HSPA12A Is a Novel Player in Nonalcoholic Steatohepatitis via Promoting Nuclear PKM2-Mediated M1 Macrophage Polarization

Nonalcoholic fatty liver disease (NAFLD) affects ∼22–28% of the global population (1). The spectrum of NAFLD begins with isolated steatosis, with subsequent development of steatohepatitis (NASH) and steatofibrosis, eventually leading to cirrhosis and hepatocellular carcinoma. Patients with NAFLD, particularly NASH, show reduced survival, primarily due to cardiovascular and liver-related diseases (2). NASH livers are more injured than livers with isolated steatosis, which more likely leads to progressive fibrosis and eventual liver-associated illness and death (3). The molecular players that modulate disease pathogenesis are starting to be identified, enabling translation of research findings to clinical trials (4,5). However, no optimum therapy is yet available, suggesting that a more comprehensive understanding of the onset and progression of NAFLD is needed.

The shift from isolated steatosis to NASH is defined histologically as emergence of liver inflammation and hepatocellular injury. The most common theory explaining this shift is the “double-hit hypothesis” (3,5,6). The first hit consists mainly of lipid accumulation, whereas the second comprises sequential innate immune responses (6–8). Compelling evidence indicates that liver macrophages play a key role in regulating the second hit. Macrophages are a highly heterogeneous, plastic cell population that undergoes pleiotropic and coordinated responses to the immunological environment; two common phenotypes are “proinflammatory” M1 and “immunoregulatory” M2 macrophages (7,9,10). Polarization of macrophages toward the M1 phenotype plays a critical role in pathogenesis of several chronic inflammatory disorders, including NASH. By contrast, M2 macrophages promote resolution of inflammation (10). Indeed, suppressing M1 polarization alleviates diet-induced NASH (7,10–12).

The M2 isoform of pyruvate kinase (PKM2) is a rate-limiting enzyme of glycolysis. Cis-trans isomerization and conversion of PKM2 from a tetramer to a dimer or monomer lead to nuclear translocation and regulate expression of target genes that play roles in the Warburg effect and cell cycle progression (13,14). Intriguingly, nuclear PKM2 is a critical determinant of macrophage M1 activation and is involved in inflammatory diseases such as coronary artery disease (13,15,16). Nuclear PKM2 increases macrophage M1 activation in response to lipopolysaccharides (LPS), mainly through TLR4 signaling (13,15,17,18). A recent study shows that PKM2 expression is associated with liver mass (19); however, the role of PKM2 in NASH pathogenesis is largely unknown.

HSPA12A is a novel and atypical member of the HSP70 family (20). Hspa12a mRNA is expressed at high levels in the brain of humans and mice under normal conditions but is decreased in the prefrontal cortex of patients with schizophrenia (20,21). Of particular interest, expression of HSPA12A is associated with shortened survival of patients with hepatocellular carcinoma (22), suggesting possible involvement of HSPA12A in modulating hepatic homeostasis. However, the role of HSPA12A in any disorders, including liver disease, is unknown.

Here, we report that NASH patients showed increased hepatic HSPA12A expression and serum HSPA12A contents. Intriguingly, knockout of HSPA12A (Hspa12a^−/−) in mice attenuated high-fat diet (HFD)–induced NASH pathogenesis and M1 macrophage activation. Further molecular analyses revealed that regulation of de novo lipogenesis in hepatocytes was regulated by the paracrine effects of macrophage HSPA12A via promotion of nuclear PKM2-mediated M1 polarization. Taken together, these findings suggest that HSPA12A acts as a novel regulator of macrophage polarization and NASH pathogenesis. Strategies designed to inhibit macrophage-specific expression of HSPA12A might have therapeutic potential for patients with NASH.

Liver specimens were collected from patients who underwent hepatic hemangioma surgery at the First Affiliated Hospital of Nanjing Medical University. Patients provided informed consent at the time of recruitment. Normal or NASH livers were identified using hematoxylin-eosin (H-E) staining, Oil Red O (ORO) staining, and serum alanine aminotransferase (ALT) and AST activity measurements, according to the NASH Clinical Research Network–modified Brunt methodology and other studies (3,23). Livers with bridging fibrosis or cirrhosis were excluded using Masson’s staining (Supplementary Fig. 1A and B and Supplementary Table 1A and B).

