Suppression of ACADM-Mediated Fatty Acid Oxidation Promotes Hepatocellular Carcinoma via Aberrant CAV1/SREBP1 Signaling

Primary liver cancer is one of the top leading cause of cancer-related deaths worldwide (1). Among the different types, hepatocellular carcinoma (HCC) is the most common type accounting for approximately 90% of the cases, with its incidence rate higher in developing countries. The administration of vaccines against hepatitis B virus in newborns has contributed to the decline of cancer incidence and mortality in Asia (2), however an increasing trend is being observed in Western countries (3) along with the rise of nonviral HCC cases.

The alterations in fatty acid metabolism are increasingly recognized for their role in inducing carcinogenesis. In rapidly proliferating cancer cells, carbons are hijacked from energy production to synthesize fatty acids, which can be sourced either exogenously or from de novo synthesis (4). Current literature cannot confirm whether the upregulation or downregulation of fatty acid oxidation (β-oxidation) contributes to HCC tumorigenesis, which may be due to the tumor heterogeneity nature of HCC. The expression of many β-oxidation-related genes was found to vary significantly between different patients (5). The upregulation of hypoxia inducible factor-1α inhibited β-oxidation, resulting in decreased reactive oxygen species and increased glycolysis to further facilitate HCC development (6). The evidence highlights the significant potential of targeting β-oxidation to treat HCC.

Medium-chain acyl-CoA dehydrogenase (ACADM) catalyzes the first step of β-oxidation and is responsible for the breakdown of medium-chain fatty acids in the mitochondria. Medium-chain acyl-CoA dehydrogenase deficiency, caused by mutations in the ACADM gene, is the most common inherited metabolic disorder in Caucasians and thus signifying the profound influence of ACADM in metabolic diseases (7). Alteration of ACADM expression is found in subjects with cardiovascular, metabolic, and nonalcoholic fatty liver diseases (8–10). However, the role of ACADM in human cancers has not been thoroughly studied. To our knowledge, the only functional study of ACADM performed in hepatoma cells was carried out where ACADM knockdown enhanced tumor growth (6). The regulation of ACADM expression has not been fully defined in other cancers (6, 11–13), and little is known regarding this enzyme and HCC in literature.

This study aims to unravel the mechanisms involved with aberrant lipid metabolism and to define novel biomarkers of HCC progression. The results presented here provide innovative insights into the dysregulation of ACADM-mediated β-oxidation by oncogenic Caveolin-1 (CAV1) and sterol regulatory element-binding protein-1 (SREBP1) to facilitate cancer development in HCC. With ACADM being a critical functional component in promoting β-oxidation, the establishment of its tumor suppressor role indicates its promising potential as a biomarker for HCC proliferation and metastasis. The mode of interplay between ACADM, SREBP1, and CAV1 inspired an enhanced synergized treatment model that will not only benefit patients with HCC, but also for treating other diseases arising from metabolic disorders.

Fifty pairs of human HCC and their corresponding nontumorous samples were obtained during surgical resections from patients at Queen Mary Hospital (QMH), Hong Kong, and were selected for analyzing ACADM mRNA expression in this study. Paraffin-embedded HCC specimens were obtained from the archives of the Department of Pathology, Sun Yat-sen University Cancer Center (SYSUCC), Guangzhou, China. The patients neither received any chemotherapy nor radiotherapy prior to the surgery. Informed written consent was obtained from all the patients. The use of human samples was approved by the Institutional Review Board of The University of Hong Kong/Hospital Authority Hong Kong West Cluster (HKU/HA HKW IRB) and the Institute Research Medical Ethics Committee of SYSUCC.

The tissue microarray slides were constructed using paired HCC and adjacent nontumorous liver tissues. First, the marked areas on each sample were punched with the MiniCore Excilone (MiniCore) to yield tissue cores of 0.6 mm in diameter. Samples were then re-embedded, fixed with 4% paraformaldehyde before embedding in paraffin wax. Subsequently, the paraffin-embedded sections were cut into 4 μmol/L-thick slices and mounted onto glass slides. After dewaxing, the slides were treated with 3% hydrogen peroxide in methanol and blocked using a Biotin-Blocking Kit (DAKO). The slides were incubated with primary antibodies after blocking: ACADM (Abcam, catalog no. ab92461, RRID:AB_10563530) at 1:600 dilution for 50 minutes and SREBP1 (Santa Cruz Biotechnology, catalog no. sc-13551, RRID:AB_628282) at 1:25 dilution overnight at 4°C. The slides were washed three times with 1× PBS before incubation with the biotinylated secondary antibodies for 1 hour, then the slides were stained with 3,3′-diaminobenzidine tetrahydrochloride in solution (DAKO). Finally, the slides were counterstained with Mayer's hematoxylin and observed under a microscope for analysis.

