Alteration of NCoR Corepressor Splicing in Mice Causes Increased Body Weight and Hepatosteatosis without Glucose Intolerance

Abstract

Alternative mRNA splicing is an important means of diversifying function in higher eukaryotes. Notably, both NCoR and SMRT corepressors are subject to alternative mRNA splicing, yielding a series of distinct corepressor variants with highly divergent functions. Normal adipogenesis is associated with a switch in corepressor splicing from NCoRω to NCoRδ, which appears to help regulate this differentiation process. We report here that mimicking this development switch in mice by a splice-specific whole-animal ablation of NCoRω is very different from a whole-animal or tissue-specific total NCoR knockout and produces significantly enhanced weight gain on a high-fat diet. Surprisingly, NCoRω^−/− mice are protected against diet-induced glucose intolerance despite enhanced adiposity and the presence of multiple additional, prodiabetic phenotypic changes. Our results indicate that the change in NCoR splicing during normal development both helps drive normal adipocyte differentiation and plays a key role in determining a metabolically appropriate storage of excess calories. We also conclude that whole-gene “knockouts” fail to reveal how important gene products are customized, tailored, and adapted through alternative mRNA splicing and thus do not reveal all the functions of the protein products of that gene.

INTRODUCTION

A given transcription factor, by physically exchanging corepressor complexes for coactivator complexes, can convert from a repressor to an activator of gene expression (1). Coactivators include a wide assortment of acetyltransferases, methyltransferases, ATP-dependent chromatin remodeling complexes, and Mediator subunits (1). Coactivators enhance transcription by covalently modifying chromatin (i.e., changing the “histone code”) or by recruiting components of the general transcriptional machinery to a target promoter. Corepressor complexes, conversely, can strip the activation marks from chromatin or can interfere with recruitment of the preinitiation complex (1). The NCoR and SMRT families of corepressors derive from an ancestral gene duplication and divergence event (2, 3). NCoR and SMRT mediate distinct but overlapping functions and physically interact with, and help mediate repression by, a broad variety of vertebrate transcription factors (2–6). The nuclear receptor family of hormone-regulated transcription factors, including thyroid hormone receptors (TRs), retinoic acid receptors, peroxisome proliferator-activated receptors (PPARs), liver X receptors (LXRs), estrogen receptors, and estrogen-related receptors, bind to their target genes and recruit the NCoR or SMRT corepressors through “receptor interaction domains” (RIDs) primarily clustered within the corepressor C-terminal regions; differences in amino acid sequence within and flanking the RID motifs help define differences in the abilities of different RIDs to interact with different nuclear receptors (2, 7). The N-terminal and central regions of SMRT and NCoR bind, in turn, additional proteins, such as histone deacetylase 3 (HDAC3), TBL/TBLR-1, and GPS2, that mediate the molecular events required for transcriptional repression (2, 3, 7). It is therefore the specificity of the SMRT and NCoR RIDs that determines whether the corepressor complex as a whole can interact or not with a given nuclear receptor partner (e.g., see reference 8). Naturally occurring nuclear receptor mutations that disrupt proper NCoR or SMRT recruitment or release cause metabolic, endocrine, and neoplastic diseases (2, 7, 9). Changes in expression, modification, or polymorphisms in the corepressors themselves have been linked to cancer and to type 2 diabetes (e.g., see reference 10).

Interestingly, genome-level sequence analysis has led to a surprising realization: relatively few genes are necessary to produce multicellular organisms, and the difference in gene number between nematodes and humans is quite modest (11). One explanation for how <25,000 genes can construct entities as complex as human beings originates in a previous, equally unexpected discovery: multiple distinct mRNAs can be generated from a single genetic locus through alternative mRNA splicing (12). Therefore, alternative mRNA splicing diversifies the functions of a single gene by creating multiple protein variants, each possessing distinct domains and distinct properties. In this manner, a limited number of genetic loci has been elaborated by evolution to perform a multitude of roles. It recently became recognized that alternative mRNA splicing of both SMRT and NCoR generates an extensive series of divergent corepressor splice variants that differ in their RIDs, in their affinity for different nuclear receptors, and in their tissue-specific expression patterns (reviewed in references 8 and 13).

Lipid and glucose metabolism is known to involve the coordinated actions of multiple transcription factors and their coregulators, including PPARs, TRs, and LXRs (14, 15). We reported previously that there is a change in NCoR splicing from an ω splice variant that possesses three RIDs, which predominates in the preadipocyte and inhibits adipose differentiation in culture when overexpressed, to a δ splice form that possesses two RIDs, predominates in the mature adipocyte, and enhances adipose differentiation when overexpressed (16). We proposed that this change in NCoR splicing serves as a key natural switch in differentiation and helps to establish the correct transcriptional program in the mature cell.

To examine the role of this NCoRω-to-NCoRδ switch in the whole organism, we created a splice-specific knockout of NCoRω in mice, resulting in constitutive and exclusive expression of the NCoRδ variant. We report here that these mice exhibit an enhanced adiposity phenotype yet are completely protected against diet-induced glucose intolerance despite a number of otherwise prodiabetic phenotypic changes; this suggests that NCoR mRNA splicing is hierarchical and dominant over many other events that result in diabetes. Notably, our splice-specific knockout of NCoRω yields a very different phenotype than a whole-gene NCoR or an NCoRδ splice-form-specific knockout (17–21; M. L. Goodson and M. L. Privalsky, unpublished results). We also note that the phenotypic changes in our NCoRω^−/− mice occur despite normal SMRT expression and splicing in these animals. We propose that alternative splicing of NCoR during normal organismal development permits appropriate adipocyte formation and glucose homeostasis and is distinct from the functions of SMRT and its splicing in this differentiation process. Our studies also demonstrate that whole-gene knockouts, currently very commonly used to determine function in biology (22), fail to reveal all the actions of the alternatively spliced versions of these gene products.

MATERIALS AND METHODS

Plasmid cloning and protein expression.

