Hurthle cell carcinomas (HCCs) are refractory to radioactive iodine and unresponsive to chemotherapeutic agents, with a fatality rate that is the highest among all types of thyroid cancer after anaplastic thyroid cancer. Our previous study on the genomic landscape of HCCs identified a high incidence of disruptions of mTOR pathway effectors. Here, we report a detailed analysis of mTOR signaling in cell line and patient-derived xenograft mouse models of HCCs. We show that mTOR signaling is upregulated and that targeting mTOR signaling using mTOR inhibitors suppresses tumor growth in primary tumors and distant metastasis. Mechanistically, ablation of mTOR signaling impaired the expression of p-S6 and cyclin A2, resulting in the decrease of the S phase and blocking of cancer cell proliferation. Strikingly, mTOR inhibitor treatment significantly reduced lung metastatic lesions, with the decreased expression of Snail in xenograft tumors. Our data demonstrate that mTOR pathway blockade represents a novel treatment strategy for HCC.

Although accounting for 5% of all malignancies arising from thyroid follicular cells, Hurthle cell carcinomas (HCCs) were not included in The Cancer Genome Atlas thyroid cancer study and are overall poorly understood (1, 2). Characterized by an abundance of mitochondria, which form more than 75% of the cell volume (3), HCCs have a spectrum of clinical behavior: The minimally invasive phenotype has a more indolent course, whereas the widely invasive phenotype, characterized by vascular invasion, is more aggressive (4–7). Currently, there is no effective chemotherapy to treat the widely invasive subtype of HCC, which has the highest fatality of all thyroid cancers after anaplastic thyroid cancer. There is, therefore, an urgent need to better understand the biology of HCCs and to identify specific biomarkers and pathways that can be targeted for therapeutic treatment.

We and others recently published comprehensive genomic characterization studies of primary HCC tumors and identified novel genomic mutations, unique chromosomal alterations, mitochondrial mutations, recurrent fusion genes, and signaling pathway dysregulation (8–10). In particular, we found that many critical genomic mutations, chromosome alterations, and gene expression pattern alterations were related to the mTOR signaling pathway. Unfortunately, there have been few models to study HCCs, which has significantly delayed the exploration of novel therapeutic treatments for patients with HCC in the clinic. Here, we have successfully developed cell line and patient-derived xenograft (PDX) models to mimic HCCs and found that inhibition of the mTOR signaling pathway suppressed proliferation and tumor growth. Mechanistically, mTOR signaling regulates cell proliferation in HCCs through cell-cycle regulation, specifically through cyclin A. Strikingly, mTOR inhibitors suppressed cell invasion and metastasis through epithelial–mesenchymal transition (EMT)–associated regulation in mice. Our findings suggest that the mTOR signaling pathway may present efficient therapeutic targets for HCCs.

Cell culture and reagents

Nthy-ori 3.1 cells were cultured in RPMI-1640 supplemented with 10% FBS, 1X penicillin–streptomycin, and 1X l-glutamine. XTC.UC1 cells (11, 12) were cultured in DMEM/F12 medium supplemented with 10% FBS, 25 mmol/L HEPES, 0.055 g/L sodium pyruvate, 0.365 g/L l-glutamine, 10 μg/mL insulin, 0.36 ng/mL hydrocortisone, 10 IU thyroid-stimulating hormone, and 1X penicillin–streptomycin. Cells were maintained in a tissue culture incubator with 5% CO2 at 37°C. The following reagents were used: rapamycin (SelleckChem; cat. No. S1039), AZD8055 (SelleckChem; cat. No. S1555), AZD2014 (AstraZeneca), MK2206 (Merck), and PD325902 (Pfizer).

Cell assays

For the cell proliferation assay, 2 × 104 cells were seeded in 6-cm dishes in triplicates. Cells were trypsinized and counted on indicated time points by cell counter. Cell proliferation curves were generated with GraphPad Prism software v.5.02.

For the soft agar assay, 1 × 105 cells were seeded onto a 6-cm dish containing a top layer of 0.3% noble agar and a bottom layer of 0.6% noble agar base in triplicates. Cells were fed with media every 3 days; 11 days after seeding, colonies with a diameter larger than 100 μmol/L were scored. Three independent experiments were performed. Scale bar is 100 μm. Data presented are mean ± SD.

For the invasion assay, 2 × 104 cells were seeded in each transwell of 24-well (8 μm) Corning invasion chamber in culture medium with 0.1% FBS. Full culture medium with 10% FBS was added in each well of the plate as the attractant. After 24-hour incubation at 37°C, cells in each transwell were removed, and the membrane was fixed in 10% formalin and stained with DAPI solution. The invaded cell number was counted in nine random views from triplicates under fluorescence microscope. Data presented are mean ± SD.

For the migration assay, 2 × 104 cells were seeded in each transwell of 24-well (8 μm) chamber in culture medium with 0.1% FBS. Full culture medium with 10% FBS was added in each well of the plate as the attractant. After 24-hour incubation at 37°C, cells in each transwell were removed, and the membrane was fixed in 10% formalin and stained with DAPI solution. The invaded cell number was counted in nine random views from triplicates under fluorescence microscope. Data presented are mean ± SD.

For the cell-cycle assay, cells were seeded in 6-well plates and incubated at 37°C overnight. Then fresh medium was changed with DMSO control and drugs as indicated in the figures. Cells were trypsinized at different time points, washed with PBS, treated with 20 μg/mL RNase A, and stained with 25 μg/mL propidium iodide (PI) for 20 minutes before being subjected to cell-cycle analyses. Flow cytometric analyses were performed using a Fortessa flow cytometer (Becton Dickinson) to measure DNA contents. Fifty-thousand events were collected per each experiment, and the data were analyzed with FCS Express 6 (De Novo Software). Data presented are mean ± SD from triplicates.

For the cell death assay, cells were treated the same way as cell-cycle assay. Cell death was quantified by Annexin V (BD Pharmingen) staining, followed by flow cytometric analysis using a Fortessa flow cytometer (Becton Dickinson). The data were analyzed with FCS Express 6 (De Novo Software). Data presented are mean ± SD from triplicates.