The stained samples were quantified without prior knowledge of corresponding patients' clinicopathologic data. To quantify ACADM expression, each specimen was individually scored as 0 for negative, 1 for weak positive, 2 for moderate positive, and 3 for strong positive. The proportion of positive-stained cells in total tumor cells was in percentages of 0, 50%, 80%, and 100%. The histoscore (H-score) was calculated by multiplying the percentage of positive cells with the score of intensity, giving a range of 0 to 300. To assess the nuclear immunoreactivity of SREBP1, the percentage and intensity of nuclear signal were scored.

The human HCC cell lines used in this study were either purchased from the ATCC [Hep3B (ATCC, catalog no. HB-8064, RRID:CVCL_0326) and PLC/PRF/5 (ATCC, catalog no. CRL-8024, RRID:CVCL_0485)] or were sourced from the Cancer Institute, Fudan University, China [LM3 (RRID:CVCL_6832) and MHCC97 L (RRID:CVCL_4973)], the Japanese Collection of Research Bioresources [Huh7 (JCRB, catalog no. JCRB0403, RRID:CVCL_0336) and HLE (JCRB, catalog no. JCRB0404, RRID:CVCL_1281)] and from Jayanta Roy-Chowdhury, Albert Einstein College of Medicine, New York [MIHA (RRID:CVCL_SA11)]. All cell lines were cultured in DMEM, high glucose supplemented with 10% FBS (Gibco), with the medium's final pH adjusted to 7. Cell cultures were kept in humidified incubators maintained at 37°C with 5% CO₂. Mycoplasma detection in cell cultures were carried out by PCR screening with primers Myco5 and Myco3 (Supplementary Table S1). Cell line authentication was performed by short tandem repeat DNA profiling (Bio-Gene).

To establish stable knockdown cells of CAV1 and ACADM, MISSION short hairpin RNAs (shRNA) targeting CAV1 and ACADM and nontarget control (CTL) were purchased from Sigma-Aldrich. To establish double knockdown cells of CAV1 and ACADM, shRNA targeting ACADM was subcloned into pLKO.1-Blast vector (RRID:Addgene_26655) via AgeI and EcoRI sites. To establish Acadm knockout in murine cells, single guide RNAs (sgRNA) targeting murine Acadm and the control sgRNAs (Integrated DNA Technologies) were synthesized and subcloned into the pX330-U6-Chimeric_BB-CBh-hSpCas9 vector (RRID:Addgene_42230). The sequences of oligos used in the study are provided in Supplementary Table S1. For stable overexpression of SREBP1 cells, the human open reading frame (ORF) cDNA of SREBP1 was purchased from Sino Biological (catalog no. HG17512-UT). The plasmids were subjected to DNA sequencing carried out by the Centre for Genomic Sciences, HKU, to confirm the correct orientation and sequence of the insert in the plasmid. Expression constructs were transfected into 293FT cells (ATCC, catalog no. PTA-5077, RRID:CVCL_6911) using the GeneCopoeia Lenti-Pac HIV expression packaging system; detailed procedures are described elsewhere (14).

The EnzyChrom Triglyceride Assay Kit and Phospholipid Assay Kit (BioAssay Systems) were used to detect the level of triglycerides and phospholipids in cells. The manufacturer's protocol was followed, the experiments were done in triplicates, and 1 × 10⁶ cells were used per sample per well. Total cell lysate was used for sample normalization.

Nile Red is a phenoxazone dye used to detect intracellular lipid droplets. Cells were prepared as follows: a clean coverslip was placed into a 6-well plate. An optimal number of cells was seeded into the well and incubated overnight at 37°C. The cells were gently rinsed with 1× PBS before fixing with 4% paraformaldehyde, staining with Nile Red (10 μg/mL, Sigma-Aldrich), then counterstaining with DAPI (Invitrogen). The coverslips were carefully mounted onto a clean glass slide with mounting medium (Vectashield). Cells were then visualized with confocal microscopy (Carl Zeiss LSM-700). Quantification of the signals was performed with the ImageJ software (ImageJ, RRID:SCR_003070) by normalizing the Nile Red signal intensity with the DAPI signal, the latter of which indicate the nuclear staining level of the cells.

