Pan-mTOR inhibitor MLN0128 is effective against intrahepatic cholangiocarcinoma induced in mice by AKT and Yap co-expression

Background & Aims— Intrahepatic cholangiocarcinoma (ICC) is a lethal malignancy without effective treatment options. MLN0128, a second-generation pan-mTOR inhibitor, shows efficacy for multiple tumor types. Methods— We established a novel ICC mouse model via hydrodynamic transfection of activated forms of AKT (myr-AKT) and Yap (YapS127A) protooncogenes (that will be referred to as AKT/ YapS127A). Genetic approaches were applied to study the requirement of mTORC1 and mTORC2 in mediating AKT/YapS127A driven tumorigenesis. Gemcitabine/Oxaliplatin and MLN0128 were administered in AKT/YapS127A tumor-bearing mice to study their antitumor efficacy in vivo. Multiple human ICC cell lines were used for in vitro experiments. Hematoxylin and eosin staining, immunohistochemistry and immunoblotting were applied for characterization and mechanistic study. Results— Co-expression of myr-AKT and YapS127A promoted ICC development in mice. Both mTORC1 and mTORC2 complexes were required for AKT/YapS127A ICC development. Gemcitabine/Oxaliplatin had limited efficacy in treating late stage AKT/YapS127A ICC. In contrast, partial tumor regression was achieved when MLN0128 was applied in the late stage of AKT/YapS127A


Introduction
Intrahepatic cholangiocarcinoma (ICC) is the second most common liver tumor, whose incidence is rapidly rising in the Western countries [1]. Patients with ICC are commonly diagnosed at advanced stages. This excludes them from surgical resection, the only possible curative treatment [2]. Gemcitabine/Oxaliplatin based regimen is used as a standard chemotherapy for advanced ICC, but its efficacy is rather limited [3]. Therefore, there is an urgent need to develop novel therapeutic strategies for the treatment of this lethal malignancy.
MLN0128 is a selective, highly potent and orally active mTOR inhibitor with Ki for mTOR at 1.4nM, and Ki for PI3K greater than 200nM [12]. It belongs to the second generation of mTOR inhibitors, which are ATP analogues and exert their inhibitory effects on the catalytic activity of both mTORC1 and mTORC2 by binding to the kinase domain of mTOR [4,13]. Because of its promising pharmacological effects, MLN0128 is under intensive investigation. MLN0128 has already entered many phase I and II clinical trials as a single agent or in combined therapy, mostly for solid tumors [14]. In preclinical settings, it shows significant therapeutic effects in various tumor types including leukemia, sarcoma, breast cancer, renal cell carcinoma, and colorectal cancer [14][15][16][17][18]. Compared to allosteric inhibitors such as rapamycin, MLN0128 suppresses mTORC1 activity more completely, as indicated by robust and lasting inhibition of 4EBP1, and it also targets mTORC2 to inhibit AKT signaling, thus leading to superior therapeutic effect [14,15,18,19]. However, to date, no study has been conducted investigating the therapeutic effect of second generation mTOR inhibitors in ICC cell lines and animal models.
Here, we established a novel ICC mouse model via hydrodynamic transfection of activated forms of AKT and Yes-associated protein (Yap) oncogenes. Using genetic approaches, we demonstrated that both mTORC1 and mTORC2 are required for ICC development in this model. Noticeably, we found that while Gemcitabine/Oxaliplatin had limited efficacy in this model, especially in late stage ICC, MLN0128 showed significant therapeutic effects. Our study therefore suggests a role for mTOR inhibitors in treating ICC.

Human Tissue Samples
A collection of formalin-fixed, paraffin-embedded ICC (n=94) samples was used in the present study. The clinicopathological features of liver cancer patients are summarized in Supplementary Table 1. ICC specimens were collected at the Medical Universities of Greifswald (Greifswald, Germany) and Sassari (Sassari, Italy). Institutional Review Board approval was obtained at the local Ethical Committee of the Medical Universities of Greifswald and Sassari. Informed consent was obtained from all subjects.

Statistical analysis
Data analysis was performed with Prism 6 (GraphPad, San Diego, CA). Data are presented as Means ± SE. Comparisons between two groups were performed with two-tailed unpaired t test. P values < 0.05 were considered statistically significant.
Additional information can be found in Supplementary Methods.