Immunoblotting

For Western blot assay, cells and tissues were lysed in standard RIPA buffer. Antibodies for p-AKT, p-S6K, p-S6, P-4EBP1, 4EBP1, cyclin A, Snail, and CD31 were purchased from Cell Signaling Technology. Antibody for actin was from Sigma. Antibodies were detected using the enhanced chemiluminescence method (Western Lightning, PerkinElmer). Immunoblot signals were acquired with the LAS-3000 Imaging system (FujiFilm) and were analyzed with ImageJ software.

DNA and RNA extraction

DNA and RNA extraction and quantification were as described previously (8).

RNA sequencing libraries and read processing

Preparation of RNA sequencing (RNA-seq) libraries and read processing was as described previously (8).

Mutation analysis

Single-nucleotide variants (SNVs) and insertions or deletions were identified using a customized pipeline as described previously (8).

Somatic mitochondrial mutations

Aligned reads were analyzed using a custom informatics pipeline for mtDNA analysis as described previously (8).

Copy-number analysis by FACETS and FISH

To characterize allele-specific somatic DNA copy-number alterations, we applied FACETS to tumor and normal tissue pairs of bam files (13). Copy-number alterations, tumor purity, ploidy, and cellular fractions were estimated and reported as described previously (8). FISH analysis was performed on formalin-fixed, paraffin-embedded sections using a three-color probe designed to confirm the copy-number changes of chromosomes 2, 5, and 7 detected by FACETS (8).

Single sample gene-set enrichment analysis

Single sample gene-set enrichment analysis (ssGSEA) computes an overexpression measure based on the genes of given gene signature (e.g., MTORC1_SIGNALING) relative to all other genes in the transcriptome. The sample gene signature ssGSEA scores were quantified by the ssGSEA implementation in the R package gsva (14) with the fragments per kilobase of transcript per million mapped reads gene expression values from each sample as the input. The MTORC1_SIGNALING signature was among the curated MSigDB Hallmark gene sets (15) and downloaded from the Molecular Signatures Database (http://software.broadinstitute.org/gsea/msigdb).

Quantitative real-time PCR

Tumor tissue was homogenized with a FastPrep bead beating grinder (MP Biomedicals). Total RNA was extracted with TRIzol reagent (Thermo Fisher Scientific) and purified by QIAGEN RNeasy mini columns (QIAGEN). cDNA was synthesized from purified RNA using reverse transcriptase. Each target gene was amplified in triplicates and detected by SYBR Green (Applied Biosystems) on a StepOnePlus real-time PCR system (Applied Biosystems). Primers are listed in Supplementary Table S1. The 2ΔΔCT method was used to calculate ΔΔCT values. mRNA expression levels of target genes were normalized against GAPDH.

Animal study

All animal work was performed in accordance with institutional guidelines and Institutional Animal Care and Use Committee approval. Tumor generation, measurement, and harvesting were performed as previously described (16). Briefly, we used 6-week-old male immunocompromised NOD-SCID IL2Rg_/_ (NSG) mice (The Jackson Laboratory) for the study. 4 × 106 cells in 100-μL PBS/Matrigel (1:1) were injected subcutaneously per tumor. When each individual tumor reached 150 mm3, mice were randomly divided into three groups and treated with vehicle and rapamycin at indicated dose every other day. Tumor volume was measured by caliper twice per week and calculated with formula 0.5 × L × W2. Tumor growth curve was generated by GraphPad Prism software. Mice were sacrificed at endpoint, and tumors were collected for assays. PDX mouse model of HCC was established by implantation of fresh tumor under the renal capsule in NSG mice. Tumor was taken from a patient with widely invasive Hurthle cell cancer and implanted within 2 hours into each mouse. Once established, PDX mice were then expanded in NSG mice for preclinical treatment.

For in vivo treatment in PDX mice, tumor-bearing mice were randomly divided into two groups and administered with vehicle or 7.5 mg/kg rapamycin every other day. Mice treated with lenvatinib were orally administered 30 mg/kg once daily. Tumor volume was monitored with MRI scanning at indicated time points. Mice were euthanized and necropsied at the time of onset of clinical symptoms or at a predetermined endpoint. The tumor tissue was removed and prepared for immunohistochemistry staining and biochemical assay.

Histologic examination

Tissue was processed, and hematoxylin and eosin (H&E) staining was performed according to standard protocols. For immunohistochemistry staining, p-AKT, p-S6K, p-S6, P-4EBP1, 4EBP1, cyclin A, CD31, and TTF1 were used, and slides were read by board-certified pathologists (R. Ghossein and B. Xu). For mitotic index, mitotic cells were counted in 10 random high-power fields from each tumor; eight to 10 tumors were counted in each group. For lung metastasis lesion count, metastasis lesion number was counted from each H&E slide; four to five mice were counted from each group. The slides were scanned, and lung tissue area was measured and calculated by CaseViewer (3DHISTECH) and ImageJ (National Institutes of Health and the Laboratory for Optical and Computational Instrumentation). Then, the metastasis lesion numbers were normalized by the lung tissue area. Data presented are mean ± SD.

MRI scanning

Mice kidney MRI was carried out on 400 MHz Bruker 9.4T Biospec scanners (Bruker Biospin MRI GmbH) equipped with a 530 mT/m ID 114-mm gradient. RF excitation and acquisition were achieved by a Bruker quadrature birdcage resonator with ID of 40 mm. The mice were immobilized with 2% isoflurane (Baxter Healthcare Corp.) gas in oxygen. Animal respiration was monitored with a small animal physiological monitoring system (SA Instruments, Inc.). Scout images along three orthogonal orientations were first acquired for animal positioning. For mouse kidney imaging, coronal T2-weighted images using respiratory gated fast spin-echo RARE sequence (Rapid Acquisition with Relaxation Enhancement) were acquired with TR 2s, TE 33 ms, RARE factor of 8, slice thickness of 0.4 mm, FOV 30 mm, in-plane resolution of 117 × 117 μm, and 10 averages.

