Hepatic HAX-1 inactivation prevents metabolic diseases by enhancing mitochondrial activity and bile salt export

Increasing hepatic mitochondrial activity through pyruvate dehydrogenase and elevating enterohepatic bile acid recirculation are promising new approaches for metabolic disease therapy, but neither approach alone can completely ameliorate disease phenotype in high-fat diet–fed mice. This study showed that diet-induced hepatosteatosis, hyperlipidemia, and insulin resistance can be completely prevented in mice with liver-specific HCLS1-associated protein X-1 (HAX-1) inactivation. Mechanistically, we showed that HAX-1 interacts with inositol 1,4,5-trisphosphate receptor-1 (InsP3R1) in the liver, and its absence reduces InsP3R1 levels, thereby improving endoplasmic reticulum–mitochondria calcium homeostasis to prevent excess calcium overload and mitochondrial dysfunction. As a result, HAX-1 ablation activates pyruvate dehydrogenase and increases mitochondria utilization of glucose and fatty acids to prevent hepatosteatosis, hyperlipidemia, and insulin resistance. In contrast to the reduction of InsP3R1 levels, hepatic HAX-1 deficiency increases bile salt exporter protein levels, thereby promoting enterohepatic bile acid recirculation, leading to activation of bile acid–responsive genes in the intestinal ileum to augment insulin sensitivity and of cholesterol transport genes in the liver to suppress hyperlipidemia. The dual mechanisms of increased mitochondrial respiration and enterohepatic bile acid recirculation due to improvement of endoplasmic reticulum–mitochondria calcium homeostasis with hepatic HAX-1 inactivation suggest that this may be a potential therapeutic target for metabolic disease intervention.

The prevalence of metabolic diseases, including obesity, diabetes, and nonalcoholic fatty liver disease, is approaching pandemic proportions worldwide (1)(2)(3). These numbers continue to escalate, and traditional strategies currently used for diabetes and nonalcoholic fatty liver disease management have limited success with efficacy, safety, and tolerability issues (4 -8). The most successful approach for metabolic disease intervention is bariatric surgery (9). The metabolic improvement of bariatric surgery has been attributed to numerous mechanisms, including altering nutrient transport to bypass the proximal intestine with rapid delivery to the distal intestine, changes in the gut microbiota, and increasing bile salt transport in the enterohepatic circulation (10 -14). The metabolic benefits of enhanced bile salt circulation are due to the activation of the membrane receptor Takeda-G-protein receptor-5 to increase energy expenditure in adipose tissues, the promotion of glucagon-like peptide-1 release by L-cells to stimulate ␤ cell insulin secretion (15), and the activation of fibroblast growth factor-19 (FGF19; 2 FGF15 in mice) synthesis in the distal intestinal ileum to improve glucose metabolism in hepatic and extrahepatic tissues (16,17).
Emerging evidence indicates that mitochondrial dysfunction may be a key driver for metabolic disease manifestation (18), and improving calcium homeostasis through cross-talks between ER and mitochondria in the liver has been suggested as a less invasive approach to maintain glucose and lipid homeostasis for metabolic disease intervention (19,20). The importance of exquisitely regulated hepatic calcium homeostasis in metabolic disease modulation is highlighted by studies showing that increased ER calcium uptake through overexpression of the sarco(endo)plasmic reticulum Ca 2ϩ ATPase-2 (SERCA2) in the liver or reducing ER calcium release through suppression of InsP3R1 not only alleviates ER stress but also improves glucose tolerance, lowers blood glucose levels, and reduces liver steatosis (21)(22)(23).
A target with the potential to reap the dual benefits of enhanced enterohepatic bile salt circulation and improved calcium homeostasis and mitochondrial respiration is the hema-topoietic cell-specific protein-associated protein X-1 (HAX-1). Despite its name, HAX-1 is ubiquitously expressed, with the highest levels observed in liver, kidney, testis, and heart (24). This protein is associated with the cytosolic face of the mitochondria and the ER (25), via sequences found in the aminoand carboxyl-terminal domains, respectively (26), where it modulates calcium cycling between these organelles and regulation of mitochondrial potential (27)(28)(29)(30). HAX-1 also has tissue-specific functions that depend on the presence or absence of specific HAX-1-interacting partners in each cell type. For example, in the heart, HAX-1 interacts with phospholamban and heat shock protein 90 to regulate Ca 2ϩ cycling and prevent ER stress, respectively, to protect against ischemic heart injury (31,32). In the liver, where phospholamban is not expressed, HAX-1 binds to the bile salt export protein (BSEP), and its absence increases BSEP localization to the bile canaliculus membrane to enhance enterohepatic bile acid recirculation (33).
Whether HAX-1 expression in the liver influences mitochondrial respiration and bile salt homeostasis, and the consequential effects on diet-induced metabolic diseases, have not been explored. This study compares glucose and lipid homeostasis in WT and liver-specific HAX-1 knockout mice after Western diet feeding with the goal of determining whether hepatic HAX-1 is a target for metabolic disease intervention.

HAX-1 is localized at mitochondria and ER in liver, and its inactivation has no deleterious effects in liver
HAX-1 is highly expressed in the liver (24), and its level of expression was similar between chow-and Western diet-fed mice. Its level of expression was also similar regardless of fasting or fed conditions (Fig. 1A). Initial characterization of HAX-1 distribution in the liver revealed its predominant location in the mitochondria and the ER (Fig. 1B). To explore the physiological functions of HAX-1 in the liver, we injected 5 ϫ 10 11 genome copies of recombinant AAV8 virus encoding the Cre recombinase gene under the control of the strong liver-specific thyroxin-binding globulin (TBG) promoter into Hax1 flox/flox mice (30,34), which effectively eliminated HAX-1 expression specifically in the liver, whereas injection of the control vector expressing the GFP, AAV-TBG-GFP, had no effect on HAX1 expression (Fig. 1C). Histological examination as well as plasma alanine aminotransferase measurements revealed no abnormalities, and HAX-1 inactivation does not cause liver injury, fibrosis, or toxicity ( Fig. 1, D and E). Thus, all subsequent experiments comparing the impact of liver-specific Hax-1 inactivation were performed by using Hax1 flox/flox mice injected with AAV-TBG-GFP as control mice and those injected with AAV-TBG-Cre as liver-specific Hax1 knockout (L-Hax1 Ϫ/Ϫ ) mice.

