Co-activation of AKT and c-Met triggers rapid hepatocellular carcinoma development via the mTORC1/FASN pathway in mice

Activation of the AKT/mTOR cascade and overexpression of c-Met have been implicated in the development of human hepatocellular carcinoma (HCC). To elucidate the functional crosstalk between the two pathways, we generated a model characterized by the combined expression of activated AKT and c-Met in the mouse liver. Co-expression of AKT and c-Met triggered rapid liver tumor development and mice required to be euthanized within 8 weeks after hydrodynamic injection. At the molecular level, liver tumors induced by AKT/c-Met display activation of AKT/mTOR and Ras/MAPK cascades as well as increased lipogenesis and glycolysis. Since a remarkable lipogenic phenotype characterizes liver lesions from AKT/c-Met mice, we determined the requirement of lipogenesis in AKT/c-Met driven hepatocarcinogenesis using conditional Fatty Acid Synthase (FASN) knockout mice. Of note, hepatocarcinogenesis induced by AKT/c-Met was fully inhibited by FASN ablation. In human HCC samples, coordinated expression of FASN, activated AKT, and c-Met proteins was detected in a subgroup of biologically aggressive tumors. Altogether, our study demonstrates that co-activation of AKT and c-Met induces HCC development that depends on the mTORC1/FASN pathway. Suppression of mTORC1 and/or FASN might be highly detrimental for the growth of human HCC subsets characterized by concomitant induction of the AKT and c-Met cascades.


Overexpression of activated AKT and c-Met induces rapid liver tumor development in mice.
To determine whether c-Met cooperates with activated AKT to induce hepatocarcinogenesis in vivo, we stably expressed pT3-EF1α -HA-myr-AKT1 and/or pT3-EF1α -V5-c-Met, along with the sleeping beauty transposase (SB), in the mouse liver using hydrodynamic transfection. As we described previously 19,20 , overexpression of AKT alone induced hepatic steatosis and proliferation, leading to HCC development after 24 weeks post hydrodynamic injection. Overexpression of c-Met alone did not result in any liver anomaly in mice up to 12 weeks post injection, but eventually led to dysplastic foci formation over long term, as previously reported 22,23 . In striking contrast, co-expression of AKT and c-Met triggered rapid liver tumor development in mice, and all mice developed lethal burden of liver tumor within 6 to 8 weeks post-injection (Fig. 1).
Histopathological analysis was performed to assess the time course of tumor development in AKT/c-Met mice. Four weeks after hydrodynamic transfection, livers of AKT/c-Met mice (n = 3) displayed numerous preneoplastic hepatocytes that contained elevated amounts of glycogen (as assessed by positive PAS reaction; not shown) and lipids (Fig. 2a,b), thus resembling preneoplastic hepatocytes developed in AKT-overexpressing mice 19,20 . Approximately 30-40% of the liver volume consisted of preneoplastic lesions, but no tumors were detected (Fig. 2a,b). Six weeks after injection, numerous hepatocellular tumors consisting mainly of lipid-rich cells were found in AKT/c-Met mice (n = 3; Fig. 2c). AKT/c-Met mice (n = 9) developed lethal burden of liver tumor and were required to be euthanized 8 weeks post injection. Hepatocellular tumors further progressed in size, occupying almost completely the liver parenchyma ( Fig. 2f). At this time point, AKT/c-Met tumors displayed additional signs of malignancy and aggressiveness, such as confluent areas of necrosis and increase in cytologic atypia as well as a loss of lipid content (Fig. 2f). Indeed, tumors were of mixed cell type (admixture of clear, lipid-rich and lipid-poor, basophilic cells), with basophilic cells being smaller than lipid-rich hepatocytes and sometimes showing an oval-cell phenotype (Fig. 2f). Of note, different from AKT mice, no cholangiocellular lesions were detected in AKT/c-Met mice. To confirm that the liver tumors were indeed induced by the ectopically injected oncogenes, we performed immunohistochemistry (IHC) in preneoplastic and neoplastic lesions from AKT/c-Met mice using an anti-HA and anti-V5-tag antibody, which labelled the ectopically injected AKT and c-Met, respectively. As expected, strong immunolabeling for HA-tag and V5-tag, co-localizing in preneoplastic and neoplastic lesions from AKT/c-Met mice, was detected (Fig. 2d,e; Supplementary Figure 1).
