Genome-Wide Analysis of ChREBP Binding Sites on Male Mouse Liver and White Adipose Chromatin

Abstract

Glucose is an essential nutrient that directly regulates the expression of numerous genes in liver and adipose tissue. The carbohydrate response element–binding protein (ChREBP) links glucose as a signaling molecule to multiple glucose-dependent transcriptional regulatory pathways, particularly genes involved in glycolytic and lipogenic processes. In this study, we used chromatin immunoprecipitation followed by next-generation sequencing to identify specific ChREBP binding targets in liver and white adipose tissue. We found a large number of ChREBP binding sites, which are attributable to 5825 genes in the liver, 2418 genes in white adipose tissue, and 5919 genes in both tissues. The majority of these target genes were involved in known metabolic processes. Pathways in insulin signaling, the adherens junction, and cancers were among the top 5 pathways in both tissues. Motif analysis revealed a consensus sequence CAYGYGnnnnnCRCRTG that was commonly shared by ChREBP binding sites. Putative ChREBP binding sequences were enriched on promoters of genes involved in insulin signaling pathway, insulin resistance, and tumorigenesis.

Glucose is perhaps a most important and, in some tissues, the preferred fuel for cellular metabolism. After a meal rich in carbohydrates, blood glucose increases sharply, which induces pancreatic β-cells to release insulin. The insulin stimulates glucose uptake in the peripheral tissues such as muscle, liver, and adipose tissues via insulin receptor-mediated signaling pathways while it represses hepatic glucose production. Insulin also up-regulates sterol regulatory element binding protein-1c (SREBP-1c), the major lipogenic transcription factor in the liver that orchestrates the conversion of glucose into triglyceride for storage and subsequent export (1). When the blood glucose level drops during fasting, glucagon released from the pancreatic β-cells stimulates the liver (and kidney to a lesser extent) to produce glucose especially for the brain, which prefers to use glucose for metabolism, maintaining whole-body glucose homeostasis.

In addition to its effects on hormones such as insulin and glucagon, glucose can also directly modulate the expression of key enzymes in glycolytic and lipogenic pathways independent of the insulin-SREBP-1c axis (2). Studies of these glucose-regulated genes identified a sequence motif called carbohydrate response element (ChoRE) that is present in the promoter of many glycolytic and lipogenic genes (3, 4). Using synthetic oligonucleotides containing ChoRE sequence, Yamashita et al (5) isolated a new transcription factor, the carbohydrate response element–binding protein (ChREBP) from rat liver extracts (5). ChREBP, a 98-kDa protein, is a basic helix-loop-helix/leucine zipper transcription factor that belongs to the Mondo family.

The activity of ChREBP is determined by its phosphorylation state and subcellular localization (6). In addition to phosphorylation/dephosphorylation regulation, our structure-function analysis of ChREBP protein revealed an intramolecular inhibition in low glucose conditions through a glucose-sensing module that renders ChREBP responsive to glucose (7–9). The glucose-sensing module is conserved across all Mondo family genes and encompasses 2 subdomains: a low-glucose inhibitory domain (LID) and a glucose-response activation conserved element. At low glucose concentrations, the LID inhibits the transactivation activity of glucose-response activation conserved element, whereas in the presence of abundant glucose, the inhibition of LID-mediated inhibition is released, probably through the glucose metabolite, glucose-6-phosphate, and ChREBP is activated (8).

Recently, it has been proposed that in pancreatic β-cells, sorcin, a penta-EF-hand protein, binds to ChREBP and sequesters it in the cytosol under low glucose concentrations, but when stimulated with glucose and accompanied by calcium ion influx in β-cells, the conformation of sorcin changes, releasing ChREBP to enter into the nucleus (10, 11). Thus, the activity of ChREBP appears to be regulated by multiple factors in different types of cells. ChREBP does not work alone; it regulates glucose-responsive genes in the form of a heterodimeric complex with a partner, Max-like protein X (Mlx) (3, 12, 13). For this reason, ChREBP is also known as Mlx-interacting protein-like (Mlxipl), and the transactivation and inhibitory activities of ChREBP appear to be associated with the heterodimer complex (12–15).

The functional importance of ChREBP was further highlighted by the phenotype of ChREBP-knockout mice (16–18). Although these mice survive without a functional ChREBP, they exhibit suppressed lipogenesis and glycolysis; they are intolerant to high sucrose and high fructose diets and have reduced epididymal and brown fat mass compared to their wild-type littermates (16, 18).

Because liver and adipose tissue are the 2 major lipogenic tissues in mice, we reasoned that ChREBP must play a major function in these 2 organs/tissues. In this study, we have identified the putative genes controlled by ChREBP in these 2 tissues by a genome-wide analysis using chromatin immunoprecipitation followed by next-generation sequencing (ChIP-seq) technology. We believe that use of this unbiased global approach will enable us to further define the general and distinctive roles of ChREBP in these 2 tissues.

Materials and Methods

Animals

C57BL/6 male mice were purchased from The Jackson Laboratory and were maintained in the accredited pathogen-free Baylor College of Medicine transgenic mouse facility. All experiments were performed after approval of the protocol by the Institutional Animal Care and Use Committee of Baylor College of Medicine.

ChIP-seq

Three-month-old male C57BL/6J mice were fasted for 24 hours followed by feeding with a high-carbohydrate fat-free diet (catalog number 901683; MP Biomedicals) for 18 hours. Mice were killed, and liver and white adipose tissue (WAT) were removed and frozen immediately in liquid nitrogen. Frozen tissues were shipped to Genpathway, Inc for further processing and ChIP-seq analysis as outlined below.