The coding sequences for amino acids 1395 to 2453 of murine NCoRω and amino acids 1395 to 2334 of murine NCoRδ, which contain the C-terminal HDAC3 interaction domain and the entire receptor interaction domains, were cloned into the EcoRI and XhoI restriction sites of a modified pSG5 expression vector (Stratagene, La Jolla, CA), introducing appropriate restriction endonuclease sites and a synthetic start codon by PCR. The complete coding sequence of human HDAC3 was cloned into the BamHI and XhoI restriction sites of a modified pGEX-KG vector (23), introducing appropriate restriction endonuclease sites by PCR. The coding sequences for amino acids 1817 to 2453 of NCoRω and amino acids 1817 to 2334 of NCoRδ were cloned into the SfgI and PmeI restriction sites of the Halo tag pFN18A vector (Promega, Madison, WI) by using PCR to introduce appropriate restriction endonuclease sites. Radiolabeled NCoRω and NCoRδ proteins were produced by in vitro transcription and translation using the TNT T7 quick coupled transcription/translation system (Promega, Madison, WI) according to the manufacturer's protocols. Halo tag-purified NCoRω and NCoRδ receptor interaction domain proteins were expressed in KRX cells and purified according to the manufacturer's protocols.

Generation of NCoRω splice-specific knockout mice.

Splice-specific NCoR knockout mice were generated by the Mouse Biology Program at the University of California at Davis (UC Davis). Briefly, we used a targeting vector containing a 16.7-kb fragment of the murine NCoR gene (exons 34 to 41 and the intervening introns) harboring a GT-to-CC mutation that ablates the 5′ splice donor site required for NCoRω splicing. This construct was electroporated into C57BL/6 embryonic stem cells (JM8.F6). Following genotyping of the stem cells to confirm a single homologous integration of the targeted mutation, cells were electroporated with a vector expressing FLP flippase to remove the FLP recombination target (FRT)-flanked neomycin resistance (neoR) cassette that was cloned within the intron downstream of exon 37, thereby resulting in the removal of approximately 2.1 kb of noncoding intronic sequence between exons 37 and 38. Clones were genotyped to verify the removal of the neoR cassette and to confirm that there were no undesired integration or rearrangement events. Blastocysts (BALBc) were injected with the manipulated embryonic stem (ES) cells and were implanted in pseudopregnant female mice. Chimeric progeny were then crossed with wild-type (WT) C57BL/6N mice (Taconic Farms). Germ line transmission was preliminarily assessed by screening for the ES cell-derived black coat color. Mice were fully genotyped by PCR using genomic tail DNA and using primers that specifically recognize the targeted splicing mutation (see Table S1 in the supplemental material). Animals were housed at up to four mice per cage in a ventilated isolator cage system with a 12-h light/12-h dark cycle at approximately 22°C and with ad libitum access to water and regular mouse chow (Purina 5058 PicoLab MouseDiet 20). All studies were approved by the University of California, Davis, Institutional Animal Care and Use Committee.

Immunoblot analysis of NCoR protein expression.

Rabbit polyclonal antisera directed against peptides encoding amino acids 2321 to 2334 (pan-NCoR) or amino acids 1953 to 1966 (NCoRω-specific) in murine NCoR were produced by Sigma Genosys (Woodlands, TX). Mouse liver extracts were produced by homogenizing 10 mg of tissue in 2× Laemmli SDS-PAGE sample buffer (62 mM Tris-Cl [pH 6.8], 10% glycerol, 2% sodium dodecyl sulfate, 5% β-mercaptoethanol, 0.25 mg/ml bromophenol blue). Immunoblotting was performed as previously described (16).

High/low-fat feeding studies.

Age-matched male mice were used for all experiments. Mice were maintained on either a high-fat diet (HFD) (60% of calories from fat; catalog no. D12492; Research Diets) or a low-fat diet (LFD) (10% of calories from fat, but total calories otherwise the same as for the HFD; catalog no. D12450B; Research Diets) for 8 to 26 weeks, with food and water ad libitum. Food and water were changed and mice were weighed weekly.

Glucose and insulin tolerance.

Mice from HFD and LFD feeding studies were fasted for 6 h and weighed prior to intraperitoneal administration of dextrose (2 g/kg of body weight) or Novolin-R insulin (2 U/kg) (Novo-Nordisk) in sterile normal saline. Glucose was measured in blood from the tail vein prior to glucose or insulin administration (time zero) and at intervals thereafter by using a TRUEresult glucometer (Nipro Diagnostics, Fort Lauderdale, FL).

Insulin signaling pathway analysis.

Akt and IRS1 phosphorylation analyses were performed by the Mouse Metabolic Phenotyping Center at UC Davis (catalog no. D3491). Briefly, mice from HFD and LFD feeding studies were fasted 6 h prior to intraperitoneal injection of saline or 10 U/kg insulin. Mice were sacrificed, and livers were harvested at 10 min postinjection, immediately frozen in liquid nitrogen, and stored at −80°C. Tissues were ground in the presence of liquid nitrogen and lysed by using radioimmunoprecipitation assay (RIPA) buffer (10 mM Tris-HCl [pH 7.4], 150 mM NaCl, 0.1% sodium dodecyl sulfate, 1% Triton X-100, 1% sodium deoxycholate, 5 mM EDTA, 1 mM NaF, 1 mM sodium orthovanadate, and protease inhibitors). Lysates were clarified by centrifugation at 13,000 rpm for 10 min, and protein concentrations were determined by using a bicinchoninic acid protein assay kit (Pierce Chemical, IL). Proteins (1,000 μg) were immunoprecipitated with anti-IRS1 antibodies (Millipore, MA). Immune complexes were collected on protein G-Sepharose beads (GE Healthcare) and washed three times with lysis buffer. Proteins were resolved by SDS-polyacrylamide gel electrophoresis and transferred onto nitrocellulose membranes. Immunoblots were performed with anti-phospho-IRS1 (Y608) (Millipore, MA) or antibodies against pAkt, Akt (Cell Signaling), and tubulin (Santa Cruz). Proteins were visualized by using enhanced chemiluminescence (ECL; Amersham Biosciences), and pixel intensities of immunoreactive bands were quantified by using the FluorChem 9900 system (Alpha Innotech, CA).

Liver triglyceride analysis.

Liver triglyceride (TG) analysis was performed by the Mouse Metabolic Phenotyping Center at UC Davis (catalog no. D3302). Briefly, mice from HFD and LFD feeding studies were fasted 6 h prior to termination. Livers were harvested, immediately frozen in liquid nitrogen, and stored at −80°C. Weighed liver samples were homogenized in methanol-chloroform. After extraction overnight, 0.7% sodium chloride was added. The aqueous layer was aspirated, and duplicate aliquots of the chloroform-lipid layer were dried under nitrogen gas. The lipid was reconstituted in isopropyl alcohol and assayed for TG spectrophotometrically by using enzymatic reagents from Fisher Diagnostics (Middletown, VA).

