Hepatocyte glucose production is a complex process that integrates cell-autonomous mechanisms with cellular signaling, enzyme activity modulation, and gene transcription. Transcriptional mechanisms controlling glucose production are redundant and involve nuclear hormone receptors and unliganded transcription factors (TFs). Our knowledge of this circuitry is incomplete. Here we used DNA affinity purification followed by mass spectrometry to probe the network of hormone-regulated TFs by using phosphoenolpyruvate carboxykinase (Pck1) and glucose-6-phosphatase (G6pc) in liver and primary hepatocytes as model systems. The repertoire of insulin-regulated TFs is unexpectedly broad and diverse. Whereas in liver the two test promoters are regulated by largely overlapping sets of TFs, in primary hepatocytes Pck1 and G6pc regulation diverges. Insulin treatment preferentially results in increased occupancy by the two promoters, consistent with a model in which the hormone’s primary role is to recruit corepressors rather than to clear activators. Nine insulin-responsive TFs are present in both models, but only FoxK1, FoxA2, ZFP91, and ZHX3 require an intact Pck1p insulin response sequence for binding. Knockdown of FoxK1 in primary hepatocytes decreased both glucose production and insulin’s ability to suppress it. The findings expand the repertoire of insulin-dependent TFs and identify FoxK1 as a contributor to insulin signaling.
The liver produces glucose during fasting in order to maintain euglycemia. This process becomes altered in diabetes, when resistance to or a lack of insulin increases glucose production, contributing to hyperglycemia and its complications (1). Hepatocyte glucose production is a complex, multilayered process that involves contributions from other organs in the form of glucogenic substrates, and an intracellular shift of the glucose-6-phosphate pool from use via glycolysis or glycogen-triglyceride synthesis to dephosphorylation and release as glucose (2). The latter process integrates cellular signaling, enzyme activity modulation, and gene transcription (1).
Transcriptional mechanisms controlling glucose production are famously redundant, with nuclear hormone receptors and posttranslationally modified transcription factors (TFs) without ligands participating in the process (3–5). But loss-of-function experiments in mice have shown that many genes continue to be properly regulated in the absence of these factors, indicating that our knowledge of the circuitry is incomplete (5–8). An interesting example of this partial knowledge is a comparison of two genes of historical interest in hormone action research: glucose-6-phosphatase (G6pc) and phosphoenolpyruvate carboxykinase (Pck1). Irrespective of their actual contributions to hepatic glucose production, these two genes are useful as a model for understanding how the two competing branches of glucose metabolism—the anabolic insulin-dependent branch and the catabolic glucagon/glucocorticoid-dependent branch—act transcriptionally. Pck1 and G6pc share transcriptional regulators such as glucocorticoid receptor (GR), FoxO1, CREBP, CRCT2, and C/EBPα/β (4,5,9–12). However, they respond differently to these TFs. Thus, when FoxO is ablated, hormonal modulation of G6pc is completely abolished while regulation of Pck1 is largely retained (5,7,13). These findings highlight a gap in our knowledge of the broader set of TFs acting on these promoters that can account for the complexity of the hormonal response to nutrients.
DNA affinity purification coupled with mass spectrometry (MS) has been used to identify TFs and chromatin-bound complexes required for gene regulation (14). Here we used it to probe the network of hormone-regulated TFs, with Pck1 and G6pc promoters (Pck1p and G6pcp) in whole liver and primary hepatocytes as the targets, and to unravel their unexpectedly vast repertoire.
Research Design and Methods
To identify the network of TFs regulating expression of the gluconeogenic genes Pck1 and G6pc, we combined DNA affinity pull down and MS analysis. In the DNA pull-down experiments, we used nuclear proteins isolated from liver from mice that were deprived of food (for 16 h overnight) and mice that were refed (deprived of food for 16 h then refed over 4 h) or primary hepatocytes that received various hormone treatments. We sorted and analyzed TFs based on how they responded to refeeding, dexamethasone (Dex) and cAMP (D/C), and Dex, cAMP, and insulin (D/C/I). Comparison of the two promoters further defined common and selective insulin-regulated TFs. Comparison of in vivo and in vitro models identified conserved insulin-regulated TFs. To validate the findings, we performed functional studies of one selected candidate in whole liver and primary hepatocytes, demonstrating its role in modulating glucose production.
