A genetic basis of mitochondrial DNAJA3 in nonalcoholic steatohepatitis-related hepatocellular carcinoma

Abstract

Background & Aims:

Nonalcoholic fatty liver disease (NAFLD) is the most common form of liver disease worldwide but only a subset of NAFLD individuals may progress to nonalcoholic steatohepatitis (NASH). While NASH is an important etiology of hepatocellular carcinoma (HCC), the underlying mechanisms responsible for conversion of NAFLD to NASH, and then to HCC are poorly understood. We aimed to identify genetic risk genes that drive NASH and NASH-related HCC.

Approach and Results:

We searched genetic alleles among the 24 most significant alleles associated with body fat distribution from a genome-wide association study of 344,369 individuals and validated the top allele in three independent cohorts of American and European patients (N = 1,380) with NAFLD/NASH/HCC. We identified a rs3747579-TT variant significantly associated with NASH-related HCC and demonstrated that rs3747579 is eQTL of a mitochondrial DNAJA3. We also found that rs3747579-TT and a previously identified PNPLA3 as a functional variant of NAFLD to have a significant additional interactions with NASH/HCC risk. HCC patients with rs3747579-TT had a reduced expression of DNAJA3 and had an unfavorable prognosis. Furthermore, mice with hepatocyte-specific Dnaja3 depletion developed NASH-dependent HCC either spontaneously under a normal diet or enhanced by diethylnitrosamine. Dnaja3-deficient mice developed NASH/HCC characterized by significant mitochondrial dysfunction, which was accompanied by excessive lipid accumulation and inflammatory responses. The molecular features of NASH/HCC in the Dnaja3 deficient mice were closely associated with human NASH/HCC.

Conclusion:

We uncovered a genetic basis of DNAJA3 as a key player of NASH-related HCC.

Introduction

Liver cancer is the third deadliest cancer and hepatocellular carcinoma (HCC) is the main liver cancer type (1). In recent years, HCC incidence continues to rise globally, despite our ability to effectively prevent hepatitis B or C virus-related HCC (2). Changing trends in etiologies such as diet-related liver diseases may contribute to the rising incidence. For example, unhealthy diets may induce nonalcoholic fatty liver disease (NAFLD), which affects approximately 25% of the world’s population. However, only one in five individuals with NAFLD progress to nonalcoholic steatohepatitis (NASH) with some advancing to liver cirrhosis (3). Individuals with metabolic syndrome associated NASH have a high risk of developing HCC, which may contribute to the rising incidence of HCC in the United States (4). The genetic basis for what drives the transition of NAFLD to NASH and then to HCC is unknown.

Several recent prospective cohort studies including familial aggregation and twin studies have provided strong evidence that NAFLD/NASH are heritable traits (5, 6). Subsequent epidemiological and genome-wide association studies (GWAS) have linked several candidate genetic variants to NAFLD (7, 8). One variant PNPLA3 I148M has been shown as a significant genetic risk factor of NAFLD by altering lipid droplet dynamics (9). Yet, the existing risk variants account for only a fraction of NASH/HCC heritability, highlighting the paradox that while not every obese or NAFLD individual develops NASH, some non-obese individuals do (10). The key question remains as to why only some of the individuals with NAFLD-related variants will progress from NAFLD to NASH and then HCC. Crucially, several studies pinpoint regional body fat distribution, even when adjusted for BMI, as an inheritable risk for detrimental metabolic outcomes (11–13). This distribution is intertwined with metabolic syndrome, playing a pivotal role in NASH/HCC development (14–16). Thus we hypothesized that some of the body fat distribution-associated genetic variants may be linked to the risk of NASH/HCC.

Here, by searching a database of body fat distribution-related variants (13) to be associated with NASH-related HCC and by generating a genetically engineered mouse model, we aimed to determine genes that drive NASH and NASH-related HCC. We identified a genetic variant rs3747579 linked to the reduced expression of mitochondrial chaperone Hsp40, also known as DnaJ Heat Shock Protein Family (Hsp40) Member A3 (DNAJA3) (17), and its association with NASH-related HCC. Experimentally, mice with hepatocyte-specific deletions of Dnaja3 developed NASH and HCC. Mice with Dnaja3 ablation progressively developed fatty liver, NASH, and cirrhosis/HCC phenotypes. We also demonstrated that hepatic Dnaja3 deficiency impaired mitochondrial function in hepatocytes, which led to compensatory excessive lipid accumulation, resulting in inflammation and tumor progression. Our results may provide a genetic basis of DNAJA3 may be useful as a risk marker for NASH and that DNAJA3-mediated signaling may be therapeutically exploited for NASH-related HCC.

Materials and Methods

NCI-UMD cohort

The demographic characteristics of participant groups of the NCI-UMD cohort are summarized in Table S1 and S2. All clinic measurements of the NCI-UMD cohort were covered by NCT00913757 (clinicaltrials.gov) as previously described (18). Serum, whole blood, plasma, or Cheek Swab were collected at the time of interview. Sera and saliva were stored at −80°C for research. The study was approved by the NCI institutional review boards. All participants provided written informed consent as previously described (18).

Animal care and generation of Dnaja3^−/− mice

All animal work was approved by the Institutional Animal Care and Use Committee (IACUC) and conducted according to guidelines established. We used the ARRIVE1 reporting guidelines (19). The details of generation of Dnaja3^flx/flx mice on a C57BL/6N background were available in Supplemental Methods. Liver-specific Dnaja3^−/− (Alb-Cre; Dnaja3^flx/flx) mice were generated by crossing Dnaja3 floxed (Dnaja3^flx/flx) mice with Albumin-Cre mice (a gift from Dr. Ann-Ping Tsou, National Yang-Ming Chiao Tung University). Dnaja3^−/− and Dnaja3^flx/flx mice were littermates. Mice were sacrificed at a time point indicated in the figure legends using CO₂ euthanasia, and blood and tissue samples were harvested for morphological, biochemical, and functional measurements. If mice showed severe weakness or 20% weight loss, mice were euthanized with CO₂. Mice numbers of five were used unless stated. All mice used were males unless stated. The age and gender of the mice were indicated in the figure legends. The details for generation of animal models and treatments as well as other experimental protocols are available in Supplemental Methods.