Serological analyses.

Mice from HFD and LFD feeding studies were fasted for 6 h. Sera for insulin, free fatty acid, triglyceride, and adiponectin analyses were collected by retro-orbital bleeding prior to exsanguination by cardiocentesis under isoflurane anesthesia. Insulin levels were measured by using an electrochemiluminescence assay (Meso Scale Discovery, Rockville, MD). Triglyceride; total, high-density lipoprotein (HDL), and low-density lipoprotein (LDL) cholesterol; and free fatty acid levels were measured by enzymatic assays (Fisher Diagnostics, Middletown, VA, and Wako Diagnostics, Richmond, VA). C-reactive peptide levels were measured by an enzyme-linked immunosorbent assay (ELISA) (AbCam, Cambridge, MA). Adiponectin levels were measured by a radioimmunoassay (RIA) (EMD Millipore, Billerica, MA). All serological assays were performed by the Mouse Metabolic Phenotyping Center at UC Davis.

Energy expenditure analysis.

Respiration, food and water intake, and movement of HFD- and LFD-fed mice were measured by using an Oxymax comprehensive laboratory animal monitoring system (CLAMS; Columbus Instruments). Mice were adapted to monitoring cages for 24 h and then analyzed for 48 h. For the refeeding study, mice were fasted for 16 h and subsequently refed for 6 h. All CLAMS analyses were performed by the Mouse Metabolic Phenotyping Center at UC Davis.

Magnetic resonance imaging.

Mice were fed a HFD for 14 weeks prior to imaging on a Bruker Biospec 70/30USR instrument (Bruker Biospin, Woodlands, TX). Mice were imaged by using a respiratory gated multislice multiecho (MSME) spin echo (single-echo) sequence with an echo time (TE) of 15 ms and a repetition time (TR) of 2,100 ms to collect 70 35-μm coronal slices. All magnetic resonance (MR) imaging was performed by the Center for Molecular and Genomic Imaging at UC Davis.

Gene expression analysis by quantitative reverse transcription-PCR (qRT-PCR).

Mouse embryonic fibroblasts (MEFs) were isolated from postcoital day 12 embryos (24). RNA was isolated from 3 WT and 3 NCoRω^−/− MEF cell lines by using Qiazol reagent and was reverse transcribed by using the QuantitectRT kit according to the manufacturer's protocol (Qiagen, Hilden, Germany). Expression was quantified by using the primers listed in Table S1 in the supplemental material and SsoFast EvaGreen Supermix in an MJ Research Opticon2 real-time thermal cycler (Bio-Rad, Hercules, CA). Expression analysis was performed by using Opticon Monitor software (version 3.1) and Prism (version 5.0c; Graphpad Software, La Jolla, CA).

RNA-seq (transcriptome sequencing) analysis of gene expression.

Total RNA was isolated from liver and epididymal white adipose tissue from HFD- and LFD-fed mice. cDNA libraries were prepared from 75 to 150 ng of RNA by using the Encore Complete RNA-seq DR multiplex system (Nugen, San Carlos, CA). Libraries were prepared according to the manufacturer's protocol, with the modification that the double-stranded cDNA was fragmented by using a Branson 250 sonifier with a 1/8-in. double-step microtip (eight 30-s cycles at power setting 3 with a 50% duty cycle). Twelve libraries per lane were sequenced by using an Illumina HiSeq2000 system, generating 1.1 × 10⁷ to 2.3 × 10⁷ 100-base sequence reads per library. Preliminary quality metrics were assessed by using FastQC (version 0.10.0; Babraham Bioinformatics, Cambridge, United Kingdom). Sequences were aligned to Genome Reference Consortium Mouse Build 38 (mm10) by using Tophat (version 2.0.9) (25). Transcripts were assembled from the resulting alignments, and differential gene expression was computed by using Cufflinks (version 2.1.1) (26). Data analysis was performed by using the Bioconductor package CummeRbund (version 2.4.0) (26). Gene expression graphs were prepared by using Prism.

Protein-protein interaction assay.

Radiolabeled NCoRω and NCoRδ proteins containing the C-terminal HDAC3 interaction domain were assayed for their ability to interact with a glutathione S-transferase (GST)–HDAC3 fusion protein, as previously described (27).

Rev-Erbα supershift assay.

Halo tag-purified NCoRω and NCoRδ receptor interaction domain proteins were used in a Rev-Erbα electrophoretic mobility supershift assay as previously described (28).

Data analysis.

All data analysis was performed by using R (version 3.0.2; R Foundation for Statistical Computing) and RStudio (version 0.98; Rstudio, Boston, MA). Graphs (excluding gene expression graphs) were prepared by using the R package ggplot2 (version 0.9.3.1) (29). Statistical significance was computed by using either a Welch's two-sided t test or Cuffdiff for the analysis of differential gene expression by RNA-seq analysis.

General assurances.

All creation and study of animals were done with approval of the University of California at Davis Animal Welfare and Environmental Health and Safety committees and met all applicable AAALAC guidelines. All DNA, cell lines, and animals generated as a result of this publication will be fully available upon request. Any additional cost of breeding and shipping reflecting a request may be charged per se, especially for genetically engineered mice. All protocols for generation of the data in this publication are available upon request. All data reported here, or relevant to the conclusions of this publication, are available at GEO.

RESULTS

NCoRω and NCoRδ share a splice acceptor but utilize two distinct splice donors (Fig. 1A) (13). We inactivated the NCoRω splice donor by site-directed mutagenesis in mice without disrupting the NCoRδ splice donor or the NCoRδ protein sequence (Fig. 1A). Mice homozygous for the NCoRω knockout (NCoRω^−/−) were recovered at the expected Mendelian ratios and were fully viable, in contrast to the early-embryonic-lethal phenotype reported for a pan-splice (total-embryo) NCoR knockout (17, 18). Analysis confirmed that NCoRω mRNA was ablated in all tissues examined (Fig. 1B and data not shown). NCoRω protein expression was also eliminated, no aberrant splice forms were detected, and total levels of NCoR expression were not altered, but rather, expression was diverted, as expected, from the mix of NCoRω and NCoRδ seen in the wild type to the exclusive expression of NCoRδ. SMRT expression and splicing were not affected (Fig. 1B and C).