Human blood samples were collected from bariatric surgery patients with NASH diagnosed after clinic and B-type ultrasonographic assessment (Supplementary Table 2A and B). Patients were fasted for 12 h before blood sampling. The study was approved by the Ethical Board of the First Affiliated Hospital of Nanjing Medical University (2016-SR-122 and 2016-SR-123). All human studies were conducted according to the World Medical Association Declaration of Helsinki.

Conditional Hspa12a knockout mice were generated using the loxP and Cre recombinant system. The Hspa12a targeting vector was constructed using the bacterial artificial chromosome (BAC) retrieval method. In brief, the region of the Hspa12a gene containing exons 2–4 was retrieved from a 129/sv BAC clone (BACPAC Resources Center, Oakland, CA) using a retrieval vector containing two homologous arms. Exons 2 and 3 were replaced by loxP sites flanking a PGK-neo cassette (a positive selection marker) (Supplementary Fig. 2A). The breeding strategy to generate wild-type (WT) (Hspa12a^flox/flox) and Hspa12a^−/− (Hspa12a^flox/flox, Cre^Tg) mice is shown in Supplementary Fig. 2B. Mice were bred at the Model Animal Research Center of Nanjing University and maintained on a 12-h light/dark cycle at 23 ± 1°C with access to food and water ad libitum. All experiments conformed to the Guide for the Care and Use of Laboratory Animals published by the U.S. National Institutes of Health (8th edition, 2011). The animal care and experimental protocols were approved by the Committee on Animal Care at Nanjing University.

For the HFD experiments, mice were fed an HFD (60% of kcal from fat, D-12492; Research Diets, New Brunswick, NJ) for 14 weeks beginning at age 5 weeks. Mice that were fed a normal chow diet and received only 6% of kcal from fat served as controls.

Cytosolic and nuclear proteins were prepared from livers or cells using the NucBuster protein extraction kit according to the manufacturer’s instructions (Novagen, Darmstadt, Germany). Western blotting was performed according to our previous methods (24,25). To control for lane loading, the membranes were probed with anti-GAPDH or anti–α-tubulin antibodies for cytosolic proteins and anti-histone H3 or anti-lamin A/C antibodies for nuclear proteins.

For analyzing the interaction between HSPA12A and PKM2 by immunoprecipitation-immunoblotting, Raw264.7 macrophages were overexpressed with flag-tagged HSPA12A. After challenge with LPS or vehicle for 16 h, cells were collected for protein extraction. Aliquots of equal volume and protein content were precipitated with anti-flag antibodies, followed by Western blotting with anti-PKM2 and anti-HSPA12A antibodies, as described previously (26). The antibodies used in the experiments are listed in Supplementary Table 3.

Quantitative real-time PCR was performed as described previously (27). In brief, total RNA was extracted for cDNA synthesis using the oligo (dT) primer. After cDNA synthesis, the expressions of indicated genes were estimated by real-time PCR using the SYBR Green master mix (Roche, Indianapolis, IN). The PCR results of β-actin served as internal controls. The primers used for PCR are listed in Supplementary Table 4.

Levels of HSPA12A in human serum and insulin in mouse serum and secretion of HSPA12A, IL-1β, IL-6, IL-12, and TNF-α from macrophages were detected using ELISA kits (Supplementary Table 5) according to the manufacturer’s instructions. Serum ALT and AST activity, glucose, LDL cholesterol, HDL cholesterol, total cholesterol, and triglyceride (TAG) were measured using a Beckman Coulter AU5800 Chemistry System analyzer (Brea, CA).

Lipid content was evaluated using both ORO (Sigma-Aldrich, St. Louis, MO) staining and a TAG assay. ORO staining was performed on 4% paraformaldehyde-fixed frozen liver sections or hepatocyte cultures. Quantification was performed using spectrophotometry at a wavelength of 510 nm after extraction of the stained ORO in cell cultures. TAG was evaluated in liver tissues or hepatocyte cultures using a TAG assay kit according to the manufacturer’s instructions (Jiancheng Biotech, Nanjing, China).