Primary Hepatocyte Studies
Primary hepatocytes were isolated from 8- to 12-week-old male C57/B6 or FoxO1 flox/flox mice, as described previously (15). All animal studies were approved by the Columbia Institutional Animal Care and Use Committee. Isolated primary hepatocytes were resuspended in hepatocyte plating medium (Medium 199 supplemented with 10% FBS [Thermo Fisher Scientific, Waltham, MA], 100 units/mL penicillin-streptomycin [Thermo Fisher Scientific], and 10 μg/mL gentamicin [Sigma-Aldrich, St. Louis, MO]), and 107 cells were placed in a 150-mm (in diameter) dish, or 5 × 105 cells were applied in each well of a 12-well collagen-coated plate. After culturing for 24 h, cells were incubated overnight in hepatocyte starvation medium comprising Medium 199 supplemented with 1% BSA (Sigma-Aldrich), penicillin-streptomycin, and gentamicin. Thereafter either the vehicle (0.002% methanol) or 1 μmol/L Dex and 0.1 mmol/L cAMP were added and left for 6 h. Then 100 nmol/L insulin was added and left for 30 min. Cells were then lysed in order to isolate protein.
Cells were ready 2 h after plating for siRNA transfection or adenovirus infection. Wild-type primary hepatocytes were transfected with control (AM4611) or Foxk1 siRNA mix (Thermo Fisher Scientific) using the Viromer BLUE transfection reagent (Lipocalyx, Halle, Germany). Foxk1 siRNA mix includes two siRNAs: FoxK1 silencer siRNA 67984 (sense) (GGAGCCUCACUUCUAUCUUtt) and siRNA2 68712 (sense) (GGGCUCUUUUUGGCGAAUAtt). FoxO1 flox primary hepatocytes were infected with adenovirus-Cre (Welgen, Worcester, MA) at 10 multiplicities of infection. After transfection (48 h), we replaced the medium with hepatocyte starvation medium, left it overnight, and then treated cells with D/C or D/C/I, as described above. Glucose production was assayed as previously described (15).
DNA Pull Down and Western Blotting
We amplified Pck1p (−600 to +69 base pairs [bp]) and G6pcp (−509 to +53 bp) from mouse genomic DNA using primers labeled with 5′-biotin: Pck1p forward, AGCTTACAGCCACTCCTAATCTCTG; Pck1p reverse, CAGAGATCGCTGAGCGCCTTG; G6pcp forward, ACGTGAACTTGGTGAAAGTCCA; G6pcp reverse, TACCTCAGGAAGCTGCCAGC. The Pck1pΔIRS promoter sequence (−600 to −417 bp, −401 to +69 bp) was amplified from the Pck1pΔIRS-MXS-mCherry plasmid by using the same biotin-labeled Pck1p primers. We cloned Pck1p into pMXS-mCherry through the use of In-Fusion Cloning with the following primers: Pck1p-MXS forward, TCAGTGAGCCATGATAGCTTACAGCCACTCCTAATCTCTG; Pck1p-MXS reverse: CATTCGAGTTACGCGGATCGCTGAGCGCCTTG. The pMXS-mCherry plasmid was linearized with EcoRV and MluI for In-Fusion recombination, followed by site-directed mutagenesis in order to generate the Pck1pΔIRS-MXS-mCherry plasmid by using two ΔIRS primers: forward, CAGCAGCCACCGGCACAC; reverse, CAGCTGTGAGGTGTCACTCCC. We sequenced each plasmid at Genewiz (South Plainfield, NJ) to confirm the mutation.
Biotinylated Pck1p or G6pcp fragments were immobilized on Dynabeads M-280 Streptavidin (Invitrogen, Carlsbad, CA) with DNA binding buffer (5 mmol/L Tris [pH 7.5], 0.5 mmol/L EDTA, 1 mol/L NaCl) and washed with BC-150 buffer (20 mmol/L Tris [pH 7.3], 150 mmol/L NaCl, 0.2 mmol/L EDTA, 20% glycerol). We prepared nuclear extracts using NE-PER Nuclear and Cytoplasmic Extraction Reagents containing Halt protease and a phosphatase inhibitor cocktail (Thermo Fisher Scientific). Before adding Pck1p Dynabeads, the nuclear extract was adjusted to 200–250 mmol/L total salt with BC-0 (20 mmol/L Tris [pH 7.3], 0.2 mmol/L EDTA, 20% glycerol) and 1 mmol/L EGTA/EDTA. Pck1p Dynabeads were incubated with rotation with 0.5 mg nuclear extract at 4°C overnight. The supernatant was discarded and the beads were washed twice with NETN buffer (50 mmol/L NaCl, 1 mmol/L EDTA, 20 mmol/L Tris [pH 8.0], 0.5% NP-40) and three times with PBS. Beads were resuspended in 50 μL sample buffer, boiled for 5 min, and fractionated on 10% SDS-PAGE. Gels were stained with EZBlue Gel Staining Reagent (Sigma-Aldrich). Half the volume of pull-down samples or 30 μg of raw protein lysate from each sample were used for Western blotting. Membranes were prestained with Ponceau Red solution (0.1% [w/v]) and cut into strips on the basis of molecular weight. Information about antibodies is in Supplementary Table 7.
