HGFAC is a ChREBP-regulated hepatokine that enhances glucose and lipid homeostasis

Carbohydrate response element–binding protein (ChREBP) is a carbohydrate-sensing transcription factor that regulates both adaptive and maladaptive genomic responses in coordination of systemic fuel homeostasis. Genetic variants in the ChREBP locus associate with diverse metabolic traits in humans, including circulating lipids. To identify novel ChREBP-regulated hepatokines that contribute to its systemic metabolic effects, we integrated ChREBP ChIP-Seq analysis in mouse liver with human genetic and genomic data for lipid traits and identified hepatocyte growth factor activator (HGFAC) as a promising ChREBP-regulated candidate in mice and humans. HGFAC is a protease that activates the pleiotropic hormone hepatocyte growth factor. We demonstrate that HGFAC-KO mice had phenotypes concordant with putative loss-of-function variants in human HGFAC. Moreover, in gain- and loss-of-function genetic mouse models, we demonstrate that HGFAC enhanced lipid and glucose homeostasis, which may be mediated in part through actions to activate hepatic PPARγ activity. Together, our studies show that ChREBP mediated an adaptive response to overnutrition via activation of HGFAC in the liver to preserve glucose and lipid homeostasis.


Introduction
Carbohydrate Responsive Element Binding Protein (ChREBP, also known as Mlxipl) is a transcription factor expressed in key metabolic tissues including liver, adipose tissue, kidney, small intestine, and pancreatic islets (1, 2). It is activated by sugar metabolites, and in the liver and small intestine, its activity is robustly increased by consumption of sugars containing fructose (3,4). Upon activation, ChREBP stimulates expression of genomic programs that contribute to both adaptive and maladaptive metabolic responses (1). Hepatic ChREBP activity is increased in human obesity and diabetes as indicated by increased expression of the potent ChREBP-beta isoform (5,6). Knockdown or knockout of hepatic ChREBP protects against dietinduced and genetic forms of obesity and accompanying metabolic disease (3,7,8).
ChREBP plays a significant role in human metabolic physiology as common genetic variants in the ChREBP locus associate with pleotropic anthropomorphic and metabolic traits at genome-wide significance including circulating lipids and cholesterol, BMI, waist-hip ratio, height, diverse hematological parameters, serum urate, liver enzymes, and blood pressure (9). However, the complement of ChREBP transcriptional targets that participate in regulating these diverse traits is incompletely understood. To date, thousands of genes have been proposed as putative ChREBP targets via Chromatin Immunoprecipitation-sequencing (ChIP-seq) assays and global gene expression analysis, a minority of which are known to be involved in metabolic processes (10,11). For example, it is well established that ChREBP regulates glycolysis and fructolysis, hepatic and adipose lipogenesis, and hepatic glucose production via regulation of key enzymes involved in these metabolic pathways (4,(12)(13)(14). At the same time, a large number of putative ChREBP transcriptional targets have either no known or poorly defined function and uncertain impact on metabolism.
Here we performed ChIP-seq for ChREBP in mouse liver and integrated this with human genetic data to identify novel ChREBP-dependent hepatokines that might participate in the regulation of systemic metabolism. Through this screen we identified Hepatocyte Growth Factor Activator (HGFAC) as a promising candidate. HGFAC is a liver-secreted, circulating protease that activates Hepatocyte Growth Factor (HGF) which has pleiotropic biological activities including regulation of morphogenesis, cell migration, transition between cell states, and proliferation in epithelial and other cell types throughout the body (15)(16)(17). Here, we demonstrate that HGFAC is indeed nutritionally regulated in a ChREBP-dependent manner and participates in an adaptive response to preserve carbohydrate and lipid homeostasis.

