|Home | About | Journals | Submit | Contact Us | Français|
Hepatocellular carcinoma (HCC) associated mortality is increasing at an alarming rate and there is a readily identifiable cohort of at risk patients with cirrhosis, viral hepatitis, non-alcoholic fatty liver disease, and diabetes. These patients are candidates for chemoprevention. Metformin is an attractive agent for chemoprevention since it is inexpensive, has a favorable safety profile and is well tolerated over long time periods.
We studied the efficacy of metformin as a prevention agent in a clinically relevant rat model of HCC, where tumors develop in the setting of chronic inflammation and cirrhosis. We used repeated injections of diethylnitrosamine to induce sequential cirrhosis and HCC and administered metformin either at the first signs of fibrosis or cirrhosis.
Prolonged metformin exposure was safe and associated with decreases in fibrotic and inflammatory markers especially when administered early at the first signs of fibrosis. In addition, early metformin treatment led to a 44% decrease in HCC incidence, while tumor burden was unchanged when metformin was administered at the first signs of cirrhosis. Interestingly, activation of the hepatic progenitor/stem cell compartment was first observed at the onset of cirrhosis and therefore only early metformin treatment suppressed receptor for advanced glycation end products and inhibited the activation of hepatic progenitor cells.
Our results are the first to demonstrate an effect on progenitor/stem cells in the setting of chemoprevention and provide further rationale to explore metformin as an early intervention in clinical trials of patients with chronic liver disease at high risk for HCC.
Hepatocellular carcinoma (HCC) is the third leading cause of cancer related mortality worldwide.1 Within the United States, the incidence of HCC is rising, along with a more rapid increase in mortality than any other malignancy.2 Treatment options are limited and outcomes remain poor, particularly in the setting of advanced disease. Hepatocellular transformation generally occurs in the presence of chronic liver injury, often as a sequelae of liver fibrosis and cirrhosis. Increasingly, diabetes and obesity have emerged as important risk factors for non-alcoholic fatty liver disease (NAFLD) and non-alcoholic steatohepatitis (NASH), and together are thought to account for 37% of HCC cases not associated with viral hepatitis.3 In contradistinction to several other malignancies, there is a readily identifiable cohort of patients at high risk for HCC who are candidates for chemoprevention as a means to reduce HCC incidence and mortality.
Metformin is the most commonly prescribed drug for the treatment of type 2 diabetes. Over the last decade, several pharmacoepidemiologic studies have strongly suggested a decreased incidence of multiple cancer types, including HCC, in diabetic patients treated with metformin,4–5 as well as improved outcomes.6–7 The mechanism of action of metformin as an anti-hyperglycemic appears to be mediated by indirect effects leading to inhibition of complex I of the respiratory transport chain.8–9 However, a growing body of pre-clinical data has introduced several other potential mechanisms for its anti-cancer and cancer preventive effects.
Importantly, the primary site of action of metformin is within the liver, where it functions to promote a state of energetic stress, thereby decreasing gluconeogenesis and glycogenolysis, as well as systemic insulin and glucose levels, if they are elevated at baseline. Diabetics harbor at least a 2–3× increased risk of developing hepatocellular carcinoma,10–11 and clinical data support an association between long-term metformin exposure and reduction of hepatocellular transformation to levels approaching those observed in non-diabetic patients.12–17 This inexpensive and extremely well-tolerated agent has been reported to exert anti-neoplastic effects in cell culture and animal models via both direct effects on the cancer cells and indirect host-mediated and often insulin-dependent mechanisms. Common themes include induction of metabolic stress, which may or may not be mediated by 5′ adenosine monophosphate-activated protein kinase (AMPK), inflammatory and immune mediated effects,18–19 and more recently inhibitory effects on cancer stem cells.20–21
Despite an explosion of pre-clinical data, only a few studies have explored the effects of metformin in the context of HCC. Some studies have reported direct effects on HCC cell lines, mainly AMPK mediated cell cycle arrest at G0/G1.16, 22 However, these experiments were performed at a very high doses (10–20 mM), which would be roughly 500–1000 times more than peak plasma concentrations observed in diabetic patients taking metformin for glycemic control. Another study utilized a single-dose diethylnitrosamine (DEN) mouse model, which induces liver tumorigenesis in the absence of cirrhosis, and reported an impressive 57% decrease in the number of liver tumors in animals treated with metformin.23 These indirect effects appeared to be mediated by AMPK-independent inhibition of hepatic lipogenesis. It is unknown whether these findings are reproducible when HCC develops in the setting of chronic liver disease and/or cirrhosis, as is typically the case in humans. Therefore, in the following study, we explore metformin as an agent of chemoprevention when administered either at the first signs of fibrosis or cirrhosis in a rodent model of sequential fibrosis, cirrhosis and HCC that more closely resembles human disease at the biochemical, histological and molecular level.24
Metformin (Glucophage) was freshly dissolved in water before treatment.