Oil Red O (Sigma) is a lysochrome diazo dye used for staining neutral triglycerides and lipids on frozen tissue sections. Briefly, fresh frozen tissue was cut into 5- to 10 μmol/L-thick sections and mounted on slides. The slides were then air-dried, fixed in ice-cold 10% formaldehyde, rinsed in distilled water, and allowed to air-dry again before being placed in absolute propylene glycol to avoid carrying water into the Oil Red O. The staining was performed in prewarmed Oil Red O solution, then differentiated in 85% propylene glycol solution. The slides were rinsed in distilled water, stained with hematoxylin, washed thoroughly under running tap water, placed in distilled water, then finally mounted with mounting medium.

The EpiQuik Chromatin Immunoprecipitation Kit (Epigenetik) was used to determine the interaction between SREBP1 and the endogenous promoter of ACADM. The manufacturer's protocol was followed and PCR was performed to observe for any protein–DNA interactions. The sequences of primers ACADM-540F and ACADM+41R flanking the SREBP1 binding site, and the primers ACADM-1450-F and ACADM-950-R flanking the -1450 to -950 region of the ACADM promoter are shown in Supplementary Table S1.

The hydrodynamic tail vein injection was employed to induce transfection of foreign DNA inside the livers of mice. The following plasmids were used in this study: pT3-EF1-NRAS, pX330-TP53, Sleeping Beauty (SB) transposon and Acadm-KO. Plasmids were amplified using the GenElute HP Endotoxin-Free Plasmid Maxiprep Kit (Sigma) according to the manufacturer's protocol. To prepare the plasmid solution for injection, 20 μg of each oncogene and 1.6 μg of SB transposon were used for each mouse. The plasmids were added to 1× PBS, resulting in a total volume of 2 mL per mouse. The plasmid solution was then filtered through a 0.22 μmol/L filter before use.

Immunocompetent male 6-week-old FVB/N mice (MGI, catalog no. 3528175, RRID:MGI:3528175) were selected for this procedure. Each mouse was placed in a mouse restrainer, then its tail was swabbed with 70% ethanol prior to injection. The plasmid solution was injected via the tail vein within 5 to 8 seconds. The mouse was then returned to its cage. Mouse weight was recorded biweekly until the mouse had reached its humane end point, at which point, the liver was harvested and fixed in 10% formaldehyde for IHC staining. Antibodies used in this study are listed in Supplementary Table S2.

Etomoxir (ETO), an irreversible carnitine palmitoyl transferase 1 (CPT1)-specific inhibitor that inhibits fatty acid entry into the mitochondrial matrix for β-oxidation, was used to treat MHCC97 L CAV1 knockdown cells. Cells were incubated with 100 μmol/L of ETO (Cayman Chemical) in the culture medium for 48 hours before being subjected to further experiments.

Male BALB/cAnN-nu (Nude) mice of approximately 4 to 5 weeks of age were selected for subcutaneous implantation of HCC cells to observe their tumorigenic ability. Various stably-transfected HCC cell lines were subcutaneously injected into the flank of mice, with the optimal cell number of each cell line suspended in 100 μL of either 1× PBS or Matrigel per injection. The tumor sizes were measured using a caliper and monitored at the indicated time points throughout the experiment, with the tumor volume calculated with the formula: 1/2 (largest diameter) × (smallest diameter)². At the experimental endpoint, the mice were sacrificed before tumors were excised and weighed.

For drug treatment, subcutaneous xenografts of 5 × 10⁶ MHCC97 L cells were injected into 4-week-old nude mice. When the tumors reached 5 mm in diameter, mice were randomly assigned to one of four groups and drugs were administered in the following combinations: (1) vehicle, (2) eicosapentaenoic acid (EPA; 200 mg/kg), (3) sorafenib (30 mg/kg), and (4) EPA + sorafenib. Both drugs were fed to mice via oral gavage in volumes of 100 μL, with sorafenib fed daily for 21 days and EPA fed every other day for 10 days. Mouse weights and tumor sizes were recorded daily. At the experimental endpoint, the tumors were excised and weighed.