HGFAC is a ChREBP genomic target that associates with metabolic traits in humans
To identify ChREBP transcriptional targets that participate in the regulation of ChREBP associated metabolic programs and phenotypes, we performed ChIP-seq analysis for ChREBP in livers of two strains of male mice gavaged with either water or fructose. We identified 4,860 distinct genomic sites enriched for ChREBP binding (Supplementary Table 1) which include well-defined loci in canonical ChREBP targets involved in glycolysis, glucose production, fructolysis, and lipogenesis such as liver pyruvate kinase (PKLR), glucose-6-phosphatase (G6PC), fatty acid synthase (FASN), and ketohexokinase (KHK), respectively ( Figure 1A). Although fructose gavage can acutely induce ChREBP-dependent changes in gene expression, ChREBP ChIP-seq peaks were readily detectable in fasted mice, and fructose gavage did not enhance ChREBP ChIP-seq peak height even at a liberal false discovery rate of 0.20. This indicates that increased chromatin occupancy is not essential for fructose to induce ChREBP-dependent gene transcription. Most ChREBP ChIP peaks occurred within 10 kb of transcriptional start sites ( Figure 1B). Consistent with ChREBP's known functions, Genomic Region Enrichment Analysis (GREAT) of putative ChREBP binding sites demonstrated enrichment for numerous metabolic processes including carbohydrate and lipid metabolism ( Figure 1C) (18).
Variants in the ChREBP locus are strongly associated with hypertriglyceridemia in human populations (19,20). However, the complement of ChREBP transcriptional targets that mediate its effects on circulating lipids is uncertain. We sought to determine whether genomic loci containing human homologues of mouse ChREBP target genes are enriched for variants that associate with hypertriglyceridemia in human populations.
Via Meta-Analysis of Gene-set ENrichmenT of variant Associations (MAGENTA), we confirmed that loci in proximity to human homologues of mouse genes that are within 20 kb of ChREBP binding sites are enriched for SNPs that associate with hypertriglyceridemia in humans (Adjusted P-val = 0.003). 87 loci/genes contributed to this enrichment with an FDR of 0.05 (Table 1 and Supplementary Table 2) (21). This list includes known ChREBP transcriptional targets such as GCKR, TM6SF2, KHK, and ChREBP (MLXIPL) itself, all previously implicated in regulating carbohydrate and triglyceride metabolism. Of these 87 loci, seven encoded putative secretory proteins including several lipoproteins (APOC2, APOE, and APOA5), VEGFA which is most wellknown for its role in angiogenesis, but also implicated in metabolic control, and HGFAC (Supplementary Table 2) (22). To our knowledge, HGFAC has not been identified as a ChREBP transcriptional target nor studied extensively in the context of systemic fuel metabolism.
HGFAC is a serine protease expressed predominately in hepatocytes and secreted as a zymogen into circulation where it is found in a single chain pro-HGFAC form (23,24). In-vitro studies have identified thrombin and kallikrein-related peptidases KLK-4 and KLK-5 to be potent activators of pro-HGFAC (25,26). Once activated, HGFAC cleaves and activates Hepatocyte Growth Factor (HGF) which can then bind and activate the c-Met receptor tyrosine kinase (MET) (23). HGF and c-MET have pleiotropic biological activities as mitogens and motogens in organogenesis, tissue repair, and cell migration, and also function as anti-inflammatory, apoptotic, and cytoprotective signals depending on the context (15). Variants in c-MET also associate with circulating triglycerides at genome wide significance in humans consistent with a potential role for HGFAC in regulating triglyceride levels through activation of HGF (27).
Moreover, increased levels of circulating HGF in people associate with features of cardiometabolic disease including obesity, risk for type 2 diabetes, and risk for cardiovascular disease (28)(29)(30)(31). Circulating HGF levels are influenced by variants in the HGFAC locus (32). A missense variant in HGFAC, rs3748034, that associates with increased circulating HGF also associates with increased circulating triglycerides in GWAS aggregate data at genome wide significance (beta = 0.0302, p < 5e-28) as well as other cardiometabolic risk factors and pleiotropic biological traits (33,34). The Ala218Ser mutation encoded by rs3748034 is predicated to be "possibly damaging" by PolyPhen-2 (35). Furthermore, another missense variant in HGFAC, rs16844401, that associates with increased circulating triglycerides also associates with increased coronary artery disease risk (36). These associations motivated further investigation to determine whether ChREBP regulates HGFAC expression and whether this interacts with nutritional status to regulate systemic fuel metabolism and cardiometabolic risk factors.

Nutritional regulation of HGFAC is ChREBP-dependent.