Results

Identification of the DNAJA3 rs3747579-TT variant linked to NASH and HCC

Using the Illumina OmniExpress SNP array to analyze the NCI-UMD case-control cohort (18), which includes 484 healthy volunteers, 499 individuals without NASH (non-NASH, either with chronic liver diseases due to other etiologies or with HCC), and 26 individuals with confirmed NASH (including at-risk and HCC) (Fig. 1A, Table S1 and S2), we identified the association of 24 SNPs between the individuals with and without NASH from the NCI-UMD cohorts. Among these, four coding variants (rs3747579, rs8052655, rs3764002, rs897453) had a Bonferroni-corrected p-value, with rs3747579 showing the strongest association in four genetic models (codominant, dominant, recessive, and log-additive models) (Figs. 1B and S1A). The associations were assessed using the max-statistic or standard statistic (Tables S5–S6). The significance persisted after adjusting for age and gender (Table S7). Among the four NASH-associated variants, rs897453 was a strong eQTL for PEMT (Table S5), which is hepatic integral membrane protein localized to the endoplasmic reticulum (ER) and a risk allele for lean NASH (20). In addition, rs3764002 was a eQTL for CMKLR1 known to be associated with protection against NASH (21). These results provide further confidence to our initial approach. We focused on rs3747579 as it had the strongest association. We found that rs3747579-TT was significantly associated with NASH with odds ratio of 16.1 (95% CI: 2.0 - 128.3; p = 1E-03) (Fig. S1B and Table S8). Furthermore, rs3747579-TT was associated with NASH in the UCSF cohort with a similar trend but it was statistically no significant, some comparisons were limited by available data and therefore may not have been adequately powered to detect meaningful differences (Table S9). We then combined both NCI-UMD cohort and UCSF cohort as the U.S. cohort and found consistent data that rs3747579-TT was associated with NASH (p = 7E-04) (Table S10). We also examined a French-Belgian cohort of 334 NAFLD patients to further determine whether rs3747579-TT is associated with NAFLD/NASH-related HCC, and found consistent results (Table S11). Noticeably, rs3747579-TT was also significantly associated with HCC (odds ratio: 2.14; p = 4.9E-03) (Table S12). Collectively, these data indicate that rs3747579-TT may be linked to NASH and HCC.

Figure 1. Identification of Candidate Genes for NASH and NASH-Related HCC Risk Variants.

(A) Schema of an integrated analysis of lipid homeostasis SNP (Nat Gent, 2019) and NASH risk. The NASH risk effect of allele observed from NASH disease cases (n = 26), including at-risk (AR) of NASH (NASH-AR, n = 19) and NASH-related HCC (NASH-HCC, n = 7) individuals compared with population controls (PC, n = 484) or chronic or alcoholic liver diseases cases (non-NASH, n = 499) including at-risk (non-NASH-AR, n = 350) and HCC (non-NASH-HCC, n = 149) individuals. In addition, to determine the nature of the relationship between lipid homeostasis SNPs and the clinicopathological characteristics of NASH-related HCC, patients with HCC (n=156) were divided into subgroups according to lipid homeostasis SNP status to character the NASH feature, such as cholesterol; the resulting were validated in Dnaja3 knockout mice.

(B) Manhattan plot based on the dominant genetic model for the association of NASH-related HCC risk coding variants that overlapped with 24 common variants of lipid homeostasis. Red dotted line indicates the threshold for NASH group (n=26) compared with non-NASH group (n= 499) after Bonferroni correction. Blue dotted line shows the nominal level of significance.

(C) Odds ratios and 95% confidence intervals of rs3747579 genotype among non-NASH (n = 499) or NASH (n = 26) relative to the control (n = 484) group. Further details are available in Table S8.

(D) The associations between rs3747579 and NASH clinical features. The box plot shows −log10 (p-values) based on the effect of the different rs3747579 genotypes in the HCC samples (n=156). Further details are available in Table S14.

(E) The Phenome-wide association (PheWAS) plot shows the top ten significant associations of rs3747579 for all available traits in the Common Metabolic Diseases Knowledge Portal (CMDKP), generated by bottom-line meta-analysis across all datasets. n refers to sample size.

(F) Kaplan-Meier curves for the overall survival according to the rs3747579 genotype of HCC patients from UMD (CC+CT: n=60; TT: n=28) cohort adjusting for age, sex and race/ethnicity.

(G) Scatter plot of the NASH-specific eQTL association between each SNP’s genotype and DNAJA3 m-value in the liver from the GTEx database (n=208). The p-values display association with disease featural phenotype of the NCI-UMD cohort among non-NASH (n = 499) (left panel), and NASH group (n=26) (right panel) relative to the control (n = 484) group. Each dot represents an eQTL of DNAJA3 (n =21). The m-value indicates the posterior probability that the eQTL affects expression in liver tissue. The eQTL affects expression in liver tissue if the m-value is ≥0.9. Further details are available in Table S15.

(H) Odds ratios and 95% confidence intervals of DNAJA3 expression among rs3747579 genotype. Summary statistics of DNAJA3 eQTL was obtained from the IEU OpenGWAS (n= 31,300).

(I) RT-PCR analysis of relative mRNA expression of DNAJA3 gene in liver cancer cell lines carrying different rs3747579 genotypes. Data were first normalized to GAPDH to get ΔCt. Relative mRNA was then calculated by 2(-ΔCt).

(J) Immunoblots showing DNAJA3 and α-Tubulin protein levels in liver cancer cell lines with different rs3747579 genotypes.

Data are represented as mean ± SD.