ChIP

Frozen tissue samples were submersed in PBS with 1% formaldehyde, cut into small (∼1 mm³) pieces with a razor blade, and incubated at room temperature for 15 minutes. Fixation was stopped by the addition of 0.125 M glycine (final). The tissue pieces were then treated with a Tissue-Tearor and finally spun down and washed twice in PBS. Chromatin was isolated from the sample by addition of 5 to 10 mL of lysis buffer containing PIPES, Igepal, phenylmethylsulfonylfluoride, and protease inhibitor cocktail, followed by disruption with a Dounce homogenizer. Samples were pelleted by centrifugation and resuspended in a buffer containing sodium deoxycholate, SDS, and Triton X-100. Lysates were sonicated using a Misonix Sonicator 3000 equipped with a microtip to shear the DNA to an average length of 300 to 500 bp. Lysates were cleared by centrifugation, and the chromatin suspensions were transferred to new tubes and stored at −80°C. The preceding procedure was directly applied to liver tissues. The fat tissue was processed following a similar protocol, with additional modifications as described (19).

For preparation of input genomic DNA, 2 aliquots of 10 to 25 μL each (approximately $\raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$50$}\right.$ of each chromatin preparation) were removed and treated with RNase for 1 to 2 hours at 37°C, proteinase K for 3 hours at 37°C, and 65°C heat for at least 6 hours to overnight for de-cross-linking. DNAs were purified by phenol-chloroform extraction and ethanol precipitated. Pellets were resuspended in $\raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$5$}\right.$ TE (10mM Tris–1mM EDTA) buffer. The resulting DNAs were quantified on a Nanodrop spectrophotometer. Extrapolation to the original chromatin volume allowed determination of the yield for each chromatin preparation (as measured by the DNA content).

Before use in ChIP, protein A agarose beads (Invitrogen) were preblocked using blocking proteins and nucleic acids for 3 hours. For each ChIP reaction, an aliquot of chromatin (20–30 μg) was precleared with 30 μL of preblocked protein A agarose beads for 1 to 2 hours. ChIP reactions were set up using precleared chromatin and ChREBP antibody (NB400-135; Novus Biologicals) in a buffer containing sodium deoxycholate and were incubated overnight at 4°C. Preblocked protein A agarose beads were added, and incubation at 4°C was continued for another 3 hours. Agarose beads containing the immune complexes were washed 2 times each with a series of buffers consisting of the deoxycholate sonication buffer, high salt buffer, LiCl buffer, and TE buffer. An SDS-containing buffer was added to elute the immune complexes from the beads, and the eluates were subjected to RNase treatment at 37°C for 20 minutes and proteinase K treatment at 37°C for 3 hours. Cross-links were reversed by incubation overnight at 65°C, and ChIP DNAs were purified by phenol-chloroform extraction and ethanol precipitation. The quality of ChIP enrichment was assayed by quantitative PCR (qPCR) using primers against candidate binding sites (Supplemental Table 1). Input DNA was queried at the same sites in parallel. Statistical analysis was performed using t test and a significance level of .05.

ChIP-seq

ChIP DNA was amplified by following the Illumina ChIP-seq DNA Sample Prep Kit protocol. In brief, DNA ends were polished and 5′ phosphorylated using T4 DNA polymerase, Klenow polymerase, and T4 polynucleotide kinase. After addition of 3′-A to the ends using the Klenow fragment (3′–5′ exonuclease minus), Illumina genomic adapters were ligated, and the sample was size fractionated (200–250 bp) on a 2% agarose gel. After a final PCR amplification step (18 cycles, Phusion polymerase), the resulting DNA libraries were quantified and tested by qPCR at the same specific genomic regions as those for the original ChIP DNA to assess the quality of the amplification reactions. DNA libraries were sequenced on a Genome Analyzer II. Sequences (36-nucleotide reads) were aligned to the mouse genome (National Center for Biotechnology Information [NCBI] Build 37.1/mm9) using ELAND (Illumina pipeline) software. The analysis only uses unique mapped reads with no more than 2 mismatches.

Analysis of ChIP-seq result

We used model-based analysis of ChIP-seq (MACS) to do the peak detection after the alignment. MACS produced the tag size shifting and used a dynamic Poisson distribution to simulate the local biases (20). MACS used the predicted fragment size to shift and extend the tags to generate wiggle format files, which stored the signal profiling information and were visualized by the Affymetrix Integrated Genome Browser (21) or UCSC Genome Browser (22). The maximal distance for protein-DNA interaction was set at 10 kb. Target genes were reported and annotated. In addition, the distance from the peak center to the transcription start site (TSS) and the score of the peak were also listed.

We conducted GO (Gene Ontology) slim classification of genes by the web-based GEne SeT AnaLysis Toolkit (WebGestalt) (23). Pathway analysis was performed using the National Institutes of Health DAVID (Database for Annotation, Visualization, and Integrated Discovery) (24). For motif finding, we used MEME Suite (25) with the default settings except that the expected motif site is any repetitions and the find uncentered regions option is selected. Logos of different motifs were generated from MEME-ChIP analysis. We used Regulatory Sequence Analysis Tools to generate the final consensus sequence logo and predict the ChoRE sequence in ChIP-seq intervals (26).

Results

Genome-wide analysis of ChREBP binding sites in liver and WATs

Male C57BL/6J mice were fasted for 24 hours and refed a high carbohydrate diet for 18 hours before they were killed. Liver and epididymal WAT were processed for ChIP-Seq using an antibody against ChREBP protein as described in Materials and Methods. This antibody recognizes both the α and β (27) isoforms of ChREBP. The total number of ChREBP DNA binding sites or ChIP interval was 17 802 in the liver and 10 477 in the WAT (Figure 1A). We found that the frequency of ChREBP binding sites is highest on chromosome 2 (7.39% of WAT ChIP intervals and 7.16% of liver ChIP intervals) and lowest on chromosome Y (0.01% of WAT ChIP intervals and 0.02% of liver ChIP intervals) (Figure 1B).

Figure 1.

Characteristics of anti-ChREBP ChIP-enriched DNA fragments in mouse liver and WAT after fasting followed by refeeding with a high-sucrose diet. A, Number of intervals presented in ChIP-seq of liver and WAT. B, Percentage of intervals located on each mouse chromosome in liver and WAT. C, Percentage of intervals located upstream of gene and downstream of gene in liver and WAT. D, Frequency of ChIP intervals in the liver specimen from the position −10 kb to 10 kb relative to the TSS of each gene. E, Frequency of ChIP intervals in the WAT specimen from the position −10 kb to 10 kb relative to the TSS of each gene. F, Number of genes related to intervals in liver and WAT.