All mice in this study were fed with standard chow and raised in individually ventilated cages equipped with local woodchip for bedding. All experiments involving live animals were performed according to the Animals (Control of Experiments) Ordinance (Hong Kong), and the Institute's guidance from Laboratory Animal Unit on animal experimentation was strictly followed.

For the genes investigated in this study, their mRNA levels were correlated to various clinicopathologic parameters in HCC patients using IBM SPSS statistics 20 (SPSS, RRID:SCR_002865). The parameters were determined and analyzed by clinical pathologists upon surgical resection. Student t test, Kaplan–Meier analysis, log-rank test, Chi-squared test, and Mann–Whitley U test were incorporated into the statistical analyses of data. All other statistical analyses were performed by GraphPad Prism 7 (GraphPad Prism, RRID:SCR_002798), where Student t test was used for the functional in vitro assays unless otherwise stated. P < 0.05 was considered as statistically significant.

The mRNA expression of ACADM was analyzed in 50 paired cases of HCC and their corresponding nontumorous tissues from the QMH cohort. It was observed that ACADM was underexpressed in 64% (32/50) of HCC cases (Fig. 1A). The overall ACADM level was significantly lower in tumorous versus nontumorous tissues; ACADM expression data from The Cancer Genome Atlas (TCGA) database also revealed a significant decrease in ACADM expression in tumorous samples (Fig. 1B). ACADM expression was found to be reduced with further HCC progression (Fig. 1C); its underexpression significantly correlated with the aggressive pathologic features such as bigger tumor size, presence of venous invasion, advanced HCC tumor stage, and poor cell differentiation (Fig. 1D; Supplementary Table S3). Receiver operating characteristic curves indicate the significant discrimination of ACADM expression between nontumorous and early stage, but not between early and late stage HCC (Supplementary Figs. S1A and S1B). Furthermore, ACADM expression was analyzed in the SYSUCC cohort using tissue microarray. IHC staining of ACADM protein expression in paired HCC tissues was scored as strong (45/51) and moderate (6/51) positive in 100% of nontumorous liver tissues, when compared with strong and moderate positive in 52.9% (27/51) of tumorous tissues. Weak positive and negative were observed in 47.1% (24/51) of tumorous tissues (Fig. 1E). Among 51 cases, ACADM was downregulated in 86.3% (44/51) of HCC tumors when compared with the adjacent nontumorous liver tissues (Fig. 1F). Together, these findings indicate that the reduced ACADM level in tumorous tissues could be a valuable biomarker of HCC.

To functionally characterize ACADM in HCC, it was suppressed in nonmetastatic Huh7 cells. Stable nontarget control and ACADM knockdown clones (shACADM; sh61 and sh65) were established and verified with Western blot analysis (Fig. 2A). ACADM knockdown enhanced cellular lipid content, triglycerides, and phospholipids in shACADM compared with the shCTL cells (Fig. 2B), with similar results observed in immortalized normal liver cell line MIHA (Supplementary Figs. S2A and S2B). As reflected by the cellular respiration rate, shACADM cells displayed reduced fatty acid oxidation when compared with shCTL cells (Fig. 2C). Fatty acid profiling revealed the overall increase in fatty acids in shACADM cells, with the same being observed in both palmitic acid and oleic acid levels (Fig. 2D). The abundance of other saturated and unsaturated fatty acids was found to be elevated in shACADM cells compared with shCTL (Supplementary Figs. S3A and S3B). Diminution of ACADM in Huh7 and MIHA cells significantly encouraged cell growth, anchorage independent growth, migration, and invasiveness of the shACADM cells, indicating their enhanced cell aggressiveness compared with the shCTL (Fig. 2E; Supplementary Figs. S4A and S4B). The tumor development rate significantly increased in subcutaneous xenografts derived from shACADM cells (Fig. 2F), with IHC staining indicating the higher proliferation rate of shACADM cells compared with the shCTL (Fig. 2G).