ChREBP activity in the liver is responsive to diets high in fructose. To examine the role of hepatic ChREBP in the regulation of HGFAC in rodents, we measured hepatic Hgfac mRNA and HFGAC protein in the liver and plasma of mice with liver specific deletion of ChREBP after 8 weeks on high fructose (HFrD) or control diet. High fructose feeding increased hepatic Hgfac mRNA expression 1.7-fold (p<.0001) in control mice, and this induction was abrogated in liver-specific ChREBP KO mice (LiChKO) (Figure 2A). Fructose-induced increases in hepatic Hgfac mRNA expression were accompanied by 4-and 2-fold increases in hepatic and circulating pro-HGFAC protein levels ( Figure 2B). Basal liver and circulating HGFAC protein levels tended to be decreased in chow fed LiChKO mice and were not induced with fructose feeding.
Circulating HGFAC also increased in mice fed a high fat/high-sucrose (HF/HS) diet and in genetically obese Zucker fatty rats on chow diet (Supplementary Figure 1), where hepatic ChREBP activity is also robustly increased independently of an obesogenic diet (37). Altogether, these data show that hepatic ChREBP mediates diet and obesity induced increases in circulating HGFAC.
Next, we sought to determine whether ChREBP-mediated regulation of HGFAC might be conserved in humans. To that end, we analyzed hepatic mRNA expression levels of HGFAC and other ChREBP transcriptional targets in the Genotype-Tissue Expression (GTEx) Biobank (38). Expression of the potent ChREBP-β isoform is an excellent surrogate marker of tissue ChREBP activity (14). However, it is expressed at low levels which are typically below the sequencing depth of most RNA-seq experiments. Consistent with this, GTEX RNA-seq data does not distinguish between ChREBP-β and -α isoforms. Due to the lack of ChREBP-β specific expression data, we used a composite expression vector comprised of 5 well-validated ChREBP target genes (FASN, PKLR, KHK, ALDOB, and SLC2A2) and found that this composite vector strongly correlates with the expression of HGFAC (Pearson correlation r 2 =0.44, p <0.0001) ( Figure 2C). Transcription factor enrichment analysis of the 5% of hepatic genes that best correlated with hepatic HGFAC expression in the GTEx Biobank showed strong enrichment for genes co-expressed with ChREBP (Adjusted P-val = 1.75E-25) indicating conservation of the ChREBP-mediated regulation of HGFAC in humans ( Figure 2D) (39). Additionally, hepatic HGFAC mRNA expression is upregulated in patients with obesity and uncontrolled diabetes ( Figure 2E), conditions that are associated with increased hepatic ChREBP activity (5,40). Collectively, these data support the hypothesis that hepatic HGFAC expression and circulating levels of HGFAC are regulated by ChREBP activity both in rodents and in humans, and hepatic HGFAC expression is increased in obesity and diabetes.

Murine Hgfac KO recapitulates the phenotype of putative human LOF HGFAC variants.
To study the roles of HGFAC in systemic metabolism, we generated whole body Hgfac KO mice that lack a portion of exon 1 and all of exon 2 ( Figure 3A). The deletion was confirmed by genomic PCR, by the absence of detectable circulating HGFAC protein, and by quantification of hepatic Hgfac mRNA ( Figure 3B-D). Hgfac KO mice were born at normal Mendelian ratios and did not appear to have any gross abnormalities when compared to their littermate controls. Activated HGFAC activates HGF and c-MET signaling. However, there is redundancy in this system and other proteases including Hepsin (HPN) and coagulation factors XIa and XIIa are also capable of activating HGF (41,42). We sought to determine whether the ability to activate endogenous HGF is impaired in Hgfac KO mice. c-MET signaling was assessed in HepG2 cells incubated with thrombin treated sera obtained from WT and Hgfac KO mice. Thrombin is one of the proteases that is capable of activating HGFAC in-vitro (26). Serum from control mice increased c-MET phosphorylation 1.9-fold when compared to DMEM alone, while this induction was attenuated with serum from KO mice ( Figure 3E). These results demonstrate that sera from Hgfac KO mice has reduced capacity to activate HGF and c-MET signaling.
A putative loss of function variant in HGFAC (rs3748034) strongly associates with increased circulating triglycerides, albumin, and platelets among other traits ( Figure 3F) (33). We determined whether Hgfac KO mice have similar phenotypes.