AKT/c-Met co-expression promotes activation of the AKT/mTOR and Ras/MAPK pathways in the mouse liver. To elucidate the molecular mechanisms underlying hepatocarcinogenesis in AKT/c-Met mice, we assessed the activation of AKT/mTOR and Ras/MAPK pathways in these mice by Western blotting (Fig. 3). Levels of total AKT were equivalent in liver lesions from AKT and AKT/c-Met mice and higher than in wild-type and c-Met mice, whereas expression of activated/phosphorylated AKT, including p-AKT(T308) and p-AKT(S473), was highest in AKT/c-Met tumors (Fig. 3). Activation of the Ras/MAPK cascade, as indicated by p-ERK1/2 levels, was also highest in AKT/c-Met liver tumors (Fig. 3).
Next, we examined the levels of the major downstream effector cascade of AKT, namely the mTOR pathway. While levels of activated/phosphorylated mTOR and ribosomal protein S6 (RPS6) as well as inactivated/phosphorylated 4E binding protein one (4EBP1) were equivalent in AKT, c-Met, and AKT/c-Met livers, upregulation of mTORC1 downstream effectors involved in de novo lipogenesis (stearoyl-CoA desaturase 1 or SCD1) and glycolysis (pyruvate kinase M1 or PKM1, PKM2, and lactate dehydrogenase A/C or LDHA/C) was most pronounced in AKT/c-Met samples (Fig. 3). Altogether, these results indicate that activation of AKT/mTOR and Ras/MAPK cascades is a molecular feature of AKT/c-Met driven hepatocarcinogenesis.
Aggressive hepatocarcinogenesis induced by AKT and c-Met co-expression is abolished by suppression of mTORC1/FASN pathway in mice. Next, since lipogenesis and glycolysis are primarily regulated in the liver by mTORC1, we determined whether an intact mTORC1 axis is needed for AKT/c-Met hepatocarcinogenesis in mice. To achieve this goal, we applied miR-30 based shRNA to silence Raptor 24 , the unique subunit of mTORC1 complex in vivo (Fig. 4). A previously described shRaptor sequence that showed efficient silencing of mouse Raptor was selected and used for the experiments 25 . We further validated the efficiency of shRaptor in the AKT/Ras mouse liver tumor cell line (Fig. 4e), which has been previously described 20 . The shRaptor sequence was cloned downstream of AKT in the pT3-EF1a vector (AKT-shRaptor). As a control, shRNA against Renilla Luciferase was also cloned into the pT3-EF1a-AKT plasmid (AKT-shLuc) (Fig. 4d). AKT-shRaptor or AKT-shLuc was hydrodynamically injected into mice together with c-Met (Fig. 4a). Consistent with results obtained in AKT/c-Met mice, all AKT-shLuc/c-Met injected mice developed lethal burden of liver tumors and needed to be euthanized by 10 weeks post injection. In striking contrast, liver tissues from AKT-shRaptor/c-Met injected mice appeared to be completely normal at the same time point (Fig. 4b,c). Thus, the present data indicate that AKT/c-Met driven hepatocarcinogenesis depends on functional mTORC1.