Statistical analysis

D'Agostino–Pearson omnibus normality test was performed to determine if datasets follow a Gaussian distribution in each comparison. If the data were Gaussian, parametric tests were performed (two-tailed unpaired t tests, one-way ANOVA with Tukey correction for multiple comparisons, or Pearson correlation). If the data were non-Gaussian, nonparametric tests were applied (Mann–Whitney U test, Kruskal–Wallis one-way ANOVA with Dunn correction for multiple comparisons, or Spearman correlation). The a priori definition of statistical significance for all hypothesis testing was two-tailed (α < 0.05). Outcomes analysis was done using the Kaplan–Meier method and compared using the log-rank test.

Informed consent

Written informed consent from patients for use of their tumor material was obtained, and the studies were approved by the Institutional Review Board at Memorial Sloan Kettering Cancer Center. Studies were conducted in accordance to recognized ethical guidelines in the Belmont Report.

Data availability

The data generated in this study are available upon request from the corresponding author. Whole-exome sequencing data and RNA-seq data have been deposited in SRA and are available under accession number SRP136351.

mTOR signaling pathway is associated with poor clinical outcomes in HCCs

Using our institutional thyroid cancer database of 6,262 patients treated from 1985 to 2015, we show that the aggressive widely invasive HCC subtype, herein referred to as HWIDE, has poorer survival compared with minimally invasive HCC, referred to as HMIN, and papillary thyroid cancer (Fig. 1A). Our previous work suggested that mTOR signaling may be involved in HCC tumorigenesis, especially in the aggressive HWIDE subtype (8). To further address this, we analyzed the gene expression data of our HCC tumors using ssGSEA. ssGSEA computes an overexpression measure based on the genes of a given gene signature. The MTORC1_SIGNALING signature was among the curated MSigDB Hallmark gene sets and downloaded from the Molecular Signatures Database (http://software.broadinstitute.org/gsea/msigdb). We show that the expression of the mTORC1 signature was significantly elevated in the tumors of patients with the HWIDE subtype who had structural recurrence (Fig. 1B and C). These data provide the rationale for investigating the mTOR pathway in HCCs and the possible use of mTOR inhibitor drugs to treat this cancer.

Model systems to study mTOR pathway in HCCs

XTC.UC1 caused tumorigenesis and metastasis in vivo

One of the major limitations of studying the pathogenesis of HCC is a lack of available model systems. In fact, XTC.UC1 is the only HCC cancer cell line ever established, and there is no genetically engineered mouse model of the disease. The XTC.UC1 cell line was established from a metastatic breast lesion of a patient with HCC (11, 12). We performed RNA-seq analysis on XTC.UC1 and compared this with our data of patients with HCC. XTC.UC1 was more similar to recurrent HCC tumors with high expression of the mTORC1 signature (Fig. 1B). To determine whether XTC.UC1 shares the major genomic features of HCC tumors from patients that we recently reported (8), we performed genomic and biochemical analyses of XTC.UC1. FISH data showed that there is whole chromosomal duplication of chromosomes 5 and 7 with a pattern of 2–3–3 (two copies of chromosome 2, three copies of chromosome 5, and three copies of chromosome 7; Fig. 1D). Copy-number analysis using FACETS (13) also showed uniparental disomy of several other chromosomes. These chromosomal alterations are very similar to what is seen in HWIDE (8). We also found that 56.5% of the significantly mutated genes we reported in HCC tumors (8) were also identified by whole-exome sequencing in XTC.UC1, and three in-frame coding rearrangements—CHCHD10_VPREB3 (chromosome 22), HEPHL1_PANX1 (chromosome 11), and BCAP29_SLC26A4 (chromosome 7)—were also found in XTC.UC1 (Fig. 1E; Supplementary Fig. S1). Mitochondrial mutations, similar to those found in our patient data (8), were also identified in XTC.UC1 cells (ref. 17; Fig. 1F). Lastly, we examined the signaling activity in the PI3K–AKT–mTOR pathway by Western blot in XTC.UC1. Consistent with previous results (18), XTC.UC1 has a loss of PTEN expression. In addition, expressions of p-AKTT308, p-AKTS473, p-S6K, RICTOR (a component of mTORC2), and RAPTOR (a component of mTORC1) were increased in XTC.UC1 compared with the immortalized thyroid cell line NTHY-ori (Fig. 1G). Upregulation of the mTOR pathway in XTC.UC1 is therefore by a multipronged mechanism involving loss of PTEN expression, whole chromosome duplication of chromosome 7 as well as multichromosome uniparental disomy causing massive loss of heterozygosity.

To investigate the tumorigenesis of XTC.UC1, we show XTC.UC1 cells formed significantly more colonies in anchorage-independent conditions compared with NTHY-ori (Fig. 2A). In addition, XTC.UC1 shows higher migration and invasiveness in vitro (Fig. 2B and C). Next, we evaluated whether XTC.UC1 could cause tumor growth in vivo. XTC.UC1 cells were injected into NSG mice subcutaneously and monitored for 140 days. XTC.UC1 caused tumorigenesis in vivo, and the xenografts had histologic features of HWIDE, such as capsular and vascular invasion (Fig. 2D and E). Moreover, we found that XTC.UC1-injected mice developed metastases in the lungs and kidneys (Fig. 2D and E). Immunohistochemistry staining showed that XTC.UC1 cells express the thyroid differentiation proteins PAX8, NKX2-1, and thyroglobulin (Fig. 2F). Lastly, both the xenograft tumors and metastases showed whole chromosomal duplication in chromosomes 5 and 7 (pattern 2–3–3; Fig. 2G). These results demonstrate that XTC.UC1 maintains the major genomic features of HCC tumors, has upregulation of the mTOR pathway, shows tumorigenesis in vivo, and displays metastatic potential. This cell line is therefore a representative model of HCC.