Paraffin-embedded liver sections were stained with H-E to evaluate the histological changes and averaged hepatocyte areas. Masson’s staining was also performed to indicate fibrosis, as described previously (24). Immunofluorescence staining of frozen liver sections or Raw264.7 macrophages was performed as described previously (24,25) using appropriate antibodies (Supplementary Table 3). In brief, after incubation with the indicated primary antibodies overnight at 4°C, Cy3- or FITC-conjugated appropriate secondary antibodies were added to the sections to visualize the staining. Hoechst 33342 reagent was used to counterstain the nuclei. The staining was observed and quantified in 7–10 randomly selected areas of each sample using a fluorescence microscope with cellSens Dimension 1.15 software (Olympus, Tokyo, Japan).

The adenoviral vector containing a three flags–tagged coding region of mouse Hspa12a (NM_175199) was generated by GeneChem (Shanghai, China). A schematic overview of virus construction is shown in Supplementary Fig. 3.

Primary hepatocytes and liver macrophages (Kupffer cells) were isolated from 7- to 10-week-old mice by digestion with 0.06% collagenase type IV, followed by centrifugation on a 25–50% Percoll gradient, as described previously (28). Primary hepatocytes were grown in DMEM supplemented with 10% FBS and 0.01 mmol/L dexamethasone. Mouse Raw264.7 macrophages and human hepatocellular carcinoma HepG2 cells were grown in DMEM supplemented with 10% FBS. Mouse AML-12 hepatocytes were cultured in DMEM/F12 supplemented with 10% FBS, ITS (5 µg/mL insulin, 5 µg/mL transferrin, and 5 ng/mL selenium), and 40 ng/mL dexamethasone.

Cells were infected for 48 h with adenovirus carrying the Hspa12a expression sequence. Cells infected with empty adenovirus served as normal expression controls (Hspa12a^n/e).

Hspa12a^o/e or Hspa12a^−/− hepatocytes were incubated with 200 μmol/L oleic acid (OA) for 24 h.

Hspa12a^o/e or Hspa12a^n/e Raw264.7 macrophages were treated for 16 h with LPS (500 ng/mL) to induce M1 polarization. The medium was then collected (referred to here as Hspa12a^o/e conditioned medium [CM] or Hspa12a^n/e CM).

AML-12 hepatocytes were incubated for 24 h with Hspa12a^o/e CM or Hspa12a^n/e CM in the presence of OA (200 μmol/L).

After they were grown for 48 h, WT and Hspa12a^−/− Kupffer cells were incubated with LPS (200 ng/mL) for 16 h. Kupffer cells were harvested for immunoblotting analysis, and the medium were collected and used as WT CM and Hspa12a^−/− CM, respectively.

RAW 264.7 macrophages were treated with 500 ng/mL LPS for 16 h. Cross-linking was performed using 500 μmol/L disuccinimidyl suberate (Thermo Fisher Scientific, Waltham, MA) for 30 min. Lysates were analyzed by Western blot as previously described (15).

To examine the effects of macrophage HSPA12A on HFD-induced NASH, we transplanted the bone marrows of Hspa12a^−/− mice to WT mice according to previously described methods (29). In brief, the recipient C57BL/6 WT mice (6 weeks old) were given acidified and antibiotic water (100 mg/L neomycin and 10 mg/L polymyxin) for 1 week before irradiation. At the day of transplantation, recipient mice were given a lethal irradiation (9 Gy). Four hours after irradiation, a total of 5 × 10⁶ bone marrow cells isolated from donor WT or Hspa12a^−/− mice (4 weeks old) were intravenously injected into a recipient WT mouse via tail vein, referred to as Hspa12a^−/−/WT and WT/WT mice, respectively. After receiving acidified and antibiotic water for 4 weeks after transplantation, the bone marrow transplantation (BMT) mice were fed with HFD for 14 weeks.

Primary mouse adipocytes differentiated from preadipocytes were used in the experiments. The primary mouse preadipocytes were used as controls. The proteins that specifically interact with HSPA12A were pulled down by HSPA12A antibody and were separated by SDS-PAGE followed by Coomassie blue staining, digestion in gel with trypsin, and analysis by liquid chromatography–tandem mass spectrometry. In brief, peptides were dissolved in solvent A (2% formic acid in 3% acetonitrile) and loaded directly onto a reversed-phase trap column (Chrom XP C18-CL-3m 120A; Eksigent). Peptide separation was performed using a reversed-phase analytical column (3 μm, 120A) (3C18-CL-120; Eksigent). Eluting peptides from the column were analyzed using an AB Sciex 5600+ TripleTOF system. Tandem mass spectrometry data were processed using ProteinPilot Software 4.5 (AB Sciex). Tandem mass spectra were searched against the UniProt mouse database (16,923 sequences) concatenated with a reverse decoy database. Trypsin/P was specified as the cleavage enzyme, allowing up to three missing cleavages, four modifications per peptide, and five charges.