Nano Liquid Chromatography/MS Analysis
Samples were analyzed with an Orbitrap Fusion mass spectrometer coupled with an Easy-nLC 1000 nanoflow liquid chromatography system (Thermo Fisher Scientific). The quantities of identified proteins were estimated by using the intensity-based absolute quantification (iBAQ) method. TFs were annotated from identified nuclear proteins according to the DBD database (16). We subtracted the iBAQ value of negative control samples from those of all treatment groups. ΔiBAQ <0 were assigned an arbitrary value of 0. We generated heat maps on the basis of a normalized iBAQ value by using One Matrix Clustered Image Maps (https://discover.nci.nih.gov/cimminer/oneMatrix.do). We analyzed functional annotation of TF lists and pathways of TFs by using the ConsensusPathDB gene set analysis tool (http://cpdb.molgen.mpg.de) for overrepresentation analysis. More details are available in the Supplementary Data.
RNA Isolation and Quantitative PCR Analysis
We lysed primary hepatocytes or liver in 1 mL TRIzol reagent. We purified RNA further using an RNeasy Mini Kit (Qiagen, Germantown, MD). For reverse transcription, we used a qScript cDNA Synthesis Kit (QuantaBio, Beverly, MA). RNA (1 μg) was used for each reverse transcriptase reaction. The 20 μL cDNA solution was diluted with RNase-free water to a final volume of 0.2 mL. We used GoTaq qPCR Master Mix (Promega, Madison, WI) for subsequent quantitative (q)PCR analysis. Primer information is available in Supplementary Table 8.
pRL3 basic was linearized with KpnI and HindIII, and promoter fragments were cloned with an In-Fusion cloning kit. Primers used for promoter amplification included the pGL3-Pck1p forward primer TTTCTCTATCGATAGAGCTTACAGCCACTCCTAATCTCT and the Pck1p-luc reverse primer CCGGAATGCCAAGCTCAGAGATCGCTGAGCGCCT. We generated a pcDNA3.1 Foxk1-Flag mutant by removing 40 amino acids from the N-terminal using the Foxk140Δ forward primer GCGCAACCTCCACCCGGG and the Foxk140Δ reverse primer TTCGGCCATGGTGGCGGATC, and a Q5 Site-Directed Mutagenesis Kit (New England BioLabs). pRL2-3xIRS has been described (17). HEK293 cells (0.5–1 × 106) were seeded in each well of a 12-well plate. DNA (200 ng of pGL3-Pck1p, pGL3-G6pcp, Foxk1-Flag, or Foxk140del, or FoxO1, or red fluorescent protein [control]) and 20 ng of pRL-CMV plasmids were mixed with Lipofectamine 3000 transfection reagents in 0.1 mL Opti-MEM medium (Thermo Fisher Scientific) and added to each well after being incubated at room temperature for 15 min. After 36 h, we aspirated the culture medium, washed cells once with PBS, and lysed them in 0.35 mL lysis buffer. We performed luciferase assays using the Dual-Luciferase Reporter Assay System (Promega). The signals were read and recorded with an Orion L microplate luminometer. Plasmids are available upon request.
We used a ChIP-IT High Sensitivity kit (Active Motif, Carlsbad, CA) for chromatin immunoprecipitation (ChIP) assay following the manufacturer’s protocol. We sonicated 300 mg of liver using an S220 Focused-ultrasonicator (Covaris) to obtain sheared chromatin. Immunoprecipitation was performed using 4 μg of anti-FOXK1 antibody (Abcam, Cambridge, MA) for 10 µg of sheared chromatin. Real-time ChIP–qPCR were carried out using GoTaq qPCR Master Mix (Promega). The signals of binding events were normalized against input DNA for primer efficiency according to the protocol of a ChIP-IT qPCR Analysis Kit (Active Motif). Negative primers for qPCR analysis were purchased from Active Motif. Primers used to detect FOXK1 binding sites on G6pc or Pck1 promoters are listed in Supplementary Table 9.