To further investigate the importance of metabolic components regulated by mTORC1 in AKT/c-Met hepatocarcinogenesis, we determined the effect of disrupting de novo lipogenesis in these mice. For this purpose, we inactivated FASN, the major player in aberrant lipid biosynthesis 14,15 , using conditional FASN knockout mice (FASN fl/fl mice). Thus, two approaches were applied. In the first approach, AKT, c-Met and Cre plasmids were co-injected into FASN fl/fl mice (AKT/c-Met/Cre; n = 3), thus allowing the simultaneous expression of AKT and c-Met oncogenes while deleting FASN in a subset of mouse hepatocytes. As a control, AKT, c-Met and pT3-EF1α (empty vector) were co-injected into FASN fl/fl mice (AKT/c-Met/pT3, n = 3) (Fig. 5a). We found that all AKT/c-Met/pT3 mice developed liver tumors by 8 weeks post injection (Fig. 5b), whereas no macroscopic or histopathological alterations were detected in the livers of AKT/c-Met/Cre mice up to 15 weeks post hydrodynamic injection (Fig. 5c). At the molecular level, AKT/c-Met/Cre mice displayed very faint or no immunoreactivity for AKT (HA) and c-Met (V5) tags as well as for FASN, p-AKT, SCD1, LDHA/C, p-RPS6, p-4EBP1, and p-ERK1/2 (

FASN post-transcriptionally regulates the levels of c-Met in human hepatoma cell lines.
Since previous studies demonstrated that inhibition of FASN suppresses c-Met expression in lymphoma cells 26 , we investigated whether the same mechanism may contribute to the suppression of liver tumor development AKT/c-Met mice when FASN is deleted. For this purpose, FASN expression was modulated in human hepatoma cell lines and protein levels of c-Met were assessed (Fig. 6). Silencing of FASN via specific siRNA resulted in a pronounced downregulation of c-Met protein in HLF and HepG2 cell lines (Fig. 6a,b). To ascertain whether FASN inhibition affects c-Met transcription, thus accounting for the loss of c-Met protein, we performed real-time quantitative reverse-transcription PCR on RNA prepared from HLF and HepG2 cells untreated and subjected to scramble and FASN siRNA. Intriguingly, we found that mRNA levels of c-Met gene were unmodified by FASN silencing when compared with those from untreated and scramble-treated liver tumor cells (Fig. 6a,b), indicating that regulation of c-Met by FASN occurs at the post-transcriptional level. Next, we assessed whether cap-dependent translation is responsible for the regulation of c-Met levels in HCC cells. For this purpose, the HepG2 cell line was subjected to the treatment with the cap-dependent translation inhibitor, 4EGI-1 (Supplementary Figure 5). Administration of 4EGI-1 did not result in a decrease but rather in an upregulation of c-Met levels (Supplementary Figure 5a). The latter findings seem to exclude that c-Met is positively regulated by cap-dependent translation in HepG2 cells. To further investigate the possible mechanism(s) responsible for c-Met downregulation in FASN-depleted cells, we compared the rate of c-Met loss on treatment with the de novo protein synthesis inhibitor, cycloheximide (CHX), either alone or in combination with the FASN inhibitor, C75. Noticeably, loss of c-Met protein was equivalent in CHX-and C75-treated HepG2 cells 48h after the treatment started ( Supplementary Figure 5b), whereas combined treatment with C75 and CHX did not result in a synergistic or additive effect on reducing c-Met protein stability (Supplementary Figure 5b). These results suggest that FASN inhibition might reduce c-Met protein stability in HepG2 cells. Furthermore, we determined whether protein degradation via the proteasome system is involved in c-Met downregulation in HCC cells. However, no increase, but rather decrease, in the ubiquitinylated levels of c-Met was detected following FASN silencing (Fig. 6a,b) in both HLF and HepG2 cells, thus excluding that FASN regulates c-Met levels via the proteasome system.
Altogether, our data indicate that FASN might contribute to preserve c-Met protein stability in HCC cells.

Discussion
Mounting evidence underlines the role of AKT and c-Met proto-oncogenes in human HCC [6][7][8][9][10][11][12]21 . However, whether AKT and c-Met functionally cooperate in liver cancer remains poorly delineated. In the present study, we have addressed this issue for the first time in mice. Our results show indeed that concomitant overexpression of AKT and c-Met in the mouse liver results in a synergistic activity of the two proto-oncogenes, leading to rapid tumor development. In accordance with the mouse data, we have found that combined suppression of AKT and c-Met signaling cascades is highly detrimental for the in vitro growth of human HCC cell lines. Importantly, suppression of AKT activity by its specific inhibitor, MK2206, resulted in the downregulation of activated c-Met in both HLE and HLF cells. Although the mechanisms whereby AKT regulates c-Met activity require additional investigation, the present data uncover a previously unrevealed crosstalk between AKT and c-Met proteins in HCC cells.