FIG 1.

Generation of NCoR splice-specific knockout mice. (A) Schematic representation of the NCoR exon 37 region alternatively spliced and joined onto exon 38 to generate the NCoRω and NCoRδ splice variants. Two different 5′ splice donor (sd) sites and one splice acceptor (sa) are used, as indicated. Assembly of the final NCoRω and -δ mRNAs (below) from the DNA (above) is indicated. The location of the N3 and N2 RIDs, with N3 specific for NCoRω, is also indicated (above). (B) Expression of NCoRω and NCoRδ (top) and SMRTα and SMRTγ (bottom) RNAs in wild-type (WT) and NCoR (ω^−/−) mice. Analysis was performed by RT-PCR and gel electrophoresis using primers that flanked the alternatively spliced region, as previously described (16). The background band indicated with an asterisk has been sequenced and represents NCoRω by all criteria employed (and is likely a single-strand PCR artifact). WAT, white adipose tissue. (C) Expression of NCoR proteins in WT and NCoRω^−/− mouse livers determined by immunoblotting with anti-pan-NCoR antibodies or anti-NCoRω-specific antibodies, as noted. These antibodies react with the corresponding antigens expressed as proteins synthesized from appropriate recombinant baculovirus/Sf9 cells but not with recombinant baculoviruses expressing control proteins; preimmune sera from the same rabbits do not exhibit any NCoR-derived reactivity (data not shown).

The NCoRω^−/− animals were anatomically and developmentally normal by gross inspection but gained weight more quickly than did wild-type animals when maintained on a standard chow diet postweaning (Fig. 2A). The weight gain and enlarged-liver phenotype were greatly enhanced on a high-fat diet (Fig. 2B and C). Both visceral and subcutaneous fat deposits appeared to be increased in NCoRω^−/− males by magnetic resonance spectroscopy and by dissection compared to the wild-type controls and were very distinct from the lean phenotype seen with NCoRδ^−/− mice (Fig. 2D and E). These results in vivo are apparently reciprocal to the inhibition of adipocyte differentiation observed when NCoRω is overexpressed in 3T3-L1 cells in culture (16) and demonstrate that the NCoRω splice variant plays a unique role in regulation of adipogenesis and adipocyte function that is distinct from that of the NCoRδ and SMRT splice variants, all of which are expressed in these animals.

FIG 2.

Phenotypic analysis of NCoRω^−/− mice. (A) Average weights of 6- to 8-week-old wild-type (WT) and NCoRω knockout (ω^−/−) male (M) and female (M) mice maintained on standard chow. Means ± standard errors of the means are shown (n = 17 to 51) (***, P < 0.001). (B) Weight gain on a high-fat diet (HFD) and on a low-fat diet (LFD). WT or NCoRω^−/− male mice were fed either a HFD or a LFD and weighed weekly. The means ± standard errors of the means are shown (n = 15 to 24); error bars are always presented but may be smaller than the symbols. (C) Liver wet weights. WT or NCoRω^−/− mice were fed either a HFD or a LFD for 25 weeks. Livers were dissected and weighed. Means ± standard errors of the means are shown (n = 3 to 6) (**, P < 0.01). (D) Adipose tissue wet weights. WT or NCoRω^−/− mice were fed a HFD for 25 weeks. Visceral (epididymal) white adipose tissue deposits were dissected and weighed. Means ± standard errors of the means are shown (n = 3 to 6). (E) Magnetic resonance image of coronal sections of a WT mouse and an NCoRω^−/− mouse fed a HFD for 13 to 15 weeks. Sections were 3.5 mm from the ventral end of each mouse. Adipose tissue appears white.

At least part of the enhanced adiposity observed in the NCoRω^−/− animals may be cell autonomous and is paralleled by an increase in adipogenic differentiation by cultured NCoRω^−/− mouse embryonic fibroblasts (MEFs) compared to MEFs derived from wild-type animals (16), although in theory, the two phenomena may, of course, be independent. A subset of, but not all, nuclear receptor target genes were derepressed in NCoRω^−/− MEFs compared to wild-type MEFs and were seen in the absence of overt adipocyte differentiation (Fig. 3 and data not shown). This finding indicates that the NCoRω splice variant plays a selective role in the repression of these target genes that differs from pan-NCoR taken as a whole (30, 31). Interestingly, the levels of expression of the Tnf, Ifng, and Nr1h3 genes, which are involved in macrophage-mediated inflammation, were reduced (Fig. 3), although we have not determined if this alteration is relevant to a biological phenotype.

FIG 3.

Gene expression in mouse embryonic fibroblasts (MEFs). Gene expression in MEFs isolated from WT and NCoRω^−/− mice was analyzed by using quantitative RT-PCR. The x axis indicates the Mouse Genome Informatics (MGI) gene symbol for each gene analyzed. Most of the genes represented are known targets of PPARγ. Shown are the log₂ fold changes in expression in NCoRω^−/− versus WT MEFs ± standard errors of the means (n = 3 to 15) (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

We utilized a comprehensive animal activity monitoring system to compare mice bearing the two genotypes. As expected, the respiratory exchange rate was reduced on the high-fat diet versus the low-fat diet for both NCoRω^−/− and wild-type mice (32) (Fig. 4A). However, this effect was more pronounced for the NCoRω^−/− animals, especially during the light cycle, suggesting a greater shift away from a carbohydrate-based metabolism to a fat-based one. There was a reduction in total energy expenditure for both genotypes on both the high-fat and low-fat diets, but this effect disappeared if calculated on a lean-mass basis, indicating that it may be driven by the increased adiposity of the NCoRω^−/− mice rather than a reduction in tissue-specific thermogenesis per se (Fig. 4B). The NCoRω^−/− animals exhibited a notable reduction in the size of the shift to a higher respiratory exchange rate when fasted and refed compared to wild-type animals (Fig. 4C). The increased adiposity of NCoRω^−/− mice compared to wild-type mice was more evident when the animals were kept at 22.2°C versus 21.4°C (Fig. 4D), suggesting an additional possible sympathetic central contribution to this phenotype. Food consumption of the NCoRω^−/− animals was not significantly different from that of the wild-type animals (Fig. 4E). We conclude that the ablation of the NCoRω splice variant led to alterations in multiple metabolic parameters, both adipocyte based and central, that are likely to contribute to the enhanced adiposity of these animals but that no single parameter was individually responsible.