Male Wistar rats (Charles River Laboratories, Wilmington, MA) were housed in accordance with the guidelines of the Massachusetts General Hospital Subcommittee on Research Animal Care and received humane care according to the criteria outlined in the “Guide for the Care and Use of Laboratory Animals” of the National Academy of Sciences. Rats were subjected to either control PBS or low-dose 50 mg/kg DEN (Sigma, St. Louis, MO) injected IP once per week over the course of 18 weeks. PBS control rats began treatment with metformin by oral gavage at 8 weeks while DEN-injured rats began treatment with metformin by oral gavage either at 8 weeks or 12 weeks (n=9 for each group). At the time of sacrifice rats were anesthetized and sedated. A terminal blood collection was performed by cardiac puncture and livers were removed for measurement of weight, snap frozen for further analysis or fixed in formalin for histology.
A cardiac terminal blood withdrawal was performed at the time of sacrifice. Blood was allowed to clot for 2 h at room temperature before centrifugation at 2,000 rpm for 10 min at 4° C. Serum was isolated and stored at −80° C prior to use. Serum levels of several biochemical markers including alkaline phosphatase (ALP), aspartate transaminase (AST), total bilirubin (TBIL), albumin (Alb), glucose (Glu) and triglycerides were measured as previously described (DRI-CHEM 4000 Analyzer, Heska, Switzerland).24
Formalin-fixed samples were embedded in paraffin, cut into 5 μm-thick sections and stained with hematoxylin-eosin (H-E), Masson’s trichrome and Sirius red according to standard procedures. Trichrome stained sections were analyzed to score the amount of fibrosis according to the method of Ishak.25 The collagen proportional area (CPA) was morphometrically quantified on Sirius red stained sections with image processing software (ImageJ, NIH). Additional sections were stained with an antibodies specific for receptor for advanced glycation end products (RAGE), delta-like homologue 1 (DLK-1) (both from Abcam, Cambridge, MA) and OV-6 (R&D Systems, Minneapolis, MN). All slides were reviewed blindly by the same liver pathologist.
Livers were homogenized and protein was extracted in radioimmunoprecipitation assay (RIPA) buffer (Boston BioProducts, Ashland, MA) containing protease and phosphatase inhibitors (Sigma). Protein concentrations were normalized to 40μg using bicinchoninic acid (BCA) method (Pierce Chemical Co., Rockford, IL). Samples were separated by SDS-PAGE and transferred to a polyvinylidene difluoride membrane (Millipore, Billerica, MA). Immunoblotting for total AMPK, phospho(Thr172) AMPK, total acetyl-coA carboxylase (ACC), phospho(Ser79) ACC (this antibody detects Ser79 in ACC1 and its equivalent site Ser218 in ACC2) (all from Cell Signaling Technology, Beverly, MA), receptor for advanced glycation endproducts (RAGE) (Abcam), solute carrier family 22, member 1 (SLC22A1) (Sigma) and solute carrier family 22, member 3 (SLC22A3) (Abcam) was performed. β-actin (Abcam) was used as a control. Blots were incubated with appropriate secondary antibodies conjugated to horseradish peroxidase (HRP; GE Healthcare, United Kingdom) and developed with a chemiluminescent HRP substrate (Perkin-Elmer, Waltham, MA). Western blots were repeated at least three times to ensure reproducibility.
Liver lysates from PBS control animals and DEN-injured animals that were treated ± metformin were prepared as described above. Protein concentrations were determined by the BCA method and 3 samples from each group were pooled together. 200 μg of pooled protein from each group were analyzed with a commercially available Rat Adipokine Array (ARY016; R&D Systems) according to the manufacturer’s instructions. This array determines the relative expression of 30 proteins implicated in adipogenesis.