We further compared the strength of association between rs3747579-TT and NASH with previously reported functional SNPs of PNPLA3 and TM6SF2 (22) to be linked to NAFLD in the NCI-UMD cohort. The rs3747579-TT significance persisted in NASH-HCC after adjusting for genetic variants of PNPLA3 and TM6SF2 (Table S13). We also included data in all 24 lipid homeostasis-related SNPs for comparison and again rs3747579-TT had the strongest association independent of PNPLA3 and TM6SF2 (Table S13). Noticeably, rs3747579 and genetic variants of PNPLA3 showed highly significant additional interactions with NASH-HCC risk (Fig. S1C). The rs3747579-TT was also significantly associated with pathological and clinical features of NASH, e.g., increased cholesterol, and triglycerides in the NCI-UMD cohort (Fig. 1D and Table S14). Using phenome-wide association studies (PheWAS), we identified waist-hip ratio, triglyceride levels, and blood counts as significant features correlated with rs3747579 (Fig. 1E).

To understand if rs3747579-TT was associated with a subgroup of HCC with different tumor biology, we examined overall survival of HCC patients based on the rs3747579 genotype status in the NCI-UMD cohorts. We found that the overall survival time of HCC patients with rs3747579-TT was significantly shorter than that of those with CC and CT alleles in UMD-NCI cohort after adjusting for age, sex and race/ethnicity (Fig. 1F). Noticeably, we found similar results when the analysis was restricted to the Caucasian ancestry in both UMD-NCI and TCGA cohorts (Fig S2A). We could not evaluate African ancestry separately in the UMD-NCI cohort since cases with rs3747579-TT were not included. While the TCGA cohort also included Asian ancestry, we did not find a significant survival association linked to rs3747579-TT in Asian HCC patients (Fig. S2B).

We performed eQTL analysis located upstream of the rs3747579 locus using the GTEX-Liver and TCGA-LIHC datasets. Intriguingly, despite its location in exon 7 of the CORO7 gene, rs3747579 exhibited a pronounced association with DNAJA3 expressions, compared to CORO7 (Fig. S3A). Moreover, we found that rs3747579-TT was significantly associated with decreased expression of DNAJA3, but not CORO7, compared to rs3747579-CC or CT allele in TCGA-LIHC cohorts (Fig. S3B). Furthermore, using the human protein atlas database (23), we found that DNAJA3, but not CORO7, is highly expressed in the liver and in hepatocytes (Figs. S3C–D). Taken together, we concluded that rs3747579 is an eQTL linked to DNAJA3 expression rather than CORO7.

We further explored the influence of rs3747579 on DNAJA3 expression in NASH-related characteristics in the UMD-NCI cohorts. These SNPs were located within a 100 kb region upstream of the DNAJA3 transcription start site and mapped to eQTL using data from the GTEx-liver cohort. We found a positive association between the m-value of DNAJA3-related eQTL and NASH-HCC, but not with non-NASH-related features (Figs. 1G). Consistently, rs3747579 in DNAJA3 eQTL had the strongest association with NASH-HCC (Fig. 1G and Table S15). Moreover, a large eQTL database (eQTLgene) confirmed that the expression pattern of DNAJA3 is dependent on the rs3747579 genotype (Fig. 1H). To further determine whether DNAJA3 rs3747579-TT is a bona fide allele linked to DNAJA3 expression in hepatoma cells, we searched the CCLE database for rs3747579 variants in liver cell lines (24) that are available in our cell repository. We found and validated rs3747579-CC in HepG2 and Hep3B, and rs3747579-TT in SNU449 and SNU475 (Fig. S4A–S4B). We confirmed that the mRNA and protein levels of DNAJA3 were significantly lower in cells with the TT allele compared to the CC allele (Fig. 1I–J). Similar results were obtained when assessing HCC cell lines that mimic the NASH phenotype (Fig. S4C). To assess if SNP rs3747579 affects transcription factor binding sites, we probed RegulomeDB for REMs within its genomic region. Chip-seq analysis revealed that rs3747579-CC variant is a part of a binding site of RBFOX2 (Fig. S5A), a transcriptional activator and regulates cholesterol homeostasis (25). Cells with RBFOX2 knockdown displayed considerably lower DNAJA3 expression than control cells (Fig. S5B). Furthermore, a luciferase reporter assay, controlled by the DNAJA3 promoter and encompassing rs3747579, demonstrated elevated luciferase activity for the rs3747579-CC allele relative to its rs3747579-TT counterpart (Fig. S5C).

Next, we conducted a functional study to explore the role of DNAJA3 in NASH-HCC. We performed a gene-based PheWAS and found a strong association between DNAJA3 and serum lipids, including triglyceride, and hepatic damage marker alkaline phosphatase (Fig. S6A). In addition, we found the expression of DNAJA3 was significantly associated with fatty liver disease (Fig. S6B). Knockdown DNAJA3 led to a marked increase in lipid droplet accumulation compared to control cells, whereas CORO7 knockdown did not produce such an effect. However, neither knockdown resulted in cell death (Figs. S6C–S6F). DNAJA3 expression was significantly decreased in NASH compared to normal (Fig. S7A), and in NASH-HCC compared to non-NASH-HCC (Fig. S7B–S7C). Furthermore, DNAJA3 expression was significantly decreased in HCC compared to the adjacent non-tumor liver tissues in all examined cohorts (Figs. S7D–S7G). In addition, DNAJA3 expression was significantly reduced in cirrhotic livers compared to non-cirrhotic livers (Figs. S7H–S7I). Consistent with the risk allele data, DNAJA3 expression in HCC was significantly associated with overall survival (Figs. S7J–S7K). Taken together, rs3747579-TT may act as a disease allele associated with a reduced DNAJA3 expression linked to NASH and NASH-related HCC.