Of these binding sites, about one third (31.11% in the liver and 31.21% in the WAT) are located upstream of the transcription initiation site (Figure 1C). About half (49.45% in the liver and 49.00% in the WAT) are found within the genes. The rest of the binding sites are located in the distal intergenic region (19.44% in the liver and 19.79% in the WAT). Binding sites are especially enriched in the close vicinity region of the TSSs in both liver (Figure 1D) and WAT (Figure 1E). Approximately one fifth (21.82% in the liver and 20.90% in the WAT) of them fall on the proximal promoter region roughly defined as 250 bp around the transcription initiation site.

These binding sites are attributable to 8337 and 11 744 genes in the WAT and liver, respectively, which include 5919 genes that are common between these 2 tissues (Figure 1F). Because there are multiple binding sites in a gene, the total number of binding sites is much larger than the total number of genes.

Validation of ChIP-seq data

We determined whether previously identified ChoREs of target genes are present in our ChIP interval sequences. The peaks of ChREBP binding in 5 representative genes are shown in Figure 2. ChREBP binds to the promoters of thyroid hormone–responsive (Thrsp; also known as S14) and thioredoxin-interacting protein (Txnip) genes in both the liver and WAT with similar peak distribution (Figure 2, A and B), whereas ChREBP preferentially binds to the promoters of fibroblast growth factor 21 (Fgf21) in both liver and WAT but with a higher preference for the liver (Figure 2C). Interestingly, the promoters of liver-type pyruvate kinase (Pklr) and glucokinase regulatory protein (Gckr) show binding by ChREBP exclusively in the liver and not at all in WAT (Figure 2, D and E). As expected, all of these ChoRE sequences reside in ChIP-seq peak areas.

Figure 2.

Validation of anti-ChREBP ChIP-seq results in mouse liver and WAT. A–E, The presence of known ChoRE sequences on the ChIP-seq peaks. We explored the ChIP-seq data using the Integrated Genome Browser and demonstrated the presence of known ChoRE sequences at the summit of ChIP-seq peaks of Thrsp (A), Txnip (B), Fgf21 (C), Pklr (D), and Gckr (E). F and G, Fold enrichment of 29 ChIP DNA fragments in liver (F) and 27 fragments in WAT (G). We performed anti-ChREBP ChIP in a separate group of mice and randomly evaluated ChIP-enriched DNA fragments located within the promoter (15 genes in liver and 14 genes in WAT) or in the gene (14 genes in liver and 13 genes in WAT) using qPCR. The error bars are SD. *, P < .05. Ab, antibody.

To more thoroughly validate the ChIP-seq results, we fasted and refed another group of C57BL/6J mice (n = 4) and performed ChIP with the same ChREBP antibody in the isolated liver and WAT samples, using preimmune rabbit IgG as an additional control. We randomly selected 13 to 15 ChREBP binding regions each from the liver and adipose tissues, a total of 56 genomic regions, to validate the ChIP-seq results using qPCR (Figure 2, F and G). In the promoter regions of the hepatic genes chosen, 11 of 15 ChREBP binding sites are enriched in the DNA samples immunoprecipitated with ChREBP antibody (Figure 2F) compared with that with control rabbit IgG. In the 14 intronic or intergenic regions of the liver genes that we tested, all but 1 of the ChREBP-binding sites were enriched in the DNA samples immunoprecipitated with ChREBP antibody compared with that with control rabbit IgG (Figure 2F). Although most of the 27 regions of the adipose genes were more than 2-fold enriched, only 5 regions reached statistical significance because of substantial variability (Figure 2G).

Pathway analysis

Gene set analysis revealed that most ChIP-based ChREBP target genes are involved in the metabolic process (72.02% and 70.66% of classifiable genes in the liver and WAT, respectively) and biological regulation (60.84% and 61.24% of classifiable genes in the liver and WAT, respectively) (Figure 3A and Supplemental Figure 1, A and B).

Figure 3.

Classification and pathway analysis of genes related to ChREBP binding. A, Frequency of genes identified by anti-ChREBP ChIP-seq that are involved in biological processes. We generated the list of biological processes from WebGestalt using GO slim classification. B and C, Enriched pathways identified by DAVID for the top 1000 and top 3000 genes with high peak intensity in liver and WAT, respectively. D, Genes in the KEGG (Kyoto Encyclopedia of Genes and Genomes) insulin signaling pathway that are related to ChREBP binding (★).

We next performed pathway analysis of the genes that exhibited ChREBP binding to obtain a global picture of biological pathways that may be regulated by ChREBP. Table 1 shows biological pathways relevant to ChIP-seq–based ChREBP target genes in liver and WAT. Three pathways with significant numbers of genes exhibiting ChREBP binding that are among the top 5 pathways in both adipose and liver tissues are the insulin signaling pathway, the adherens junction, and pathways in cancers. Pathways involved in metabolism (citrate cycle, propanoate, and purine metabolism) are among the ChREBP-regulated pathways in both liver and WAT. Various signaling pathways, including MAPK, ErbB, Wnt, neurotrophin, and adipocytokine signaling, were also among the list in both tissues. Genes involved in RNA and protein degradation were also represented, which suggests possible regulation by ChREBP.

Table 1.