Mice (8- to 12-weeks old) were deprived of food overnight, and a subset was allowed to refeed for 1 h. We killed the mice, collected liver tissue, and fixed it in 4% paraformaldehyde for 2 h. We rinsed the tissue in PBS and then transferred it to 30% sucrose for dehydration overnight at 4°C, after which we embedded it in OCT compound (Sakura, Torrance, CA) and froze it at –80°C. Tissue blocks were sectioned into 8-μm-thick slices, which were rinsed with PBS three times and incubated in HistoVT One buffer (Nacalai USA, San Diego, CA) at 70°C for 30 min. After being blocked with 10% donkey serum in PBST (1X PBS supplemented with 0.1% Triton X-100) at room temperature for 30 min, tissue sections were incubated with FoxK1 antibody (1:500) at room temperature for 2 h and then with Alexa 488 donkey antirabbit antibody (1:400; Thermo Fisher Scientific) for 1 h. Frozen sections were counterstained with Hoechst stain (1:1,000; Thermo Fisher Scientific) for 5 min. Slides were mounted with Dako Glycergel and imaged with an Olympus fluorescent microscope.
Each experiment was replicated at least three times for each condition. The liver tissue underwent MS once. We used the two-tailed Student t test and one-way ANOVA for statistical analysis; P < 0.05 indicated statistical significance. All data are presented as the mean ± SEM.
Identification of Hormone-Regulated TFs Through DNA Pull Down and MS
To interrogate the functional diversity of hormone-regulated TFs, we leveraged the observation that hepatic Pck1 and G6pc share transcriptional regulators such as GR, FoxO1, CREBP, and C/EBPs α and β (1). However, when we ablated FoxO1 in primary hepatocytes with >99% efficiency using adenovirus-Cre, G6pc induction by D/C and by D/C/I were abolished, whereas Pck1 induction by D/C decreased by only ∼50% (5,18) (Fig. 1A–C). The findings illustrate the functional diversity in the hormonal regulation of these two genes, which we set out to explore.
To this end, we combined DNA affinity purification (“pull down”) with MS. We prepared nuclear extracts from the liver or primary hepatocytes of mice that were deprived of food or refed; the liver or cells were treated with various combinations of D/C/I and then incubated with biotin-labeled PCR fragments encompassing the insulin-responsive promoters of Pck1 (Pck1p) (19) or G6pc (G6pcp) (20). After affinity purification on streptavidin beads and gel electrophoresis, we performed nano liquid chromatography–MS to identify bound proteins (21) (Fig. 1D and Supplementary Fig. 1A). We ranked identified TFs using the iBAQ method (22). To determine hormone responsiveness, we established an arbitrary iBAQ ratio (>1.5 or <0.5) for use when comparing intact liver from refed mice with liver from mice deprived of food; we used D/C versus the vehicle and D/C/I versus D/C when comparing primary hepatocytes. We divided hormone-regulated TFs into those enriched by hormone treatment (iBAQ ratio >1.5) and those depleted by hormone treatment (iBAQ ratio <0.5). We excluded from the analysis TFs with iBAQ ratios >0.5 and <1.5. Further analyses determined common TFs found in both promoters in primary hepatocytes and in liver, as well as promoter-specific TFs (Fig. 1E). Finally, we performed functional studies of primary hepatocytes to validate FoxK1, a TF candidate that emerged from this stepwise analysis (Fig. 1E).
To test the method, we first used Western blotting to examine candidate TFs in cytoplasmic and nuclear extracts of primary hepatocytes. Consistent with the literature, we found that D/C treatment induced time-dependent nuclear entry of GR (23,24), whereas insulin promoted the nuclear exit of FoxO1 (25) (Supplementary Fig. 1B). Silver staining of polyacrylamide gels demonstrated enrichment of the initial material (Supplementary Fig. 2A). Accordingly, we detected FoxO1 and GR on Western blots of Pck1p pull-down samples (Supplementary Fig. 1C). These data indicate the feasibility of this method for use in identifying hormone-regulated TFs. The number of total and unique peptides identified was comparable between and reproducible in all MS determinations (Fig. 1F, Supplementary Fig. 2C and D, and Supplementary Table 1). TFs accounted for ∼5% of all nuclear proteins identified, and 100–150 proteins were included in each MS experiment (Fig. 1G) in liver and in primary hepatocytes.