The oncogenic effect of ACADM knockdown was further analyzed by hydrodynamic injection mouse model. ACADM expression was found to correlate to NRAS/p53 gene alterations according to the TCGA database (Supplementary Fig. S5A). The establishment of the Acadm-KO plasmids (Acadm-KO1 and Acadm-KO2) were confirmed in murine NIH3T3 cells with Western blot analysis (Supplementary Fig. S5B). Plasmids used in the experiment and the approximate timeline of actions were depicted, with the mice split into three groups to observe the effect of different oncogenic plasmid combinations on HCC development (Fig. 2H). At week 4 post-injection, larger tumors were formed in mice injected with the N-RasV12, p53-, and Acadm-KO plasmids versus mice without Acadm-KO plasmid (Supplementary Fig. S5C). At the end of the experiment, it was observed that injection with Acadm-KO plasmid resulted in significantly more tumor formation and larger tumor sizes compared with either the control group or the mice only injected with RasV12 and p53-KO plasmids (Fig. 2H). Nevertheless, significant differences were not observed between the end-point liver weight, liver:body weight ratio and mouse weight of mice injected with RasV12 and p53-KO plasmids with or without Acadm-KO plasmid (Supplementary Fig. S5D). IHC staining of the excised livers revealed the Acadm knockout expression and N-Ras overexpression in mice that received injection of Acadm-KO and RasV12 plasmids (Fig. 2I). The expression of p53 was not detected in the normal liver; in mice injected with p53-KO plasmid, p53 was not detected in tumor cells but observed in adjacent lymphoid cells, suggesting the successful knockout of p53 in tumors.

To determine how ACADM is regulated in HCC, the activity of ACADM promoter in cells was investigated. Submitting the ACADM promoter sequence to MatInspector (Genomatrix) revealed two putative transcription factor binding sites upstream of the ACADM transcription site, peroxisome proliferator response element (PPRE), and sterol regulatory element (SRE). To confirm whether these two sites play a significant role in regulating ACADM transcription, both sites were mutated. In MHCC97 L cells, high ACADM promoter activity was observed for the SRE-mutant versus the wild-type ACADM promoter (-1450), with the nonsignificance between the PPRE-mutant and the wild-type promoter. SRE is the binding site of SREBP1; RT-qPCR using primers flanking SRE but not primers amplifying unrelated region of ACADM promoter revealed the increased copy number of ACADM promoter fragments pulled down by anti-SREBP1 antibody in MHCC97 L cells (Fig. 3A). These findings suggested that SREBP1 is a negative regulator of ACADM.

To determine the clinical relevance of SREBP1 and ACADM in HCC, we examined their expressions in a cohort of HCC tissues from SYSUCC (n = 41). The SREBP1 expression was inversely correlated with ACADM expression with significance; in cases with SREBP1 overexpression, reduced levels of ACADM were detected in tumorous tissues (Fig. 3B). In tissue microarray comprising of 46 HCC specimens, strong and moderate positive staining of SREBP1 were detected in 41.3% (19/46) of tumorous tissues whereas only weak positive and negative were detected in nontumorous tissues (Fig. 3C; Supplementary Fig. S6). The nuclear immunoreactivity of SREBP1 was significantly higher in tumorous tissues as indicated by the higher H-score. The overall SREBP1 expression was higher in tumorous tissues when compared with nontumorous tissues, and SREBP1 overexpression was found in 56.5% (26/46) of the cases (Fig. 3C). These data established the association between ACADM underexpression and SREBP1 upregulation in human HCC.

Because SREBP1 has previously been reported to interact with CAV1 in other diseases (15, 16), this warrants further investigation into their interplay in HCC. It was revealed that the ACADM promoter activity was significantly elevated in shCAV1 cells (Fig. 3D). It was observed that a lower CAV1 level prevented the nuclear accumulation of SREBP1 and enhanced ACADM expression in MHCC97 L shCAV1 cells, with treatment of filipin, a CAV1 inhibitor (17), inducing the same effect in MHCC97 L shCTL cells (Fig. 3E). The results implicate that CAV1 facilitates nuclear accumulation of SREBP1, leading to the negative transcriptional regulation of ACADM. Restoration of SREBP1 in MHCC97 L shCAV1 cells reduced ACADM expression (Fig. 3F). Conversely, knockdown of SREBP1 upregulated ACADM level in MHCC97 L cells (Fig. 3G). Our data also showed that CAV1 interacted with full-length SREBP1 in MHCC97 L cells, implicating the potential role of CAV1 to mediate the maturation and nuclear accumulation of SREBP1 (Supplementary Fig. S7). Further investigation will be needed to delineate how CAV1 regulates the activation of SREBP1 leading to ACADM downregulation.