HGFAC KO Mice develop impaired glucose homeostasis.
To examine the potential role of HGFAC in systemic metabolism, we challenged 8-week-old Hgfac KO mice and their littermate controls with high-fat/high-sucrose (HFHS) diet for 18 weeks. We did not observe any differences in body weight or fat mass during the study ( Figure 4A and B). However, a modest reduction in lean body mass was observed in Hgfac KO mice ( Figure 4C). To assess glucose homeostasis, we performed glucose and glycerol tolerance tests in Hgfac KO mice and controls at time points throughout the study. Glycerol is a preferred gluconeogenic substrate and glycerol tolerance tests reflect hepatic glucose production capacity (43). After 4 weeks on HFHS diet, Hgfac KO mice are glycerol intolerant with a 1.4-fold increase in glycemic excursion (p<0.05) ( Figure 4D). At this age, there was no difference between KO mice and controls with respect to glycemic excursion during a glucose tolerance test ( Figure   4E), suggesting that young Hgfac KO animals may have a selective impairment in hepatic insulin sensitivity. However, after 13 weeks of HFHS diet, Hgfac KO mice developed glucose intolerance with a 1.6-fold increase in incremental AUC (p<0.005) as well as insulin resistance with a 30% decrease in area above the curve (p<0.05), as measured by IP glucose and insulin tolerance tests, respectively (Figure 4F and G).
HGF has been proposed to regulate pancreatic beta-cell development and insulin secretory capacity (44). To test insulin secretory capacity in Hgfac KO mice, we performed an oral mixed meal tolerance test which triggers more robust and sustained insulin secretion compared to IP glucose administration. Basal insulin and glucose levels were no different between Hgfac KO mice and controls ( Figure 4H). At 10 min, insulin levels were 1.6-fold higher in Hgfac KO mice compared to controls (3.37+/-0.48 ng/ml Hgfac KO vs 2.1+/-0.4 ng/ml controls, p<0.05) with only a modest increase in glycemia at this time point. Altogether, these data indicate that Hgfac KO mice subjected to HFHS diet develop early hepatic insulin resistance followed by systemic insulin resistance with intact insulin secretory capacity.
We also examined whether HFHS diet might exacerbate the increase in circulating triglyceride levels observed in chow-fed Hgfac KO mice. In contrast with the data in chow diet, we did not observe differences in circulating triglycerides after 7 weeks of HFHS diet. Similarly, triglyceride levels were not different between Hgfac KO and control after IP administration of poloxamer 407 which inhibits lipoprotein lipase and peripheral triglyceride clearance (Supplementary Figure 3) indicating that VLDL production is similar between genotypes in this dietary context (45).

Hgfac KO Downregulates Hepatic PPARγ Expression.
To define mechanisms that might contribute to altered triglyceride and carbohydrate metabolism in Hgfac KO mice, we performed RNA-seq analysis on liver from chow and HFHS-fed Hgfac KO mice and littermate controls after 4 weeks on diet.
Hgfac was the most significantly downregulated mRNA on both diets, confirming successful KO ( Figure 5A). By pathway enrichment analysis ( Figure 5B), genes involved in cell cycling were the most downregulated set in chow-fed Hgfac KO mice. This is consistent with HGF's known effects to stimulate hepatocyte proliferation (46). Pathway analysis also suggested changes in lipid metabolism with reduced "PPAR signaling pathway" and "Fatty acid degradation" in KO mice on both diets. Upregulation of genes involved in ribosomal function were observed in the Hgfac KO mice potentially consistent with reduced cell cycling and enhanced differentiated function as a result of reduced HGF signaling. Gene sets associated with complement and coagulation pathways were also upregulated in Hgfac KO mice. Upregulation of complement and coagulation pathways is notable as putative loss of function variants in the HGFAC locus also associate with increased circulating fibrinogen levels (47).