Concomitant upregulation of Yap and AKT in human ICC samples
First, to evaluate the importance of the Yap and AKT pathways in human cholangiocarcinogenesis, we determined the protein levels of activated/phosphorylated AKT in a collection of human ICC samples by immunohistochemistry. As a surrogate marker of Yap activation, its nuclear accumulation was investigated. We found that activated/ phosphorylated levels of AKT are upregulated, when compared with nontumorous surrounding counterparts, in 68 of 94 (72.34%) ICC specimens (Fig. 1). Equivalent levels of AKT immunoreactivity in ICC and corresponding non-neoplastic livers were detected in the remaining samples. In accordance with previous findings [21,22], nuclear accumulation of Yap was almost ubiquitously detected in ICC (90/94; 95.74%), whereas only faint nuclear and/or cytoplasmic immunoreactivity of Yap was observed in surrounding non-neoplastic counterparts. Noticeably, 63 of 94 (67.02%) ICC displayed simultaneous activation of phosphorylated AKT and Yap, thus suggesting the relevance of the two concomitant oncogenic events along cholangiocarcinogenesis. No association between the status of AKT or Yap and clinicopathologic features of the patients, including age, gender, etiology, presence of cirrhosis, and tumor size was detected (Supplementary Table 2 and 3).

Coordinated activation of AKT and Yap induces ICC in mice
Due to the high frequency of simultaneous activation of AKT and Yap in human ICC, we stably expressed activated forms of human AKT1 (myristoylated AKT1 or myr-AKT1, with C-terminal HA tag) and an unphosphorylatable form of Yap that cannot be targeted for proteasomal degradation by the Hippo kinase pathway (YapS127A) in the mouse liver via hydrodynamic injection (which will be referred to as AKT/YapS127A). As we previously reported, overexpression of myr-AKT alone led to hepatic steatosis and HCC formation ∼30 weeks post injection [23]. Overexpression of YapS127A did not result in any liver abnormality [21]. In contrast, when mice were injected with AKT/YapS127A, as early as 3 weeks post injection, tumor lesions could be observed macroscopically and histologically. Tumors grew rapidly and all mice became moribund, requiring to be euthanized by 5.5 to 7.5 Zhang  weeks post injection ( Supplementary Fig. 1A). Extra-hepatic metastases were not observed in these mice, presumably due to the rapid progression of liver tumors leading to mouse mortality. Histological analysis showed that the tumors displayed a ductular phenotype, resembling human ICC. Consistently, all tumor cells were positive for CK19, a bile duct marker ( Fig. 2A). Furthermore, AKT/YapS127A tumors exhibited positive staining for markers of tissue desmoplasia, such as vimentin and smooth muscle actin (α -SMA) in the stromal cells, a prominent feature of human ICC, which was further proven by Sirius Red staining ( Supplementary Fig. 1B). Western blotting was performed to determine the activation of AKT and Yap signaling in wild-type livers and AKT/YapS127A ICC lesions. As expected, total AKT and phosphorylated AKT at S473 and T308 sites as well as key downstream components of mTORC1 and mTORC2 signaling (p-mTOR, p-PRS6, p-4E-BP1, p-PRAS40 and p-NDRG1) were elevated in AKT/YapS127A mouse cancer tissues. Levels of Yap and its downstream target Jag1 were also increased in AKT/YapS127A tissues. Ras/MAPK pathway was activated, as shown by increased level of p-ERK. Furthermore, we detected high levels of Hexokinase 1/2 (HK1/2) and pyruvate kinase M1/M2 (PKM1/2), thus indicating increased glycolysis, while increased survivin suggested a stronger tendency to survive of AKT/YapS127A tumors (Fig. 2B). In summary, coexpression of activated AKT and Yap induces ICC development in mice.