Hepatic HAX-1 deficiency reduces plasma triglyceride levels in chow-fed mice
Significantly lower fasting plasma triglyceride and VLDL levels, but not total cholesterol, IDL/LDL, or HDL levels, were observed in chow-fed L-Hax1 Ϫ/Ϫ mice compared with control mice (Fig. 2, A and B). Fasting plasma nonesterified fatty acid levels were also similar between chow-fed control and L-Hax1 Ϫ/Ϫ mice (Fig. 2C). The difference in plasma triglyceride and VLDL levels in control and L-Hax1 Ϫ/Ϫ mice was not due to differences in VLDL synthesis and secretion (Fig. 2D). However, when control and L-Hax1 Ϫ/Ϫ mice were administered a bolus lipid meal after an overnight fast, the L-Hax1 Ϫ/Ϫ mice showed a trend toward less postprandial hypertriglyceridemia, indicative of accelerated clearance, although area under the curve analysis revealed that the difference did not reach statistical significance (Fig. 2E).

L-Hax1 ؊/؊ mice sustain euglycemia with lower insulin levels
Fasting blood glucose levels were similar and within normal range in both age-matched control Hax1 flox/flox mice and L-Hax1 Ϫ/Ϫ mice (Fig. 3A). Interestingly, whereas fasting insulin levels in the control mice were similar and within the normal range observed with WT mice, fasting insulin levels were significantly lower in L-Hax1 Ϫ/Ϫ mice compared with the control mice (Fig. 3B), leading to a lower homeostatic model assessment-insulin resistance (HOMA-IR) index suggestive of elevated insulin sensitivity in mice with liver HAX-1 inactivation (Fig. 3C). Indeed, an oral glucose tolerance test administered after an overnight fast showed decreased glucose excursion in L-Hax1 Ϫ/Ϫ mice (Fig. 3D). Taken together, these results indicate that the L-Hax1 Ϫ/Ϫ mice are capable of maintaining euglycemia with lower insulin levels by a mechanism that is consistent with improved glucose tolerance and insulin sensitivity.

Hepatic HAX-1 inactivation protects against Western diet-induced hyperlipidemia, hyperinsulinemia, and hepatosteatosis
The low plasma triglyceride levels and their ability to maintain euglycemia with low basal fasting insulin levels in chow-fed L-Hax1 Ϫ/Ϫ mice prompted us to investigate the influence of liver HAX-1 inactivation on metabolic disease manifestation in response to chronic feeding of a high-fat/cholesterol-type Western diet. The L-Hax1 Ϫ/Ϫ mice displayed a slightly lower body weight and adiposity compared with control mice after 12 weeks of Western diet feeding, but the difference did not reach statistical significance (Fig. 4A). The marginal differences in body weight cannot be attributed to differences in food intake. In fact, food intake appeared to be higher in L-Hax1 Ϫ/Ϫ mice compared with control mice (4.3 Ϯ 0.35 g/day versus 3.6 Ϯ 0.4 g/day). Both the control and the L-Hax1 Ϫ/Ϫ mice were capable of maintaining euglycemia throughout the 12-week Western diet feeding period, and neither group developed overt hyperglycemia (Fig. 4B). However, the control mice developed hyperinsulinemia, indicative of prediabetes and insulin resistance, whereas insulin levels in the L-Hax1 Ϫ/Ϫ mice remained at low or normal levels, indicative of resistance to diet-induced hyperinsulinemia (Fig. 4B). Estimation of insulin resistance by HOMA-IR is consistent with the interpretation that the control mice became insulin-resistant after 12 weeks of Western diet feeding, but the L-Hax1 Ϫ/Ϫ mice remained insulin-sensitive (Fig. 4B). Glucose tolerance test after oral administration of a glucose solution as well as insulin sensitivity test administered by insulin injection confirmed that the Western dietfed control mice were glucose-intolerant and insulin-resistant but the L-Hax1 Ϫ/Ϫ mice remained glucose-tolerant and Hepatic HAX-1 inactivation prevents metabolic diseases

Hepatic HAX-1 inactivation prevents metabolic diseases
insulin-sensitive 12 weeks after feeding the Western-type diet (Fig. 4, C and D).
Interestingly, the L-Hax1 Ϫ/Ϫ mice displayed lower fasting plasma nonesterified fatty acid (NEFA) levels compared with control mice under Western diet-feeding conditions (Fig. 4E). Importantly, plasma triglyceride and cholesterol levels throughout the 12-week period also remained at normal levels of Ͻ100 mg/dl in the L-Hax1 Ϫ/Ϫ mice in contrast to the hyper-lipidemia observed in control mice (Fig. 4F). Differences in plasma triglyceride and cholesterol levels were reflected by the lower plasma VLDL and LDL levels in L-Hax1 Ϫ/Ϫ mice (Fig.  4F). Examination of the livers from Western diet-fed control and L-Hax1 Ϫ/Ϫ mice revealed significant steatosis with more triglyceride and cholesterol in the livers of control mice compared with that observed in L-Hax1 Ϫ/Ϫ mice (Fig. 5). Thus, in addition to protection against diet-induced hyperlipidemia and glucose intolerance, hepatic HAX-1 deficiency also protects against fatty liver disease.