Development of PDX model to study HCC

Due to the lack of HCC in vivo models, we also developed a PDX model by implanting fresh tumor from a patient with the HWIDE subtype under the renal capsule in NSG mice (Fig. 3A and B). One of the critical challenges of the HCC study is the lack of proper mouse model. In our study, we have tried different means to generate mouse models using the HCC cell line and tumor tissue. The PDX model generated by subrenal capsule implantation was the only technique which worked. Due to the rich blood supply, mouse kidneys provide a good microenvironment for tumor tissue growth. Furthermore, relative larger tumor tissue can be implanted in mouse kidney to facility PDX tumor growth. To the best of our knowledge, this is the first report of a PDX model for HCC. This PDX model (named TCm1) could serve as a useful tool for studies on HCC. Pathologic and genomic analyses showed that the PDX model TCm1 maintains the major features of the primary tumor from the patient (named TC1). First, a histologic analysis showed that the TCm1 tumor is HWIDE, similar to the primary tumor TC1 (Fig. 3C). Second, next-generation sequencing showed the PDX model TCm1 had a mutation profile comparable with the primary tumor TC1, with 33 of 38 SNV mutations and six of 11 INDELs (Fig. 3D), and these included mutations in the TP53 gene, which occurs in 7% of HCCs (8). Third, copy-number analysis showed TCm1 and TC1 had similar major chromosomal alterations, including whole chromosome amplification of chromosomes 5, 6, 7, 12, and 17 and uniparental disomy of other chromosomes (Fig. 3E). Although this PDX tumor did not have any mutations in the PIK–AKT–mTOR pathway, our previous report showed that amplifications seen on chromosomes 5, 7, 12, and 17 were associated with the activation of the PI3K–AKT–mTOR pathway by upregulation of RICTOR (Chr 5), BRAF (Chr 7), RHEB (chr 7), AGAP2 (Chr 12), and RAPTOR (Chr 17; ref. 8). Lastly, the tumor harbored two missense mitochondrial DNA mutations (one in the ND4 gene of complex I and another in the cytochrome B of complex III). The PDX model TCm1 was therefore representative of the HWIDE phenotype.

mTOR inhibitors disrupt expression of cyclin A and proliferation of XTC.UC1 cells

To determine whether XTC.UC1 cells are sensitive to AKT signaling inhibition, we first treated NTHY-ori and XTC.UC1 cells with MK2206, a pan-AKT inhibitor. MK2206 showed minor inhibition of both NTHY-ori and XTC.UC1 cell proliferation (Supplementary Fig. S2A and S2B). To assess whether mTOR signaling is required for XTC.UC1 cell proliferation, we then treated XTC.UC1 cells with three mTOR inhibitors: rapamycin (mTORC1-specific inhibitor) and AZD8055 and AZD2014 (dual inhibitors of mTORC1 and mTORC2; ref. 19). Our rationale for using AZD8055 and AZD2014 was based on the observation that components of both mTORC1 and mTORC2 are overexpressed in HCC. Our data showed that all three mTOR inhibitors significantly reduced the cell number according to cell proliferation assay (Fig. 4A and B; Supplementary Fig. S3). As reported, we found that inhibition of rapamycin was not dose dependent, whereas both AZD8055 and AZD2014 showed dose-dependent inhibition. AZD8055 and AZD2014 produced greater growth inhibition than rapamycin, suggesting that combined inhibition of mTORC1 and mTORC2 may be more efficient than inhibition of mTORC1 alone.

The ability of cancer cells to form colonies in soft agar is an in vitro surrogate of tumorigenicity. Soft agar assays assess the capacity of tumor cells to not only proliferate but also resist anoikis under three-dimensional culture conditions. As shown in Fig. 2A, XTC.UC1 cells formed colonies in soft agar. To assess whether mTOR inhibitors could suppress colony formation, we treated XTC.UC1 cells seeded in soft agar with DMSO and the three mTOR inhibitors. As expected, all three inhibitors significantly suppressed colony formation (Fig. 4C). Importantly, mTOR inhibitors also significantly suppressed the capacity of invasion of XTC.UC1 cells (Fig. 4D). Apoptosis assays confirmed that there was no significant difference in programmed cell death between the DMSO-treated and mTOR inhibitor–treated XTC.UC1 cells (Fig. 4E). Cell-cycle analysis showed that inhibition of mTOR signaling significantly decreased the S phase population and increased the G1 phase population (Fig. 4F). These data suggest that mTOR signaling inhibition blocks cell-cycle progression.

To examine the inhibition of mTOR signaling at a molecular level, we treated XTC.UC1 cells with mTOR inhibitors at different doses for 2 hours and performed Western blot assays for several effectors in the signaling pathway. The cellular EC50s for rapamycin were calculated as 0.1248 and 0.6844 nmol/L for p-S6K T389 and p-S6 S235/236, respectively. AZD8055 and AZD2014 showed higher EC50 than rapamycin (11.2 and 29.43 nmol/L for AZD8055; 144.2 and 397.6 nmol/L for AZD2014; Fig. 4G and H). In all tested doses, rapamycin failed to inhibit the phosphorylation of 4E-BP1 on S65 and AKT on S473 (Fig. 4H and I). This is likely due to the different mechanisms of inhibition of these mTOR inhibitors.

We next sought to understand the mechanism by which mTOR signaling regulates cell cycle and invasion in HCCs. To address this, we treated XTC.UC1 cells with mTOR inhibitors, collected cells at different time points, and checked the marker gene expression for mTOR signaling, cell cycle, and EMT by Western blot. Indeed, phosphorylation of S6K on T389 and S6 on S235/236 was impaired after treatment (Fig. 4I). Importantly, treatment of XTC.UC1 cells with rapamycin, AZD8055, and AZD2014 resulted in reduced cyclin A2 expression at all time points (Fig. 4I). In Fig. 4H, we observe some diminished 4EBP1 phosphorylation, suggesting some of the decrease in cyclin A is from a global inhibition of protein synthesis. However, it is possible cyclin A2 inhibition may be a specific mechanism because we do not observe any decrease in other cyclins. Cyclin A, which is required for both the S phase and the M phase in the cell cycle, is central in cell-cycle regulation (20, 21). Cyclin A loss can prevent DNA replication, resulting in inhibition of the cell cycle and cellular proliferation. Interestingly, we found that treatment with mTOR inhibitor also decreased Snail, which plays important roles in EMT and cancer cell invasion (Fig. 4I). Consistent with our invasion assay results, this suggests that mTOR inhibitors may potentially inhibit processes important for HCC cancer metastasis.