Consistent with the pathway analysis, Pparg was in the top 10 most differentially expressed genes comparing chow-fed HGFAC KO mice and controls (Supplementary Table 3). To confirm this, we quantified hepatic mRNA gene expression by qPCR which revealed that Pparg but not Ppara is downregulated in livers of chow and HFHSfed Hgfac KO mice compared to controls ( Figure 5C). Furthermore, PPARγ target genes were also downregulated. These results were replicated in a second cohort (Supplementary Figure 4). Hepatic PPARγ is reported to enhance liver fat accretion yet preserve hepatic and systemic insulin sensitivity (48,49). HFHS-feeding increased the levels of hepatic triglycerides by 49% and 34% in Hgfac KO mice and controls, respectively ( Figure 5D). However, hepatic triglyceride levels were reduced by 40% and  (48,49). This may be in part mediated by the effects of PPARγ on hepatic PDHA activity.

HGFAC overexpression enhances glucose homeostasis.
As HGFAC deficiency decreased expression of hepatic Pparg and its targets, we examined whether HGFAC overexpression has reciprocal molecular and metabolic effects. Adenoviral (ADV) mediated overexpression of HGFAC resulted in a robust increase of circulating HGFAC over a period of two weeks compared to ADV-GFP controls ( Figure 6A). This was associated with markedly improved glucose tolerance with a 30% reduction in incremental AUC (p<0.005) ( Figure 6B) and a 50% reduction in glycemic excursion during a glycerol tolerance test performed in second cohort (p<0.0005) (Supplementary Figure 5A). Additionally, fasting glucose levels were slightly but significantly lower in ADV-HGFAC mice in fed and fasted conditions (Supplementary Figure 5B). Analysis of hepatic gene expression revealed that HGFAC overexpression induced expression of Pparg but not Ppara, as well as PPARγ target genes such as Cd36 and Fabp4 as well as Pdk4 which may participate in regulation of PDHA phosphorylation ( Figure 6C). Furthermore, HGFAC overexpression increased phosphorylation of PDHA (S293) as well as Proliferating Cell Nuclear Antigen (PCNA) levels, indicating increased proliferation ( Figure 6D). Whereas short-term overexpression of HGFAC was sufficient to produce glycemic and gene expression phenotypes reciprocal to Hgfac KO, we did not observe changes in hepatic or circulating triglyceride levels in this time frame ( Figure 6E). Thus, HGFAC overexpression can induce changes in hepatic PPARγ expression and glucose homeostasis independently of its effects on hepatic lipids.
To assess whether HGFAC's effect to induce Pparg expression is likely mediated through its ability to activate HGF and c-MET signaling, we treated murine AML12 hepatocyte-like cells with recombinant, active HGF. HGF treatment increased c-MET phosphorylation and increased Pparg mRNA expression by 30% ( Figure 6F). These effects were inhibited by pre-treatment with PHA665752, a c-MET inhibitor ( Figure 6G) (51). Altogether, these results support a model whereby overnutrition enhances ChREBP-dependent upregulation of HGFAC which activates an HGF-PPARγ signaling axis to preserve systemic glucose homeostasis.

Discussion:
ChREBP is a critical nutrient sensing transcription factor that is activated by cellular carbohydrate metabolites and mediates genomic and physiological responses to overnutrition in key metabolic tissues including the liver. The precise mechanisms by which carbohydrates active ChREBP remain controversial (reviewed in (1)). Putative mechanisms include carbohydrate mediated translocation of ChREBP protein from the cytosol to the nucleus, alterations in ChREBP post-translational modifications, and/or allosteric effects of specific carbohydrate metabolites on ChREBP itself to enhance transactivation. We previously demonstrated that fructose gavage acutely and robustly activates ChREBP-dependent gene expression in mouse liver following short-term fasting (4). Here, we performed ChIP-seq for ChREBP following fructose gavage after a 5 hour fast to map the induction of ChREBP binding in mouse liver chromatin. We identified ~ 4000 ChREBP binding sites in livers from two mouse strains that appears similar to previous efforts (10). To our surprise, chromatin-bound ChREBP was readily detectable in fasted animals and no significant increase in binding was observed following fructose gavage. These results suggest that while fructose can acutely activate ChREBP-dependent gene transcription, this transactivation is achieved by ChREBP that is already bound to chromatin and indicates that carbohydrate-stimulated nuclear translocation and accumulation of nuclear ChREBP is not essential for the ability of carbohydrates to enhance ChREBP's transcriptional activity. These results favor models suggesting that either carbohydrate-mediated post-translational modification or allosteric activation are key mechanisms that might stimulate ChREBP's transcriptional activity.