Dnaja3 deficient Mice Develop NAFLD, NASH, and HCC

To validate the function of Dnaja3, mice with a conditional knockout of Dnaja3 in hepatocytes (Dnaja3^+/− and Dnaja3^−/−) were constructed to examine the influence of reduced gene expression on progression of NASH to HCC (Fig. 2A–2B). The details for liver conditional knockout mice of Dnaja3 generation are available in Supplemental Methods and Fig. S8. In brief, a genetically modified mouse line with exons 1 and 2 of Dnaja3 flanked by loxP elements (Dnaja3^f/f) was generated and crossed with Albumin-Cre (Alb-Cre) to generate the Dnaja3 mice (Fig. 2A). PCR genotyping, mRNA expression, and protein levels analyses revealed an efficient depletion of Dnaja3 in hepatocytes of Dnaja3^−/− mice (Figs. S9A, 2C and S9B). Interestingly, we observed that Dnaja3^−/− mice were undergoing retarded growth during the first twelve months under normal diet, especially in the first month (Figs. S9C–S9D). The Dnaja3^−/− mice showed elevated serum levels of aspartate aminotransferase (AST) and alanine aminotransferase (ALT) (Figs. S9E and 2D). We found evidence of fatty liver in 1-month-old Dnaja3^−/− mice and NASH in 13 months old Dnaja3^−/− mice under normal diet (Figs. 2E–2H and S9F), with a similar tendency of the body weight change (Fig. S9D). In evaluating the commonly used NASH activity score (NAS, a composite score of steatosis, hepatocellular ballooning and lobular inflammation) for histological staging and grading system of human NASH, we found an elevated NASH score increases a 6-fold in the liver of Dnaja3^−/− mice compared to in heterozygous or Dnaja3^f/f mice (p < 0.005) (Figs. 2G). In addition, the serum cholesterol levels were elevated in Dnaja3^−/− mice (Fig. 2I).

Figure 2. Mice with Hepatocyte-Specific Dnaja3 Depletion Developed NASH-Dependent HCC Spontaneously.

(A) Generation of mice with Dnaja3 deletion in hepatocytes. Two loxP (locus of X-over P1) sites were inserted (Black arrowhead) at exons 1 and 2 of the Dnaja3 gene as the targeted locus (red solid boxes). Cross-breeding of Dnaja3^f/f mice with Alb-Cre mice led to hepatic Dnaja3 deficiency. The details of generation of Dnaja3^flx/flx mice on a C57BL/6N background were available in Supplemental Methods.

(B) Pathological progression of male Dnaja3^−/− mice for DEN-induced and spontaneous mouse models.

(C) mRNA levels of Dnaja3 in liver tissues from Dnaja3^f/f, Dnaja3^+/−, and Dnaja3^−/− mice (n = 5).

(D) Enzymatic activity of ALT was monitored in serum of spontaneous male Dnaja3^f/f, Dnaja3^+/−and Dnaja3^−/− mice (n = 5).

(E) Representative gross liver images from 1-month-old spontaneous Dnaja3^f/f and Dnaja3^−/− male mice (left panel). Images of hematoxylin-eosin (H&E) staining of representative liver sections obtained from 1-month-old male mice (middle panel). Sirius Red staining of liver sections from 1-month spontaneous Dnaja3^f/f and Dnaja3^−/− mice (right panel).

(F) H&E staining of liver sections from 13-month spontaneous male mice depicting the individual component of steatohepatitis: Steatosis, Hepatocyte Ballooning and Inflammation as indicated by the yellow arrow.

(G) Steatosis, Hepatocyte Ballooning and Inflammation scores were quantified from Fig. 2F. Results were expressed as mean ± SD.

(H) Representative images of Oil Red O staining of liver sections from 13-month spontaneous male Dnaja3^f/f and Dnaja3^−/− mice.

(I) Serum level analysis of cholesterol was monitored at 13-month spontaneous male mice (Dnaja3^f/f and Dnaja3^−/−: n=7; Dnaja3^+/−: n=8).

(J) Representative gross liver images from 11, 13, and 21-month-old spontaneous Dnaja3^f/f, Dnaja3^+/− and Dnaja3^−/− male mice.

(K) Summary of the incidence of Dnaja3-mediated hyperplastic nodules, HCC (advanced and early-stage), and the presence of no over pathological lesions in the livers from 11, 13, and 21-month-old spontaneous male mice (n = 5).

Histologically confirmed the male Dnaja3^−/− mice developed hepatocellular nodules at the age of 11 months under normal diet (Figs. 2J, 2K and S9G). Furthermore, 100% of the Dnaja3^−/− and heterozygous male mice spontaneously developed liver tumors by 13 and 21 months, respectively (Figs. 2J and 2K). In addition, Ki67-positive cells in the liver were elevated in Dnaja3^−/− mice as a cell growth marker (Fig. S9H). By 21 months, some of the knockout mice also developed lung tumor lesions (Fig. S9I). These results indicate that hepatic Dnaja3 deficiency may be associated with the NASH phenotype and tumorigenesis.

Dnaja3 deficient mice sequentially developed a fatty liver, NASH, advanced fibrosis, and HCC and enhanced by diethylnitrosamine

To create DNA damage-induced metabolic changes resembling human NASH-HCC, we subjected Dnaja3 deficient mice to diethylnitrosamine (DEN), a known carcinogen that triggers DNA damage, inflammation, and hepatocarcinogenesis (26). We found that DEN accelerated HCC development in Dnaja3 deficient male mice with a significantly shorter time compared to heterozygous or Dnaja3^−/− mice without DEN (p <0.0001)(Figs. 2J–2K and 3A–3C). Hepatocellular nodules could be observed in the 3-month DEN-treated Dnaja3^−/− mice (Figs. 3A–3C). Histological examination with H&E staining showed that these macroscopic tumors represented early-stage to advanced HCC (Fig. 3B). At 6 months, the DEN-treated Dnaja3^−/− mice exhibited more tumors and larger tumor size than DEN-treated Dnaja3^+/− or Dnaja3^f/f mice (Figs. 3A–3C). Similar results could be found in female mice (Fig. S10). However, a sex disparity was observed in our transgenic mice models with or without DEN-treatment, where the earlier onset of HCC occurred in male mice than in the female mice (Fig. 2J–2K, 3A–3C and S10). Induction of NASH-HCC was accompanied by elevated levels of an HCC marker alpha-fetoprotein (AFP) in DEN-treated Dnaja3^−/− mice (Fig. 3D). Moreover, 40% of the DEN-treated Dnaja3^−/− mice developed lung metastasis, while no metastatic tumor was found in the DEN-treated Dnaja3^f/f mice (Fig. 3E). Furthermore, the overall survival of Dnaja3^−/− mice was significantly worse than that of the heterozygous or Dnaja3^f/f mice both without and with DEN treatment (Fig. 3F). Notably, DEN-treated Dnaja3^−/− mice had a 60% decrease in median survival compared to the untreated Dnaja3^−/− mice (Fig. 3F).