Insulin-like growth factor 1 (IGF-1) protein expression was quantified in both liver tissue and serum using commercially available ELISA systems (R&D systems) according to the manufacturer’s instructions. Each sample was quantified in triplicate and experiments were repeated to ensure accuracy.
RNA was isolated from liver tissue using TRIzol (Life Technologies, Grand Island, NY) according to the manufacturer’s instructions and subsequently treated with DNAse I (Promega, Madison, WI). Total RNA (250 ng) from each sample was used to synthesize cDNA by single strand reverse transcription (SuperScript III First-Strand Synthesis SuperMix; Life Technologies). Expression of alpha smooth muscle actin (α-SMA), CD44, chemokine (C-C motif) ligand 2 (CCL2), chemokine (C-X-C motif) ligand 16 (CXCL16), cluster of differentiation 68 (CD68), collagen, Type 1, alpha 1 (COL1A1), delta-like 1 homolog (DLK-1), RAGE, secreted phosphoprotein 1 (SPP1, osteopontin), sterol regulatory element-binding protein 1c (SREBP-1c), TIMP metallopeptidase inhibitor 1 (TIMP1), and transforming growth factor beta 1 (TGF-β1) in liver tissue was analyzed by quantitative real-time PCR using TaqMan gene expression assays (Life Technologies) on Applied Bioscience 7900HT Fast Real Time PCR system using 96 well plates with a reaction volume of 20 uL. The 2−ΔCT method was used for relative quantification of mRNA with normalization to 18S. The Taqman assays were as follows: ACTA2 Rn01759928_g1, CD44 Rn00563924_m1, CCL2 Rn00580555_m1, CXCL16 Rn01496393_m1, CD68 Rn01495634_g1, COL1A1 Rn01463848_m1, DLK1 Rn00587011_m1, RAGE Rn01525753_g1, SPP1 Rn01449972_m1, SREBP-1c Rn01495769_m1, TIMP1 Rn01430873_g1 and TGF-β1 Rn00572010_m1.
Data are represented as mean ± standard deviation. Statistical significance of difference between groups was calculated using an unpaired two-tailed t test.
We utilized a carcinogen based model of recurrent liver injury to investigate the effects of metformin on the development of HCC in a cirrhotic microenvironment. Based on our previous results,24 rats subjected to weekly DEN at a low dose (50 mg/kg) develop fibrosis after 8 weeks, followed by cirrhosis at 12 weeks with substantial HCC development after 18 weeks (Fig. 1a). Therefore, metformin (250 mg/kg once daily by oral gavage) was administered to PBS control rats beginning after 8 weeks and to DEN-injured rats after either 8 weeks or 12 weeks. While metformin significantly (p < 0.05) decreased body weights by 12.3% in PBS controls, there was no change in total body weight in DEN-injured rats treated with metformin (Fig. 1b). Importantly, metformin afforded a significant (p < 0.01) 44% decrease in the number of surface tumors when administered after the first signs of fibrosis (metformin early) (Fig. 1c). By comparison, metformin did not decrease liver tumor development when administered after the first signs of cirrhosis (metformin late). Overall, metformin was well-tolerated and did not increase markers of liver injury including ALP (Fig. 1D) and AST (Fig. 1E).
Macroscopically, the livers from PBS controls treated with or without metformin appeared normal, while the livers from DEN-injured rats appeared cirrhotic (Fig. 2a) and contained many HCCs (Fig. 2b). Interestingly, only the livers from DEN-injured rats that received metformin early were less fibrotic (Fig. 2a). In fact, explanted livers from DEN-injured rats that received metformin early had consistently diminished collagen deposition as assessed by Sirius red staining (Fig. 2a). Morphometric quantification of these stains revealed that early metformin treatment significantly (p < 0.01) decreased collagen deposition by 57% (Fig. 3a). However, there was no measurable difference in Ishak scores as assessed by Masson’s trichrome staining (Fig. 3b). This discrepancy between quantitative collagen deposition and Ishak score after metformin treatment is consistent with a previous report.26 In addition, only DEN-injured rats that received metformin at the early time point showed a marked improvement in serum total bilirubin (Fig. 3c) and albumin (Fig. 3d) levels which would also suggest some mitigation in the burden of chronic liver disease. Finally, there was also a trend for less inflammation after early metformin treatment (Fig. 3e) as assessed by H&E staining (Fig. 2a) but this difference did not reach significance (p = 0.06). We did, however, observe that metformin, especially when administered early, decreased the mRNA expression of several known fibrotic and inflammatory markers including TGF-β1, α-SMA, COL1A1, TIMP1, CD68, SPP1, CCL2 and CXCL16 (Supp. Fig. S1).