FIG 4.

Analysis of energy utilization. (A) Respiratory exchange ratio (RER) in the dark or light period for WT or NCoRω^−/− mice fed either a HFD or a LFD. The means ± standard errors of the means are shown (n = 7 to 9 mice) (†, P ≤ 0.1; *, P < 0.05). (B) Energy expenditure (EE) normalized to lean body mass for WT or NCoRω^−/− mice on a HFD or a LFD, during the light or dark cycle. Means ± standard errors of the means are shown (n = 7 to 9 mice). LM, lean body mass. (C) Respiratory exchange ratios of WT and NCoRω^−/− mice fasted and then refed either a HFD or a LFD. Means ± standard errors of the means are shown (n = 3 to 5 mice). The dark period was from 1800 to 0600 h. (D) WT or NCoRω^−/− mice were fed either a HFD or a LFD for 14 weeks and weighed weekly. At 9 weeks, mice were moved from a 21.4°C room to a 22.2°C ± 0.26°C room. Means ± standard errors of the means are shown (n = 3 to 6 mice); error bars are always presented but may be smaller than the symbols. (E) Food consumption of mice in panels A and D. Means ± standard errors of the means are shown (n = 7 to 9 mice).

Wild-type C57BL/6N mice develop severe glucose intolerance when maintained on a high-fat diet (33) (Fig. 5A). Remarkably, the NCoRω^−/− mice were extremely glucose tolerant on a low-fat diet and maintained this strong glucose tolerance on the high-fat diet despite their greater adiposity (Fig. 5A). In fact, the NCoRω^−/− mice were more glucose tolerant on the high-fat diet than were the wild-type controls on a low-fat diet (Fig. 5A). Interestingly, the NCoRω^−/− mice exhibited more insulin resistance on the HFD than on the LFD, as was also seen with the wild-type controls (Fig. 5B), although this enhanced insulin resistance was nonetheless largely independent of genotype; this may also explain the slightly lower glucose tolerance that we observed for the NCoRω^−/− mice on the HFD than for those on the LFD. We conclude that the loss of the NCoRω splice variant prevents the glucose intolerance, but not the increased insulin resistance, associated with a HFD.

FIG 5.

Glucose and insulin sensitivity, adipocyte sizes, and liver histology. (A) Intraperitoneal glucose tolerance test for WT and NCoRω^−/− mice previously maintained on either a HFD or a LFD and fasted as described in Materials and Methods. Blood glucose was measured after an intraperitoneal injection of 2 g/kg dextrose at the indicated time intervals. Means ± standard errors of the means are shown (n = 12 to 21 mice). (B) Intraperitoneal insulin tolerance test for WT and NCoRω^−/− mice previously maintained on either a HFD or a LFD and fasted as described in Materials and Methods. The blood glucose (BG) level was measured at the intervals noted after injection with 2 U/kg of insulin and was normalized to untreated fasting blood glucose levels for each mouse. Means ± standard errors of the means are shown (n = 6 to 15 mice). (C) Visceral (epididymal) adipocyte size distribution for WT and NCoRω^−/− mice fed a HFD. Formalin-fixed, paraffin-embedded sections of visceral white adipose tissues were stained with hematoxylin and eosin. A cross-sectional area of adipocytes was measured from digital images by using Image J (version 1.47; National Institutes of Health); the histogram represents the distribution of adipocyte sizes for each genotype (n = 4 to 7 mice). (D) Liver histology. Formalin-fixed, paraffin-embedded sections of livers isolated from WT and NCoRω^−/− mice maintained on either a HFD or a LFD for 25 weeks were stained with hematoxylin and eosin. Representative micrographs are shown. (E) IRS1 and Akt phosphorylation in response to insulin. WT or NCoRω^−/− mice (2 each) were injected intraperitoneally with 10 U/kg insulin (+) or saline (−). Immunoblots of phosphorylated (pIRS1^Y608) and total IRS1, phosphorylated (pAkt^S473) and total Akt, as well as tubulin (loading control) are shown. (F) Liver triglyceride levels in WT and NCoRω^−/− mice.

There is a wide panel of additional markers that are considered to be glucose intolerance/diabetes risk factors in the causation of metabolic disease, including increased adipose hypertrophy in visceral fat and enhanced liver steatosis. Levels of these markers were elevated in the wild-type mice on the HFD compared to those on the LFD, as expected (34–38), and, remarkably, were still more elevated in the NCoRω^−/− mice on the HFD despite the glucose tolerance of the latter (Fig. 5C, D, and F). Notably, the visceral adipose hypertrophy seen with the HFD was less evident in either genotype for subcutaneous adipocytes or in either adipose deposit on the LFD (data not shown). Much of the in vivo increase in adipose tissue mass was observed as an increase in cell size rather than cell number. The response to insulin in the NCoRω^−/− mice, as exhibited by IRS1 and Akt phosphorylation in liver, was comparable to or slightly higher than that in the WT mice (Fig. 5E) and, together with data from our WT-like insulin tolerance assays, appears unlikely to fully account for the strong glucose tolerance in the NCoRω^−/− animals. The total serum cholesterol level was elevated in NCoRω^−/− mice on either diet, and total serum LDL complexes were clearly increased in the NCoRω^−/− splice-specific knockout mice on the HFD (Fig. 6A to C). On the other hand, circulating adiponectin, C-reactive peptide, and insulin levels in NCoRω^−/− mice were not detectably changed on the HFD versus the LFD, despite the increased adiposity (Fig. 6F to H). Interestingly, there were reductions in the levels of circulating triglycerides and free fatty acids in the NCoRω^−/− mice compared to the wild-type mice on both the low-fat and high-fat diets (Fig. 6D and E). Our results indicate that the retention of the NCoRω splice variant therefore counteracts what is otherwise a generally prodiabetic physiology in the animals on the high-fat diet, consistent with the theory that this alternative corepressor splicing form operates epistatic to these signals and plays a dominant role in the establishment of normal glucose metabolism.

FIG 6.

Serological analysis of WT and NCoRω^−/− mice on HFD or LFD maintenance. (A) Total cholesterol (TC). (B) HDL cholesterol. (C) LDL cholesterol. (D) Triglycerides (TG). (E) Nonesterified free fatty acids (FFA). (F) C-reactive peptide (CRP). (G) Fasting insulin. (H) Total adiponectin (ADPN). The mean levels of each protein in the circulation ± standard errors of the means are shown (n = 3 to 5 mice for each genotype); statistical significance is indicated as described in the legends of Fig. 2 to 4.