mTORC1 signaling is required for AKT/YapS127A-induced cholangiocarcinogenesis
To investigate whether mTORC1 signaling is necessary for ICC development induced by AKT/YapS127A, we constructed AKT-shRaptor plasmid and control AKT-shLuc (shRNA against Renilla Luciferase) plasmid [24], which were hydrodynamically co-injected with YapS127A into mice (AKT-shRaptor/YapS127A or AKTshLuc/YapS127A) (Fig. 3A). Silencing of Raptor by shRNA substantially delayed the development of ICC by AKT/ YapS127A expression in mice. Indeed, while control AKTshLuc/YapS127A mice developed lethal tumor burden by 6.5 weeks post injection, only few ICC lesions occurred in AKT-shRaptor/YapS127A mice at this time point (Fig. 3B). Even as late as 13 weeks post injection, the tumor burden as indicated by liver weight in AKT-shRaptor/YapS127A mice was still significantly lagging behind, and the survival was greatly improved (Fig. 3C). This suggests that mTORC1 is required for AKT/YapS127A-induced cholangiocarcinogenesis. Consistently, clonogenic capacity of the HuCCT1 and KMCH human ICC cell lines was significantly suppressed after inhibiting Raptor via lentivirus transfection of shRNA ( Supplementary Fig. 2). Altogether, our data indicate that mTORC1 signaling plays a key role in cholangiocarcinogenesis.

Therapeutic efficacy of Gemcitabine/Oxaliplatin in the treatment of AKT/YapS127A ICC
Since Gemcitabine/Oxaliplatin regimen is the standard chemotherapy for advanced ICC [3], we first tested the therapeutic effect of Gemcitabine plus Oxaliplatin in the AKT/YapS127A ICC model. We started the therapy 3.5 weeks post injection when the tumor nodules start to emerge ( Supplementary Fig. 1). Within 5.2 to 6.5 weeks post injection, all vehicle-treated mice (n=5) developed lethal burden of tumor and had to be euthanized, while the Gemcitabine/Oxaliplatin-treated mice were in good conditions. The ICC burden was significantly decreased in Gemcitabine/Oxaliplatin-treated mice than that in vehicle-treated mice based on liver weight and mouse survival (Supplementary Fig. 4B and C). Histological examination confirmed the results, indicating that Gemcitabine/Oxaliplatin is effective against early stage AKT/YapS127A ICC. Since most ICC patients are diagnosed at late stage [2], we next determined the efficacy of this treatment for advanced ICC in AKT/YapS127A mice. Administration of Gemcitabine plus Oxaliplatin was started 5.5 weeks post injection, when mice have moderate tumor burden, for a total of 3 weeks (Supplementary Fig. 5A). ICC continued to grow, although at a slower pace (Supplementary Fig. 5B and C), and mouse survival was prolonged by ∼2 weeks upon chemotherapy ( Supplementary Fig. 5C). Thus, Gemcitabine/Oxaliplatin has relatively mild efficacy in treating late stage AKT/ YapS127A ICC.

MLN0128 inhibits ICC cell growth in vitro
Since both mTORC1 and mTORC2 cascades are crucial for AKT/YapS127A induced cholangiocarcinogenesis ( Fig. 3 and 4), we investigated the therapeutic effect of the second generation mTOR inhibitor MLN0128 in antagonizing ICC development. We screened 8 human CCA cell lines and selected out 6 cell lines with high expression of both mTORC1 (p-RPS6 and p-4EBP1) and mTORC2 (p-AKT) cascades (Fig. 5A). Next, we determined the in vitro effect of MLN0128 and treated CCA cells with MLN0128 for 48 hours. MLN0128 inhibited the growth of all 6 cell lines at nanomolar concentrations (Fig. 5B). At the biochemical level, as fast as 2 hours after MLN0128 treatment, AKT/mTOR signaling was suppressed, as indicated by decreased expression of p-AKT, p-mTOR, p-RPS6 and p-4E-BP1 in ICC cells (Fig. 5C). While MLN0128 could suppress the expression of Cyclin D1/E in the KMCH cell line, it had no effect on the expression of other proliferation-associated proteins, including PCNA and Cyclin A/B1/D1/E, in HuCCT1 cells. However, the expression of cleaved caspase-3/7 and PARP, well known markers of apoptosis induction, were elevated in both cell lines starting 24 hours after MLN0128 treatment (Fig. 5C). Fig. 6). Altogether, our data suggest that MLN0128 effectively inhibits ICC cell growth, mainly due to induction of apoptosis.