There is variability within the literature concerning whether AMPK, the metabolic checkpoint and energy sensor, is activated by metformin treatment, particularly when metformin is used as an anti-neoplastic agent in non-diabetic models. AMPK controls several metabolic processes including fatty acid and protein biosynthesis. With respect to HCC, one group reported AMPK activation (phosphorylation) in HepG2 and Hep3B cells treated with very high metformin dosing (20 mM),16 while AMPK activation actually decreased in the liver after metformin treatment in a single-dose DEN mouse model of HCC in the absence of cirrhosis.23 In the current study, we also note decreased AMPK activation in rat livers from PBS controls after treatment with metformin (Fig. 4a). By comparison, in the setting of DEN-induced cirrhosis, p-AMPK increases dramatically with metformin treatment (Fig. 4b). Total AMPK levels remained stable in all animals and do not account for differences between groups. While the published work in a non-cirrhotic mouse model demonstrated no metformin mediated activation of AMPK, it did show decreased ACC and SREBP-1c expression. It was therefore suggested that metformin prevents HCC development in these non-cirrhotic animals by inhibiting pathways driving hepatic lipogenesis.23 We also observed decreased mRNA expression of SREBP-1c, the key transcription factor driving expression of fatty acid synthesis enzymes, but not protein levels of ACC, the rate limiting enzyme involved in fatty acid synthesis, in livers from our PBS control rats treated with metformin (Fig. 4a,c). However, SREBP-1c mRNA and ACC protein levels were also downregulated in DEN-injured livers suggesting that lipogenesis might already be diminished during chronic liver injury (Fig. 4a,c). Importantly, metformin treatment did not further decrease their expression in cirrhotic rat liver. Metformin did increase pACC, the inactive form of the enzyme, which is thought to be phosphorylated by activated AMPK. However, we did not observe any significant differences in serum triglycerides levels between the different groups (Fig. 4c). Taken together, our results are different from those reported in non cirrhotic livers,23 and suggest that mechanisms other than diminished hepatic lipogenesis are also responsible for the decreased incidence of HCC.
We therefore examined whether DEN-induced chronic liver injury has any effect on organic cation transporters 1 (OCT1 or SLC22A1) and 3 (OCT3 or SLC22A3) which are responsible for the uptake of metformin. Western blot analysis of SLC22A1 and SLC22A3 revealed that their expression was not altered by DEN (Fig. 4a). In addition, metformin treatment did not alter their expression.
Given the numerous proposed mechanisms of action for metformin, we next took a broad approach and examined the expression of several growth factors, chemokines and cytokines using a commercially available adipokine array. We analyzed protein expression in normal livers from PBS controls, cirrhotic livers from DEN-injured rats, and livers from DEN-injured rats treated with metformin either at the early or late time point. Several known mediators of inflammation and fibrogenesis were up-regulated in DEN-injured rats, and in general early administration of metformin reduced the expression of these proteins. However, the major observed effects of early metformin treatment were on the expression of RAGE and DLK-1 (Supp. Fig. S2). Interestingly, RAGE is a known soluble mediator of oval cell activation and DLK-1 is a prototypical marker for oval cells. Oval cells represent a class of bipotential hepatic progenitor cells, which are activated during severe hepatic injury, have the potential for self-renewal, and can differentiate into both hepatocytes and cholangiocytes. In humans, these bipotential hepatic progenitor cells are markedly elevated in the setting of chronic liver disease and cirrhosis27 and their activation appears to correlate with HCC incidence,28 tumor progression,29 and resistance to chemotherapy.30
Previously, RAGE ablation has been shown to significantly inhibit hepatocarcinogenesis in an Mdr2−/− model of chronic liver injury through suppression oval cell activation.31 We therefore decided to verify the effects of metformin on RAGE expression in the DEN rat model. Consistent with the adipokine array, RAGE mRNA expression was elevated in DEN-injured livers and significantly (p < 0.01) decreased by 86% or 59% after early or late treatment with metformin, respectively (Fig. 5a). Interestingly, we also stained liver sections for RAGE and observed increased expression in DEN-injured livers which was only inhibited by early metformin treatment (Fig. 5d). Since RAGE exists in transmembrane and soluble forms, we also examined protein expression in both liver lysates and serum by western blot analysis. Again, RAGE expression was elevated in DEN-injured animals and suppressed by early metformin treatment (Supp. Fig. S2).