We also analyzed the expressions of key metabolic target genes in the liver and adipose tissue by qRT-PCR or RNA-seq (Fig. 7A and data not shown). A subset, but not all, of the target genes for nuclear receptors were more strongly expressed (“derepressed”) in the NCoRω^−/− animals: these include some but not all known TR, PPARγ, and LXRα target genes. As a consequence, we suggest that NCoRω and -δ are differentially utilized by different nuclear receptors and at different target genes for the same nuclear receptor.

FIG 7.

Differential gene regulation between WT and NCoRω^−/− mice on HFD and LFD. (A) RNA-seq analysis of nuclear receptor target gene expression from liver or visceral white adipose tissue. Many of the genes represented are known targets for multiple nuclear receptors: Me1 is a TR and LXR target; Srebf1 is a PPAR, LXR, TR, and Rev-Erb target; and Scd1 is an LXR, TR, and Rev-Erb target. SPOT14 (Thrsp) is a well-known TR target but is also a target for LXR. The x axis indicates the MGI gene symbol. Shown is the fold change in expression between NCoRω^−/− (n = 3) and WT (n = 3) mice ± standard errors of the means. Asterisks indicate statistical significance based on Cuffdiff analysis of differential expression. (B) Mitochondrial DNA (mtDNA) from skeletal muscle of normal-chow-fed WT and NCoRω^−/− mice. Levels of mitochondrial DNAs encoding the 16S ribosomal DNA and NADH dehydrogenase 4 (mtND4) DNA were quantified by quantitative PCR and normalized to the amount of nuclear DNA at the Cebpb locus, as calculated by the same method. The means ± standard errors of the means are shown (n = 4 to 5 mice). (C) Comparison of gene regulation in NCoRω mice to that in previously reported adipocyte (AKO) and skeletal muscle (skMKO) tissue-specific pan-NCoR knockout mice. Relative gene expression was analyzed in either visceral white adipocyte tissue (“Visc. WAT”) or skeletal muscle (Skel. Muscle) by either RNA-seq or qRT-PCR, respectively, for mice on the HFD. Relative gene expression data for the muscle and the adipocyte pan-NCoR knockouts were obtained from data reported previously (12, 14); mRNA levels were quantified, except for Retn and Adipoq, which represent protein expression levels, by using GraphClick (Arizona Software, Neuchatel, Switzerland). Relative expression levels are shown as heat maps, with colors representing expression levels relative to those of the WT. (D) Expression analysis of genes known to be associated with liver steatosis. (E) Expression analysis of genes known to be associated with cholesterol metabolism and transport. (F) Expression analysis of genes known to be involved in regulation of glucose sensitivity as well as gluconeogenesis. (G) Expression analysis of genes regulated by PPARγ in response to Cdk5 phosphorylation. The fold changes in expression levels (“Rel. Expr.”) between NCoRω^−/− (n = 3) and WT (n = 3) mice ± standard errors of the means determined by using RNA-seq are shown for liver and visceral fat as described above for panel A.

Notably, the gene expression in white adipose tissue and muscle in the NCoRω^−/− mice differs from the panels that are derepressed by the corresponding pan-splice tissue-specific NCoR knockouts that have been reported previously, and there was no sign of mitochondrial hyperplasia (Fig. 7B and C) (19, 21, 30, 31), indicating that the NCoRω splice variant plays a selective role in the repression of these genes that is very different from that played by the NCoRδ or by any of the SMRT splice variants (30, 31).

Consistent with the increased liver size and steatosis, we saw abnormal expression of liver steatotic genes (Fig. 7D) and of genes involved in determining circulating free fatty acid and serum cholesterol levels (Fig. 7E) in the livers of NCoRω^−/− mice on the HFD, as expected from the serology (see above). We also observed changes in the levels of expression of genes involved in improved glucose sensitivity and gluconeogenesis in the NCoRω^−/− mice on the HFD (Fig. 7F) that help to explain the overall strong glucose tolerance and relatively low levels of circulating glucose in these animals. Notably, the ability of PPARγ to alleviate glucose intolerance in response to certain ligands is associated with the activation of a specific subset of PPARγ target genes; interestingly, certain, but not all, antidiabetic PPARγ target genes were also derepressed in our NCoRω knockout (Fig. 7G). In contrast to the findings described above, we saw a highly mixed pattern of proinflammatory gene derepression, or not, in MEFs, livers, and adipose tissues, indicating that no simple link of corepressor splicing to inflammation explains our glucose sensitivity phenotype (Fig. 3 and 7 and data not shown).

In many ways, the phenotype of our NCoRω^−/− mice resembles the phenotype of liver-specific ablation of Rev-Erb, NCoR, or histone deacetylase 3 (HDAC3) in mice (39–41). Notably, this role of HDAC3 in liver may not require HDAC3 enzymatic activity per se but instead reflects the ability of HDAC3 to be physically tethered to chromatin by a second, nonactivating interaction domain in the C-terminal region of NCoR (42). Importantly, the alternative splicing event that distinguishes NCoRω from NCoRδ maps close to this second HDAC3 interaction domain, and it is theoretically possible that our NCoRω^−/− phenotype was due to a loss of the NCoR/HDAC3 interaction in the NCoRδ splice form. However, we found that both corresponding NCoRδ and NCoRω domains bound HDAC3 in an in vitro GST pulldown assay although to nonidentical extents (Fig. 8A); these experiments rule out the possibility that our NCoRω knockout simply eliminates the second HDAC3 recruitment site but of course does not rule out that there is some unknown aspect of the NCoRδ/HDAC3 interaction that somehow differs functionally from that of the NCoRω/HDAC3 interaction. The orphan nuclear receptor Rev-Erb is known to regulate liver lipogenesis and is thought to be an important target of NCoR and a mediator of circadian regulation in the liver. Notably, Rev-Erb bound to NCoRδ at least as well as NCoRω (Fig. 8B) (43). It is therefore similarly unlikely that a failure of Rev-Erb to physically interact with NCoRδ explains our NCoRω^−/− phenotype.

FIG 8.

