, T., Ravasi, T., Hasegawa, Y., de Hoon, M.J., et al.; FANTOM Consortium; Riken Omics Science Center (2009). The transcriptional network that controls growth arrest and differentiation in a human myeloid leukemia cell line. Nat. Genet. 41, 553–562. Tan, C.Y., and Vidal-Puig, A. (2008). Adipose tissue expandability: the metabolic problems of obesity may arise from the inability to become more obese. Biochem. Soc. Trans. 36, 935–940.

Cell Metabolism 19, 981–992, June 3, 2014 ª2014 Elsevier Inc. 991

Cell Metabolism EBF1 Regulates Adipocyte Morphology and Lipolysis

Treiber, T., Mandel, E.M., Pott, S., Gyo¨ry, I., Firner, S., Liu, E.T., and Grosschedl, R. (2010). Early B cell factor 1 regulates B cell gene networks by activation, repression, and transcription- independent poising of chromatin. Immunity 32, 714–725.

Wang, Q.A., Tao, C., Gupta, R.K., and Scherer, P.E. (2013). Tracking adipogenesis during white adipose tissue development, expansion and regeneration. Nat. Med. 19, 1338–1344.

van Harmelen, V., Skurk, T., and Hauner, H. (2005). Primary culture and differentiation of human adipocyte precursor cells. Methods Mol. Med. 107, 125–135.

Weyer, C., Foley, J.E., Bogardus, C., Tataranni, P.A., and Pratley, R.E. (2000). Enlarged subcutaneous abdominal adipocyte size, but not obesity itself, predicts type II diabetes independent of insulin resistance. Diabetologia 43, 1498– 1506.

Virtue, S., and Vidal-Puig, A. (2010). Adipose tissue expandability, lipotoxicity and the Metabolic Syndrome—an allostatic perspective. Biochim. Biophys. Acta 1801, 338–349.

Zhang, Y., Liu, T., Meyer, C.A., Eeckhoute, J., Johnson, D.S., Bernstein, B.E., Nusbaum, C., Myers, R.M., Brown, M., Li, W., and Liu, X.S. (2008). Modelbased analysis of ChIP-Seq (MACS). Genome Biol. 9, R137.

992 Cell Metabolism 19, 981–992, June 3, 2014 ª2014 Elsevier Inc.