DLK-1 is a prototypical marker of oval cells. In fact, DLK-1(+) cells have been identified in sub-populations of hepatic progenitor cells and in human HCC. They have been demonstrated to exhibit stronger proliferation and tumorigenicity, possess the potential for self renewal, and possess stronger properties of chemoresistance compared with DLK-1(−) counterparts.32 We confirmed the effects of metformin on oval cell activation by examining the mRNA expression of DLK-1 and CD44, another marker of hepatic progenitor cells. DLK-1 was barely detectable in normal livers but its expression increased dramatically in DEN-injured livers and decreased by 84% or 89% after early or late treatment with metformin, respectively (Fig. 5b). Likewise, CD44 was also barely detectable in normal livers and its expression increased dramatically in DEN-injured livers and decreased by 78% or 35% after early or late treatment with metformin, respectively (Fig. 5c). We also performed IHC for oval cells using antibodies for DLK-1 and OV-6, a well-characterized oval cell antigen. As expected, DLK-1 and OV-6 both stained oval cells in the portal tracts and the staining increased with DEN-induced fibrosis (Fig. 5d). Interestingly, similar to the results with RAGE, these regions of marked ductular reaction were only decreased by early administration of metformin.
Since late administration of metformin also decreased the mRNA expression of DLK-1 and to a lesser extent RAGE and CD44 (Fig. 5a–c), we decided to examine the expression of these markers at the time points when metformin treatment was initiated. Interestingly, while small increases in RAGE and CD44 mRNA levels were seen after 8 weeks of DEN treatment (fibrosis), the expression of RAGE, DLK-1 and CD44 increased by 3.4 fold, 188 fold, and 129 fold, respectively, after 12 weeks of DEN treatment (cirrhosis). We next examined the expression of RAGE, DLK-1 and OV-6 by IHC in livers from rats injured with DEN for either 8 weeks or 12 weeks. While RAGE staining and activation of oval cells, as assessed by DLK-1 and OV-6, were minimal at 8 weeks, they dramatically increased at 12 weeks with the onset of cirrhosis (Fig. 6d). These data therefore suggest that even though late administration of metformin can still reduce the mRNA expression of DLK-1 and CD44 as well as some markers of fibrosis and inflammation (Supp. Fig. S1), its ability to prevent HCC requires early administration before the dramatic activation of hepatic progenitor cells with the onset of cirrhosis. Overall, our results suggest that early metformin treatment suppresses the expression of RAGE a soluble mediator of oval cell activation and thereby reduces HCC development.
Current therapeutic options for HCC remain limited and an effective chemoprevention strategy is one of the most viable means of curbing rapidly increasing mortality rates worldwide. Epidemiology studies have shown that each year of metformin use is associated with an incremental decreased risk of HCC in diabetic patients.16 Interestingly, we found that metformin reduced HCC incidence by 44% when administered for 10 weeks (from weeks 9 to 18) but no reduction was seen when metformin was administered for 6 weeks (from weeks 13 to 18). Of note, Tajima et al. found that metformin decreased HCC incidence when administered concurrently with high fat diet for 60 weeks; however, metformin failed to protect against liver tumorigenesis if treatment was delayed until week 30.33 While the efficacy of metformin is probably dependent upon both the amount of underlying disease present at the time of treatment in addition to the duration of treatment, our study definitely provides further evidence that early and long-term treatment with metformin is a promising strategy in the prevention of tumors within the liver, its main site of action.34
In the majority of cases, primary liver cancer arises in a cirrhotic microenvironment. It is therefore crucial to consider the field effects of a chronically injured liver when evaluating the efficacy of any proposed prevention agent. Of the animal models currently available,35 we have shown that the low-dose DEN rat model with recurrent DEN administration closely recapitulates human hepatocarcinogenesis with the development of fibrosis (8 weeks), cirrhosis (12 weeks), and ultimately HCC (18 weeks).24
The mechanism of metformin mediated decrease in HCC development is almost certainly multifactorial and includes modulation of inflammatory and fibrotic mediators within the surrounding liver mileu. Although metformin did not have a measureable effect on Ishak cirrhosis score, there was an unquestionable decrease in the volume of both periportal fibrosis and web-like collagen deposition between portal tracts visible by staining in animals treated with metformin. This effect was quantified by Sirius red staining and is supported by metformin mediated decreases in the mRNA expression of COL1A1 as well (Supp. Fig. S1).