Feeding-Regulated Pck1p and G6pcp TFs in Liver
We first performed MS once with hepatic nuclear extracts from mice deprived of food and from refed mice. As a control, we showed that depriving mice of food promoted FoxO1 binding to Pck1p and G6pcp, whereas refeeding inhibited it, with a stronger effect on G6pcp than on Pck1p (Fig. 2A). Heat maps of TFs identified by MS showed extensive similarities between the two promoters (Fig. 2B). Indeed, we identified 128 common TFs for G6pcp and Pck1p, 30 TFs selective to Pck1p, and 11 selective to G6pcp (Fig. 2C and Supplementary Table 2). Analysis of TFs found on both promoters in the Kyoto Encyclopedia of Genes and Genomes highlighted pathways associated with metabolic diseases such as maturity-onset diabetes of the young, circadian rhythm, nonalcoholic fatty liver disease, hepatocellular carcinoma, and insulin resistance (Supplementary Table 6). Analysis of Pck1p-selective TFs in the Kyoto Encyclopedia of Genes and Genomes identified insulin resistance, transcriptional dysregulation in cancer, and cellular senescence as top functional pathways (Supplementary Table 2).
Next we determined which of these TFs were regulated by refeeding (Supplementary Table 2). Refeeding enriched 95 TFs on G6pcp and 92 on Pck1p; this accounts for 60% of the totals. Of these TFs, 63 were common between the two promoters. Refeeding depleted 15 TFs from G6pcp and 23 from Pck1p (Fig. 2D). Eight of these were common to both promoters. These data begin to explain differences in the responses of the two promoters to insulin: about one-third of refeeding-regulated TFs are selective to G6pcp or Pck1p. It is notable that refeeding results in an overall enrichment of TFs bound to the two promoters, consistent with active repression by insulin rather than—or in addition to—clearance of activators. In addition, LXRβ (NR1H2) and peroxisome proliferator–activated receptor α (PPARα) stand out as novel candidates among Pck1p-specific TFs; transcription factor 7-like 2 (TCF7L2), the premier genetic marker of susceptibility to diabetes, stands out as a shared TF (Supplementary Table 2).
Hormone-Responsive Pck1p and G6pcp TFs in Primary Hepatocytes
Next we performed similar experiments in primary hepatocytes in order to exclude the contributions from other cell types present in whole liver and to fine-tune the hormonal response by comparing three states: basal (no hormone), D/C (to represent a fasted state), and D/C/I (to represent a fed state) (Supplementary Fig. 2 and Supplementary Table 1). Western blotting of pull-down samples from primary hepatocytes demonstrated that D/C increased the binding of FoxO1 to Pck1 and G6pc, whereas insulin decreased it, as expected (Fig. 3A). Heat maps of TF profiles identified in two separate experiments illustrate that TF decoration of Pck1p and G6pcp differs substantially (Fig. 3B). From three independent MS experiments performed for each promoter, we identified 79 common TFs on G6pcp and 82 on Pck1p (Fig. 3C and Supplementary Table 3). A total of 54 TFs were common to both promoters, 25 bound selectively to G6pcp, and 28 bound selectively to Pck1p (Fig. 3C and Supplementary Table 3). Pathway analysis of the 54 shared TFs identified three disease processes: maturity-onset diabetes of the young, hepatocellular carcinoma, and thermogenesis; this is consistent with the results of the liver study (Supplementary Table 3). Of the TFs shared between the two promoters, 46 were also among those TFs common to both promoters in the whole liver (Fig. 3C and Supplementary Table 3).
The number of TFs detected in each experiment and their regulation by D/C remained constant in the various experiments (Supplementary Fig. 3A and B). On average, ∼40% of TFs (26 for G6pcp and 43 for Pck1p) were affected by D/C treatment (Supplementary Table 4). Six TFs became enriched at both promoters in response to D/C: ARID5B, CEBPB, NFIL3, MTA, NRF1, and SALL1 (Fig. 3D and Table 1). In contrast, JUNB and SMAD3 were the only two common TFs depleted by D/C treatment (Table 1).
|Pck1p TFs .||G6pcp TFs .||Common TFs .|
|D/C-regulated .||.||Insulin- regulated .||IRS-dependent .||D/C-regulated .||Insulin-regulated .||D/C-regulated .||Insulin-regulated .|
|Pck1p TFs .||G6pcp TFs .||Common TFs .|
|D/C-regulated .||.||Insulin- regulated .||IRS-dependent .||D/C-regulated .||Insulin-regulated .||D/C-regulated .||Insulin-regulated .|
Hormone-enriched TFs are italicized, and hormone-depleted TFs are underlined. TFs also found in the liver of refed animals are set in boldface type.