Histologically, preneoplastic and neoplastic lesions developed in AKT/c-Met mice, mainly consisting of lipid-rich cells, closely resemble those from mice overexpressing AKT alone 19,20 . However, different from AKT mice, in which liver lesions with hepatocytic, ductular, and mixed differentiation developed 19,20 , AKT/c-Met mice exhibited only lesions with hepatocytic features. Thus, these data imply that overexpression of c-Met promotes the development of liver lesions characterized by a commitment toward the hepatocyte lineage. Although the mechanisms whereby c-Met drives development of liver tumors with hepatocytic differentiation remain to be elucidated, our present data are in agreement with a recent report using liver cell lines in which the expression of c-Met and epidermal growth factor receptor (EGFR) has been modulated 27 . In the latter study, the authors showed in fact that c-Met is a strong inducer of hepatocyte differentiation, whereas EGFR promotes cholangiocyte specification while concomitantly suppressing hepatocyte commitment via NOTCH-dependent mechanisms 27 .
At the molecular level, we found that simultaneous overexpression of AKT and c-Met in the liver triggers the sustained activation of the AKT/mTOR and Ras/MAPK cascades. Of note, AKT/c-Met lesions displayed the selective induction of mTOR targets involved in glycolysis and de novo lipogenesis, whereas the levels of other canonical effectors of this pathway, such as p-RPS6 and p-4EBP1, were not upregulated when compared with AKT corresponding lesions. Together with the metabolic effects resulting from FASN suppression, however, we cannot exclude that FASN plays also additional roles on AKT/c-Met cells. For instance, FASN depletion either in vivo or in vitro resulted in the downregulation of c-Met protein, with no changes in c-Met mRNA levels, in accordance with previous data in human breast, prostate, and lung cancer cell lines 28,29 . In addition, our data speak against a major role played by the proteasome system in the regulation of c-Met by FASN, as ubiquitination of c-Met was not increased following FASN silencing in hepatoma cell lines. Of note, the negative regulation of c-Met protein levels by FASN independent of the proteasome system has been previously reported in the DU145 prostate cancer cell line following the treatment with the FASN inhibitor, luteolin 28 , further indicating that downregulation of c-Met occurs via mechanisms that are ubiquitin-independent in cancer. Nonetheless, our present data suggest that FASN might be implicated in the regulation of c-Met protein stability, as the loss of c-Met protein in HepG2 cells was equivalent following the treatment with the FASN inhibitor C75 and the protein synthesis inhibitor CHX, whereas the two inhibitors did not act synergistically to downregulate c-Met when used in combination. Concerning the precise mechanisms whereby FASN regulates c-Met stability and activity, it has been recently hypothesized that FASN activity maintains lipid rafts, which may help to stabilize the levels of c-Met 28 . Lipid rafts are plasma membrane regions that regulate cellular signaling, at least partly through the compartmentalization of growth factor receptors 30,31 . Since it has been shown that the active form of c-Met resides in lipid rafts 32 , it is possible that disruption of lipid rafts following FASN suppression might trigger the inhibition of c-Met signaling. However, other mechanisms might also play an important role in FASN-mediated control over c-Met levels. For instance, FASN may regulate c-Met levels via microRNA modulation. In accordance with this hypothesis, recent studies showed that fatty acids are important regulators of microRNAs in the liver 33 . Thus, it would be important to further test whether FASN influences the activity of microRNAs that regulate the expression of c-Met in HCC.