Hepatic HAX-1 deficiency suppresses glucose-induced lipogenesis
To examine the mechanism underlying the differences in plasma lipid levels and hepatosteatosis susceptibility between control and L-Hax1 Ϫ/Ϫ mice, primary hepatocytes isolated from these animals were incubated in media containing 5 and 25 mM glucose to mimic normal and hyperglycemic conditions and then assayed for triglyceride content in cell lysate and in the extracellular medium. In the presence of 5 mM glucose, no difference in intracellular triglyceride content was observed between control and L-Hax1 Ϫ/Ϫ hepatocytes (Fig. 6A). However, in the presence of 25 mM glucose, an ϳ2.5-fold increase in intracellular triglyceride content was observed in control hepatocytes but the L-Hax1 Ϫ/Ϫ hepatocytes did not display glucoseinduced steatosis, and their intracellular triglyceride levels remained similar to those observed in control and L-Hax1 Ϫ/Ϫ hepatocytes cultured under normoglycemic conditions (Fig.  6A). The L-Hax1 Ϫ/Ϫ hepatocytes also secrete less triglyceride into the medium under normoglycemic conditions (Fig.  6A). Whereas hyperglycemic conditions increased the triglyceride levels in the medium of L-Hax1 Ϫ/Ϫ hepatocytes to levels similar to those observed when control hepatocytes were cultured under normoglycemic conditions, significantly more triglycerides were found in the media of the

Hepatic HAX-1 inactivation prevents metabolic diseases
control hepatocytes when cultured with 25 mM glucose (Fig.  6A). These in vitro data were corroborated with in vivo experiments showing reduced de novo lipogenesis in the liver of L-Hax1 Ϫ/Ϫ mice (Fig. 6B). Furthermore, although no statistically significant differences were observed in VLDL secretion between control and L-Hax1 Ϫ/Ϫ mice (Fig. 2E), postprandial VLDL secretion after a bolus glucose meal was significantly reduced in L-Hax1 Ϫ/Ϫ mice (Fig. 6C). Taken together, these results indicate that hepatic HAX-1 inactivation reduces glucose-induced lipogenesis and thereby lowers plasma triglyceride levels and protects against diet-induced hepatosteatosis in L-Hax1 Ϫ/Ϫ mice.

Hepatic HAX-1 deficiency increases mitochondrial respiration and fatty acid oxidation
The reduced glucose-induced lipogenesis with the consequential effects of lower plasma triglyceride levels and suppression of hepatosteatosis observed in L-Hax1 Ϫ/Ϫ mice suggested that hepatic HAX-1 deficiency may increase glucose utilization. To test this possibility, primary hepatocytes isolated from control and L-Hax1 Ϫ/Ϫ mice were incubated with normal medium containing 25 mM glucose, and oxygen consumption rates were monitored over a 2-h period. Oligomycin was added to inhibit ATP synthesis, followed by carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP) Cre-injected mice. B, fasting blood glucose and insulin levels were measured from eight mice in each group, and the data were used for HOMA-IR estimation of insulin resistance. C, blood glucose levels were monitored after the administration of a bolus glucose solution (2 g/kg body weight) to Western diet-fed control (solid symbols) and L-Hax1 Ϫ/Ϫ mice (open symbols) (n ϭ 7). The data were used to calculate AUC for statistical analysis. D, blood glucose levels were monitored after insulin injection (0.75 units/kg body weight) into Western diet-fed control (solid symbols; n ϭ 6) and L-Hax1 Ϫ/Ϫ mice (open symbols; n ϭ 9), and AUC was calculated based on the data. E, fasting plasma NEFA levels were measured from four GFP-and six Cre-injected mice. F, fasting plasma triglyceride and cholesterol levels were measured from eight mice in each group. Samples were pooled from three mice in each group for FPLC analysis of lipoprotein distribution. All data represent the mean Ϯ S.D. (error bars) with the indicated p values determined by two-tailed Student's t test.

Hepatic HAX-1 inactivation prevents metabolic diseases
for mitochondria uncoupling and measurement of mitochondrial respiratory capacity, and then rotenone and antimycin A were added to inhibit complex 1 for measurement of nonmitochondrial respiration. Results revealed a significant increase in basal and maximal respiration rates as well as ATP production by hepatocytes from L-Hax1 Ϫ/Ϫ mice compared with hepatocytes isolated from control mice (Fig.  7A). However, spare respiration capacity and coupling efficiency were not different between L-Hax1 Ϫ/Ϫ and control hepatocytes (Fig. 7A). The latter observations indicated that HAX-1 deficiency improves normal mitochondrial respiration that is unrelated to reserve respiratory capacity in response to stress or energy dissipation as heat.
When oxygen consumption rates were measured in response to 150 M palmitate in the presence or absence of the fatty acid oxidation inhibitor etomoxir to assess fatty acid-specific oxidation, we found that the L-Hax1 Ϫ/Ϫ hepatocytes also displayed significantly higher oxygen consumption and ATP production in response to fatty acids compared with control hepatocytes (Fig. 7B). Moreover, HAX-1-deficient hepatocytes also displayed increased conversion of [ 3 H]palmitate to 3 H 2 O (Fig. 7B). Consistent with the higher fatty acid oxidation rates of L-Hax1 Ϫ/Ϫ hepatocytes, NEFA levels were found to be lower in the livers of L-Hax1 Ϫ/Ϫ mice (Fig. 7B).
The increased mitochondrial oxidation observed in HAX-1deficient hepatocytes was not due to differences in mitochondria number as indicated by similar ratio of mitochondrially encoded NADH:ubiquinone oxidoreductase (ND1) gene to nuclearly encoded hexokinase-2 (HK2) gene in the livers of control and L-Hax1 Ϫ/Ϫ mice (Fig. 8). Analysis of hepatic RNA expression revealed that HAX-1 inactivation altered the expression of selected fatty acid oxidation genes, particularly under Western diet conditions, including increased expression of carnitine palmitoyltransferase-1 and -2 (Cpt1 and Cpt2), which are responsible for fatty acid transport into the mitochondria, enoyl-CoA hydratase-1 (Echs1), which participates in mitochondrial ␤-oxidation, and cluster of differentiation 36 (Cd36), which is responsible for fatty acid uptake, and the reduced expression of acetyl CoA carboxylase-2 (Acc2), an enzyme that inhibits fatty acid oxidation and promotes fatty acid synthesis (Fig. 8).