Variants in the human ChREBP locus associate with pleiotropic biological traits at genome-wide or near genome-wide significance with a particularly strong association with hypertriglyceridemia. The transcriptional targets that mediate ChREBP's pleiotropic biological effects remain incompletely defined. By mapping ChREBP genomic binding sites in mouse liver and integrating this with human genetics data, we identified candidate transcriptional targets that might contribute to ChREBP-mediated regulation of circulating lipids. While genes and loci in proximity to ChREBP binding sites were enriched for variants that associated with hypertriglyceridemia, of the thousands of hepatic ChREBP binding sites, in this analysis, only ~ 2% of such sites contributed to the enrichment. We anticipate that a relatively small and partially overlapping subsets of ChREBP gene targets may contribute to its regulation of other metabolic traits.
Of the candidate genes identified here, a small minority comprised liver derived circulating factors or "hepatokines" that might regulate metabolism systemically. We elected to focus further attention on HGFAC as a putative ChREBP-regulated hepatokine and demonstrated that circulating HGFAC is indeed nutritionally regulated in a ChREBP-dependent manner. Moreover, we have shown that it participates in an adaptive metabolic response to obesogenic diets in part through its effects to stimulate hepatic Pparg expression and transcriptional activity (Figure 7).
To test the role of HGFAC in metabolism, we used a Crispr/Cas9 strategy to generate global Hgfac knockout mice. The ability of Hgfac KO serum to activate HGF and facilitate c-MET signaling was impaired. That this activity was attenuated, but not fully abrogated is consistent with known redundancy in enzymes capable of HGF activation (42,52). Alternative proteases including kallikreins, urokinases, matriptase, and hepsin may partially compensate for loss of HGFAC activity (52)(53)(54)(55). While the vast majority of HGFAC expression is in the liver, it is also expressed at much lower levels in other tissues including the intestines and possibly the pancreatic islets (23,24). Therefore, we cannot exclude specific roles for HGFAC outside of the liver in this study.
The reduction in HGFAC activity and attenuation in HGF signaling did indeed produce metabolic phenotypes. We observed that Hgfac KO reduced and increased hepatic and circulating triglyceride, respectively and this was associated with impaired hepatic and systemic glucose tolerance. ADV-HGFAC overexpression produced a reciprocal phenotype with respect to glucose homeostasis but did not alter liver or circulating lipids in the short time frame of this experiment. Our results contrast with reported effects of acute treatment with recombinant, active HGF in rodents to reduce steatosis and with inconsistent effects on circulating triglycerides (56,57). Additionally, marked and sustained transgenic overexpression of HGF under a metallothionein promoter also reduced steatosis in contrast with our observations (58). The differences observed in these publications and our experiments may be due to differences in gainversus loss-of-function experiments, differential effects in acute versus chronic paradigms, and the degree of changes in HGF activity and signaling.
The specific mechanism by which pro-HGF is activated, either by HGFAC versus other proteases, also appears to have marked impact on where HGF signaling may be enhanced and on the resultant systemic metabolic effects. As an example, hepsin (HPN) is a membrane bound protease expressed in multiple tissues that is also capable of HGF activation. Hepsin KO which also reduces HGF-c-MET signaling produces a vastly different metabolic phenotype compared with Hgfac KO mice. Global hepsin KO mice are resistant to diet induced obesity and this lean phenotype is associated with enhanced glucose and lipid homeostasis (59). Profound changes in energy homeostasis in hepsin KO mice and its lean phenotype appear to be due to extensive expansion of brown fat and increased thermogenesis, features which we did not observe in Hgfac KO mice.