Data are expressed as the mean ± SD. Groups were compared using an unpaired, two-tailed Student t test or using one-way or two-way ANOVA followed by Tukey post hoc test. A P value of <0.05 was considered significant.

The expression profile of HSPA12A protein in liver and cells is unclear. Western blotting revealed low-level expression of HSPA12A in mouse liver (Supplementary Fig. 4). In mouse liver, expression of HSPA12A by macrophages was 17.2-fold higher than that in hepatocytes (Fig. 1A). Macrophages in human livers also showed abundant expression of HSPA12A, whereas expression by hepatocytes was barely detectable (Fig. 1B and Supplementary Fig. 5).

To investigate the clinical significance of HSPA12A in NASH, we evaluated expression of HSPA12A in NASH patients. Expression of Hspa12a mRNA in NASH livers was significantly higher than that in non-NASH controls (Fig. 1C). Moreover, we observed a positive correlation between circulating HSPA12A levels and ALT and AST activity in NASH patients (Fig. 1D and Supplementary Fig. 6). Unexpectedly, expression of HSPA12A decreased in the cytosolic fraction but increased in the nuclear fraction, suggesting nuclear translocation of HSPA12A in the human NASH liver (Fig. 1E). Similar distribution of HSPA12A was observed in the livers of mice fed an HFD (Fig. 1F). Therefore, NASH is associated with upregulated expression of HSPA12A.

Because expression of HSPA12A was upregulated in NASH livers, it is conceivable that HSPA12A knockout may have therapeutic potential. Therefore, we generated HSPA12A knockout mice (Hspa12a^−/−) using the Cre-loxP recombinant system (Supplementary Fig. 2A and B). Successful knockout of HSPA12A in the liver was verified by protein analysis (Fig. 2A). The body weight of adult Hspa12a^−/− mice, along with serum levels of glucose and lipids, was similar to that in WT mice with access to food and water ad libitum, as were fasting serum insulin levels (Supplementary Fig. 7A–C). Liver weight, hepatocyte size, and lipid amount were similar in Hspa12a^−/− and WT mice fed a normal chow diet (Fig. 2B–D). However, HFD-induced increases in liver weight and hepatocyte size were decreased in Hspa12a^−/− compared with WT mice (Fig. 2B and C). Furthermore, HFD-induced steatosis was less marked in Hspa12a^−/− mice, as indicated by reduced TAG content, histological alteration, and ORO staining (Fig. 2C and D). Notably, HFD-induced liver injury, as indicated by elevated serum ALT and AST activities, was abolished in Hspa12a^−/− mice (Fig. 2E). No significant fibrosis, indicated by Masson’s staining, was observed in both genotypes after HFD feeding for 14 weeks (Supplementary Fig. 8). Taken together, these findings strongly suggest that HSPA12A deficiency prevents progression of NAFLD to NASH.

To gain insight into the mechanisms underlying HSPA12A deficiency–mediated inhibition of NAFLD progression, we examined expression of genes linked to lipid metabolism in the livers of HFD-fed mice. We noted a significant decrease in transcription of lipogenic transcription factors (Chrebp, Srebp-1c, Ppara, and Pparg) in Hspa12a^−/− livers compared with WT livers (Fig. 2F). In agreement, the expression of target genes, including Gck, Lpk, Acc, Fas, Scd1, Gpat, Dgat2, and Elovl6, involved in de novo fat synthesis were reduced in Hspa12a^−/− livers compared with WT livers (Fig. 2F). We also noticed that genes regulating lipid storage and transportation (Cidea, Plin1, and Apob) and lipolysis (Hsl and Cgi58) were downregulated in Hspa12a^−/− livers (Supplementary Fig. 9A and B). As the most prominent changes in expression were observed for transcription factors and target genes involved in de novo fat synthesis, we examined expression of PPARγ and SCD1 proteins. In line with the aforementioned results, expression of both proteins was reduced in the liver of Hspa12a^−/− compared with WT controls upon HFD feeding (Supplementary Fig. 9C). The findings suggest that HSPA12A regulates de novo fat synthesis and other lipid metabolic signaling in the liver upon HFD feeding.