Hepatic HAX-1 deficiency increases PDH activity
In view of the importance of the PDH complex in glucose oxidation, we determined PDH activities in livers of L-Hax1 ϩ/ϩ and L-Hax1 Ϫ/Ϫ mice and found that Hax-1 inactivation increased PDH activity in the liver (Fig. 9A). Whereas this increase was not due to changes in PDH kinase Hepatic HAX-1 inactivation prevents metabolic diseases 2 and 4 (PDK2 and -4) expression (Fig. 9B), HAX-1 inactivation significantly reduced PDHE1␣ phosphorylation under both chow-and Western diet-fed conditions (Fig. 9C). Increased mitochondrial respiration due to higher PDH activity has also been reported to elevate LDL receptor expression in hepatocytes (35). Indeed, increased LDL receptor protein levels were observed in Western diet-fed L-Hax1 Ϫ/Ϫ mice compared with those observed in control mice (Fig. 9D). The increased level of LDL receptor may partly account for the lower plasma lipid levels observed in L-Hax1 Ϫ/Ϫ mice.

Hepatic HAX-1 inactivation improves ER-mitochondria calcium homeostasis via InsP3R1 reduction
Increased PDH activity due to higher levels of nonphosphorylated PDH in the absence of differences in PDK2 and PDK4 suggested that HAX-1 inactivation may increase the dephosphorylation of PDH, a process that is mediated by calcium-dependent phosphatases in the mitochondria (36). In view of emerging evidence showing that ER and calcium homeostasis is an important determinant of mitochondrial respiration and metabolic diseases (19,20,37), and HAX-1 is present in both

Hepatic HAX-1 inactivation prevents metabolic diseases
the ER and the mitochondria, we determined whether HAX-1 inactivation influenced ER and mitochondria calcium homeostasis in the liver. Primary hepatocytes isolated from L-Hax1 Ϫ/Ϫ mice were found to display significantly higher ER calcium levels compared with control mice under both normoglycemic (5 mM glucose) and hyperglycemic (25 mM glucose) conditions (Fig. 9E). In contrast, mitochondrial calcium levels were found to be lower in L-Hax1 Ϫ/Ϫ hepatocytes compared with control cells (Fig. 9F). Importantly, HAX-1 inactivation also prevented palmitate-and hyperglycemia-induced mitochondrial calcium overload (Fig. 9F).
Mitochondrial dysfunction due to overnutrition is caused by excessive ER calcium transport to the mitochondria, a process that is mediated in part by InsP3R1 (37,38). Interestingly, we found lower levels of InsP3R1 protein in the liver of L-Hax1 Ϫ/Ϫ mice compared with control mice (Fig. 9G). To determine how HAX-1 may regulate InsP3R1 level in the liver, we performed co-immunoprecipitation experiments with HAX-1-specific antibodies and discovered that HAX-1 interacts with InsP3R1 (Fig. 9G). The finding that HAX-1 interacts with InsP3R1 and its ablation reduces InsP3R1 levels in hepatocytes is reminiscent of HAX-1's interaction with SERCA2 to suppress its degradation in the heart (30). Thus, the reduced InsP3R1 levels in L-Hax1 Ϫ/Ϫ mice are likely due to its enhanced degradation. Importantly, the reduced InsP3R1 level due to HAX-1 deficiency limits ER calcium transport to the mitochondria, thereby preventing excessive nutrient-induced mitochondrial calcium overload and dysfunction.

Hepatic HAX-1 inactivation increases hepatic BSEP level and bile acid-responsive gene expression in intestine and liver
Another liver protein that interacts with HAX-1 and has activity that is sensitive to cellular calcium modulation is the bile salt export protein BSEP encoded by the ABCB11 gene (33,39). Consistent with results observed in kidney cells (33), BSEP levels in whole liver lysates as well as bile canalicular membranes were also higher in L-Hax1 Ϫ/Ϫ mice compared with control mice (Fig. 10A). As described previously (40), the elevated BSEP expression also resulted in increased bile acid pool size, as reflected by higher levels of bile acids in plasma and liver (Fig. 10B). However, fecal bile acid levels and bile acid levels in the intestine were not different between these animals, suggesting that hepatic HAX-1 deficiency promotes the conservation of bile acids within the enterohepatic circulation, similar to that observed in liver-specific Abcb11 transgenic mice (40,41). Interestingly, we also observed increased excretion of cholesterol in the feces of both chow-and Western diet-fed L-Hax1 Ϫ/Ϫ mice (Fig. 10C).
Similar to results reported for the Abcb11 transgenic mice (40), hepatic HAX-1 deficiency did not alter intestinal expression of the bile acid transporter Asbt but robustly increased the expression of bile acid-responsive genes, including Fgf15, Ost␣/␤, and the short heterodimeric partner Shp (Fig. 11). The elevated bile acid pool size in L-Hax1 Ϫ/Ϫ mice also led to farnesoid X receptor activation in the liver, as reflected by increased expression of Shp, Abcg5, and Abcg8 (42) (Fig. 12). However, Cyp7a expression was found to be similar between control and L-Hax1 Ϫ/Ϫ mice (Fig. 12). We also found that expression of sterol regulatory element-binding protein-1-responsive genes, which may be down-regulated by bile acids (43), were also lower, as reflected by reduced mRNA levels of acetyl-CoA carboxylase-1 (Acc1), fatty acid synthase (Fasn), and stearoyl-CoA desaturase (Scd1) in the livers of L-Hax1 Ϫ/Ϫ mice (Fig. 12). Additionally, expression of the mitochondria cholesterol transport protein steroidogenic acute response protein (StAR) and the mitochondrial bile acid synthesis gene Cyp27a was also increased in the livers of Western diet-fed L-Hax1 Ϫ/Ϫ mice (Fig. 12). The increased expression of Abcg5 and Abcg8 is consistent with higher fecal cholesterol content in L-Hax1 Ϫ/Ϫ mice, whereas the increased expression of StAR, which mediates cholesterol transport into the mitochondria (44), and Cyp27a, which promotes the alternative acidic pathway of bile acid synthesis (45), may also partly account for the reduced hepatic cholesterol level and increased bile acid pool despite the decreased Cyp7a expression observed in L-Hax1 Ϫ/Ϫ mice. Taken together, these data indicate that hepatic HAX-1 inactivation, via increased BSEP levels in bile canaliculus membrane and enterohepatic bile acid circulation (40), also protects against diet-induced metabolic disease by activation of bile acid-responsive genes in the intestine and liver.