Next we sorted G6pcp and Pck1p TFs on the basis of their response to insulin. The percentage of insulin-responsive TFs was constant at ∼30% in each G6pcp experiment, whereas it varied more widely among Pck1p experiments (Supplementary Fig. 3C and D). We found 14 insulin-regulated TFs on G6pcp and 33 on Pck1p. Only two were common to both: BPTF and SMAD3; the others were promoter-specific (Table 1 and Fig. 3E). Liver and primary hepatocytes had 13 insulin-responsive TFs in common (Table 1 and Supplementary Table 2).
The insulin response sequence (IRS) on Pck1p (–416 to –402 bp from the transcription start site) mediates insulin inhibition of Pck1 (19). To pare down the list of insulin-regulated TFs on Pck1p, we performed pull-down experiments using a Pck1pΔΙRS mutant (Supplementary Table 1). Whereas p-RNA Pol II bound equally to wild-type and ΔIRS mutant Pck1p, FoxO1 bound to wild-type Pck1p but not ΔIRS Pck1p (Fig. 3F); this is consistent with a critical role of this sequence in insulin’s regulation of this promoter. Comparison of TFs detected by using wild-type and ΔIRS Pck1p identified 12 TFs that require an intact IRS to bind to Pck1p in an insulin-dependent manner (Fig. 3G, Table 1, and Supplementary Table 5). This list includes two forkhead proteins, FoxA2 and FoxK1.
In summary, MS pull down from liver and primary hepatocytes identified an array of Pck1 and G6pc TFs, but only a fraction of them were hormone-responsive: in whole liver, ∼60% were affected by refeeding; in primary hepatocytes, ∼40% were affected by D/C and ∼30%, by D/C/I. Interestingly, in intact liver the two promoters have in common the majority of TFs (∼70%), whereas in primary hepatocytes the promoters share only 5–10% of the TFs. Nine insulin-responsive TFs are bound to Pck1p in both intact liver and primary hepatocytes, and four of them—FoxA2, FoxK1, ZFP91, and ZHX3—depend on an intact Pck1p IRS sequence for binding (Table 1).
Functional Characterization of FoxK1
The forkhead protein FoxK1 displayed features of an insulin-regulated Pck1p TF in primary hepatocytes and intact liver. Multiple analyses of DNA pull down and MS indicate that insulin increased its recruitment to Pck1p more than threefold in primary hepatocytes, and refeeding increased it approximately threefold in intact liver (Fig. 4A). Binding of FoxK1 required an intact Pck1p IRS (Table 1). FoxK1 is consistently associated with Pck1p, but we also detected it on G6pcp in a single experiment with primary hepatocytes (Fig. 4B). Western blotting of cytoplasmic and nuclear extracts from primary hepatocytes treated with D/C/I showed that insulin increased nuclear accumulation of FoxK1, depleting it from the cytoplasm (Fig. 4C). We detected an enrichment of FoxK1 in Pck1p, but not in Pck1pΔIRS pull-down samples, after they were treated with insulin (Fig. 4D). These findings confirm the MS data and indicate that insulin regulates the binding of FoxK1 to Pck1p through the IRS. A 3xIRS luciferase assay revealed that both FoxK1 and FoxO1 bind to IRS and increase luciferase activity (Supplementary Fig. 4). Interestingly, when cotransfected with FoxO1, FoxK1 dose-dependently inhibited FoxO1’s activity. The effect was abrogated by deleting the 40-amino-acid SIN3-interacting domain at the N-terminal end of FoxK1 (Supplementary Fig. 4). These data suggest that FoxK1 competes with FoxO1 to bind to the IRS and that the N-terminal domain is required for FoxK1’s repressor function. Furthermore, the effect of insulin on FoxK1 was not due to increased transcription (at least in vitro), as Foxk1 mRNA levels actually decreased by ∼50% in primary hepatocytes treated with insulin (Fig. 4E).