Blockade of mTORC1 signaling suppresses tumorigenesis and metastasis in vivo in XTC.UC1

To characterize the effects of mTOR inhibition on HCCs in vivo, we established an XTC.UC1 xenograft mouse model and treated mice with the mTOR inhibitor rapamycin, which is approved for clinical use in patients. Due to severe toxicity reported in preclinical studies and phase I clinical trials, neither AZD8055 nor AZD2014 is approved for clinical use; therefore, for in vivo studies, only rapamycin was used. Strikingly, rapamycin significantly blocked tumor growth in vivo (Fig. 5A and B). Histologic analysis showed that mitotic cells were significantly decreased in rapamycin-treated tumors compared with vehicle-treated tumors (Fig. 5C). This clearly indicates the tumor cell proliferation in vivo was suppressed by rapamycin treatment. Immunohistochemistry staining showed that phosphorylation of S6 level was reduced dramatically (Fig. 5D). Altogether, these data demonstrate that mTOR signaling promotes tumor growth in HCCs, and inhibition of the mTOR signaling pathway by mTOR inhibitors results in decreased tumor growth in HCCs.

Approximately 30% to 40% of patients who have HCC with the HWIDE subtype show lung metastasis and ultimately die from it (4–7). The mechanism driving lung metastasis in HCC is unknown. Our xenograft model showed that XTC.UC1 cells can metastasize to the lungs in mice, providing us a powerful tool to address this. Our investigation into the impact of rapamycin on tumor metastasis development found that lung metastases were significantly decreased in rapamycin-treated mice (Fig. 5E). Furthermore, FISH staining showed that the copy number for chromosomes 2, 5, and 7 in lung metastases was 2–2–2, in contrast to the primary tumors, where the copy-number profile is 2-3–3 (Fig. 5F). This finding suggests that the high-polyploid cells are more susceptible to rapamycin treatment in lung metastases.

To validate the previously observed mechanisms in our xenograft model, we collected tumors from vehicle- and rapamycin-treated mice and performed Western blot analysis using tumor lysates. Three tumors were chosen randomly from each group. As expected, mTOR signaling was suppressed as indicated by p-S6 (S235/236) expression level (Fig. 5G). Similarly, cyclin A2 expression was impaired in all tumors from two rapamycin groups treated at different doses. This decrease in cyclin A2 could be due to a global inhibition of protein synthesis by mTOR inhibition as indicated by our in vitro studies where we see inhibition of 4EBP1 phosphorylation (see Fig. 5H). However, cyclin D1 and cyclin E1 expressions were not affected (Fig. 5G). This therefore suggests that the inhibition of cyclin A2 by mTOR inhibitors may be a specific mechanism by which these drugs inhibit cell proliferation and growth. Furthermore, Snail expression was suppressed in rapamycin-treated tumors, which is consistent with the finding that lung metastases were significantly decreased in rapamycin-treated mice (Fig. 5D and G). Finally, both in vitro and in vivo results showed that mTOR inhibitor treatment did not induce significant apoptosis (Figs. 4E and I, and 5G).

Rapamycin suppresses HCC tumor growth in PDX mouse models

To further validate our findings, we next conducted in vivo experiments in our unique PDX model, TCm1. After tumor growth was confirmed, mice were treated with vehicle and rapamycin every second day over a 5-month period. Tumor growth was monitored by serial MRI scans. Strikingly, rapamycin treatment significantly suppressed tumor growth in the TCm1 PDX model, which is consistent with the findings in XTC.UC1 xenograft models (Fig. 6AD). PDX tumors from each group were collected for further analysis. Histologic analysis confirmed the HWIDE phenotype in both vehicle- and rapamycin-treated tumors. Importantly, immunohistochemistry staining showed that the expression of p-S6 and cyclin A2 was reduced dramatically, which is consistent with our previously observed mechanisms (Fig. 6E). Furthermore, Snail expression was also decreased in rapamycin-treated tumors (Fig. 6E). Western blot and quantitative real-time PCR on randomly selected tumors confirmed that expression of p-S6, cyclin A2, and Snail was significantly decreased (Fig. 6F and G). The results of the PDX experiment therefore validated the findings of XTC.UC1. Overall, these preclinical results demonstrate that mTOR signaling is required to drive HCC carcinogenesis, which involves cell-cycle regulation through cyclin A2. The decrease of Snail expression suggests that mTOR plays a role in HCC metastasis through EMT; this needs further investigation.

Combination treatment of mTOR inhibitors with AKT/MEK inhibitors and VEGF inhibitors

We next investigated the effect of combining mTOR inhibitors with AKT/MEK inhibitors and VEGFR inhibitors which act on the PIK/AKT/mTOR pathway. Inhibition of mTORC1 can result in a feedback stimulation of AKT and MEK, both of which could potentially dampen the inhibition of the downstream effector proteins pS6. Combination of mTORC1 inhibitors with an AKT inhibitor or MEK inhibitor could potentially result in increased efficacy. To study this, we conducted experiments with rapamycin in combination with the AKT inhibitor MK2206 and MEK inhibitor PD325901 (Supplementary Fig. S4A). The addition of MK2206 did not dramatically further reduce cell proliferation (Supplementary Fig. S4B), despite a decrease in p-AKT (Supplementary Fig. S4C). A similar phenotype was observed in combination treatment of rapamycin with PD325901.