MLN0128 leads to stable disease in early stage AKT/Yap127A ICC mice
Next, the therapeutic efficacy of MLN0128 in vivo was investigated. We treated AKT/ YapS127A mice with MLN0128 at 1mg/kg/day according to previous studies [12,14,17,18,25]. Our experiment confirmed that mTOR signaling was effectively suppressed after 5 days of MLN0128 administration, with no obvious toxicity ( Supplementary Fig. 7). We first started the MLN0128 therapy in early stage of ICC development, namely at 3.5 weeks post injection. Vehicle or MLN0128 was orally administered by gavage to the mice (Fig. 6A). Lethal tumor burden occurred 6.5 weeks post injection in vehicle-treated mice, when MLN0128-treated mice appeared instead healthy with normal abdominal size (Fig. 6B). ICC burden in MLN0128-treated mice was significantly smaller than that in vehicle-treated mice as shown by gross and histological evidence, and by liver weight (Fig. 6C and D). MLN0128 administration significantly improved AKT/YapS127A mouse survival (Fig. 6E). CK19 staining also provided consistent evidence (Fig. 6F). Importantly, MLN0128 showed significantly superior efficacy to that of Gemcitabine/Oxaliplatin in early stage AKT/YapS127A ICC when comparing liver weight data, as the measurement of tumor burden ( Supplementary Fig. 4D). Histological analysis revealed that AKT/YapS127A mice after MLN0128 treatment for 3 weeks had a similar tumor burden to AKT/YapS127A mice at 3.5 weeks post injection, the time point when MLN0128 therapy started. In addition, there was no significant difference in liver weight between the two groups (P = 0.518; Supplementary Fig. 8), indicating that MLN0128 treatment leads to stable disease when applied to early stage tumors. Unexpectedly, tumor cells still were highly proliferative upon MLN0128 treatment, as shown by Ki-67 staining, with only mild decrease of proliferation index from vehicle treated tumors (Fig. 6F). In striking contrast, MLN0128 administration induced massive apoptosis in AKT/YapS127A ICC cells, as detected by TUNEL assay. Specifically, apoptosis index was ∼5 times higher in MLN0128 treated tumor samples when compared with the vehicle treated group (Fig. 6F).
To further investigate the molecular mechanisms underlying therapeutic efficacy of MLN0128, we evaluated the protein expression of major signaling pathways in vehicle or MLN0128 treated tumor samples. We first used tumor samples treated with vehicle or MLN0128 harvested after 3 weeks of treatment. However, because of the significant differences of tumor burden (Fig. 6), we found difficult to determine whether the changes were due to tumor burden difference or changes in signaling pathways (Supplementary Fig.  9). We therefore chose to use ICC tumor samples from mice at 3.5 weeks post injection and compare them with tumor harvested after 3 weeks of MLN0128 therapy, as we have shown that tumor burdens are similar in the pre-treatment and MLN0128 treated cohorts ( Supplementary Fig. 8). Indeed, HA tag represented the expression of transfected AKT, and there was no difference in HA expression, indicating equal tumor burden from the two groups of ICC tissues.
We found that MLN0128 could effectively inhibit AKT/mTOR signaling, while had no influence on Yap and Jag1 expression. There was no difference in p-ERK expression between the two groups, suggesting that no feedback activation of the Ras/MAPK pathway occurs after AKT signaling inhibition by MLN0128 (Fig. 7A and Supplementary Fig. 10A). Consistent with in vitro findings, MLN0128 treatment had no effects on cell proliferation as indicated by equal expression level of PCNA and Cyclin A/B1/D1/E in tissues from the two groups ( Fig. 7B and Supplementary Fig. 10B). The expressions of cleaved caspase-3/7 increased in ICC after MLN0128 treatment, indicating induction of apoptosis by MLN0128 ( Fig. 7C and Supplementary Fig. 10C). Next, we investigated the mechanisms underlying the increased apoptosis. The expression of major survival and apoptosis related genes, including Survivin, MCL-1, BCL-XL, BIM and BCL-2 did not show significant difference in control and MLN0128 treated tumor samples ( Fig. 7C and Supplementary Fig. 10C). BAD proapoptotic protein was mildly upregulated in MLN0128 samples ( Fig. 7C and Supplementary Fig. 10C). Importantly, ER stress markers, including Bip and CHOP, were highly upregulated in MLN0128 treated ICC tumor samples ( Fig. 7C and Supplementary  Fig. 10C). Overall, the data suggest that MLN0128 achieves disease control in early stage AKT/YapS127A ICC by triggering apoptosis, most likely via inducing ER stress.