To document the in vivo function of FoxK1, we performed Western blotting on proteins isolated from DNA pull-down experiments; we detected FoxK1 on both Pck1p and G6pcp in the livers of refed mice (Fig. 4F). Next we immunostained liver sections from mice deprived of food and refed mice. When mice were deprived of food, FoxK1 localized to both the nucleus and the cytoplasm, whereas refeeding induced its accumulation in the nucleus (Fig. 4G). To confirm further the binding of FoxK1 to the promoters of these genes, we conducted ChIP–qPCR on liver samples using the FoxK1 antibody. We found that refeeding or insulin greatly increased FoxK1 binding to the Pck1 promoter (−509 to −393 bp) and the G6pc promoter (−266 to −135 bp) (Fig. 4H and Supplementary Fig. 5). However, only insulin robustly promoted FoxK1 binding. Moreover, the Pck1 promoter luciferase assay revealed that FoxK1 inhibits Pck1 transcription (Fig. 4I). Removing 40 amino acids from the N-terminus of FoxK1, corresponding to the SIN3-interacting domain, fully restored Pck1 transcription. These data confirm that FoxK1 is an insulin-regulated TF in vivo and may function as a transcription repressor in controlling Pck1 and G6pc expression.
FoxK1 Regulates Hepatocyte Glucose Production
The studies described above demonstrate that insulin and refeeding promote the binding of FoxK1 to Pck1p and G6pcp. Next we explored the role of FoxK1 in glucose production by primary hepatocytes. By transfecting two siRNAs, we reduced FoxK1 levels by ∼80% (Fig. 5A). Glucose production assays showed that FoxK1 knockdown not only blunted insulin’s ability to suppress this process but also partly reduced its induction by D/C (Fig. 5B). It should be noted that some FoxK1 exists in the nuclei of D/C-treated cells. Accordingly, D/C failed to induce PCK1 and G6PC protein levels (Fig. 5C) and mRNA levels (Fig. 5D and E). Phosphorylated Akt levels did not change, indicating that the effect of FoxK1 is not due to increased insulin signaling; we did observe slight decreases of FoxO1 and PGC1α protein levels (Fig. 5C), which might contribute to the effects on G6PC and PCK1. qPCR analysis confirmed that FoxK1 knockdown impaired D/C-induced Ppargc1a and Fbp1 expression without affecting Foxo1 or Gck (Fig. 5D–I). These data are consistent with a role of FoxK1 in the inhibition of glucose production by insulin and in integrating signals from counterregulatory hormones. The same siRNA sequences, expressed through an adenoviral vector, failed to lower FoxK1 levels in intact liver, preventing physiologic analysis. On the basis of our in vitro studies, however, we expect that FoxK1 ablation in mouse liver will reduce hepatic glucose production and improve glucose tolerance. Conditional knockout animals are being generated and related data will be reported in subsequent publications.
In this study we used MS of peptides isolated by biotin-labeled DNA affinity purification to identify a diverse group of TFs that bind to the proximal Pck1 and G6pc promoters in a hormone-regulated manner. We identified three general and seemingly novel features of this analysis: 1) The repertoire of hormone-regulated TFs is unexpectedly diverse. 2) The patterns in which TF binds to the two test promoters overlap more in intact liver than in primary hepatocytes; in the latter, Pck1 and G6pc seem to be regulated through largely divergent TFs. 3) Insulin treatment preferentially results in increased occupancy by the two promoters, consistent with a model in which the hormone’s primary role is to recruit corepressors rather than to clear activators. This is consistent with the function of FoxK1 (see below) and highlights the unique role of FoxK1 in insulin’s actions on Pck1 and G6pc.
This approach has obvious limitations: 1) It can analyze only short sequences and cannot account for enhancers or long-range interactions. 2) It is intrinsically sensitive to the stability of DNA-protein and protein-protein interactions. 3) Because the target DNA is present in excess, the method may overestimate the actual number of TFs binding to endogenous target promoters; this may explain experimental variations in the detection of TFs. For example, we readily detected FoxO1 through Western blotting of Pck1p and G6pcp pull-down samples, but its detection by MS varied. However, because this feature dovetails with FoxO1 regulation of Pck1 (i.e., it is variable), it strengthens the relevance of these results. With regard to G6pc, the failure of MS to detect FoxO1 can be explained by the presence of genomic sites that are alternatives to those included in our pull-down assays, or by the time course of FoxO1 binding to G6pcp. The variability of the various experiments may also be related to hepatocyte heterogeneity, sensitivity to the treatment steps, abundance of the target protein, or interactions between hepatocytes and other cells such as stellate, Kupffer, and endothelial cells. Systematic functional characterization of the newly identified candidate TFs will address lingering questions.