Literature has revealed the emerging roles of CAV1 in cancer metabolism (18–20). However, its mode of action remains unclear, especially in cancer. To investigate the prospective role of CAV1 in β-oxidation, an examination of the effect of CAV1 on cellular lipid levels was conducted in metastatic MHCC97 L cells. First, stable nontarget control (shCTL) and CAV1 knockdown (shCAV1) clones were established (Fig. 4A). The knockdown of CAV1 significantly reduced the cellular triglycerides, phospholipids, and lipid accumulation levels compared with the shCTL cells (Fig. 4B). The oxygen consumption rate increased in shCAV1 cells, indicating the increase in β-oxidation compared with the shCTL (Fig. 4C). The same effects were observed in nonmetastatic Hep3B cells (Supplementary Figs. S8A–S8C), suggesting that CAV1 can downregulate β-oxidation.

Because CAV1 exerts potent effects in driving HCC tumorigenesis and metastasis, it is intriguing whether an alteration in β-oxidation contributes to the oncogenic properties of CAV1. Etomoxir (ETO) treatment inhibited fatty acid breakdown and resulted in the restoration of triglycerides, phospholipids, and lipid droplets in shCAV1 cells (Fig. 4D). CAV1 knockdown significantly diminished the proliferation rate, anchorage-independent growth, migration, and invasiveness of MHCC97 L cells; however, such diminishment was partly restored in shCAV1 cells treated with ETO (Fig. 4E), with the same effect observed in Hep3B cells (Supplementary Fig. S8D). The cell aggressiveness was also augmented in nude mice intraperitoneally injected with ETO, with immunohistochemistry indicative of the CAV1 expression and Oil Red O staining revealing the build-up of oil droplets in shCAV1 tumors treated with ETO compared with the nontreated tumors (Fig. 4F). To demonstrate the functional interaction between CAV1 and SREBP1, SREBP1 was overexpressed in MHCC97 L CAV1 knockdown cells; the enhancement of fatty acid oxidation in shCAV1 cells was suppressed by SREBP1 overexpression (Fig. 4G). Functionally, the reduced promoting function of CAV1 was rescued by overexpressing SREBP1 in shCAV1 cells; these cells displayed increased anchorage independent growth, cell migration, and invasion (Fig. 4H).

To investigate whether CAV1 correlates with ACADM in HCC, their mRNA and protein expression levels were analyzed. In the HCC cell line panel, ACADM and CAV1 expressions were found to be negatively correlated (Fig. 5A and B). IHC revealed the alternated expressions of CAV1 and ACADM in tumors derived from MHCC97 L cells implanted into the mouse liver (Fig. 5C). The enhanced transcriptional and protein levels of ACADM were also validated in both Hep3B and MHCC97 L shCAV1 cells (Fig. 5D and E). The mRNA expressions of ACADM and CAV1 in 25 paired clinical samples of HCC and nontumorous tissues from QMH were determined; consistent with their association observed in cell lines, their negative correlation was also observed to be significant in data obtained from the TCGA database (Fig. 5F).

To validate whether CAV1 mediates β-oxidation via ACADM in HCC, ACADM was suppressed in shCAV1 MHCC97 L and Hep3B cells to recapitulate the functional effect of CAV1 overexpression (Fig. 6A). The resulting effects were that the levels of triglycerides, phospholipids, intracellular lipid contents, as well as fatty acid oxidation were elevated in double knockdown cells of CAV1 and ACADM (shCAV1/sh61 and shCAV1/sh65; Fig. 6B,–D). The enhancement of HCC anchorage independent growth, cell growth, and motility were also observed in the double knockdown cells compared with the shCAV1 cells (Fig. 6E,–G). The same effect was observed in animal models, with the shCAV1/sh61 tumors proliferating much quicker than the shCAV1 only tumors (Fig. 6H); IHC staining revealed the darker staining of CD31 and Ki67 in shCAV1/sh61 tumors, highlighting the elevated angiogenesis and cell proliferation rates compared with the shCAV1 only cells (Fig. 6I).