MLN0128 leads to tumor regression in late stage AKT/YapS127A ICC
Next, we evaluated the efficacy of MLN0128 in the treatment of advanced AKT/YapS127A ICC. MLN0128 administration was started 5.5 weeks post injection. Mice were euthanized 3 days, 1, 2 or 3 weeks after MLN0128 administration (Fig. 8A). During the first 3 days, a striking shrinkage of the mouse abdomen was observed ( Supplementary Fig. 11A). The liver burden and liver weight gradually reduced after 1, 2 or 3 weeks of MLN0128 therapy ( Fig.  8B and Supplementary Fig. 11B). MLN0128 treatment greatly improved survival, with mice at 8.5 weeks post injection being in good conditions (Fig. 8B), although low to moderate tumor burden could still be observed ( Fig. 8B and Supplementary Fig. 11B). Consistent with the observation in early stage ICC treatment, MLN0128 had limited influence on cancer cell proliferation as indicated by Ki-67 staining, but promoted massive apoptosis, as shown by TUNEL assay (Fig. 8C and D). Overall, the data demonstrate that MLN0128 induces cancer cell apoptosis and leads to partial tumor regression in late stage AKT/YapS127A ICC mice.

Discussion
ICC is a deadly malignancy with limited treatment options. Murine models are critical to study the molecular mechanisms of tumorigenesis and to evaluate drug therapeutic potential. However, ICC mouse models are still limited and difficult to generate, which limits the use of these preclinical models. Here, we show that hydrodynamic transfection of activated forms of AKT (myr-AKT) and Yap (YapS127A) into the mouse liver rapidly induces ICC formation. Yamada et al. established an ICC murine model by intrabiliary transduction of AKT and Yap that was coupled with bile duct ligation followed by IL-33 administration (AKT/Yap/IL-33) [26]. Histological analyses indicated that the two ICC models are highly similar morphologically (Supplementary Fig. 12A). We performed microarray analysis of AKT/YapS127A ICC and normal liver, and identified 1320 and 799 mRNAs that were upregulated and downregulated, respectively, in AKT/YapS127A ICC. We compared the differentially regulated genes in AKT/YapS127A ICC with those identified in AKT/Yap/ IL-33 ICC, and found that up-and down-regulated genes overlapped significantly in the two datasets (p < 1E-200 for both up-and down-regulated genes) (Supplementary Fig. 12B). Thus, the myr-AKT/YapS127A mouse model represents a reliable and reproducible system to study ICC pathogenesis in vivo.
It is important to note that we applied hydrodynamic transfection of AKT and YapS127A plasmids to induce ICC in mice. Hydrodynamic injection delivers gene(s) into hepatocytes around the pericentral region of the liver [20,27]. Lineage tracing experiment demonstrated that AKT/YapS127A ICCs derived from hepatocytes (Wang J, unpublished data). Recent studies suggest that hepatocytes can reprogram into BECs in various chronic liver diseases, such as chronic HCV infection [28]. In Western and East Asian countries, where biliary tract inflammation is rare, HBV and HCV infection and alcohol abuse are the major risk factors for ICC, as for HCC [29]. Thus, ICC derived from hepatocytes might be a highly relevant model to study human cholangiocarcinogenesis, especially in the context of chronic liver injury where hepatocytes are presumably the targets of transformation. The precise mechanisms whereby activated AKT/mTOR and Yap pathways synergize to promote ICC development remain to be determined. It has been reported that Yap activation induces ductal fate differentiation in mature hepatocytes [30]. We hypothesize that YapS127A induces the de-differentiation of hepatocytes towards biliary cell lineage, and AKT/mTOR signaling promotes malignant transformation to eventually form ICC. Clearly, additional experiments are required to elucidate the functional contribution of AKT/mTOR and Yap signaling in cholangiocarcinogenesis.