Subsequently, we examined the effect of combining mTOR inhibitors with VEGFR inhibitors. Combination of the mTORC1 inhibitor everolimus with receptor tyrosine kinase inhibitors that block VEGFR, such as sorafenib and lenvatinib, has been reported to show increased efficacy compared with everolimus alone in renal cell cancer (22). There is good rationale for combining an mTORC1 inhibitor with VEGFR inhibitors in HCC because of the hallmark finding of increased vascular invasion associated with the more aggressive types of HCC. To explore this, we conducted a combination experiment of rapamycin with lenvatinib in our PDX model TC1m1 to assess the effect of these drugs alone and in combination on the tumor vasculature by staining for the angiogenic marker CD31. It has been reported that the count of intratumoral microvascular density (IMVD) of CD31 is higher in tumors from patients with a worse prognosis than in those with a good prognosis (23). To study this, NSG mice harboring TC1m1 PDX tumor were divided into four groups (n = 5 per group) and treated with vehicle (control), rapamycin (intraperitoneal, 7.5 mg/kg once every 2 days), lenvatinib (orally, 30 mg/kg once daily), and combination of rapamycin and lenvatinib, respectively, over 4 weeks (Supplementary Fig. S5). Mice were sacrificed at the 4-week time point, and tumors were collected and subjected to immunohistochemical evaluation for CD31. In our study, the results showed that the average IMVD values of CD31 in tumors from mice treated with different drugs were significantly decreased to 44.5% (rapamycin), 60.6% (lenvatinib), and 16.9% (combination of rapamycin and lenvatinib), respectively, compared with the vehicle control group. Furthermore, it is noteworthy that the size of CD31-stained vascular vessels in lenvatinib-treated tumors became much thinner and smaller, whereas no phenotypic change presented in sole rapamycin-treated counterpart. Tumor growth and survival over this 4-week period were assessed with rapamycin alone, lenvatinib alone, and rapamycin plus lenvatinib. Lenvatinib alone was able to suppress growth, though not as effective as rapamycin. Although the lenvatinib combination showed dramatic synergistic effect of rapamycin and lenvatinib on vasculature inhibition, the combination treatment showed severe toxicity to the mice. We were therefore unable to study growth or survival with the combination treatment in mice. However, our data provide the rationale for a clinical trial combining everolimus with VEGF inhibitors such as lenvatinib or sorafenib. We are currently running a phase II clinical trial of everolimus with sorafenib in patients with HCC and will report the results separately.

After anaplastic thyroid cancer, the HWIDE subtype of HCC has the highest fatality of all thyroid cancers. There is therefore a need to understand this cancer and develop new treatments. Mechanistically, HCCs are not well studied compared with other thyroid subtypes. Our previous genomic studies suggested the mTOR signaling pathway was upregulated in the more aggressive forms of HCC. This was due to mutations of genes critical to the PIK3/AKT/mTOR pathway as well as the large proportion of HCCs with extensive chromosomal polysomy, characterized by whole chromosome duplication of chromosomes 5 and 7 and extensive uniparental disomy of the remaining chromosomes (8–10). These chromosomal alterations, unique to HCCs, result in a coordinated increase in expression of several critical genes in the mTOR signaling pathway, including RICTOR (Chr 5), BRAF (Chr 7), RHEB (Chr 7), AGAP2 (Chr 12), and RAPTOR (Ch17). The mTOR signaling pathway integrates both intracellular and extracellular signals and serves as a central regulator of various cellular processes, including cell growth, proliferation, metabolism, and survival. mTORC1-mediated signaling to S6K1 and 4EBP1 contributes to mTORC1-driven cell growth and cell-cycle progression. The mTOR signaling pathway is important in several types of cancer (24, 25). This provides the rationale for investigating further the possible use of mTOR inhibitors to target these aggressive and often fatal cancers.

Unfortunately, HCCs are not well studied partially because of the lack of a good mouse model. Here, we report the development of HCC models (XTC.UC1 xenograft model and TCm1 PDX model), which provide unique and powerful resources for translational studies on HCCs. We performed comprehensive characterization and demonstrated that the cell line XTC.UC1 has all the characteristic features of the more aggressive HWIDE subtype of HCC.

Using these models, we have conducted comprehensive studies demonstrating that mTOR signaling is increased and can be targeted using mTOR inhibitor drugs. Our in vitro data in XTC.UC1 showed that rapamycin, an allosteric inhibitor of mTORC1, suppressed XTC.UC1 cell proliferation, transformation, and invasion. In vivo tumor growth was suppressed, and rapamycin was able to dramatically reduce the number of lung metastases in this xenograft model. Chromosomal analysis showed cells with polysomic chromosomal alterations were more sensitive to rapamycin. Importantly, our data revealed that rapamycin inhibited XTC.UC1 proliferation through cell-cycle regulation, and this correlated with an inhibition of cyclin A2. Cyclin A2 is expressed in dividing somatic cells (21) and is involved in the initiation and completion of DNA replication during the S phase. Consistent with this, rapamycin treatment significantly decreased the S phase and cyclin A2 expression in XTC.UC1 cells. An alternative hypothesis for reduced cyclin A expression could be a direct result of mTOR inhibition on protein synthesis and/or autophagy with cyclin A being a downstream consequence of this. However, protein levels were unaltered in other cyclins, such as cyclin D and cyclin E, suggesting this was a specific effect on cyclin A2. Expression of the EMT regulator Snail was decreased after rapamycin treatment, suggesting that rapamycin is associated with cell motility through Snail-mediated EMT processes. However, again this is correlative; an alternative explanation could be a direct result of reduced protein synthesis from mTOR inhibition. Using our PDX model (TCm1), we validated the findings from XTC.UC1 xenograft tumor models, again showing that p-S6 and cyclin A2 expressions were dramatically decreased after rapamycin treatment.