Importantly, the current study expands the observation that FASN and its mediated lipogenesis are required for AKT driven carcinogenesis. Our group 34 and others 35 have previously reported that AKT-overexpressing cells are incapable of survival and proliferation in vitro when de novo fatty acid synthesis is inhibited. It is worthwhile remarking that hepatocytes overexpressing AKT still rely on FASN when c-Met is co-expressed. Indeed, despite the strong acceleration of hepatocarcinogenesis driven by co-transfection of AKT and c-Met protooncogenes in the mouse liver, AKT/Met-overexpressing cells are still dependent on the presence of FASN to exert their oncogenic potential. Furthermore, we found that AKT/Met dependent hepatocarcinogenesis is not rescued by dietary fatty acids supplementation in AKT/Met mice depleted of FASN, indicating that AKT/Met cells are unable to compensate the inhibition of de novo fatty acid synthesis with exogenous fatty acid uptake. Thus, the present findings indicate that c-Met upregulation hastens tumor development in AKT-injected hepatocytes albeit without rendering these cells resistant to FASN depletion.
Nonetheless, the impact of FASN inhibition might be not limited to liver tumors overexpressing AKT and c-Met. In accordance with the latter hypothesis, we have recently found that ablation of FASN strongly delays c-Myc induced liver tumor development in the mouse (Che L, unpublished results). Thus, our findings together support the hypothesis that increased de novo lipogenesis is a key metabolic feature of hepatocarcinogenesis, presumably not limited to AKT/c-Met overexpressing tumors. As a lipogenic phenotype characterizes preneoplastic and neoplastic murine liver lesions as well as human HCC and predisposing conditions (NASH, NAFLD, etc.), drugs targeting de novo lipogenesis may be useful both as chemopreventive and therapeutic agents for liver cancer. Since inhibitors of FASN are already commercially available and used for the treatment of obesity with a good safety profile 36 , clinical trials using these drugs in liver cancer should be designed.
Finally, we showed that preneoplastic and neoplastic lesions from AKT/c-Met mice exhibit high levels of AKT/mTOR and Ras/MAPK cascades. In a previous study, we found the coordinated activation of AKT/mTOR and Ras/MAPK cascades in a subset of human HCCs with aggressive biological behavior 23 . Thus, the AKT/c-Met mouse model of hepatocarcinogenesis might represent a valid preclinical tool to investigate the therapeutic potential of various targeted therapies against HCC.
Hydrodynamic transfection and mouse monitoring. Wild-type FVB/N mice were obtained from Charles River (Wilmington, MA). The FASN fl/fl mouse (in the C57BL/6 background) has been previously described 37 . AlbCre mice 38 were purchased from Jackson Laboratory (Bar Harbor, ME). AlbCre mice were crossed with FASN fl/fl mice to eventually generate liver specific FASN knockout mice (AlbCre;FASN fl/fl mice). Hydrodynamic transfection was performed as described 19,20,22,23,39,40 . In brief, the plasmids encoding the gene(s) of interest along with sleeping beauty transposase (SB) in a ratio of 25:1 were diluted in 2 ml saline (0.9% NaCl), filtered through 0.22 μ m filter, and injected into the lateral tail vein of the mice in 5 to 7 seconds. For high fat diet treatment of mice, AKT/c-Met/Cre injected FASN fl/fl mice were fed high fat soft pellets with 60% fat calories (Bio-Serv, Flemington, NJ) starting from the second day after the hydrodynamic injection for 10 weeks. Mice were housed, fed, and monitored in accordance with protocols approved by the Committee for Animal Research at the University of California, San Francisco.
Immunohistochemical staining. Liver specimens were fixed in 4% paraformaldehyde and embedded in paraffin. Preneoplastic and neoplastic liver lesions were assessed by two board-certified pathologists (M.E. and F.D.) in accordance with the criteria by Frith et al. 41 , as previously described in detail 40 . For immunohistochemistry, deparaffinized sections were incubated in 3% H 2 O 2 dissolved in 1× phosphate-buffered saline (PBS) for 30 minutes to quench the endogenous peroxidase. For antigen retrieval, slides were microwaved in 10 mM citrate buffer (pH 6.0) for 12 minutes. Subsequently, slides were incubated with primary antibodies (Supplementary Table 2) overnight at 4 °C. All the primary antibodies used in the present investigation were selected among those that were previously validated by the manufacturers for immunohistochemistry. The immunoreactivity was