Our genetic studies demonstrated that both mTORC1 and mTORC2 cascades are required for AKT/YapS127A driven ICC development. As mTORC1 is a major cascade downstream of AKT, as expected, silencing of mTORC1 strongly inhibited AKT/YapS127A induced tumorigenesis. Interestingly, ablation of mTORC2 inhibited AKT/S127A ICC development, leading instead to HCC formation in the Rictor KO genetic background. Our preliminary data suggest that Notch signaling, a key determinant of ICC development, is suppressed by the loss of mTORC2 (Xu M, unpublished data). It would be important to further investigate the biochemical crosstalk between mTORC2 and Notch cascades and how they differentially regulate ICC and HCC development.
Using AKT/YapS127A mice as preclinical model, we tested the therapeutic efficacy of multiple drugs, including Gemcitabine/Oxaliplatin, MLN0128, and Everolimus. Everolimus, a Rapalogue, has been widely tested in clinical trials for cancer treatment, with unsatisfactory results. This is likely due to the fact that Everolimus is a partial mTORC1 inhibitor and has no activity against mTORC2. Indeed, when we tested Everolimus for the treatment of AKT/YapS127A ICC, we found that it has rather limited efficacy ( Supplementary Fig. 13). Currently, multiple efforts are focusing on the development of more efficient mTOR inhibitors, including pan-mTOR inhibitors or dual PI3K/mTOR inhibitors, for cancer therapeutics. Because our genetic studies suggest that both mTORC1 and mTORC2 are required for cholangiocarcinogenesis, we chose to test the therapeutic efficacy of the dual mTOR inhibitor MLN0128, which has not been tested in ICC. Here, we show that MLN0128 efficiently inhibits the growth of a panel of human ICC cells with high AKT/mTOR activity, mainly via inducing apoptosis. Furthermore, we found that MLN0128 leads to stable disease in AKT/YapS127A mice when administered in early stage of tumorigenesis and partial regression when given at late stage. Importantly, MLN0128 exhibited superior therapeutic efficacy than Gemcitabine/Oxaliplatin and Everolimus in the treatment of late stage AKT/YapS127A ICC (Supplementary Fig. 13B). Mechanistically, we demonstrated that the in vivo anti-tumor activity of MLN0128 is mainly due to apoptosis induction. Further evidence indicates that MLN0128 promotes ER stress in AKT/YapS127A ICC cells, resulting in the expression of the pro-apoptotic gene CHOP. The role of ER stress in the induction of apoptosis by mTOR inhibitors has been recently shown also in colon cancer cells, where DR5 was reported to be the potential downstream target of CHOP [31]. Differently from this study, however, no changes in DR5 mRNA expression were detected in MLN0128 treated AKT/YapS127A ICC (Supplementary Fig. 10D). ER stress is induced by several types of stressors, leading to unfolded protein response (UPR). ER stress/UPR has three different arms, one controlled by PERK, another by IRE1α, and a third by ATF6 [32]. It would be important to identify the arm of ER Stress/UPR modulated by MLN0128 in ICC tumor cells and mechanisms contributing to MLN0128 induced apoptosis.
Altogether, our in vitro and in vivo data support further investigation on MLN0128 and other pan-mTOR inhibitors for ICC treatment. Additional studies using patient derived xenograft (PDX) models will be of high importance. Furthermore, combined therapy of MLN0128 with other drugs should be evaluated to establish the usefulness of mTOR inhibitors for the treatment of human ICC.

Supplementary Material
Refer to Web version on PubMed Central for supplementary material.

•
A novel preclinical mouse model of intrahepatic cholangiocarcinoma (ICC) was established via hydrodynamic transfection of activated forms of AKT and Yap, two protooncogenes frequently activated in human ICC.
• This model recapitulates many morphological and molecular features of human ICC.

•
The anti-neoplastic potential of MLN0128 may be superior to Gemcitabine/ Oxaliplatin based chemotherapy for the treatment of ICC, especially in the ICCs exhibiting activated AKT/mTOR cascade. Strong induction of apoptosis was identified as a major mechanism triggering growth restraint of ICC cells.   were shown in the upper panel. CK19 IHC staining was quantified and demonstrated as the percentage of positive staining area of the whole area. Ki-67 and TUNEL positive cells were counted and quantified as proliferation and apoptosis index. Quantified data are presented as mean ± SE in lower panel. *P < 0.05; ***P < 0.001.