Pathway Analysis of Gene Lists by DAVID

Tissue	Term	Count	P Value	Benjamini
Liver	mmu04520: Adherens junction	54	3.0E–09	5.8E–07
Liver	mmu04910: Insulin signaling pathway	77	4.8E–06	4.7E–04
Liver	mmu05200: Pathways in cancer	156	7.8E–06	5.0E–04
Liver	mmu04120: Ubiquitin mediated proteolysis	75	1.2E–05	5.6E–04
Liver	mmu00380: Tryptophan metabolism	29	1.5E–05	6.0E–04
Liver	mmu00250: Alanine, aspartate, and glutamate metabolism	23	4.1E–05	1.3E–03
Liver	mmu04010: MAPK signaling pathway	128	5.7E–05	1.6E–03
Liver	mmu04920: Adipocytokine signaling pathway	41	8.2E–05	2.0E–03
Liver	mmu00410: β-Alanine metabolism	18	1.1E–04	2.1E–03
Liver	mmu05220: Chronic myeloid leukemia	45	1.0E–04	2.2E–03
Liver	mmu04144: Endocytosis	100	1.3E–04	2.3E–03
Liver	mmu04142: Lysosome	64	1.5E–04	2.4E–03
Liver	mmu00280: Valine, leucine, and isoleucine degradation	30	2.1E–04	3.1E–03
Liver	mmu00310: Lysine degradation	27	3.9E–04	5.3E–03
Liver	mmu04722: Neurotrophin signaling pathway	67	5.0E–04	6.5E–03
Liver	mmu04012: ErbB signaling pathway	48	5.6E–04	6.7E–03
Liver	mmu00640: Propanoate metabolism	21	7.2E–04	8.2E–03
Liver	mmu04912: GnRH signaling pathway	52	7.7E–04	8.2E–03
Liver	mmu04610: Complement and coagulation cascades	42	9.0E–04	9.2E–03
Liver	mmu00071: Fatty acid metabolism	28	1.1E–03	9.2E–03
Liver	mmu04310: Wnt signaling pathway	74	9.8E–04	9.5E–03
Liver	mmu03018: RNA degradation	35	1.0E–03	9.6E–03
Liver	mmu04330: Notch signaling pathway	30	1.5E–03	1.2E–02
Liver	mmu05215: Prostate cancer	48	1.5E–03	1.2E–02
Liver	mmu00970: Aminoacyl-tRNA biosynthesis	26	1.9E–03	1.4E–02
Liver	mmu05214: Glioma	36	2.1E–03	1.5E–02
Liver	mmu00140: Steroid hormone biosynthesis	27	2.8E–03	2.0E–02
Liver	mmu04510: Focal adhesion	92	3.2E–03	2.2E–02
Liver	mmu00230: Purine metabolism	75	3.4E–03	2.2E–02
Liver	mmu00020: Citrate cycle (TCA cycle)	20	4.1E–03	2.5E–02
Liver	mmu00982: Drug metabolism	40	4.0E–03	2.5E–02
Liver	mmu00980: Metabolism of xenobiotics by cytochrome P450	36	4.1E–03	2.5E–02
Liver	mmu05016: Huntington disease	85	4.7E–03	2.7E–02
Liver	mmu00330: Arginine and proline metabolism	30	4.8E–03	2.7E–02
Liver	mmu04530: Tight junction	65	5.3E–03	2.9E–02
Liver	mmu04110: Cell cycle	62	5.5E–03	2.9E–02
Liver	mmu00561: Glycerolipid metabolism	27	6.2E–03	3.2E–02
Liver	mmu05212: Pancreatic cancer	38	6.4E–03	3.2E–02
Liver	mmu04720: Long-term potentiation	37	7.0E–03	3.3E–02
Liver	mmu05223: Non-small cell lung cancer	30	6.9E–03	3.4E–02
Liver	mmu05412: Arrhythmogenic right ventricular cardiomyopathy (ARVC)	39	7.7E–03	3.6E–02
Liver	mmu04320: Dorso-ventral axis formation	15	8.5E–03	3.8E–02
Liver	mmu04350: TGF-β signaling pathway	44	8.4E–03	3.8E–02
Liver	mmu00650: Butanoate metabolism	22	9.1E–03	3.9E–02
Liver	mmu00510: N-Glycan biosynthesis	26	9.6E–03	4.1E–02
Liver	mmu04916: Melanogenesis	49	1.1E–02	4.5E–02
WAT	mmu04910: Insulin signaling pathway	64	2.4E–08	4.5E–06
WAT	mmu04510: Focal adhesion	81	2.4E–07	2.3E–05
WAT	mmu05200: Pathways in cancer	111	3.0E–05	1.1E–03
WAT	mmu00230: Purine metabolism	62	2.8E–05	1.3E–03
WAT	mmu04520: Adherens junction	36	2.5E–05	1.6E–03
WAT	mmu05222: Small cell lung cancer	38	6.8E–05	1.9E–03
WAT	mmu04142: Lysosome	49	6.6E–05	2.1E–03
WAT	mmu00280: Valine, leucine, and isoleucine degradation	24	1.3E–04	3.2E–03
WAT	mmu04010: MAPK signaling pathway	90	3.0E–04	6.3E–03
WAT	mmu04120: Ubiquitin-mediated proteolysis	52	3.5E–04	6.7E–03
WAT	mmu04722: Neurotrophin signaling pathway	50	3.9E–04	6.8E–03
WAT	mmu04920: Adipocytokine signaling pathway	30	4.6E–04	7.3E–03
WAT	mmu04110: Cell cycle	49	5.1E–04	7.6E–03
WAT	mmu00640: Propanoate metabolism	17	5.6E–04	7.6E–03
WAT	mmu04310: Wnt signaling pathway	55	6.3E–04	8.0E–03
WAT	mmu03320: PPAR signaling pathway	33	9.7E–04	1.2E–02
WAT	mmu04530: Tight junction	50	1.1E–03	1.2E–02
WAT	mmu04012: ErbB signaling pathway	35	1.4E–03	1.5E–02
WAT	mmu04912: GnRH signaling pathway	38	1.5E–03	1.5E–02
WAT	mmu00020: Citrate cycle (TCA cycle)	16	3.0E–03	2.8E–02

Abbreviation: TCA, tricarboxylic acid.