Interaction between NCoRω or NCoRδ, Rev-Erbα, and HDAC3. (A) “GST pulldown” protein-protein interaction assay of NCoRω or NCoRδ (amino acids 1395 to 2453 of NCoRω or equivalent) with GST only or GST-HDAC3. Proteins were resolved by SDS-PAGE and quantified by phosphorimager analysis. Mean levels ± standard errors of the means (n = 3) are shown. (B) Electrophoretic mobility supershift analysis of the interaction of increasing amounts of NCoRω or NCoRδ with Rev-Erbα bound to direct repeat with 2-base spacer (DR-2) DNA. Supershifted products were quantified by phosphorimager analysis, and the apparent binding affinity was calculated. Means ± standard errors of the means are shown (n = 3 for each splice variant; P = 0.0037).

It is intriguing, however, that our NCoRω^−/− mice not only have increased liver expression levels of fatty acid synthesis and sequestration genes, such as fatty acid synthetase (Fasn) and steroid response element binding protein (SREBP), helping to account for the steatosis phenotype (Fig. 7D), but also express elevated levels of perilipin 1 (Plin1) and Plin2 (Fig. 7D). We suggest that this elevation in perilipin levels allows sequestration of the resulting increased lipid levels into protective vesicles and thereby prevents the loss of glucose sensitivity that is otherwise associated with increased lipid accumulation. Interestingly, similar explanations have been proposed for certain other manipulations of corepressor function in liver that can result in retention of glucose sensitivity despite giving rise to hepatic steatosis (40). It is also notable that the loss of NCoRω in our experiments stimulated the expression of lipid synthesis genes but had little or no effect on gluconeogenesis genes (Fig. 7D and E). Thus, the diverting of glucose to lipid synthesis pathways may also help explain the observed phenotype of enhanced adiposity in the absence of elevated circulating glucose levels (40).

DISCUSSION

We observed what is generally considered to be a proadipogenic yet glucose-tolerant phenotype in mice that have been genetically manipulated to be unable to synthesize the ω splice form of the NCoR corepressor. Importantly, these mice make enough of the other major NCoR splice form, δ, to make the total amount of NCoR synthesized indistinguishable from the amount synthesized in wild-type mice as a sum of ω and δ. Thus, the phenotype that we saw is not due to reduced total levels of NCoR but rather reflects the specific loss of the NCoRω splice form and its replacement with equal or near-equal levels of the NCoRδ splice form. Of course, we cannot formally establish if the phenotype of our mice is due to the loss of NCoRω or the gain in NCoRδ. We favor the former interpretation, in that δ is already the major constituent of NCoR in terminally differentiated wild-type adipocytes; therefore, the loss of all remaining NCoRω from these cells in our NCoRω^−/− mice is more likely to result in the phenotypes that we observed than is the modest increase in NCoRδ levels seen for the NCoRω^−/− knockout mice. We emphasize that a change in mRNA splicing from the NCoRω to the NCoRδ form is closely tied to, and likely a driver of, normal adipocyte differentiation. Thus, our NCoRω^−/− mice query a biologically relevant process.

The proadipogenic, hepatosteatotic but glucose-tolerant phenotype of our NCoRω^−/− mice, which otherwise have normal development and normal or near-normal fecundity, is dramatically different from that seen with the comparable pan-splice, pan-tissue NCoR knockout, which is embryonic lethal due to a failure in early hematopoiesis, and the apparently normal adipogenesis reported for mice bearing artificial point mutations in NCoR RIDs 3 and 2 (17, 44; Goodson and Privalsky, unpublished). Although artificial deletion of the middle RID (RID2) of SMRT (a deletion that does not correspond to any known alternative splicing of SMRT or of NCoR) is associated with a proadipogenic phenotype (30), this nonhomologous SMRT RID2 deletion also causes enhanced insulin sensitivity and other phenotypes not observed in our own NCoRω knockout (i.e., a RID3 deletion). It is also interesting that we saw an effect of the loss of NCoRω despite normal expression and splicing of SMRT in the same animals. Notably, SMRT is also spliced to form a RID3 form much like NCoRω, yet normal synthesis of this RID3 SMRT clearly does not substitute for the RID3 NCoRω in our NCoRω^−/− phenotype (8, 13). We conclude that these RID3 variants of SMRT and NCoR must play divergent biological roles.

We have therefore shown that the different NCoR splice variants play surprisingly divergent roles in metabolism and that changes in NCoR splicing during adipogenesis (from the ω to the δ splice form) are likely to both help regulate and drive this differentiation process and are crucial for establishing the correct physiology of the mature adipocyte. Notably, other tissues involved in glucose utilization and energy homeostasis (e.g., liver, skeletal muscle, and brain) also display specific patterns of NCoR splicing. Therefore, alternative splicing of the NCoR corepressor customizes nuclear receptor/transcription factor function in response to different physiological states and in different tissues, and it also serves to program the overall metabolic response to excess caloric consumption and energy surplus at the organismal level. Future analysis of yet additional sites of action of the NCoRω splice variant, such as the brain and testis, where NCoRω is particularly strongly expressed over the δ splice form, will be particularly interesting (8).

The proadipogenic/glucose-tolerant phenotype observed in our splice-specific, pan-tissue NCoRω knockout resembles in certain aspects the proadipogenic glucose-tolerant phenotype reported for a pan-splice, adipocyte-specific NCoR knockout or the glucose-tolerant phenotype of a pan-splice, macrophage-specific NCoR knockout (19, 20). However, closer examination reveals substantial differences between these mouse models. Most notably, the glucose tolerance conferred by our splice-specific, pan-tissue NCoRω knockout is extremely profound and occurs despite other broad, prodiabetic changes in these mice. In contrast, the glucose tolerance conferred by the adipocyte-specific or macrophage-specific pan-splice NCoR knockout is more modest and occurs within the context of a wider, antidiabetic physiological context. We also know that the “adipocyte-specific” pan-splice knockout probably yields off-target effects due to tissue expression of the Cre recombinase in undesired cell types, whereas the NCoRω^−/− splice-specific pan-tissue knockout that we report here is quite targeted in its nature. Additional significant differences were observed when we compared the consequences of our pan-tissue, splice-specific NCoRω knockout with those reported for a muscle-specific pan-splice NCoR knockout (21); the latter generated a dramatic, strong, prooxidative gene expression signature that was quite different form our overall phenotype and from the effects on muscles in our NCoRω^−/− mice. Thus, we conclude that the NCoRω splice form plays a crucial role in adipogenesis development and glucose metabolism that is readily distinguishable from the role of the NCoRδ splice form. We also know that the specific loss of the NCoRω splice form produces changes in gene regulation in mouse muscles and adipocytes that differ from those previously reported for the corresponding pan-splice NCoR knockouts and that these changes contribute to the phenotypic differences that we reported here.