To investigate how HSPA12A regulates accumulation of lipids in hepatocytes, we examined the effects of loss- and gain-of-HSPA12A function on lipid deposition using three types of hepatocyte culture: primary mouse hepatocytes, mouse hepatocyte AML-12 cells, and human hepatocellular carcinoma HepG2 cells. Unexpectedly, we found that OA-induced lipid deposition in primary hepatocytes was not affected by either Hspa12a deficiency or overexpression (Fig. 3A and B). Also, overexpression of HSPA12A did not affect OA-induced lipid deposition in either AML-12 or HepG2 hepatocytes (Fig. 3B).

Because hepatic macrophages showed higher HSPA12A expression than hepatocytes (Fig. 1A and B), we designed culture experiments to evaluate the interaction between macrophage HSPA12A and hepatocyte lipid deposition. To do this, we overexpressed HSPA12A (Hspa12a^o/e) in Raw264.7 macrophages by infection with the adenovirus carrying the Hspa12a expression sequence; macrophages infected with empty virus served as normal expression controls (Hspa12a^n/e). The CM collected from these macrophage cultures were referred to as Hspa12a^o/e CM and Hspa12a^n/e CM. Notably, Hspa12a^o/e CM increased the OA-induced lipid deposition in AML-12 hepatocytes (Fig. 3C). The regulation of primary Kupffer cell HSPA12A on hepatocyte lipid deposition was also examined. The CM was collected from LPS-treated primary Hspa12a^−/− and WT Kupffer cell cultures, referred to as Hspa12a^−/− CM and WT CM, respectively. AML-12 hepatocytes incubated with Hspa12a^−/− CM accumulated significantly less lipid than cells treated with WT CM in the presence of OA (Fig. 3D). Taken together, these data indicate that macrophage HSPA12A regulates hepatocyte steatosis through paracrine effects.

The effects of macrophage HSPA12A on hepatocyte lipid deposition prompted us to focus on macrophage activation and inflammatory responses because they play critical roles in NASH pathogenesis (7,10). Indeed, upon HFD feeding, we found a significant reduction in the number of F4/80-positive cells in the liver of Hspa12a^−/− mice compared with that of WT mice (Fig. 4A). Accordingly, analysis of mRNA levels revealed markedly lower expression of proinflammatory mediators and M1 markers, including Il-1β, Il-6, Tnfα, Nfκb, Tlr4, Inos, Mcp1, Ifnγ, Il-12, Cd86, and Cd68, in livers of Hspa12a^−/− mice compared with WT controls upon HFD (Fig. 4B and C). By contrast, the livers of Hspa12a^−/− mice showed markedly higher expression of mRNA encoding M2 markers, including Cd163, Arg1, and Cd206, than that of WT mice fed an HFD (Fig. 4C). In agreement, protein analysis confirmed reduced expression of iNOS (an M1 marker) and increased expression of CD206 (an M2 marker) in livers from HFD-fed Hspa12a^−/− mice (Fig. 4D). Moreover, livers from Hspa12a^−/− mice showed lower expression of TLR4, MyD88, and phosphorylated (p-) NF-κB (p65) than those from WT controls upon HFD (Fig. 4D). Collectively, these data indicate that HSPA12A deficiency attenuates both macrophage recruitment and M1 polarization in the liver of HFD-fed mice.

To investigate how HSPA12A regulates macrophage polarization, we first examined whether HSPA12A expression is altered upon activation of M1 macrophages. Liver macrophages from NASH patients showed markedly higher protein expression and nuclear translocation of HSPA12A than controls, as illustrated by immunostaining (Fig. 5A and Supplementary Fig. 10). The same pattern of HSPA12A expression and distribution was observed in Raw264.7 macrophages after LPS treatment, as indicated by immunoblotting and immunostaining (Fig. 5B and C). We also observed increased expression of HSPA12A in Raw264.7 macrophages exposed to OA (Supplementary Fig. 11). These findings suggest that HSPA12A in macrophages is upregulated and nuclear translocated during an M1 response.