To the best of our knowledge, this is the first PDX model generated for HCCs and used for preclinical therapeutic treatment, a potentially powerful tool for HCC mechanistic studies and drug screening. Despite certain limitations—tumor growth in vivo is very slow; the assessment of drug efficacy requires serial MRI imaging—this PDX HCC model represents an important step forward in model development to study this rare cancer. Overall, these data reinforce the rationale that the mTOR pathway could be a promising therapeutic target for treatment of HCCs. However, combination treatment with other tyrosine kinase inhibitors that target VEGF will likely prove to be most efficacious. In our studies, we did not see any major effect in HCC by combining rapamycin with either AKT or MEK inhibitor. However, rapamycin in combination with the multitarget tyrosine kinase inhibitor lenvatinib produced a major inhibition of angiogenesis in our PDX model. Lenvatinib is a potent inhibitor of angiogenesis targeting VEGFR1, VEGFR2, and VEGFR3 as well as fibroblast growth factor receptors (FGFR1, FGFR2, FGFR3, and FGFR4), PDGFRα, RET, and KIT 3, 4, 5. Preclinical and clinical work in renal cell cancer has shown improved efficacy when the mTOR inhibitor everolimus was combined with lenvatinib (22). Motzer and colleagues reported that lenvatinib plus everolimus significantly prolonged progression-free survival compared with everolimus alone [median, 14.6 months; 95% confidence interval (CI), 5.9–20.1 vs. 5.5 months; (3.5–7.1); HR, 0.40; 95% CI, 0.24–0.68; P = 0.0005; ref. 22]. Because vascular invasion is such an important determinant of outcome in HCC, there is strong rationale in using VEGFR inhibitors in combination with mTOR inhibitors in HCC. Together with the recent genomic data from us and others, these data have thus led to an ongoing larger phase II clinical trial (NCT 02143726) specifically on patients with HCC using the mTOR inhibitor everolimus (26) in combination with the multitarget tyrosine kinase and VEGFR inhibitor sorafenib. Patients are randomly assigned to receive sorafenib alone or sorafenib combined with everolimus. Results are forthcoming soon. It is important to mention that resistance to mTOR inhibitors can also occur. Recent research suggests that overexpression of mEAK-7/mTORC3 may be an alternative mechanism of mTOR signaling and be responsible for resistance to mTOR inhibitors such as rapamycin (27, 28). We analyzed 47 HCC tumors and identified 9 of 47 had amplification of this gene. Of the nine tumors with amplification, six also had overexpression of mEAK-7 all of which were HWIDE. Interestingly, three of six developed recurrence. This may suggest overexpression of mEAK-7 may be associated with a more aggressive biology. Tumors which have mTOR activation and overexpression of mEAK-7 may benefit from combination treatment of mTOR inhibitor along with inhibitors of mEAK-7/mTORC3. Clearly, further research on this is therefore warranted with the development of novel inhibitors of mEAK-7/mTORC3.

In conclusion, our study establishes mTOR inhibition as a potentially effective therapeutic strategy for HCCs and reveals important therapeutic vulnerabilities in HCCs. We present cellular, genomic, biochemical, and in vivo data demonstrating that HCC tumors are especially sensitive to mTOR blockade. We provide the rationale for combining mTOR inhibitors with VEGFR inhibition. Furthermore, we have generated the first human PDX model for HCCs, a resource that will be important for therapeutic development for this rare, difficult-to-treat cancer.

V. Makarov reports a patent for EP3090066A2 issued to Determinants of cancer response to immunotherapy. T. Chan and V. Makarov are listed as inventors for products developed at the Sloan Kettering Institute for Cancer Research. J.A. Fagin reports grants from the NIH during the conduct of the study. E.J. Sherman reports personal fees from Bayer and Eisai; grants and personal fees from Eli Lilly and Regeneron; nonfinancial support from Novartis; and personal fees and nonfinancial support from Roche outside the submitted work. R. Ghossein reports grants from NIH P30CA008748 during the conduct of the study. T.A. Chan reports other support from Gritstone Bio and PGDX; grants from Pfizer; grants from Nysnobio; personal fees from Illumina; and grants and personal fees from Bristol Myers Squibb outside the submitted work. No disclosures were reported by the other authors.

Y. Dong: Conceptualization, data curation, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. Y. Gong: Data curation, investigation, methodology, writing–review and editing. F. Kuo: Data curation, formal analysis, investigation, methodology, writing–review and editing. V. Makarov: Data curation, formal analysis, investigation, methodology, writing–review and editing. E. Reznik: Data curation, investigation, methodology, writing–review and editing. G.J. Nanjangud: Data curation, investigation. O. Aras: Data curation, methodology. H. Zhao: Data curation. R. Qu: Data curation. J.A. Fagin: Writing–review and editing. E.J. Sherman: Writing–review and editing. B. Xu: Data curation, methodology, writing–review and editing. R. Ghossein: Data curation, investigation, methodology, writing–review and editing. T.A. Chan: Conceptualization, supervision, funding acquisition, investigation, project administration, writing–review and editing. I. Ganly: Conceptualization, data curation, formal analysis, supervision, funding acquisition, validation, investigation, methodology, writing–original draft.

This work was supported by LesLois Shaw Foundation, Pechter Foundation, NIH SPORE grant P50 CA172012-01A1, and NIH/NCI Cancer Center Support Grant P30 CA008748.

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.