Our NCoRω^−/− mice also resemble in certain aspects the phenotype of liver-specific ablation of Rev-Erb, NCoR, or histone deacetylase 3 (HDAC3) in mice (39–41). The glucose sensitivity of HDAC3^−/− mice has been attributed to an increase in the conversion of glucose to lipid and a decrease in gluconeogenesis, which are also consistent with our own NCoRω^−/− studies. Interestingly, the role of HDAC3 in mouse liver may not require HDAC3 enzyme activity per se, and the associated requirement of NCoR for the HDAC3^−/− phenotype may reflect an interaction of the HDAC3 protein with a second HDAC3 interaction domain in the C-terminal region of NCoR (42). The alternative splicing event that distinguishes NCoRω from NCoRδ maps very close to this second HDAC3 interaction domain (8); as a consequence, it was possible that our NCoRω^−/− phenotype was due to an alteration of the NCoR/HDAC3 interaction in the NCoRδ splice form. We found that both the NCoRδ and NCoRω domains bound HDAC3 in an in vitro GST pulldown assay although at somewhat different efficiencies; these experiments appear to eliminate the possibility that our NCoRω knockout simply eliminates the second HDAC3 recruitment site but of course do not rule out that there is nonetheless some undetected, functional aspect of the NCoRδ/HDAC3 interaction that differs from that of the NCoRω/HDAC3 interaction. Alternatively, the phenotype of our NCoRω^−/− mice may simply reflect the absence of the third nuclear receptor interaction domain, N3, from the NCoRδ splice form that is present in the NCoRω splice form. Notably, NCoRδ binds Rev-Erb, a known regulator of hepatic lipogenesis and circadian rhythm, at least as well as NCoRω; this appears to therefore also eliminate the possibility that a complete loss of the NCoR/Rev-Erb interaction is a likely explanation of our observed NCoRω^−/− phenotype.

It is notable, nonetheless, that our NCoRω^−/− mice not only show increased liver expression levels of fatty acid synthesis and sequestration genes, helping account for the steatosis phenotype, but also simultaneously express elevated levels of perilipins 1 and 2, proteins associated with proper sequestration and formation of lipid droplets in cells. We propose that this elevation in perilipin levels allows protective sequestration of the increased amount of lipid, synthesized in parallel due to the loss of NCoRω, and thereby prevents the loss of glucose sensitivity that is otherwise associated with increased lipid accumulation. Interestingly, similar explanations have been proposed for why certain other manipulations of all NCoR or of all HDAC3 in liver give rise to retention of glucose sensitivity despite increased liver steatosis (40, 42); our results indicate that perturbation of NCoRω function in particular may be the basis for those previous observations.

In some but not all ways, our NCoRω^−/− mice phenocopy the effects of PPARγ agonists, such as the thiazolidinediones (TZDs) used to treat human type II diabetes (19). This is consistent with the higher affinity of PPARγ for binding NCoRω than NCoRδ observed in vitro, although only a subset of the PPARγ target genes was derepressed in the NCoRω^−/− mice, and notably, not all the effects of pan-splice NCoR and SMRT release from PPARγ (e.g., by TZDs) were seen with the NCoRω^−/− mice (19, 45–47; this study; Goodson and Privalsky, unpublished). We therefore believe that the ability to use NCoRω as a corepressor is determined by the identity of the target gene and not only by the identity of the nuclear receptor per se.

Whatever the specific mechanism, our studies convincingly indicate that NCoRω and NCoRδ regulate distinct target genes and thereby mediate different biological outcomes within a given cell type and within the organism as a whole; this suggests a potential for developing novel therapies that have the positive effects of thiazolidinediones without the negative effects. In fact, we found no indication of increased blood volume when we measured hemoglobin concentrations or hematocrit in our NCoRω^−/− mice on the LFD (preliminary results not shown). Increased blood volume is one of the chief mechanistic bases behind the increased cardiac risks of the TZD compounds, which has led to them being “black-boxed” for the treatment of human type 2 diabetes (48). Given this, pharmacologically phenocopying the specific effects of the NCoRω splice-specific ablation on glucose sensitization, potentially without the risks associated with the less specific TZDs, would have significant clinical potential. We need to compare our initial observations on blood volume more directly to the actions of TZDs and under different feeding regimens in the future.

Type 2 diabetes has been linked to increased Cdk5 phosphorylation of PPARγ, impairing its ability to activate specific glucose-sensitizing genes (49); NCoR may, in fact, even help physically target Cdk5 to PPARγ for this purpose (19). We know that the expression levels of at least several of these PPARγ glucose-sensitizing target genes are also elevated in our NCoRω knockouts. We are presently comparing the levels of PPARγ phosphorylation in wild-type mice to those in our NCoR splice mutant mice to explore if the different corepressor splice variants differ in their Cdk5 targeting capacity or, reciprocally, if PPARγ phosphorylation influences the binding or release of the specific corepressor splice forms from PPARγ itself.

Thus, we conclude that mRNA alternative splicing plays a key role in diversifying the actions of important regulatory molecules and can yield products of the same genetic locus that can possess distinct domains and distinct biochemical properties and have diametrically opposite functions in differentiation, development, and in biology. In this manner, a limited number of genetic loci has been elaborated by evolution to perform a multitude of roles. We believe that this is an important basis for generating the increased complexity that is characteristic of higher metazoans. Both NCoR and SMRT have their affinity for different transcription factors adjusted “on the fly” through mRNA splicing events that either add, remove, or adjust the interaction surfaces on the corepressor, thus joining a flourishing horde of vertebrate gene products whose functions are elaborated, adjusted, and custom fitted by alternative mRNA splicing. In looking backward, it is essential to reexamine much of the data already generated on corepressor function in light of this newfound diversity and to make note of which corepressor spliced variant was examined in each study so as to evaluate which variant yielded which result. Our results also indicate that ignoring the role of alternative mRNA splicing when ablating whole genes in vertebrate systems will overlook a crucial component of regulation that helps distinguish the higher from the lower vertebrates.