Figure 3. Hepatic Dnaja3 Deficiency Promotes DEN-Induced HCC.

(A) Gross liver images from DEN-treated male Dnaja3^f/f, Dnaja3^+/−, and Dnaja3^−/− mice.

(B) H&E staining of liver sections from DEN-treated male Dnaja3^f/f, Dnaja3^+/− and, Dnaja3^−/− mice. Yellow star indicating the tumors. N, normal area; T, tumor area.

(C) Summary of hyperplastic nodules, HCC (advanced and early stage), and no overt pathological lesions in the livers from DEN-treated male mice (n = 5).

(D) Immunoblots showing DNAJA3, AFP, and GAPDH protein levels in the livers of DEN-treated mice.

(E) Representative gross lung images from 10-month-old DEN-treated Dnaja3^f/f, Dnaja3^+/−, and Dnaja3^−/− male mice (upper panel). Images of hematoxylin-eosin (H&E) staining of representative lung sections obtained from 10-month-old DEN-treated Dnaja3^f/f, DNAJA3^+/−, and Dnaja3^−/− male mice (lower panel).

(F) Kaplan–Meier analyses showing the overall survival of spontaneous (upper panel) or DEN-treated (lower panel) Dnaja3^f/f, and Dnaja3^−/− male mice (DEN-treated group: Dnaja3^f/f: n=20; Dnaja3^−/−:n=11; DNAJA3^+/−: n=6; Spontaneous group: Dnaja3^f/f and Dnaja3^−/−: n=16; DNAJA3^+/−: n=11).

Grossly and histologically, the liver of DEN-treated Dnaja3^−/− mice displayed many typical characteristics of human NASH (Figs. 4 and S11). Within a month, these mice exhibited noticeable liver enlargement and discoloration, indicating early lipid accumulation (Fig. S11A). Electron microscopy (EM) confirmed lipid droplet buildup in the liver after 1-month of DEN treatment (Fig. S11B). Histological analysis confirmed the presence of extensive microvesicular steatosis by 1 month and large droplet steatosis by 6 months (Fig. 4A). By 3 months, the DEN-treated Dnaja3^−/− mice expressed significantly higher levels of pro-inflammatory cytokines and chemokines (Fig. S11C), along with the presence of lymphocytic infiltrates and ectopic lymphoid-like structures (Fig. 4B), the typical features of NASH (27). An increase of F4/80-positive macrophage infiltrates (Kupffer cells) was evident in the liver of both 3- and 6-month-old Dnaja3^−/− mice (Fig. S11D). However, these features were absent in the DEN-treated Dnaja3^f/f mice during the same period.

Figure 4. The Histopathology of Dnaja3 Deletion Mice Associated with the Development of NAFLD and NASH Phenotype.

(A) H&E staining of liver sections from DEN-treated Dnaja3^f/f, Dnaja3^+/−, and Dnaja3^−/− male mice.

(B) H&E staining of liver sections from 6-month DEN-treated Dnaja3^f/f, Dnaja3^+/− and, Dnaja3^−/− male mice. Black arrow indicates lymphocytic infiltrate-like structures.

(C) Enzymatic activity of ALT and AST was monitored in serum of 6-month DEN-treated and spontaneous Dnaja3^f/f, Dnaja3^+/−, and Dnaja3^−/− male mice (n=6).

(D) Sirius Red staining of liver sections from 6-month DEN-treated Dnaja3^f/f, Dnaja3^+/−, and Dnaja3^−/− male mice.

(E) H&E staining of liver sections from 6-month DEN-treated male mice depicting the individual component of steatohepatitis: Steatosis, Hepatocyte Ballooning and Inflammation as indicated by the yellow arrow.

(F) The histological score for Steatosis, Hepatocyte Ballooning, and Inflammation. The score was quantified from 6-month DEN-treated male mice. Results were expressed as mean ± SD.

(G) The histological score for NASH activity. NAS scores were quantified from 6-month DEN-treated male mice. Results were expressed as mean ± SD.

The liver of DEN-treated Dnaja3^−/− mice showed a marked increase in apoptosis compared to their Dnaja3^f/f counterparts as evidenced by TUNEL assay results (Fig. S11E) and further validated by immunoblot analyses of cleaved caspase-3 (Fig. S11F). Liver damage was more pronounced in the 6-month DEN-treated Dnaja3^−/− mice, accompanied by elevated serum levels of AST and ALT (Fig. 4C). Moreover, markers of oxidative stress Nfe2l2 (28) and Grp78 (29) were significantly increased in the liver of DEN-treated Dnaja3^−/− mice (Fig. S11G). The proliferation marker Ki67 and collagen deposition (Sirius red staining) were also elevated in the liver of 6-month DEN-treated Dnaja3^−/− mice (Figs. S11H and 4D). Consistently, the NAS values were significantly elevated a 7-fold in the liver of 6-month DEN-treated Dnaja3^−/− mice (p < 0.005) (Figs. 4E–4G). Taken together, these results indicate that hepatic Dnaja3 deficiency along with DEN-induced liver damage may accelerate NASH development.