Next, we examined the direct effects of HSPA12A on the M1 macrophage response. Overexpression of HSPA12A increased mRNA expression of M1 markers (Inos, Il-1β, Tnfα, Mcp1, Ccl3, and Ccl4) in LPS-treated Raw264.7 macrophages (Fig. 5D). In agreement, expression of iNOS and TNF-α protein in Hspa12a^o/e macrophages was higher than that in Hspa12a^n/e macrophages in the presence of LPS (Fig. 5E). Expression of TLR4 and p-NF-κB, both critical for M1 macrophage activation, also increased in Hspa12a^o/e macrophages (Fig. 5E). By contrast, primary Kupffer cells from Hspa12a^−/− mice showed decreased expression of iNOS, TLR4 MyD88, and p-NF-κB protein compared with WT Kupffer cells after LPS stimulation (Supplementary Fig. 12). These findings confirm the aforementioned in vivo results showing that HSPA12A deficiency attenuates M1 macrophage polarization in the liver of HFD-fed mice (Fig. 4B–D).

Nuclear PKM2 promotes M1 macrophage polarization (15). Because mass spectrometry analysis revealed an interaction between PKM and HSPA12A in differentiated adipocytes (Supplementary Fig. 13A and B), we examined whether PKM2 mediates HSPA12A-induced M1 macrophage polarization. PKM2 expression and nuclear translocation was increased in Raw264.7 macrophages in response to LPS stimulation (Supplementary Fig. 14). Liver macrophages of NASH patients and NASH mice demonstrated marked increases in PKM2 expression and nuclear translocation (Fig. 6A and B and Supplementary Fig. 15A–C). However, HSPA12A deficiency decreased Pkm2 mRNA expression in liver and PKM2 nuclear translocation in liver macrophages of HFD-fed mice (Fig. 6C and D and Supplementary Fig. 16). By contrast, HSPA12A overexpression increased PKM2 expression and nuclear translocation in Raw264.7 macrophages (Fig. 7A and Supplementary Fig. 17). In line with this, Hspa12a^o/e Raw264.7 macrophages expressed higher levels of monomeric and dimeric PKM2 than Hspa12a^n/e control cells in the presence of LPS (Supplementary Fig. 18). Collectively, these data suggest that HSPA12A regulates PKM2 expression and nuclear translocation during M1 macrophage polarization.

To determine the role of PKM2 during HSPA12A-mediated M1 macrophage polarization in response to LPS, Hspa12a^o/e Raw264.7 macrophages were treated with a small molecule, DASA-58. DASA-58 is a specific enzyme activator of PKM2, which can prevent PKM2 nuclear translocation (15,30,31). Notably, DASA-58 reversed the HSPA12A overexpression-induced nuclear translocation of PKM2 and expression of M1 markers (iNOS, MCP-1, and TNF-α) in LPS-treated macrophages (Fig. 7A and B). Taken together, these findings suggest that nuclear PKM2 mediates the HSPA12A-induced M1 macrophage polarization.

Next, we asked whether nuclear PKM2 mediates the paracrine effects of macrophage HSPA12A on hepatocyte steatosis. We found that overexpression of HSPA12A led to increased secretion of IL-1β, IL-6, TNF-α, and IL-12 by LPS-treated Raw264.7 macrophages; however, this increase was reversed by DASA-58 (Fig. 8A and Supplementary Fig. 19). ELISA assay detected no difference in HSPA12A levels in medium from Hspa12a^o/e and Hspa12a^n/e Raw264.7 macrophages (Supplementary Fig. 20).

CM collected from DASA-58–treated Hspa12a^o/e macrophage cultures was termed DASA-58 Hspa12a^o/e CM. Lipid deposition in AML-12 hepatocytes treated with DASA-58 Hspa12a^o/e CM was significantly lower than that in AML-12 hepatocytes treated with Hspa12a^o/e CM (Fig. 8B). In agreement, we noted a marked reduction in the expression of genes associated with de novo fat synthesis and lipolysis in AML-12 hepatocytes treated with DASA-58 Hspa12a^o/e CM compared with AML-12 hepatocytes treated with Hspa12a^o/e CM (Fig. 8C).