On the basis of the in vitro findings, we hypothesized that the inhibition of fatty acid synthesis genes can potentially inhibit the proliferation of cancer cells. EPA was found to inhibit the maturation of SREBP1 protein in hepatocytes (21). Reduction of activated SREBP1 was observed in MHCC97 L cells after EPA treatment (Fig. 7A). The therapeutic effect of EPA alone and in combination with sorafenib was investigated in MHCC97 L subcutaneous xenograft mouse model (Fig. 7B). The significant efficacy enhancement of the combinational treatment of sorafenib with EPA to suppress HCC growth was observed (Fig. 7C). Significant reductions in tumor volume and weight were also observed in the combinational treatment group compared with single administrations of either drug (Fig. 7D). Animals treated with drugs did not reveal significant weight loss when compared with animals of other groups (Supplementary Fig. S9). Tumors treated with EPA alone or in combination with sorafenib resulted in the decrease in SREPB-1 and increase in ACADM expressions (Fig. 7E). Tumors formed in mice that received the combined treatment showed the least Ki67 staining among the four experimental groups.

Despite a previous report of using Hep3B cells to investigate the role of ACADM in tumor growth (6), the demonstration of its role in HCC metastasis is still deficient. In this study, ACADM knockdown remarkably enhanced lipid accumulation and cell aggressiveness in vitro, whereas ACADM knockout modulated the tumor growth in mice, suggesting that the loss of ACADM augmented HCC aggressiveness and that ACADM is a functional component in promoting β-oxidation. Although a nonsignificant difference in ACADM expression between HCC lesions and adjacent normal tissues was previously reported (6), our data showed that ACADM was underexpressed in the vast majority of patients with HCC. Altogether, the results highlighted ACADM's potential as a valuable biomarker during HCC development.

The mutation of the SRE binding site, which substantially boosted ACADM expression in HCC cells, suggested that genes binding to SRE are responsible for controlling ACADM expression. SRE is the transcription factor binding site of SREBPs, which are cholesterol sensors located in the endoplasmic reticulum (ER) that regulate intracellular cholesterol (22) and fatty acid synthesis (23). SREBP1c is one of the three isoforms of SREBPs that is mainly found in the liver, muscles, and fat tissues (24). In cancer cells, the frequent overexpression of SREBPs resulted in the accumulation of lipids to enhance cell proliferation rate (25). In this study, the previously unprecedented suppression of ACADM by SREBP1 prevented β-oxidation, leading to further lipid accumulation to fuel cancer growth; our observation corroborates with other findings that SREBP1 overexpression can suppress various lipid oxidation genes in bovine hepatocytes (26, 27).

The precursor form of SREBP1 resides in the ER; upregulation of SREBP1 leads to lipid accumulation in normal hepatic and hepatoma cells under ER stress (28). Though normally localized to the plasma membrane, CAV1 has been shown to accumulate in the ER, which causes the protein to be targeted to lipid droplets (29), therefore it is possible for CAV1 to interact with the precursor of SREBP1 in the ER. Although previously reported to interact with each other in other cancers, the interplay of SREBP1 and CAV1 remained ambiguous in HCC. Here, CAV1 was revealed to be a positive regulator of SREBP1 by acting upstream to modulate SREBP1 expression in HCC. The overexpression of SREBP1 in CAV1 knockdown cells did not affect CAV1 expression, but restored their cell aggressiveness, which confirmed the positive correlation between these two genes. However, how CAV1 facilitates SREBP1 nuclear accumulation remains unclear. The transcriptional increase in ACADM expression was detected upon the knockdown of CAV1. Pooling together these results, they indicate that CAV1 can modulate fatty acid metabolism via the activation of SREBP1 to suppress ACADM in HCC.

The oncogenic role of CAV1 in cancer has been well established; it is an important factor involved in tumorigenesis and progression of many cancers, but with context dependent functions. Our previous study demonstrated the definitive role of CAV1 in HCC metastasis, also revealing the dramatic expression of CAV1 in metastatic HCC cells (14). The upregulation of CAV1 associated with the presence of C-terminal truncated HBx in HCC, which activates the transcription of CAV1 with significant functional impact on HCC tumorigenesis (30). On the basis of the knowledge that metastasis is a prominent feature in the advanced stage of HCC and that CAV1 is a potent metastasis promoter, the role of ACADM as an effector of CAV1 was explored. Compared with other β-oxidation genes, ACADM was of particular interest due to how its oxidation target, medium-chain fatty acids (C6 - C12), has the ability to diffuse unaided into the mitochondria as opposed to longer-chain fatty acids, which have to be imported by CPT1; this is due to the increased solubility of shorter-chain fatty acids into the mitochondrial membrane (31). This diversion from the carnitine shuttle may reveal a previously unrecognized pathway that bypasses CPT1 to regulate β-oxidation. We speculated that CAV1 regulates β-oxidation by inhibiting ACADM expression, which was proven by the negative correlation of CAV1 and ACADM in HCC cellular models, animal models, and patient biopsies.