Dnaja3 deficiency contributes to the mitochondrial dysfunction and excessive fat accumulation

To determine molecular alterations of hepatic Dnaja3 deficiency, we analyzed transcriptome profiles of the livers from 1-, 3-, or 6-month-old DEN-treated Dnaja3^−/− and Dnaja3^f/f mice. Among the top-20 downregulated pathways in the DEN-treated Dnaja3^−/− livers, many were related to the mitochondrial function, such as mitochondrial translation, oxidative phosphorylation (OXPHOS), and fatty acid oxidation (FAO) pathways (Fig. 5A; left panel), and these features persisted over time (Fig. 5A; right panel). In contrast, the top upregulated pathways in the livers of DEN-treated Dnaja3^−/− mice were related to immune cell infiltration and collagen organization (Fig. 5A). Furthermore, the altered pathways in the liver of DEN-treated Dnaja3^−/− mice could also be found in patients with NAFLD, NASH, cirrhosis, and HCC (Fig. S12A). Several of the most affected genes in Dnaja3^−/− livers, including LGALS3, SPP1, and IGFBP1(30), are also linked to human NASH (Fig. S12B). We found a significant concordance of the Dnaja3^−/− mice gene expression patterns at 3 or 6 months to the human NASH transcriptome (FDR 0.02 and 0.11, respectively) (Fig. 5B–5C). Moreover, by 6 months, every Dnaja3^−/− mouse had developed HCC, and the gene expression from these mice strongly correlated with human HCC data (Fig. S12C–S12E). In addition, the pattern of gene expression within the transcriptome from Dnaja3^−/− mice at 3 or 6 months demonstrated strong concordance with human cirrhosis (Fisher p-value: 0.04 and 0.04, respectively) (Fig. S12D–S12E).

Figure 5. Hepatic Dnaja3 Deficiency Causes Dysregulated Mitochondrial and Fatty Acid Metabolism.

(A) Bar plots showing the most significant gene sets enriched in the livers from DEN treated Dnaja3^−/− versus Dnaja3^f/f mice by gene set enrichment analysis (GSEA). NES indicates the normalized enrichment score in the GSEA algorithm (left panel). Heatmap showing NES of 40 gene sets (rows) across clustering. Each pathway is significantly up- or down-regulated in the liver tissue from Dnaja3^−/− mice versus Dnaja3^f/f male mice for 1, 3, 6 months (right panel).

(B) Similarity between global transcriptome of livers from DEN treated Dnaja3^−/− or Dnaja3^f/f mice for 1, 3, 6 months and global transcriptome in liver biopsy tissues from 7 human NASH patients and 26 NAFLD individuals in SA cohort using subclass mapping algorithm. FDR values are represented as colors in a heatmap.

(C) Similarity between global transcriptome of livers from DEN treated Dnaja3^−/− or Dnaja3^f/f mice for 1, 3, 6 months and global transcriptome in liver biopsy tissues from 7 human NASH patients and 26 NAFLD individuals in SA cohort using subclass mapping algorithm. A heatmap showing p-values from the Fisher test.

(D) Transmission electron micrographs analyses of liver samples from DEN-treated Dnaja3^f/f and Dnaja3^−/− mice. N, nuclei; M, mitochondria; LD, lipid droplets; ER, endoplasmic reticulum.

(E) mRNA levels of lipid droplet protein in the livers of male mice. Results were expressed as mean ± SD.

(F-G) Serum level analysis of cholesterol (E) and triglyceride (TG) (F) were monitored at 6-month DEN-treated male mice (n=7).

Given DNAJA3’s role as a mitochondrial chaperone, we utilized EM to study hepatocyte mitochondrial morphology. In spontaneous Dnaja3^−/− mice, hepatocytes exhibited swollen, round, and hypodense mitochondria, unlike their Dnaja3^f/f counterparts (Fig. S13). By the first month of DEN treatment in Dnaja3^−/− mice, pronounced mitochondrial swelling was evident, characterized by complete cristae loss and cytoplasmic material intrusion (Fig. 5D). By the third month, several mitochondria appeared diminutive, showcasing either rudimentary cristae or significant proteolysis (Fig. 5D). In contrast, DEN-treated Dnaja3^f/f livers showed mitochondria with intact membranes and preserved cristae structure (Fig. 5D). Subsequent assays revealed elevated mRNA levels of lipid droplet proteins, Plin2 and Plin3 (Fig. 5E). This was synchronized with heightened serum cholesterol and decreased triglycerides (Figs. 5F–G). These findings suggest that Dnaja3 deficiency could precipitate mitochondrial aberrations, further intensifying hepatocyte steatosis.

Dnaja3 deficiency developed HCC links to mitochondrial bile acid production and fatty acid oxidation

To gain insights into proteomic changes during NASH to HCC progression, we performed an LC-MS/MS-based proteomic analysis with livers collected from 6-month DEN-treated mice. Differentially altered proteins were subjected to pathway analysis. We found that many mitochondrial related pathways, including mitochondrial FAO, amino acid catabolism, and mitochondrial translation, were enriched, while lipid biosynthesis and cell cycle proteins were significantly upregulated in liver tumors of DEN-treated Dnaja3^−/− mice (Fig. 6A). There was a significant enrichment in the lipid biosynthesis pathways, but a significant depletion of pathways related to mitochondrial functions in tumors from DEN-treated Dnaja3^−/− mice (Fig. 6B). We found a good concordance of affected pathways between transcriptomic and proteomic analyses (Figs. 6C–D).

Figure 6. Tumor Development is Associated with Mitochondrial Dysfunction and Fatty Acid Metabolic Dysregulation in DEN-Treated Dnaja3^−/− Mice.

(A) Bar plots showing the most significant gene set enriched in protein levels of livers from DEN treated Dnaja3^−/− versus Dnaja3^f/f mice by GSEA (left panel). Each pathway is significantly up-or down-regulated in the Dnaja3^+/− or Dnaja3^−/− mice versus Dnaja3^f/f male mice for 6 months (right panel).