To further focus on genes with higher peak values at the binding sites, we selected subsets of genes for pathway analysis. In the liver, when we sorted the intervals by their peak values and selected the top 3000 intervals for pathway analysis, we identified 3 biological pathways with significant number of genes represented (Figure 3B): the signaling pathways insulin signaling, peroxisome proliferator–activated receptor (PPAR) signaling, and adipocytokine signaling. When we picked the top 1000 instead of 3000 intervals and repeated the analysis, we narrowed the choices to the insulin and adipocytokine pathways.

In the WAT, the insulin signaling pathway and citrate cycle are enriched in the top 3000 or 1000 intervals with the highest peak values (Figure 3C). Additional pathways involved in focal adhesion, pentose phosphate metabolism, pyruvate metabolism, propanoate metabolism, PPAR signaling, and cancer are enriched in the cutoff of 3000 but not 1000.

We demonstrated close relationships between potential ChREBP target genes and insulin signaling in Figure 3D. For genes enriched in the top 1000 intervals, ChREBP may regulate Prkag2, Acaca, Acacb, G6pc, Ppp1r3c, Ppp1r3b, Sorbs1, Pygl, Fasn, Pik3r1, and Shc4 genes in both tissues (Table 2). Irs2 and Gys1 were presented in the top 1000 gene list in liver and top 3000 in WAT, whereas Pck1 was in the group of the top 1000 in WAT and the top 3000 list in liver. Socs3, Cblb, Sos1, Pklr, Gys2, and Prkaa2 were not presented in the top 3000 WAT ChIP-seq result. Mknk2, Prkar2a, Eif4e, Prkaa1, Sh2b2, and Lipe were genes related to the insulin signaling pathway that were not enriched in the top 3000 liver ChIP-seq result.

Table 2.

Genes in Enriched Pathways Identified by DAVID for Top 1000 and Top 3000 Genes in Liver and WAT

Tissue	Category	Term	Genes	Count	P Value	Benjamini
Liver	Top 3000	mmu04910: Insulin signaling pathway	PRKAG2, PPP1R3C, PPP1R3B, SORBS1, SOS1, FASN, GYS1, GYS2, PRKACA, PRKAA2, INSR, RAPGEF1, PIK3R1, SHC4, SREBF1, IRS2, IRS3, SOCS2, SOCS3, ACACA, FBP1, RAF1, PDE3A, ACACB, PCK1, CBLB, G6PC, CRKL, PYGL, GSK3B, PKLR, PTPN1	32	2.5E–06	4.6E–04
Liver	Top 3000	mmu03320: PPAR signaling pathway	SCD1, ACOX2, PPARA, SLC27A1, CPT1B, PPARD, ACADM, EHHADH, DBI, CPT1A, PCK1, ACSL1, APOA1, SORBS1, CYP7A1, APOA5, UBC, ACAA1A, SCP2, ACSL5	20	8.4E–05	7.8E–03
Liver	Top 3000	mmu04920: Adipocytokine signaling pathway	PPARA, CPT1B, IRS2, IRS3, SOCS3, LEPR, PRKAG2, ADIPOR2, ACACB, CPT1A, TRADD, PCK1, TNFRSF1B, G6PC, ACSL1, PRKAA2, ACSL5	17	3.3E–04	2.0E–02
Liver	Top 1000	mmu04910: Insulin signaling pathway	IRS2, SOCS3, PRKAG2, ACACA, ACACB, CBLB, G6PC, PPP1R3C, PPP1R3B, SORBS1, PYGL, SOS1, PKLR, FASN, GYS1, GYS2, PRKAA2, PIK3R1, SHC4	19	1.2E–06	1.8E–04
Liver	Top 1000	mmu04920: Adipocytokine signaling pathway	PPARA, G6PC, IRS2, ACSL1, SOCS3, PRKAG2, ADIPOR2, PRKAA2, ACACB, TRADD	10	4.9E–04	3.6E–02
WAT	Top 3000	mmu04910: Insulin signaling pathway	PRKAG3, PRKAG2, HK2, MKNK2, PRKAR2B, PPP1R3C, PRKAR2A, EIF4EBP1, PPP1R3B, SORBS1, SOS2, FASN, GYS1, PRKAA1, SH2B2, INSR, RAPGEF1, PIK3R1, SHC4, IRS2, ACACA, RAF1, ACACB, PCK2, PCK1, PPP1CA, G6PC, EIF4E, PYGL, LIPE	30	4.2E–06	7.5E–04
WAT	Top 3000	mmu00020: Citrate cycle (TCA cycle)	SDHB, PCX, ACO1, SUCLG2, IDH1, ACLY, PDHA1, DLAT, SUCLA2, PCK2, IDH3A, PCK1	12	3.0E–05	2.7E–03
WAT	Top 3000	mmu04510: Focal adhesion	BCAR1, COL3A1, PIP5K1C, ITGB5, PTEN, PXN, LAMB3, ITGB8, SOS2, COL6A1, RAPGEF1, PIK3R1, SHC4, EGFR, COL4A2, ACTN4, RAF1, CAPN2, VAV2, COL5A3, VWF, PPP1CA, LAMA4, LAMA3, ITGA6, LAMC3, CCND2, LAMB1–1, VEGFA, GRLF1, LAMC2, MYLK, PARVB, PARVA	34	1.4E–04	8.5E–03
WAT	Top 3000	mmu00030: Pentose phosphate pathway	PGM2, TALDO1, PFKL, H6PD, PGD, PFKP, DERA, TKT, RPIA, GPI1	10	2.0E–04	8.9E–03
WAT	Top 3000	mmu00620: Pyruvate metabolism	PCX, AKR1B3, PKM2, ACYP2, ALDH2, ACACA, PDHA1, ACACB, GRHPR, DLAT, PCK2, PCK1	12	5.0E–04	1.8E–02
WAT	Top 3000	mmu00640: Propanoate metabolism	ACADM, SUCLG2, ALDH2, ACACA, ECHS1, ACACB, HIBCH, SUCLA2, PCCB, PCCA	10	6.7E–04	2.0E–02
WAT	Top 3000	mmu03320: PPAR signaling pathway	SCD1, ACOX1, SCD2, ACADM, RXRA, PPARG, PCK2, DBI, ADIPOQ, CPT1A, PCK1, ACSL1, SORBS1, UBC, FABP4, UBB, PLTP, ANGPTL4	17	9.3E–04	2.1E–02
WAT	Top 3000	mmu05200: Pathways in cancer	TRAF1, WNT5B, APC2, PPARG, ZBTB16, KIT, GLI2, PTEN, TCF7L2, ACVR1C, MAX, LAMB3, SOS2, RALB, RUNX1, AXIN2, TRAF5, PIK3R1, TRP53, EGFR, TCF7, COL4A2, EPAS1, RXRA, TGFBR2, CYCS, RUNX1T1, RAF1, SMAD3, FGF22, SMAD2, FGF21, RB1, STAT1, DAPK1, JUP, LAMA4, LAMA3, ITGA6, PIAS4, LAMC3, VEGFA, LAMB1–1, LAMC2	44	2.2E–03	4.3E–02
WAT	Top 1000	mmu04910: Insulin signaling pathway	PRKAG2, ACACA, MKNK2, ACACB, PCK1, PRKAR2A, G6PC, PPP1R3C, EIF4E, PPP1R3B, SORBS1, PYGL, FASN, PRKAA1, SH2B2, LIPE, PIK3R1, SHC4	18	4.3E–06	6.3E–04
WAT	Top 1000	mmu00020: Citrate cycle (TCA cycle)	SDHB, ACO1, SUCLG2, IDH1, PDHA1, DLAT, IDH3A, PCK1	8	6.8E–05	5.0E–03