Discussion
The current study presents the first evidence that hepatic HAX-1 inactivation in male mice increases insulin sensitivity and protects against diet-induced hyperlipidemia, hyperinsulinemia, and hepatosteatosis. This study did not examine female mice; thus, metabolic benefits of HAX-1 inactivation in females remain uncertain. Nevertheless, the impact of HAX-1 inactivation in male mice is highly potent, as low insulin concentration is sufficient to maintain euglycemia in chow-fed L-Hax1 Ϫ/Ϫ mice.

Hepatic HAX-1 inactivation prevents metabolic diseases
The metabolic benefits of hepatic HAX-1 inactivation are multifactorial due to (i) reduction of InsP3R1 levels to maintain ER-mitochondria calcium homeostasis, thereby preventing excess nutrient-induced mitochondria calcium overload and dysfunction; (ii) decreased PDH phosphorylation, thereby improving mitochondrial respiration to increase nutrient utili-zation and LDL receptor expression to enhance postprandial lipid clearance; and (iii) increased BSEP levels and its presence in the bile canaliculus membrane, thereby improving enterohepatic bile acid circulation, which leads to activation of bile acidresponsive genes in intestine and liver to augment insulin sensitivity and reduce hepatic cholesterol levels. The improvement

inactivation increased pyruvate dehydrogenase activity and reduced phosphorylation with a corresponding increase in LDL receptor (LDLR) levels in mechanisms related to improvement of ER and mitochondria calcium homeostasis. A, liver homogenates prepared from control and Western diet-fed control (GFP-injected) and L-Hax1
Ϫ/Ϫ (Cre-injected) mice (n ϭ 6) were used to measure PDH enzymatic activity. Western blot analyses were performed with the lysates with antibodies against PDK2 and PDK4 (B) and antibodies against PDH-E1␣ and antibodies specific to PDH-E1␣ phosphorylated at residue Ser-300 (C). PDH phosphorylation was determined based on the ratio of pPDHE1␣/PDHE1␣ intensity in each sample. The data represent mean Ϯ S.D. (error bars) from three biological replicates. D, the liver homogenates were also subjected to Western blot analysis with antibodies against the LDL receptor, with anti-␤-actin blots used as loading controls. The data represent mean Ϯ S.D. from four biological replicates. E, hepatocytes isolated from control (GFP, solid bars) and HAX-1-inactivated (Cre, open bars) mice were incubated with 5 or 25 mM glucose. Thapsigargin was added to induce Ca 2ϩ release. Relative ER [Ca 2ϩ ] was determined by Fluo-4 fluorescence. The data represent mean Ϯ S.D. from n ϭ 4 mice in each group. **, p Ͻ 0.01 compared with control as evaluated by Student's t test. F, hepatocytes isolated from control (GFP, solid bars) and HAX-1-inactivated (Cre, open bars) mice were incubated with buffer with or without 150 M palmitate or with 5 or 25 mM glucose. Mitochondrial calcium levels were measured by the mitochondria-specific fluorescence indicator Rhod-2. The data represent mean Ϯ S.D. from n ϭ 4 mice in each group. *, p Ͻ 0.01 compared with GFP hepatocytes; #, p Ͻ 0.01 compared with control without addition as evaluated by Student's t test. G, liver homogenates from chow-fed control (GFP) and L-Hax1 Ϫ/Ϫ (Cre) mice were immunoprecipitated with anti-HAX-1 antibodies (n ϭ 4 in each group). Representative Western blots show total liver lysates and the anti-HAX-1 immunoprecipitated IP 3 R1 using ␤-actin immunoreactivity as loading control.

Hepatic HAX-1 inactivation prevents metabolic diseases
of bile acid recycling also results in suppression of SREBP-1responsive genes in the liver, leading to reduced lipogenesis, as well as increased expression of mitochondrial oxidation genes, probably due to bile acid-induced expression of these peroxisome proliferator-activated receptor ␣ genes (46).
It is interesting to note that despite their improvement in hepatic mitochondrial respiration and fatty acid oxidation, Western diet-fed L-Hax1 Ϫ/Ϫ mice displayed similar body weight gain and adiposity as control mice. This apparent paradox may be due to a slight increase in food intake as well as the reduced adipocyte lipolysis in Western diet-fed L-Hax1 Ϫ/Ϫ mice, as indicated by their lower fasting plasma NEFA levels compared with control mice (Fig. 4E). The reduced adipocyte lipolysis in Western diet-fed L-Hax1 Ϫ/Ϫ mice may be related to their improved insulin sensitivity (Fig. 4D) (47). This hypothesis relating hepatic nutrient oxidation with insulin sensitivity and adipocyte lipolysis will require more direct experimentation before any definitive conclusions can be made.
Our data revealed that HAX-1 interacts with InsP3R1 and that its ablation reduces InsP3R1 in hepatocytes, which is reminiscent of HAX-1 regulation of SERCA2a degradation in the heart (30). Although InsP3R1 signaling is necessary for ER calcium release to the mitochondria to sustain cellular bioenergetics and its total absence diminishes PDH complex activity due to impairment of calcium-dependent dephosphorylation (48), excessive ER calcium transport to the mitochondria also causes mitochondrial dysfunction especially under overnutrition conditions (37,38). Thus, exquisite control of InsP3R1 level/activ-  Figure 12. Hepatic HAX-1 inactivation activates FXR and reduces SREBP-1 genes in liver. Six control and 8 -9 L-Hax1 Ϫ/Ϫ mice were used to measure hepatic expression of FXR-responsive genes, such as Shp, Abcg5, and Abcg8, and SREBP-1-responsive genes, such as Acc1, Fasn, and Scd1. The RNA was also used for quantitative PCR to measure expression of alternative bile acid pathway genes, such as Star, and Cyp27a. The data represent mean Ϯ S.D. (error bars). Statistical evaluation comparing mice with different genotypes on the same diet was performed by Student's t test.