(B) Volcano plots showing gene sets enriched in the protein levels of liver tissue from DEN-treated Dnaja3^−/− male mice versus Dnaja3^f/f (gray dots; Welch’s t-test). Selected enriched mitochondrial function (upper panel) and lipid biosynthesis (lower panel) pathways are colored.

(C) Bar plots showing the most significant gene set enriched in mRNA levels of livers from DEN treated Dnaja3^−/− versus Dnaja3^f/f male mice by GSEA (left panel). Each pathway is significantly up-or down-regulated in the Dnaja3^+/− or Dnaja3^−/− mice versus Dnaja3^f/f male mice for 6 months (right panel).

(D) Venn diagram depicting the overlapping pathways between the RNA-Seq and proteomic analysis with significantly altered functional categories in liver tissue from Dnaja3^−/− male mice versus Dnaja3^f/f male mice.

We then searched DNAJA3-interacting proteins using three protein-protein interaction network databases and found 17 mitochondrial proteins potentially binding to DNAJA3, and these proteins could be grouped as 4 clusters by Markov Clustering (Fig. S14A). Interestingly, we found that many DNAJA3-interacting mitochondrial proteins were reduced in the liver of DEN-treated Dnaja3^−/− mice (Fig. S14B). Pathway enrichment analysis revealed NAFLD as the top significant hit among the DNAJA3-interacting client proteins (Fig. S14C). Consistently, pathway analysis of DNAJA3 interacting proteins revealed enrichment of biological processes related to mitochondrion functions (Fig. S14D).

We next mapped our proteomic data to known human metabolic networks and found that the mitochondrial protein levels were highly downregulated in the liver of DEN-treated Dnaja3^−/−, which was linked to several metabolic pathways, such as OXPHOS, FAO, and bile acid biosynthesis (Fig. 7A). In parallel, protein levels in fatty acid and cholesterol biosynthetic pathways as well as cytosolic proteins involved in glycolysis, the pentose phosphate pathway (PPP), and the serine/glycine biosynthetic pathway were highly upregulated in the liver of DEN-treated Dnaja3^−/− mice (Fig. 7A). We validated the expression of the key genes using RT–PCR analyses (Figs. 7B–7C). We also performed metabolomic profiling and found that, consistent with proteomic data described above, both bile acids and the FAO end-product (Acetyl-CoA) were significantly decreased, while the amount of saturated (SFA) and monounsaturated fatty acids (MUFA) as well as pyruvate and lactate (products of glycolysis) and 6PG (a key intermediate in the PPP) were significantly increased in the livers of DEN-treated Dnaja3^−/− mice (Fig. 7A and Table S16 and S17).

Figure 7. Dnaja3 Deletion Induces Lipid Accumulation and Cholesterol Synthesis Through Blocking Mitochondrial Function.

(A) A diagram describing the changes in different metabolic pathways in livers from Dnaja3^−/− mice compared to Dnaja3^f/f male mice. Red square and blue square symbols represent down- and upregulated, respectively, metabolic enzymes in livers from Dnaja3^−/− mice versus Dnaja3^f/f mice. Red round and blue round symbols represent down- and upregulated metabolites in livers from Dnaja3^−/− mice versus Dnaja3^f/f male mice.

(B and C) Differential expression of OXPHOS (B) and fatty acid metabolic (C) enzymes in livers from 6-month DEN-treated Dnaja3^f/f, Dnaja3^+/−, and Dnaja3^−/− male mice. Results were expressed as mean ± SD. *p < 0.05; **p < 0.01; ***p < 0.005 (n=3).

To further explore mitochondrial functions linked to DNAJA3 deficiency-induced malignant transformation of hepatocytes, we analyzed HCC cells isolated from DEN-treated Dnaja3^−/− and Dnaja3^f/f mice. We found a reduced expression of mitochondrial OXPHOS subunits, belonging to complexes I, III, and V, in Dnaja3^−/− tumor cells compared to those of Dnaja3^f/f (Fig. S15A). We also characterized the level of MitoSOX fluorescence in the Dnaja3^−/− cells, which serves as an indicator of mitochondrial ROS levels by flow cytometry and found that Dnaja3^−/− tumors displayed markedly increased mitochondrial ROS levels (Fig. S15B). Moreover, the basal oxygen consumption rate (OCR) was significantly decreased while the extracellular acidification rate (ECAR) was significantly increased in the Dnaja3^−/− tumors (Fig. S15C). In addition, the ATP levels and ATP/ADP ratio were significantly reduced in Dnaja3^−/− tumors (Fig. S15D). We also assessed mitochondrial membrane potential (ΔΨM) using JC-1 fluorescence to evaluate the mitochondria polarization state, with lower ΔΨM being indicated by the shift to green fluorescence from red fluorescence. We found that the JC-1 green fluorescence was markedly increased in Dnaja3^−/− tumors versus Dnaja3^f/f tumors (Fig. S15E). Taken together, our results are consistent with the hypothesis that Dnaja3 deficiency may induce hepatic mitochondrial dysfunctions, which in turn induce HCC.

To clarify Dnaja3’s role against mitochondrial dysfunction and lipid accumulation, we conducted a rescue experiment with Dnaja3^−/− mice. These mice were transduced with an adeno-associated virus overexpressing Dnaja3 (pssAAV-CB-Dnaja3) or a control vector (pssAAV-CB-Control). Within 14 days, effective Dnaja3 overexpression was noted in Dnaja3^−/− mice livers, as indicated in Figs. S16A and S16B. After 12 weeks, pssAAV-CB-driven Dnaja3 significantly countered the effects of Dnaja3 deficiency on mitochondrial genes ATP5A and NDUFB8 while inhibiting mitochondrial FAO gene ACACB (Fig. S16C). Notably, the AAV-treated groups exhibited a significant reduction in serum cholesterol levels (Fig. S16D). Fatty acid profiling highlighted reduced SFA and MUFA levels in Dnaja3-present livers of of Dnaja3^−/− mice (Table. S18).