Abbreviation: TCA, tricarboxylic acid.

Motif analysis

We observed that most of the ChoREs in the promoter of well-established ChREBP target genes are located at the summit (tip) of the ChIP-seq peaks (Figure 2, A–E); we therefore extracted 50-bp sequence at the tips of the top 1000 peaks from liver ChIP-seq data and top 1000 peaks of WAT ChIP-seq data for motif discovery. Three motifs containing 17-bp ChoRE or a ChoRE-like sequence were found in a quarter of these peak sequences (506 sequences) (Figure 4A): (1) 2 CAYGYG separated by 5 nucleotides (203 sequences), (2) CAYGYK at one end and (5′-)CAYGY at the other end (330 sequences), and (3) (5′-)CABG at one end and (5′-)CAY E-box half-site at the other end (170 sequences). These ChoRE or ChoRE-like motifs are distributed throughout the 50-bp peak region (Figure 4B).

Figure 4.

Enriched ChREBP binding motifs in ChIP-seq results. A, ChREBP binding motifs recognized by analysis of the top 1000 peak sequences and their frequencies. We used MEME-CHIP to analyze the 50-bp sequences of the top 1000 peak sequences from ChIP-seq in liver and WAT and identified 3 ChoRE-like motifs. B, Frequency of distance between the discovered ChoRE-like sequence and ChIP-seq peak center. C, Evaluation of ChoRE prediction models. We used 6 different set of sequences to validate ChoRE prediction by ensuring the presence of 18 mouse ChoRE sequences on Pklr, Gckr, Pnpla3, G6pc, G0s2, Thrsp, Fasn, Acaca, Txnip, Slc2a4, Fgf21, Klf10, Pc, Hif1a, Bhlhe40, Mlxipl, and Mid1ip1 (2 sequences) genes. D, Consensus ChREBP binding sequence. We used Regulatory Sequence Analysis Tools to generate the ChoRE sequence logo with 219 sequences from the best prediction model (Motif1 + 16 known ChoREs).

We next searched for ChoREs in the rest of ChIP-seq data using enriched sequences from those top 1000 peaks. We validated our ChoRE prediction by ensuring the presence of 18 mouse ChoRE sequences, retrieved from known ChREBP target genes or conserved sequence prediction, which displayed binding peaks in our study in liver (Pklr, Gckr, and Pnpla3), WAT (G6pc and G0s2), or both tissues (Thrsp, Fasn, Acaca, Txnip, Slc2a4, Fgf21, Klf10, Pc, Hif1a, Bhlhe40, Mlxipl, and Mid1ip1 [2 sequences]) (27–38).

We compared detection of these sequences using six different models (Figure 4C): (1) motif 1 alone (203 sequences), (2) motifs 1 and 2 (426 sequences), (3) combined all 3 ChoRE motifs (506 sequences), (4) 23 experimentally verified ChoRE sequences present in mouse, rat, or human, (5) motif 1 plus 16 experimentally verified ChoREs (219 sequences; 7 of 23 known ChoREs were already present in motif 1), and (6) motif 1 and 2 plus 13 experimentally verified ChoREs (439 sequences; 10 of 23 known ChoREs were already present in motifs 1 and 2). Hif1a and Pnpla3 ChoREs are absent when we use input ChoRE sequences from motif 1 alone or motifs 1 and 2 but present in other types of analysis. Thrsp ChoRE cannot be detected by motif 1 alone. Analysis using these 23 known ChoREs failed to detect the mouse Fgf21 ChoRE sequence. The combination of sequences from motif 1 and 16 experimentally verified ChoRE sequences was able to detect all of the 18 mouse ChoRE sequences presented in our peak sequences. Addition of the sequences of motif 2 in analysis worsened the sensitivity of the ChoRE detection. We therefore use sequences from motif 1 and 16 known ChoREs to generate a CAYGYGnnnnnCRCRTG consensus sequence for ChREBP binding (Figure 4D). The selected putative ChREBP target genes and their ChoREs are listed in Table 3.

Table 3.