Simultaneous nuclear translocation of HSPA12A and PKM2 during M1 macrophage activation motivated us to investigate whether there is an interaction between them. Flag-tagged HSPA12A was immunoprecipitated from Hspa12a^o/e Raw264.7 macrophages treated with LPS or vehicle for 16 h. Precipitates were immunoblotted with antibodies specific for PKM2 and HSPA12A. PKM2 protein was recovered from the flag-tagged HSPA12A immunocomplexes, with more PKM2 recovered after LPS treatment (Fig. 8D). These data suggest that HSPA12A formed a complex with PKM2, and the formation of the HSPA12A-PKM2 complex was increased during macrophage polarization to an M1 phenotype.

To examine whether macrophage HSPA12A affects HFD-induced hepatic steatosis, we transplanted bone marrow from Hspa12a^−/− mice into recipient WT mice (Hspa12a^−/−/WT). Recipient WT mice transplanted with WT bone marrow served as controls (WT/WT). After a lethal irradiation, all the mice (15/15) without BMT died within 15 days; however, none of the BMT mice died (Supplementary Fig. 21A). HSPA12A expression in white blood cells from Hspa12a^−/−/WT mice was hardly detectable compared with that in WT/WT controls (Supplementary Fig. 21B and C). After an HFD for 14 weeks, liver weight, hepatocyte area, and lipid accumulation (as indicated by ORO staining and TAG content) in Hspa12a^−/−/WT mice were reduced compared with WT/WT mice (Supplementary Fig. 21D–F). In addition, expression of iNOS, TNF-α, TLR4, and MyD88 protein, along with NF-κB phosphorylation, in the livers of Hspa12a^−/−/WT mice was lower than in WT/WT mice fed an HFD (Supplementary Fig. 22).

Here, we identified macrophage HSPA12A as a novel regulator of hepatic inflammation and NASH pathogenesis. Thus, strategies aimed at inhibiting macrophage-specific HSPA12A may have potential as therapeutic interventions for NASH patients.

HSPA12A was first cloned from atherosclerotic lesions, and its cerebral expression is decreased in patients with schizophrenia (20,21,32). Recently, a study identified a correlation between increased HSPA12A expression and shortened survival of patients with hepatocellular carcinoma (22). However, there is no evidence that HSPA12A is the direct cause of any pathophysiological event. Here, we used human samples, mouse models, and cell culture models to show the following: 1) NASH patients display increased hepatic expression of HSPA12A and higher levels of circulating HSPA12A, which are positively associated with hepatic steatosis and injury, respectively; 2) HSPA12A is expressed preferably in liver macrophages rather than hepatocytes; 3) HSPA12A deficiency either globally or in macrophages attenuates HFD-induced hepatic steatosis, liver injury, and inflammatory M1 macrophage responses in mice; and 4) hepatocyte steatosis is regulated by the paracrine effects of macrophage HSPA12A, which are mediated, at least in part, by nuclear PKM2-dependent modulation of M1 macrophage polarization. Taken together, these results provide evidence that the paracrine effects of macrophage HSPA12A regulate hepatocyte steatosis and injury.

Activation of liver macrophages, and subsequent secretion of proinflammatory mediators, is a key event for onset and progression of NAFLD (9,10). Hepatic macrophages comprise resident (Kupffer cells) and recruited macrophages, all of which express macrophage markers such as CD68, F4/80, and CD11b. Macrophages are activated by LPS or free fatty acids, which induce an M1 phenotype and expression of cytokines, chemokines, and signaling molecules through TLR4 signaling (7,33,34). Subsequently, chemokines attract blood-derived monocytes/macrophages to the liver; these cells then release cytokines that drive NASH progression, which is characterized by increased hepatocellular lipid accumulation and damage (6,34). Indeed, diet-induced NASH is alleviated by depleting Kupffer cells and knockout of IL-1β and TLR4 (11,35,36). Thus, limiting polarization of macrophages toward an M1 phenotype is an attractive strategy for preventing progression of NAFLD (9,10). Here, we observed that deficiency of HSPA12A alleviated HFD-evoked onset and progression of NAFLD, which were concomitant with reduced macrophage recruitment and M1 polarization; HSPA12A deficiency also reduced expression of chemokines and proinflammatory cytokines in HFD-fed mice. By contrast, overexpression of HSPA12A increased M1 macrophage polarization in response to LPS, as indicated by increased expression of iNOS, IL-1β, TNF-α, CCL3, CCL4, and TLR4, and by secretion of IL-1β, IL-6, IL-12, and TNF-α. More importantly, CM from HSPA12A-overexpressing macrophages led to a marked increase in lipid deposition in hepatocytes. However, we observed no change in HSPA12A secretion by macrophages after LPS treatment, suggesting that elevated serum levels of HSPA12A in NASH patients may be due to secretion by other cell types. Considering that hepatocellular HSPA12A had no effect on its own lipid deposition, our results suggest that hepatocyte steatosis is regulated by the paracrine effects of macrophage HSPA12A. We detected only a modest increase in secretion of TNF-α, IL-1β, IL-6, and IL-12 by HSPA12A-overexpressing macrophages, suggesting that the relationship between macrophage HSPA12A and hepatocyte lipid deposition is regulated by the synergistic actions of a group of secreted cytokines and/or other as yet undetermined factors.