As a routine medication used to treat unresectable HCC, sorafenib is an important multikinase inhibitor drug that is also applied for the treatment of several other types of cancers, although its effect on prolonging the survival of patients for only a few months leaves a lot to be desired. From previous reports, it can be theorized that by targeting SREBP1 activity, it will inhibit the expression of fatty acid synthesis genes, which can potentially inhibit the proliferation of cancer cells. EPA is a major component of ω-3 polyunsaturated fatty acids that can enhance fatty acid oxidation and reduce de novo lipogenesis by modulating transcription factors to inhibit SREBP1 nuclear translocation (32, 33); it has previously been shown to inhibit SREBP1 activity by inhibiting its nuclear translocation in hepatocytes (21, 34). In normal cells, the overabundance of unsaturated fatty acids triggers a negative feedback loop, which suppresses SREBP1c expression to prevent excessive lipid accumulation (35). The administration of EPA in combination with another omega-3 fatty acid, docosahexaenoic acid (DHA), have been shown to alleviate illness and to promote general good health in mice (36), with the enhanced protective effect of EPA over DHA highlighted (37). Here, we showed that the combinational treatment of EPA and sorafenib markedly suppressed the growth of MHCC97 L subcutaneous xenografts when compared with treatments with either drug alone. This phenomenon can be attributed to the broad-spectrum protein kinase inhibitor activity of EPA previously observed in both prostate and breast cancers (38, 39), which further enhanced the multikinase inhibitor and antitumor properties of sorafenib. Therefore, our data indicate the immense potential of the co-administration of EPA and sorafenib in slowing down HCC progression to provide a better prognosis for patients.

In conclusion, this study has revealed the mode of ACADM-mediated fatty acid oxidation in HCC and how its dysregulation led to the increase in fatty acid availability for promoting the proliferation and metastatic abilities of HCC cells (Fig. 7F). We highlighted the critical function of ACADM as a tumor suppressor in its role of modulating fatty acid metabolism to inhibit tumorigenesis and HCC development, which indicate its potential as a biomarker for HCC proliferation and metastasis. Our data delineated the novel CAV1/SREBP1/ACADM axis in the regulation of fatty acid oxidation in HCC, and revealed the immense therapeutic potential of suppressing SREBP1 activity to synergize sorafenib potency in treating HCC. All in all, this study has contributed to a better understanding of the mechanistic pathways that shape the dysregulation of fatty acid metabolism in HCC, which will be beneficial to the advancement of targeted therapies for cancer patients, since tumor metabolism plays such a key role in cancer development.

A.P. Ma reports grants from Hong Kong Research Grants Council, General Research Fund, and grants from University Research Committee, Seed Fund for Basic Research during the conduct of the study. I. Ng reports grants from Hong Kong Research Grants Council during the conduct of the study. No disclosures were reported by the other authors.

A.P.Y. Ma: Conceptualization, investigation, writing–original draft. C.L.S. Yeung: Investigation. S.K. Tey: Investigation. X. Mao: Investigation. S.W.K. Wong: Investigation. T.H. Ng: Investigation. F.C.F. Ko: Investigation. E.M.L. Kwong: Investigation. A.H.N. Tang: Formal analysis. I.O.-L. Ng: Resources. S.H. Cai: Resources. J.P. Yun: Resources. J.W.P. Yam: Conceptualization, data curation, supervision, funding acquisition, writing–review and editing.

This work was supported by the Hong Kong Research Grants Council, General Research Fund (Project no.: 17100418) and the University Research Committee, Seed Fund for Basic Research (Project no.: 201711159021). I.O.L. Ng is Loke Yew Professor in Pathology. The authors thank Dr. Rakesh Sharma at the Proteomics and Metabolomics Core Facility, LKS Faculty of Medicine, HKU for conducting the fatty acid profiling. Imaging data were acquired using equipment maintained by the Imaging and Flow Cytometry Core, Center for PanorOmic Sciences, LKS Faculty of Medicine, HKU.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.