List of Selected Genes and Predicted ChoRE Sequences on Their Promoters

Gene	Chromosome	Location		Position From TSS	Sequence
		TSS	Predicted ChoRE
Pklr	3	88940544	88940374	−170	CACGGGGCACTCCCGTG
Thrsp	7	104566019	104567432	−1413	CACTTGGTGGCCCTAGG
Fasn	11	120685860	120692357	−6497	CATGCGCTGCAGGCATG
Fasn	11	120685860	120685682	179	CGCGTGGCCCGCGCGAG
Acaca	11	84008939	83986684	−22 255	CATGTGAAAACACTGTG
Txnip	3	96361879	96361794	−85	CACGAGGGCTGCACGAG
G6pc	11	101229043	101226090	−2953	CATATACCGAGCACATG
Slc2a4	11	69761691	69762434	−743	CACGCGGCCAGCACATG
G0S2	1	195099381	195099524	−143	GGCGTGCACTACTCGTG
Gckr	5	31599953	31599585	−368	CATGTGATCAAGCCATG
Klf10	15	38230461	38230605	−144	CAGCCGCTGATCACGCG
Pnpla3	15	83998245	83998006	−239	AGAGTGCCACCGGCGTG
Pnpla3	15	83998245	83998103	−142	CACGCGGGAGGGACATG
Pc	19	4510471	4510107	−364	CACGTGCTACCCGAGTG
Hif1aa	12	75008853	75008741	−112	CAGCTGGGGCTCAGCTG
Fgf21	7	52870859	52870931	−72	CACGCGTCAGGAGTGGG
Mid1ip1	X	10294490	10294124	−366	CACGTGCATCCGGCATG
Mid1ip1	X	10294490	10294113	−377	CACCCCATGTCCACGTG
Bhlhe40	6	108610622	108610365	−257	CACGTGAGGCTCATGTG
Bhlhe40	6	108610622	108609766	−856	CAGGTGCGCGCGGCGTG
Bhlhe40	6	108610622	108609425	−1197	CGCGTCCGGGCCACGTG
Gapdh	6	125115600	125117145	−1545	CACGGGCCACACAGCTG
Gapdh	6	125115600	125115846	−246	CACACGCTTGGTGCGTG
Gpd1	15	99548023	99546539	−1484	CAGGAGGTCTGCACATG
Mlxipl	5	135565759	135565916	+158	CGCGTGGAGCTCAGGTG
Mlxipl	5	135582760	135577105	−5655	CACGAGGTGGGAGGCTG
Acacb	5	114615526	114611660	−3866	CATGAGGTTTGCACGTG
Ppp1r3b	8	36438794	36438672	−142	CACGTGGTAGAGTGCTG
Ppp1r3c	19	36811093	36811282	−189	CACGTTCCAGGGGCAGG
Pygl	12	71328669	71334944	−6275	CACATGCTTGGCACATG
Gys2	6	142421628	142421467	+162	CACGCTTCATCCAGCTG
Fbp1	13	62989595	62995044	−5449	GACAGGCTAAGCACGTG
Fbp1	13	62989595	62995033	−5438	CACGTGCACTCAAGTTG
Irs2	8	11008429	11011390	−2961	CGTGAGCACAGCACGTG
Irs3	5	138086941	138094849	−7908	CTCATGCAGTTCACGTG
Srebf1	11	60034105	60036424	−2319	CGGGCGCTGGGCACGTG
Mtor	4	147822690	147822408	−282	CACACGGTCACAGCGTG
Raf1	6	115626652	115626695	−43	CACATTCCAGGCTCGAG
Adipor2	6	119367500	119368809	−1309	CATCGGGCTGTCACGTG
Adipor2	6	119367500	119368798	−1298	CACGTGTCCTGCCAGTG
Hnf1a	5	115421070	115421169	−99	CATGAGGCCTGCACTTG
Nfe2l2	2	75542697	75543365	−668	CACGCTGCAAACTCGTG
Klf15	6	90412619	90413530	912	CATTCGCCACCCCCGTG
Sult1e1a	5	88020635	88030255	−9620	CATGTCAGATCCACAAC
Rtp4a	16	23610004	23606874	−3130	CACGTGAAGGAGATGGG
Slc34a2a	5	53440591	53439394	−1197	TCCGTGGTGGTCACGTG
Lgals3bpa	11	118263244	118263395	−151	CACAGGGCGCTTGGGTG
Ehhadha	16	21787906	21788287	−381	CAGCTGGAAGGACCGTG
Cav1a	6	17257639	17257641	+3	CAGGCGCTCTCTGCCTG
Cxcl1a	5	91320270	91319967	−303	CACGTGCACATAACGCG
Junba	8	87502646	87510120	−7474	CGCGTGCGTCGTTCGTG

^aGlucose-repressed genes identified by Ma et al (29) in rat primary hepatocytes.

To explore ChREBP binding motifs on the glucose-repressed genes, we adopted the published list of 38 rat glucose-repressed genes in primary hepatocytes for additional analysis (29). Fifteen genes are present in our mouse liver ChIP-seq data. There are 11 genes with peak sequences in the promoter region, 9 having putative conventional ChoREs located at the summit of ChIP-seq peak. Information on the predicted ChoRE sequences of these 9 genes (Sult1e1, Rtp4, Slc34a2, Lgals3bp, Ehhadh, Cav1, Cxcl1, Junb, and Hif1a) is presented in Table 3.

Discussion

Liver and adipose tissues are 2 important lipid-handling tissues in mammals. Although liver is the major organ for fatty acid and triglyceride synthesis, adipose tissues are the main organs for lipid storage. It is, therefore, not surprising that of all the genes (14 162 genes) that exhibit ChREBP binding sequences by our analysis, a large majority (5919 genes) were present in both liver and WATs (Figure 1F). In comparison with a recent study in HepG2 human hepatocellular carcinoma cell model (39), our ChIP-seq experiment using mouse liver has identified a 15 times larger number of genes. This discrepancy may possibly be explained by differences in species and properties of cancer vs noncancer cells.

Gene set analysis of our data revealed that ChIP-based ChREBP target genes are involved in different metabolic processes and biological regulation (Figure 3A). These putative ChREBP target genes encompass diverse biochemical pathways. Only a handful of these pathways have been documented in the past to be regulated by ChREBP, for example, fatty acid metabolism and the citrate cycle (Table 1). Cancer-related pathways, including those involved in cell cycle and notch signaling, were significantly enriched in our study. Genes involved in the adherens junction were also enriched. Dysregulation of the cell-cell adherens junction is closely related to invasion and metastasis of cancer (40). Our findings underline the potential importance of ChREBP as a player in tumorigenesis and cancer progression (41–43).