Proteases including HGFAC and HPN are promiscuous and may activate other peptide hormones which may also contribute to their differing biological effects. For instance, HGFAC can also cleave and activate pro-macrophage stimulating protein Our results show that the ChREBP-HGFAC axis regulates hepatic PPARγ signaling in mice. We further validated this observation by showing that HGF treatment can increase Pparg expression in hepatocyte-like AML12 cells and this can be blocked by a c-MET inhibitor. While the metabolic role of PPARγ is most well recognized with respect to adipogenesis, hepatic PPARγ also appears important in regulating systemic metabolism (62)(63)(64). Liver-specific deletion of Pparg reduces steatosis, but leads to hypertriglyceridemia and glucose intolerance associated with muscle and adipose insulin resistance (48). While the beneficial effects of hepatic PPARγ have been attributed to its effects on reducing circulating lipids, recent work has demonstrated that the PPARγ agonist pioglitazone enhances hepatic insulin sensitivity independently of its effects on hepatic lipids and is instead dependent on PPARγ's ability to inhibit hepatic pyruvate dehydrogenase activity (50). Data from HGFAC KO mice are consistent with this hypothesis in that decreased PPARγ activity is accompanied by a reduction in inhibitory phosphorylation of the PDH catalytic subunit on Ser293. Adenoviral overexpression of HGFAC led to marked improvement in glucose tolerance with increased hepatic PPARγ expression and increased phosphorylation of PDH consistent with this model. Altogether, these results indicate that HGF and PPARγ may mediate some of its effects on glucose homeostasis through regulation of hepatic PDH phosphorylation.
Putative loss of function variants in human HGFAC strongly associate with increased circulating triglycerides, albumin, and platelets and these phenotypes are recapitulated in Hgfac KO mice (34). This concordance supports the hypothesis that putative HGFAC loss of function variants likely impair its catalytic activity. Moreover, these results suggest that this molecular physiology is conserved from rodents to humans. Interestingly, the rs1801282 (Pro12Ala) PPARG variant associated with increased PPARG expression and reduced risk for diabetes and circulating triglycerides also associates with reduced albumin levels (65). These effects on albumin are directionally concordant with the changes in albumin that occur in HGFAC KO mice and the reduction in hepatic Pparg. Again, this suggests that an HGF-PPARγ signaling axis is conserved in humans and that some of the beneficial effects of PPARγ on systemic metabolism could be mediated through effects in the liver in addition to adipose tissue.
Our results suggest an integrated physiology whereby carbohydrate sensing via ChREBP impacts systemic growth factor signaling (HGFAC-HGF-MET) that may mediate both adaptive and maladaptive responses through paracrine and endocrine effects. In the context of obesogenic diets, this signaling axis enhances hepatic PPARG expression which may mediate a compensatory response to preserve systemic glucose homeostasis. HGF, the principal target for HGFAC has previously been implicated in other aspects of glucose homeostasis. For example, HGF may enhance pancreatic beta cell proliferation (44,(66)(67)(68). Increased ChREBP-mediated HGFAC secretion might be a potential mechanism to increase beta cell mass in the setting of increased dietary carbohydrate burden. Additionally, within the liver, HGF has been reported to enhance insulin signaling and hepatic glucose clearance via physical interactions between its receptor, c-MET, and the insulin receptor (69). HGF also is secreted by adipocytes and can promote angiogenesis in adipose tissue and adipose angiogenesis is an integral feature of adipose tissue expansion (70)(71)(72). Therefore, elevated ChREBP-HGFAC-HGF may promote healthy expansion of adipose tissue for efficient storage of fuel during overnutrition. These observations may support a role for ChREBP mediated upregulation of HGFAC and HGF signaling as an adaptive response to increased nutritional burden, and will require further investigation. ChREBP itself has been shown to regulate mouse hepatocyte and murine and human beta cell proliferation (73)(74)(75). Putative loss of function variants in HGFAC associate with increased circulating HGF in humans and also associate with increased cardiovascular risk factors (32,34).
Increased circulating HGF itself is increasingly recognized as a cardiometabolic risk factor that may be independent of other canonical cardiovascular risk factors (30,32,(84)(85)(86). Further investigation into the relationship between ChREBP, HGFAC, and HGF signaling may define new mechanisms contributing to the pathogenesis of cardiometabolic disease in humans.

ChIP-seq and Analysis.