Hepatic HAX-1 inactivation prevents metabolic diseases
ity is necessary to optimize metabolic homeostasis. In the current study, we showed that HAX-1 inactivation lowers but does not ablate InsP3R1 expression, thereby leading to the prevention of mitochondrial calcium overload, decreased PDH phosphorylation, and improvement of mitochondrial activity in hepatocytes. These results are consistent with previous reports that silencing of InsP3R1 improves mitochondrial activity and maintains metabolic homeostasis in obese mice (37). Liver-specific InsP3R1 inactivation also reduces lipid droplet formation and protects against high-fat diet-induced hepatosteatosis (49). Additionally, our results showed that increased glucose oxidation also lowers postprandial VLDL secretion. This observation, along with increased hepatic LDL receptor expression due to higher PDH activity (35), may explain the resistance of L-Hax1 Ϫ/Ϫ mice to diet-induced hyperlipidemia. Taken together, these results indicate that PDH activation via InsP3R1 reduction is one mechanism by which L-Hax1 Ϫ/Ϫ mice were protected against Western diet-induced hyperinsulinemia, hyperlipidemia, and hepatosteatosis.
The importance of PDH activity in mitochondrial respiration is well-established in the literature. As noted above, PDH activity is inactivated by phosphorylation with PDKs and can be activated by PDH phosphatase-mediated dephosphorylation. Inactivation of the major PDK enzymes in the liver has been shown to be protective against diet-induced glucose intolerance and hepatosteatosis (50,51). These and other similar studies suggested that hepatic PDK enzymes may be therapeutic targets for diabetes management, and indeed a liver-specific pan-PDK inhibitor was effective in improving insulin signaling and lowers hepatic lipid accumulation in high-fat diet-induced obese mice (52). Nevertheless, despite these encouraging results, it is noteworthy that normalization of plasma glucose and insulin levels as well as the amelioration of hepatosteatosis were not achieved in these animal models. Hence, additional approaches to complement PDH activation may be necessary for optimal metabolic benefits.
The current study showed that normal plasma glucose, insulin, and lipid levels can be achieved in Western diet-fed mice by liver-specific inactivation of HAX-1. Hepatosteatosis was also significantly reduced in Western diet-fed L-Hax1 Ϫ/Ϫ mice compared with control mice. In addition to reducing PDH phosphorylation and improving mitochondrial activity, HAX-1 inactivation in the liver also increased BSEP/ABCB11 levels and its localization to the bile canaliculus. This result is consistent with a previous study reporting that HAX-1 interacts with BSEP/ABCB11 and that the absence of HAX-1 increases BSEP/ ABCB11 localization to the apical membranes of liver and kidney cells (33). Although the exact mechanism underlying the increased BSEP/ABCB11 remains unknown, it is tempting to speculate that the reduced expression of InsP3R1 redirects InsP3 to InsP3R2, which has been shown previously to promote BSEP/ABCB11 translocation and bile salt export activity (39). This hypothesis will be examined more rigorously in future studies. More importantly, elevated BSEP/ABCB11 has also been shown to enhance enterohepatic cycling of bile salts without significant change in fecal bile salt excretion (41).
Taken together, our results indicate that hepatic HAX-1 inactivation mimics bariatric surgery with respect to enhancing bile salt recycling through the enterohepatic circulation and increased expression of FGF15 and other bile acid-responsive genes in intestine. FGF15 and its human ortholog FGF19 are incretins secreted by the intestine and regulate glucose homeostasis by improving insulin sensitivity in hepatic and extrahepatic tissues (53). The observation that less insulin is required to maintain euglycemia in chow-fed L-Hax1 Ϫ/Ϫ mice and their protection against Western diet-induced hyperinsulinemia is consistent with the interpretation that metabolic benefits of hepatic HAX-1 inactivation are at least in part due to increased BSEP/ABCB11-mediated FGF15 expression in the intestine. However, despite these benefits of elevated hepatic BSEP/ ABCB11 levels to increase bile acid recirculation, its overexpression in the liver alone may be compromised due to an indirect effect of increasing cholesterol absorption through the intestine (54). Thus, a strategy to promote postprandial lipid clearance is necessary to complement the BSEP/ABCB11 increase for optimal metabolic benefits. In this regard, the present study shows that hepatic HAX-1 inactivation reduces postprandial VLDL secretion and increases LDL receptor expression in the liver, indicating that targeting hepatic HAX-1 may hold therapeutic promise along these lines.

Antibodies, primers, and key reagents
All antibodies and primers used in the current study were obtained commercially as listed in Tables S1 and S2.