FoxK1 in Hormone-Regulated Glucose Metabolism
A novel finding of this work is the identification of FoxK1 as an insulin-regulated TF with a role in glucose production by hepatocytes. Among forkhead TFs, FoxK1 is thought to function as a repressor with roles in myogenic differentiation, cell proliferation, and autophagy (26–29). It has been shown to be phosphorylated by mTORC1, promoting its nuclear exclusion (30), and it has been loosely associated with glucose metabolism in liver tumor cell studies (31). To our knowledge, its involvement in insulin action has not been previously disclosed. Interestingly, it seems to act as an insulin-recruited repressor of gene expression. In this regard, it possesses a Sin3A-interaction domain at the N-terminus that is important to its repressor function, possibly by recruiting Sin3a to target promoters, similar to the mechansim of action of FoxO1 (15). In vivo studies to address this question will have to take into account a second isoform, FoxK2, with potentially similar functions (26,32).
New TF Candidates and Therapeutic Targets for Type 2 Diabetes
Our study identified almost all TFs previously associated with the regulation of hepatic glucose metabolism by insulin, such as HNF4α, C/EBPβ, LXRβ, FoxA2, and RBPj (11,18,33,34). This finding increases our confidence that newly identified TFs represent physiologic mediators of hormone signaling, some of which may provide alternative targets of treatment to reduce hepatic glucose production in type 2 diabetes.
LXRβ and PPARα stand out as novel candidates among Pck1p-specific TFs. Given the resurgence of interest in ligands of these nuclear receptors, it is important to keep in mind that our studies suggest that they are specific to Pck1p and that they bind to Pck1p during food deprivation and are depleted upon refeeding. Therefore, antagonists, rather than agonists, may be useful in reducing hepatic glucose production (35,36).
An important new finding of this work is the identification of TCF7L2 as an insulin-regulated Pck1 TF. TCF7L2 is contained within the main type 2 diabetes susceptibility locus identified in multiple genome-wide association studies of different ethnic groups (37,38). Although TCF7L2 had been suggested to affect primarily pancreatic β-cell function (39–41), evidence is not definitive (42). The identification of TCF7L2 as an insulin-regulated TF on Pck1p in primary hepatocytes and liver is consistent with its role in glucose production and provides a new testable hypothesis of the mechanism of diabetes susceptibility linked to this locus.
We found 14 insulin-regulated TFs on G6pcp and 33 on Pck1p. Only two were common to both: BPTF and SMAD3. SMAD3 lies in the transforming growth factor-β pathway and has been implicated in the development of liver fibrosis (43,44). Evidence linking it to hepatic gluconeogenesis through protein phosphatase 2A, AMPK, and FoxO1 is limited (45). Our study indicates that SMAD3 is depleted by D/C and enriched by insulin on Pck1 and G6pc, consistent with a role in suppressing these genes. The TFs newly identified in this study provide candidates that might deconvolute the regulation of hepatic glucose production and may be potential targets for diabetes treatment.
NOTE ADDED IN PROOF
While this paper was under review, two additional articles describing a metabolic role of Foxk1 (and Foxk2) appeared: Sakaguchi M, Cai W, Wang CH, et al. FoxK1 and FoxK2 in insulin regulation of cellular and mitochondrial metabolism. Nat Commun 2019;10:1582; and Sukonina V, Ma H, Zhang W, et al. FOXK1 and FOXK2 regulate aerobic glycolysis. Nature 2019;566:279–283.
Acknowledgments. The authors thank KyeongJin Kim and Junjie Yu, Columbia University Medical Center, for helpful discussions.
Funding. This work was supported by New York Obesity Research Center Training Grant T32-DK755925 (to L.W.) and through National Institutes of Health and National Institute of Diabetes and Digestive and Kidney Diseases grants DK57539 (to D.A.) and DK63618 (to the Columbia Diabetes Research Center).
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. L.W. designed and performed all experiments and wrote the manuscript. Q.L. and J.H. analyzed the mass spectrometry data. T.K. performed the ChIP-qPCR experiment. J.Q. and D.A. conceived the experimental design, analyzed and discussed the data, and wrote the manuscript. D.A. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Data Availability. The data supporting the findings of this study are available within the article and the Supplementary Data, or they are available upon reasonable request to the authors. The research resources mentioned in this study are also available upon request.