The insulin signaling pathway is among the top 3 pathways in both liver and WAT. In addition, this is the only pathway identified in the top 1000 intervals with the highest peak values in both liver and WAT. A few of their members have been described as ChREBP target genes involved in glucose metabolism and lipogenesis, ie, Acaca, G6pc, Fasn, and Pklr. Here we showed that ChREBP binds to a number of additional genes within this pathway (Table 2), including Prkag2, Acacb, Ppp1r3c, Ppp1r3b, Sorbs1, Pygl, Pik3r1, Shc4, Irs2, Socs3, Cblb, Sos1, Gys1, Gys2, Prkaa2, Mknk2, Pck1, Prkar2a, Eif4e, Prkaa1, Sh2b2, and Lipe. Srebf1, which encodes the lipogenic transcription factor SREBP-1c, is another key gene in this pathway that is among those in the top 3000 intervals in the liver (Table 2). This observation is largely consistent with a recent study using a HepG2 human hepatocellular carcinoma cell model to analyze the ChREBP targets (39), which demonstrated that glucose up-regulates SREBP-1c expression and binding of ChREBP to the Srebf1 gene may mediate high glucose-induced expression of SREBP-1c. Human and mouse Srebf1 promoters in this region are conserved, and the sequence of the validated human Srebf1 promoter fragment falls within the mouse Srebf1 promoter sequence resulting from our ChIP-seq in the liver. We predicted putative ChoRE (Table 3) on the mouse Srebf1 promoter and promoters of genes in the insulin signaling pathway (Mtor, Raf1, Acacb, Ppp1r3b, Ppp1r3c, Pygl, Gys2, Fbp1, Irs2, and Irs3), as well as promoters of important genes related to insulin resistance and diabetes (Adipor2, Hnf1a, Nfe2l2, and Klf15) (44–47). The functional activity of these predicted ChoREs should be further validated because binding may not necessarily affect transcription activity. For example, ChREBP binding to Gpd1 ChoRE was previously demonstrated and also supported by our data; however, neither the natural Gpd1 promoter with this ChoRE sequence nor the artificial promoter containing 2 copies of Gpd1 ChoRE responded to the presence of high glucose (29).

The consensus sequences CAYGnGnnnnnCnCRTG and CABGTGnnCnGnGnSTG were previously proposed in rat liver and the human HepG2 cell line, respectively (29, 39). Our motif analysis of ChIP-seq data reveals similar ChoRE sequences consisting of 2 E-boxes with a sequence signature of (5′-)CAYGYG that are separated by 5 nucleotides. It has been shown that promoters containing 2 perfect CACGTG motifs separated by 5 nucleotides conferred a very strong response to high glucose in primary rat hepatocytes. Spacing of 4 nucleotides between these CACGTG motifs completely abolished induction of promoter activity by high glucose, whereas spacing by 6, 7, or 15 nucleotides resulted in a weak glucose response (4). Our recent study in the 832/13 rat insulinoma cell line further provided corroborative evidence that ChREBP does not activate promoters containing 2 E-boxes with a 4-bp spacer (32). It is not known whether spacing lengths other than 5 nucleotides could enable a response to ChREBP.

Our analysis successfully detected all previously verified or predicted ChoRE sequences (Table 3) except the Gcgr promoter (48), which did not display a binding peak in our ChIP-seq result. We also found additional ChREBP binding sequences on Fasn, Pnpla3, Bhlhe40, Gapdh, and Mlxipl. The ChREBP binding to ChoRE located on the distal promoter of rat Fasn was well characterized. It also has been shown by ChIP that ChREBP binds to the mouse proximal Fasn promoter (49). Our analysis reveals the putative ChoRE sequence located in that binding region as shown in Table 3. A functional ChoRE sequence on the mouse Pnpla3 promoter was previously reported (30). The mutated ChoRE-containing Pnpla3 promoter did not respond to ChREBP/Mlx at all. For that reason, another ChoRE-like sequence that we detected between that ChoRE and the TSS (Table 3) is probably inactive. Serial deletion experiments revealed the functional ChoRE of the mouse Bhlhe40 promoter (37). The 5′ deletion of the promoter fragment, which contains one of our putative ChoREs, diminished the responsiveness of mouse Bhlhe40 promoter activity to a dominant active form of ChREBP. Thus, this sequence may function as a ChoRE. Another ChoRE-like sequence that we predicted was not included in that study. The mouse Gapdh promoter also contained a ChoRE sequence (50). Activity of the Gapdh promoter in response to glucose was not completely abolished when that sequence was manipulated. Another Gapdh ChoRE identified in our study might be responsible for the rest of the glucose response. The β-isoform of Mlxipl or ChREBP in liver and adipose tissues as well as ChoRE on its promoter was recently described (27). The α isoform of Mlxipl may also have a ChoRE on the promoter because its expression was influenced by fasting and refeeding in mice. It is possible that our predicted ChoRE on the distal promoter of the α isoform of Mlxipl may serve this function.

The most commonly reported action of ChREBP is up-regulation of target genes; however, as noted in Table 3, in a minority of cases, the transcription factor may repress specific target genes. Close scrutiny of the negatively and positively regulated target genes has failed to reveal a consistent difference between them, suggesting that a ChoRE marks a gene target but does not necessarily determine whether the gene is subject to up- or down-regulation. It is plausible that other cis- or trans-acting factors contribute to the direction of ChREBP-mediated gene regulation, which is a subject for future investigation.

In summary, genome-wide analysis of ChREBP binding sites in the liver and WAT reveals the optimal consensus ChREBP binding sequence CAYGYGnnnnnCRCRTG and demonstrates that ChREBP is a major transcription factor that binds to large numbers of putative hepatic and/or adipose target genes involved in several cellular processes, including several key metabolic pathways, the adherens junction, the insulin signaling pathway, and tumorigenesis.