Animals
All animal experimental protocols were reviewed and approved by the institutional animal care and use committee at the University of Cincinnati (approval 17-09-08-01) in accordance with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The Hax1 flox/flox mice generated originally in the Ihle laboratory (55) were mated with C57BL/6J mice for 6 generations to establish a colony of Hax1 flox/flox mice. Liver-specific Hax1 knockout (L-Hax1 Ϫ/Ϫ ) mice were developed by injecting the Hax1 flox/flox mice with 5 ϫ 10 11 genome copies of adeno-associated virus (AAV) expressing the Cre recombinase gene under the control of the liverspecific TBG promoter (AAV-TBG-Cre). The Hax1 flox/flox mice injected with AAV-TBG-GFP were used as a control without HAX-1 deletion. Male Hax1 flox/flox littermates at the age of 10 weeks were used for AAV injection. Mice were fed either a normal chow diet (Teklad, Madison, WI) or a Western type diet containing 15.2% kcal protein, 42.7% kcal carbohydrate, 42.0% kcal fat, and 0.2% cholesterol (Teklad, TD88137). Animal characterization was performed after an overnight fast, typically after a 12-week dietary period. Body weights were determined using a Denver 300K scale, and adiposity measurements were obtained using 1 H magnetic resonance spectroscopy (EchoMRI-100, Echomedical Systems, Houston, TX) as described previously (56).

Blood chemistry
Blood was collected from mice after an overnight fast. Blood glucose levels were measured using an Accu-Check Active Glu-

Hepatic HAX-1 inactivation prevents metabolic diseases
cometer. Plasma insulin levels were determined by ELISA kits. Plasma cholesterol and triglycerides were determined using Infinity cholesterol and triglyceride kits (Table S1). Plasma nonesterified fatty acid levels were measured using calorimetric assay kits (Wako Chemicals, Richmond, VA). For lipoprotein separation, 200 l of plasma pooled from 4 -5 mice in each group were subjected to fast-performance liquid chromatography (FPLC) on two Superose 6 HR 10/30 columns connected in tandem (Amersham Biosciences), and 0.5-ml fractions were collected for cholesterol and triglyceride determinations. For the glucose tolerance test, a glucose solution (2 g/kg body weight) was administered by oral gavage after an overnight fast, and blood glucose levels were measured over a 2-h period. For the insulin tolerance test, blood glucose levels in fasted mice were monitored over a 90-min period after intraperitoneal insulin injection (0.75 units/kg body weight).

Histology
Tissues were fixed in 4% paraformaldehyde overnight, dehydrated, and then embedded in paraffin to prepare 5-m sections. The sections were stained with hematoxylin and eosin and visualized under a light microscope.

Lipid accumulation in liver
Liver tissue was homogenized in buffer containing 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, and 5 mM EDTA (0.5 g/ml) in the presence of 0.5 Ci of [ 3 H]cholesterol for extraction efficiency. The lipids were extracted twice with 1 ml of petroleum ether, mixed, and centrifuged at 1000 ϫ g for 10 min. Fifty l of the organic phase was collected, dried, and used to quantify triglyceride, cholesterol, and NEFA levels in each sample. Extraction efficiency was normalized based on recovery of the [ 3 H]cholesterol.

Hepatic triglyceride secretion and de novo lipogenesis
After an overnight fast, mice were injected intraperitoneally with a 10% solution of Poloxamer 407 (1 g/kg body weight) diluted in saline to assess triglyceride secretion. Postprandial lipid clearance was performed by administering a 200-l bolus of olive oil to each animal by oral gavage. For both experiments, blood samples (50 l) were collected from the tail vein prior to and hourly thereafter to measure triglyceride levels. For in vitro studies, primary hepatocytes, isolated as described previously (57), were incubated in serum-free M199 medium for 24 h either in nonstimulated medium (5 mM glucose and 0.1 M dexamethasone) or stimulated medium (25 mM glucose with 100 nM insulin and 0.1 M dexamethasone). Lipids were extracted from the media and cell lysates with chloroform/ methanol (2:1, v/v) for the triglyceride assay. The hepatocytes were also stained with Oil Red O to visualize lipid accumulation under a microscope. Lipogenesis in vivo was performed by feeding a [ 14 C]glucose solution (2 mg/g, 0.05 Ci) by intragastric gavage. Tissues were harvested after 2 h for lipid extraction with chloroform/methanol (2:1). The rate of glucose incorporation into newly synthesized fatty acids was calculated as nmol of [ 14 C]glucose incorporated into fatty acids/g of tissue.

Pyruvate dehydrogenase activity measurement
Lysates were prepared from the livers of GFP-and Cre-injected Hax1 flox/flox mice by homogenization in ice-cold PBS containing protease inhibitor mixture (Roche Diagnostics). Four hundred g of protein in each lysate was used to measure PDH enzymatic activity using a commercially available kit (Table S1).

Cellular oxygen consumption and fatty acid oxidation
Oxygen consumption rate (OCR) was measured in intact hepatocytes in real time using an XF24 Extracellular Flux Analyzer and the XF 24 version 1.5.0.69 software. Hepatocytes were seeded on collagen-coated XF24 plates at 10,000 cells/well. The cells were incubated for 1 h at ambient temperature and then overnight at 37°C in William's E medium supplemented with 100 units/ml penicillin/streptomycin, 2 mM L-glutamine, 5% FBS, 100 nM insulin, and 100 nM dexamethasone. The cells were rinsed twice with PBS and kept at 37°C in 500 l of Dulbecco's modified Eagle's medium supplemented with 25 mM glucose, 1 mM sodium pyruvate, and 4 mM Glutamax for 1 h in a CO 2 -free incubator in the Seahorse Extracellular Flux Analyzer. After baseline measurement, additions were delivered in the following order through the instrument's individual injection ports: oligomycin (12 g/ml), FCCP (1.5 M), and rotenone (1 M)/ antimycin A (1.5 M) (Sigma).
For fatty acid oxidation, cells were incubated for 5 h in Dulbecco's modified Eagle's medium without any substrate, washed with Krebs-Henseleit buffer (11 mM NaCl, 4.7 mM KCl, 2 mM MgSO 4 , 1.2 mM Na 2 HPO 4 , 2.5 mM glucose, 0.5 mM carnitine, and 0.1 M insulin), and then incubated with 500 l of the same buffer in a CO 2 -free incubator at 37°C for 1 h. BSAconjugated palmitate (150 M) was added to stimulate oxygen consumption. Etomoxir was added in port B to a final concentration of 50 M on the flux plate to inhibit respiration and to calculate fatty acid-specific OCR and ATP production rate.
Fatty acid oxidation in hepatocytes was also assessed by measuring the release of 3 H 2 O after incubation of hepatocytes with [9,10-3 H]palmitic acid as described (58). Briefly, the hepatocytes were incubated with 0 -400 M BSA-conjugated [ 3 H]palmitate and 0.5 mM carnitine for 2 h at 37°C. The medium was removed, and the proteins were precipitated with TCA and then neutralized with NaOH. The 3 H 2 O in the supernatant was extracted by Dowex 1X8 resin (Sigma) for scintillation counting. The amount of 3 H 2 O released was normalized to the amount of proteins in the cell lysate.