Despite a growing body of literature on the topic, the mechanism of action of metformin as an anti-neoplastic agent remains controversial. It is generally believed, that metformin exerts its effects via inhibition of complex I of the mitochondrial respiratory chain, with a resultant increase in the ratio of AMP and ADP to ATP. This is thought to trigger activation of the LKB1/AMPK pathway with downstream inhibitory effects on protein and fatty acid synthesis.18 The importance of this pathway becomes difficult to interpret in the setting of cirrhosis. We do in fact however show significant activation of AMPK in response to metformin treatment in surrounding liver tissue with subsequent inactivation of ACC, the rate-limiting step of de novo lipogenesis. While Bhalla et al. reported inhibitory effects of metformin on hepatic lipogenesis in a non-cirrhotic mouse model (single-dose DEN),23 it is unlikely that metformin-mediated effects on fatty acid synthesis meaningfully contribute to HCC prevention in the setting of chronic liver injury and cirrhosis based on our findings. Interestingly, metformin also did not change levels of lipogenioc genes, including SREBP-1c, in a high fat diet model of non-alcoholic fatty liver disease.33
Another commonly proposed mechanism of action explaining metformin’s efficacy as an anti-neoplastic agent is via indirect effects on tumor cells resulting from diminished levels of insulin, IGF-1, and glucose.18 However, based on analysis at the time of sacrifice, metformin did not alter liver or circulating levels of IGF-1 and serum glucose levels did not change in response to metformin treatment in this non-diabetic model (Supp. Fig. S3).
Expansion and activation of the hepatic progenitor cell compartment is widespread in chronic liver disease, and although a direct causal relationship linking oval cells and hepatocarcinogenesis has not been proven formally, it is generally agreed that oval cell activation initiates or promotes HCC.36–39 Intriguingly, the multiligand receptor RAGE, a known driver of chronic inflammation and diabetes associated complications, was shown by Pusterla et al. to be a key regulator of oval cell activation and tumor development in an mdr2−/− model of inflammation associated liver carcinogenesis.31 In this model, RAGE ablation led to a marked reduction in oval cell activation and impaired tumor development. In addition, several studies have implicated RAGE in the development of chronic liver damage, inflammation, and fibrosis.40–41 Our findings confirm RAGE overexpression in cirrhotic rats and we report, for the first time to our knowledge, a metformin associated decrease in RAGE signaling in a pre-clinical cancer model.
Recent evidence has linked the anti-neoplastic effects of metformin not only in the liver42 but also in the breast20, 43 and pancreas44–45 to its ability to target subpopulations of cancer stem cells within these tumors. Within the liver, multiple studies have demonstrated stem cell markers in primary HCC and these tumors are associated with increased recurrence and worse survival.46–48 Although our model is best suited to address the possibility of using metformin as a prevention agent, further work is needed to assess whether metformin could be used in combination with current chemotherapeutics in the treatment of established HCC. Preliminary data assessing combination therapy with metformin and doxorubicin in HepG2 and Hep3B xenograft models is promising.16
At present, a search for “metformin and cancer” on clinicaltrials.gov yields 199 ongoing or recently closed clinical trials assessing the efficacy of metformin both as an agent for chemoprevention, as well as a chemotherapeutic. The majority of these trials are investigating breast, pancreatic, lung, prostate, and colorectal cancer, but not a single trial is examining the effects of metformin on primary liver tumors. This is somewhat surprising since metformin’s primary site of action is within the liver. Preclinical studies focusing on metformin and liver cancer have also lagged behind other tumor types, with few studies currently published in the literature. Our data offer additional evidence in support of ongoing research efforts investigating the anti-neoplastic effects of metformin in the treatment and prevention of HCC.
Funding: This work was supported by the National Institutes of Health (grant numbers T32CA071345 (D.K.D., B.S., and K.K.T.) and K01CA140861 (B.C.F.)) and the Massachusetts General Hospital Department of Surgery (K.K.T.).
Conflicts: The authors declare no conflict of interest.