To further investigate the role of Dnaja3 in NAFLD, we fed male Dnaja3^−/−, Dnaja3^+/−, and Dnaja3^f/f mice with a high-fat diet (HFD) for 8 weeks, followed by the analysis of liver function (Fig. S17A and S17B). Interestingly, HFD altered body weight in Dnaja3^−/− mice (Fig. S17C). Comparatively, serum cholesterol, ALT, and AST levels in Dnaja3^−/− mice were significantly higher than in the Dnaja3^+/−, and Dnaja3^f/f mice under the HFD (Figs. S17D, S17E, and S17F). A pronounced elevation in the NASH score and collagen deposition was detected in the liver of Dnaja3^−/− mice compared to the heterozygous or Dnaja3^f/f mice on HFD (p=0.005) (Fig. S17G). In parallale, the Dnaja3^−/− mouse livers exhibited a marked increase in apoptosis than Dnaja3^f/f mice on the same diet (Fig. S17H). It’s noteworthy to highlight that while in vivo observations depicted discernible differences in cell death, the same could not be mirrored in vitro (Fig. S6E). Remarkably, the presence of Dnaja3 in AAV-treated groups resulted in a rescued effect (Figs. S17D, S17E, and S17F). In conclusion, our rescue experiments utilizing AAV-mediated Dnaja3 overexpression have provided valuable insights into the role of Dnaja3 in mitochondrial function and fatty acid metabolism.

Discussion

NASH is associated with obesity, metabolic syndrome, high cholesterol, and type 2 diabetes. Individuals with normal weight could develop metabolic syndrome and abnormal body fat distribution, which are risk factors of NASH and high mortality rates. It is evident that the heritability of body fat distribution is a better diagnostic criterion for the development of NASH rather than BMI (14). In this study, we identified a body fat distribution-related rs3747579-TT variant linked to a reduced expression of mitochondrial DNAJA3 and associated with NASH. Experimentally, Dnaja3 ablation leads to the development of NASH and HCC in mice with features reminiscent to human NASH and NASH-related HCC. These results are consistent with the hypothesis that individuals carrying a DNAJA3 variant may have a high risk to develop progressive liver disease.

DNAJA3 collaborates with Hsp70 to bind client proteins, ensuring their proper folding for mitochondrial proteostasis (31, 32). It also acts as a tumor suppressor in solid tumors and plays roles in cell proliferation, survival, and signaling pathways. Its associations with oncogenic proteins and key regulators like HIF-1α and VEGF make it vital in tumorigenesis exploration (33–35). Moreover, our prior study revealed that full Dnaja3 knockout leads to early embryonic lethality, with affected embryos dying between E4.5 and E7.5 (36). In this study, Dnaja3 deficient cells showed profound loss of mitochondrial structure and severe dysfunction of OXPHOS in Dnaja3-deficient cells both in vivo and in vitro. As a cochaperone, we found that DNAJA3 interacts with several mitochondrial client proteins, including the mitoribosomal proteins and mitochondrial crucial chaperone, suggesting that DNAJA3 may provide crucial chaperone activity to mediate mitoribosomal functions for the translation of OXPHOS.

Inhibiting bile acid synthesis leads to cholesterol accumulation, a substrate for bile acid, resulting in liver damage. This accumulation triggers mitochondrial dysfunction, heightening ROS production and fostering fibrogenesis. This cyclical disturbance in hepatic cholesterol and fatty acid homeostasis further accelerates cell death and injury, potentially advancing NASH/HCC progression (37, 38). Together, these insights emphasize DNAJA3’s role in NAFLD/NASH pathogenesis and validate our mouse model’s representation of the NAFLD to NASH and cirrhosis/HCC continuum, mirroring the human NASH-related HCC progression.

Our study revealed that a metabolic syndrome-related rs3747579-TT variant may serve as the risk allele linked to reduced expression of DNAJA3 in NASH-related HCC. It is interesting to note that when examining the ALFA Allele Frequency database (https://www.ncbi.nlm.nih.gov/snp/rs3747579#frequency_tab), we found that the rs3747579-TT allele frequency varies among ethnic groups, ranging from 76% (Asian), 72% (European), 58% (latino) and 21% (African). This coincides with the observation that the NAFLD/NASH prevalence also varies among ethnic groups, ranging from 32% (Middle East), 27% (other Asian), 24% (European), 31% (latino) and 13% (African) (39). It should be noted that the rs3747579 locus is located approximately 30 Kb upstream of DNAJA3 transcriptional initiation site. While a link between the rs3747579-TT variant and a corresponding reduced expression of DNAJA3 was validated, the underlying mechanism is unclear. It is known that SNPs in the regulatory regions may have enhancer functions by mediating the binding of critical transcription factors and forming allele-specific long-range chromatin loops with promoters (40). Future studies should aim to characterize the functional interplay between rs3747579 associated functional SNPs, as well as investigate potential interactions with other transcription factors and chromatin remodeling complexes.

Limitations of the study

First, the number of NASH cases is small in our cohort. It is recognized that genetic studies with confirmed NASH are extremely rare if any because NASH diagnosis requires histological confirmation. Second, We could not analyze obesity and diabetes since these data are not available in the control group. Given the cohort size and limited covariables, this is the limitation of the study, which requires future work to extend this finding.

List of Abbreviations:

NAFLD

nonalcoholic fatty liver disease

NASH

nonalcoholic steatohepatitis

HCC

hepatocellular carcinoma

mtHSP40

mitochondrial chaperone Hsp40

Alb-Cre

Albumin-Cre

DEN

diethylnitrosamine

AFP

alpha-fetoprotein

OXPHOS

oxidative phosphorylation

MUFA

monounsaturated fatty acids

PUFA

polyunsaturated fatty acids

FAO

fatty acid β-oxidation

mtDNA

mitochondrial DNA

OCR

oxygen consumption rate

ECAR

extracellular acidification rate

eQTL

expression quantitative trait loci