Calcium measurements
Changes in intracellular calcium concentrations in hepatocytes were determined using the Fluo-4 AM NW calcium assay kit. Following culture of cells with serum-free M119 medium for 16 h in the presence of 5 or 25 mM glucose, the hepatocytes were loaded with Fluo-4 AM in the presence of probenecid. Thapsigargin (450 nM) or DMSO was added, and fluorescence (490-nm excitation; 516-nm emission) was measured over a 5-min period. Calcium levels in the ER are expressed as fluorescence units in thapsigargin-treated cells minus fluorescence units in DMSO-treated cells. Mitochondrial calcium levels under basal and nutrient-stimulated conditions were assessed Hepatic HAX-1 inactivation prevents metabolic diseases as described (38). In brief, the hepatocytes were incubated with or without 150 M palmitate or with 5 or 25 mM glucose for 6 h prior to the addition of the mitochondria-specific calcium indicator Rhod-2 AM. The cells were washed three times and then replenished with fresh medium to measure fluorescence intensity (552-nm excitation; 581-nm emission).

Fecal cholesterol measurements
Feces were collected from individually housed mice over 3 days. Dried feces (0.3-0.5 g) were then minced and extracted in 2.5 ml of saline and 2.5 ml of chloroform/methanol (2:1, v/v). The organic phase that contains the extracted lipids was collected in a glass tube, air-dried, and then redissolved in 400 l of 100% ethanol, and cholesterol content was measured enzymatically using a cholesterol assay kit (Table S1).

Bile acid analysis
Bile acid levels in plasma, liver, small intestine, and feces were measured as described (57). Briefly, fasting mice were sacrificed, and fresh tissues were collected, weighed, and minced. Bile acids were extracted with 75% ethanol for 1 h at 50°C. Bile acids were measured enzymatically using the mouse total bile acid assay kit (Table S1). Bile acid levels in each tissue were expressed as mol/kg body weight of each animal.

Co-immunoprecipitation
The HAX-1-specific antibodies were cross-linked to Protein A Dynabeads with BS3 cross-linker prior to incubation for 12 h at 4°C with liver lysates that had been precleared with off-target antibody and Protein A-Sepharose beads. The samples were eluted with 2ϫ NuPAGE lithium dodecyl sulfate sample buffer (Table S1) for 5 min at 95°C and used for Western blot analysis.

Western blot analysis
Livers were excised and homogenized in ice-cold radioimmune precipitation assay buffer containing protease and phosphatase inhibitor mixture. For assessment of BSEP expression, canalicular plasma membrane was isolated as described previously (59). Samples were diluted with a 4ϫ volume of SDS sample buffer (40% (v/v) glycerol, 240 mM Tris-HCl (pH 6.8), 8% SDS, 0.04% bromphenol blue, 5% ␤-mercaptoethanol) and heated at 95°C for 5 min. The samples were electrophoresed on Express Plus PAGE gels and then transferred to polyvinylidene difluoride membranes. The membranes were blocked in TBS solution containing 5% milk and 0.1% Tween 20 for 1 h at 4°C and then incubated for 90 min with a 1:1000 dilution of primary antibodies. The membranes were washed and then incubated with horseradish peroxidase-conjugated secondary antibodies and visualized using Pierce enhanced chemiluminescence reagents.

Real-time PCR analysis to determine mitochondria number and gene expression
Total genomic DNA was extracted from 30 mg of liver tissues using the DNeasy blood and tissue kit (Qiagen, CA). Relative mitochondrial DNA copy number was determined by quantitative PCR utilizing the mitochondrial NADH dehydrogenase 1 (ND1) gene as a marker for mitochondrial DNA and hexoki-nase 2 (HK2) as a marker for the nuclearly encoded gene. To determine gene expression, RNA was extracted from tissues using Direct-Zol RNA kits and then used to generate cDNA using the qScript cDNA synthesis kit. Quantitative real-time PCR was performed on a StepOnePlus Fast thermocycler using Fast SYBR Green master mix with primer sequences as shown in Table S2.

Statistical analysis
Sample size for each experiment was estimated based on previous studies documenting that 6 mice/group are sufficient to evaluate statistically significant differences in diet-induced body weight gains, adiposity, and plasma lipid, glucose, and insulin levels (60). Mice were not randomized prior to assignment to experimental groups. No samples, mice, or data points were excluded from the reported analyses. Statistical analysis was performed using SigmaPlot version 13.0 software. Normality was examined using the Shapiro-Wilk test. Data with equal variance based on Brown-Forsythe analysis were evaluated by Student's t test for studies comparing two groups. Data with unequal variance were evaluated using the Mann-Whitney test. Differences at p Ͻ 0.05 were considered statistically significant.