Next, we sought answers to the following two questions: does M1 polarization mediate the paracrine effects of macrophage HSPA12A on hepatocyte steatosis, and how does HSPA12A modulate macrophage polarization? Recent studies show that PKM2 is a critical determinant of macrophage activation and subsequent inflammatory responses. The PKM2 tetramer acts as a powerful glycolytic enzyme that regulates glycolysis in the cytosol. However, PKM monomers or dimers are enzymatically inactive and can translocate to the nucleus, where they act as cofactors to activate expression of IL-1β et al., ultimately leading to activation of M1 macrophages (15–17). Induction of PKM2 expression by LPS is mediated by the transcription factors NF-κB and PPARγ (14,15,19). Interestingly, we observed that PKM2 expression in the liver of NASH patients was upregulated and that this upregulation was more prominent in liver macrophages showing notable levels of nuclear translocation. The livers of HFD-fed mice also showed more nuclear translocation of PKM2, which was attenuated by HSPA12A deficiency, suggesting a regulation of HSPA12A in PKM2 nuclear translocation. Importantly, preventing nuclear translocation of PKM2 in macrophages reversed HSPA12A-induced M1 polarization and blocked the paracrine effects of macrophage HSPA12A on hepatocellular steatosis. Immunoprecipitation-Western blot analyses revealed that HSPA12A forms a complex with PKM2 in macrophages. Collectively, these results indicate that HSPA12A interacts with PKM2 and increases its nuclear translocation, thereby inducing M1 macrophage polarization and secretion of proinflammatory M1 cytokines; ultimately, this leads to fat accumulation in hepatocytes via paracrine effects (Fig. 8E). However, it is unknown how HSPA12A regulates PKM2 expression and how HSPA12A interacts with PKM2. Further studies need to address these issues.

In conclusion, we show that deficiency of HSPA12A has a beneficial effect on hepatic steatosis and injury in mice with HFD-induced NASH. The underlying mechanism involves paracrine interactions between macrophages and hepatocytes. The data suggest that inhibitors of macrophage-specific HSPA12A may improve treatment and management of nonalcoholic liver disease.

Acknowledgments. The authors thank the Translational Medicine Core Facilities, Medical School of Nanjing University, for generous help in mass spectrum analysis.

Funding. This work was supported by the National Natural Science Foundation of China (81870234, 81770854, 81571378, 81571290, 81370260, and 81371450), Jiangsu Province’s Outstanding Medical Academic Leaders program (15), the Priority Academic Program Development of Jiangsu Higher Education Institutions, the Collaborative Innovation Center for Cardiovascular Disease Translational Medicine, and Jiangsu Provincial Key Discipline of Medicine (ZDXKA2016003).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Q.K., N.L., H.C., L.D., B.X., and L.F. performed all animal study procedures and most of the in vitro experiments. X.Z., X.Ca., T.Q., and X.Ch. collected and analyzed human samples. Z.Z. scored livers for NASH features in humans and mice. C.L. developed the study concept and experimental design. Y.L. interpreted the data and wrote the manuscript. L.L. and Z.D. developed the study concept and experimental design, interpreted the data, and wrote the manuscript. Z.D. 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.