Wild-type, male 8-week old C3H/HeJ and C57BL/6J mice were fasted for 5 hours and gavaged with fructose (4 g /kg BW) versus water control (n=6 / group). Mice were euthanized 90 min after gavage and tissues were harvested and snap-frozen in liquid nitrogen for further analysis. Chromatin was prepared using truChIP Chromatin Shearing Tissue Kit (Covaris) according to the manufacture's protocol with modifications. Briefly, 25-30 mg of frozen liver tissue were quickly minced with razor blades in PBS at room temperature. Tissue was crosslinked with 0.5 M disuccinimidyl glutarate in PBS for 45 min at room temperature, followed by second fixation with 1% formaldehyde in Fixing Buffer A (Covaris) for 5 min at room temperature. Crosslinking was stopped by Quenching Buffer E (Covaris). After washing, nuclei were isolated by Dounce homogenization followed by centrifugation. After washing and centrifugation, the nuclear pellet was resuspended in cold 0.25% SDS Shearing Buffer (Covaris). Chromatin was sheared in 1 ml AFA milliTUBEs (Covaris) using Covaris S220X focused ultrasonicator with the following parameters: peak incident power 140W, duty factor 5%, cycles per burst 200 for 12 min. The sheared chromatin was centrifuged at 13,000rpm for 10min at 4°C to remove debris, and a 10 ul aliquot was de-crosslinked and used for quantification with Qubit (Thermo Fisher Scientific ChIP-seq reads were demultiplexed using bcl2fastq and aligned to the GRCm38 mouse genome using Bowtie2 (87). PCR duplicates and low-quality reads were removed by Picard. Reads were processed using SAMtools and subjected to peak-calling with MACS2. SAMtools was also used to obtain 2 pseudoreplicates per sample (88,89).
Only the peaks present in both pseudoreplicates were included for further downstream analysis. The coverage for peaks was obtained using BEDtools multicov (90).
Normalization and differential analysis were performed using edgeR between fructose and water gavage conditions (91). To visualize ChIP-seq signals, reads were converted to the BigWig file format using BEDtools and bedGraphToBigWig (92).

HGFAC/HGF activation assay. Blood was collected from 3 WT and 3 HGFAC KO mice
and allowed to clot at room temperature for 1 hour, centrifuged at 7000 G for 15 min and serum was collected. Serum was incubated with 10 ug/ml dextran sulfate and 500 ng/ml of Thrombin 3 hour at 37°C. Activated serum was diluted with DMEM media (1:10). HepG2 cells were treated with this media for 5 min and then harvested.
Activation of c-MET was assessed by immunoblotting. The ADV-GFP control vector has been previously described in (94). Adenoviral vectors were produced and purified as previously described (94 (95). Transcript level count was uploaded to the BioJupies server and analyzed for differential gene expression and KEGG pathway enrichment (96)(97)(98).               Figure 4. HGFAC KO mice have impaired carbohydrate metabolism on HF/HS diet. A) Body weight of male WT and HGFAC KO mice during 18 weeks of HF/HS feeding (n=11-12/group unless otherwise specified) B) Fat and C) lean mass by NMR at 18 weeks. Glucose homeostasis was assessed at intervals throughout the study including D) IP glycerol tolerance test at 4 weeks, E) IP glucose tolerance at 5 weeks, F) IP glucose tolerance test at 13 weeks, G) IP insulin tolerance test at 14 weeks, and H) a mixed meal tolerance test to assess insulin secretion was performed at 16 weeks. Tail vein insulin levels were measured at 0 and 10 minutes. Data represent means ± SEM, Statistics were assessed by two-tailed unpaired t-test, *p<0.05; or two-way ANOVA with Sidak's multiple comparisons between individual groups, ^ p<0.05 for comparison across genotypes within time points, $ p<0.05 for comparison across time points within genotypes.    Glucose and fructose from high sugar diets enhance production of sugar metabolites (hexose-phosphates) in the liver that activate hepatic ChREBP and lead to increased Hgfac transcription and translation. HGFAC is secreted into the circulation where, once activated, it can act in a paracrine or endocrine fashion to proteolytically cleave and activate HGF. HGF binds and activates the c-MET tyrosine kinase receptor on hepatocytes and other cell types. In liver, this leads to upregulation of PPARγ expression that in turn activates transcriptional programs to promote hepatic triglyceride storage and to decreased circulating triglycerides. Additionally, hepatic PPARγ activity decreases activation of the pyruvate dehydrogenase complex and this contributes to enhance systemic glucose tolerance.