1.
Cancer Genome Atlas Research Network
. 
Integrated genomic characterization of papillary thyroid carcinoma
.
Cell
2014
;
159
:
676
90
.
2.
Hundahl
SA
,
Fleming
ID
,
Fremgen
AM
,
Menck
HR
. 
A National Cancer Data Base report on 53,856 cases of thyroid carcinoma treated in the U.S., 1985-1995
.
Cancer
1998
;
83
:
2638
48
.
3.
Máximo
V
,
Lima
J
,
Prazeres
H
,
Soares
P
,
Sobrinho-Simões
M
. 
The biology and the genetics of Hürthle cell tumors of the thyroid
.
Endocr Relat Cancer
2016
;
23
:
X2
.
4.
Ghossein
RA
,
Hiltzik
DH
,
Carlson
DL
,
Patel
S
,
Shaha
A
,
Shah
JP
, et al
Prognostic factors of recurrence in encapsulated Hurthle cell carcinoma of the thyroid gland: a clinicopathologic study of 50 cases
.
Cancer
2006
;
106
:
1669
76
.
5.
Shaha
AR
,
Shah
JP
,
Loree
TR
. 
Patterns of nodal and distant metastasis based on histologic varieties in differentiated carcinoma of the thyroid
.
Am J Surg
1996
;
172
:
692
4
.
6.
Grossman
RF
,
Clark
OH
. 
Hurthle cell carcinoma
.
Cancer Control
1997
;
4
:
13
7
.
7.
Lopez-Penabad
L
,
Chiu
AC
,
Hoff
AO
,
Schultz
P
,
Gaztambide
S
,
Ordoñez
NG
, et al
Prognostic factors in patients with Hürthle cell neoplasms of the thyroid
.
Cancer
2003
;
97
:
1186
94
.
8.
Ganly
I
,
Makarov
V
,
Deraje
S
,
Dong
Y
,
Reznik
E
,
Seshan
V
, et al
Integrated genomic analysis of Hürthle cell cancer reveals oncogenic drivers, recurrent mitochondrial mutations, and unique chromosomal landscapes
.
Cancer Cell
2018
;
34
:
256
70
.
9.
Ganly
I
,
McFadden
DG
. 
Short review: genomic alterations in Hürthle cell carcinoma
.
Thyroid
2019
;
29
:
471
9
.
10.
Gopal
RK
,
Kübler
K
,
Calvo
SE
,
Polak
P
,
Livitz
D
,
Rosebrock
D
, et al
Widespread chromosomal losses and mitochondrial DNA alterations as genetic drivers in Hürthle cell carcinoma
.
Cancer Cell
2018
;
34
:
242
55
.
11.
Zielke
A
,
Tezelman
S
,
Jossart
GH
,
Wong
M
,
Siperstein
AE
,
Duh
QY
, et al
Establishment of a highly differentiated thyroid cancer cell line of Hürthle cell origin
.
Thyroid
1998
;
8
:
475
83
.
12.
Zielke
A
,
Hoffmann
S
,
Plaul
U
,
Duh
QY
,
Clark
OH
,
Rothmund
M
. 
Pleiotropic effects of thyroid stimulating hormone in a differentiated thyroid cancer cell line. Studies on proliferation, thyroglobulin secretion, adhesion, migration and invasion
.
Exp Clin Endocrinol Diabetes
1999
;
107
:
361
9
.
13.
Shen
R
,
Seshan
VE
. 
FACETS: Allele-specific copy number and clonal heterogeneity analysis tool for high-throughput DNA sequencing
.
Nucleic Acids Res
2016
;
44
:
e131
.
14.
Hänzelmann
S
,
Castelo
R
,
Guinney
J
. 
GSVA: gene set variation analysis for microarray and RNA-seq data
.
BMC Bioinform
2013
;
14
:
7
.
15.
Liberzon
A
,
Birger
C
,
Thorvaldsdóttir
H
,
Ghandi
M
,
Mesirov
JP
,
Tamayo
P
. 
The Molecular Signatures Database (MSigDB) hallmark gene set collection
.
Cell Syst
2015
;
1
:
417
25
.
16.
Dong
Y
,
Manley
BJ
,
Becerra
MF
,
Redzematovic
A
,
Casuscelli
J
,
Tennenbaum
DM
, et al
Tumor xenografts of human clear cell renal cell carcinoma but not corresponding cell lines recapitulate clinical response to sunitinib: Feasibility of using biopsy samples
.
Eur Urol Focus
2017
;
3
:
590
8
.
17.
Bonora
E
,
Porcelli
AM
,
Gasparre
G
,
Biondi
A
,
Ghelli
A
,
Carelli
V
, et al
Defective oxidative phosphorylation in thyroid oncocytic carcinoma is associated with pathogenic mitochondrial DNA mutations affecting complexes I and III
.
Cancer Res
2006
;
66
:
6087
96
.
18.
Pradella
LM
,
Evangelisti
C
,
Ligorio
C
,
Ceccarelli
C
,
Neri
I
,
Zuntini
R
, et al
A novel deleterious PTEN mutation in a patient with early-onset bilateral breast cancer
.
BMC Cancer
2014
;
14
:
70
.
19.
Pike
KG
,
Malagu
K
,
Hummersone
MG
,
Menear
KA
,
Duggan
HM
,
Gomez
S
, et al
Optimization of potent and selective dual mTORC1 and mTORC2 inhibitors: the discovery of AZD8055 and AZD2014
.
Bioorg Med Chem Lett
2013
;
23
:
1212
6
.
20.
Woo
RA
,
Poon
RY
. 
Cyclin-dependent kinases and S phase control in mammalian cells
.
Cell Cycle
2003
;
2
:
316
24
.
21.
Loukil
A
,
Cheung
CT
,
Bendris
N
,
Lemmers
B
,
Peter
M
,
Blanchard
JM
. 
Cyclin A2: at the crossroads of cell cycle and cell invasion
.
World J Biol Chem
2015
;
6
:
346
50
.
22.
Motzer
RJ
,
Hutson
TE
,
Glen
H
,
Michaelson
MD
,
Molina
A
,
Eisen
T
, et al
Lenvatinib, everolimus, and the combination in patients with metastatic renal cell carcinoma: a randomised, phase 2, open-label, multicentre trial
.
Lancet Oncol
2015
;
16
:
1473
82
.
23.
Basilio-de-Oliveira
RP
,
Pannain
VL
. 
Prognostic angiogenic markers (endoglin, VEGF, CD31) and tumor cell proliferation (Ki67) for gastrointestinal stromal tumors
.
World J Gastroenterol
2015
;
21
:
6924
30
.
24.
Samatar
AA
,
Poulikakos
PI
. 
Targeting RAS-ERK signalling in cancer: promises and challenges
.
Nat Rev Drug Discov
2014
;
13
:
928
42
.
25.
McCubrey
JA
,
Steelman
LS
,
Chappell
WH
,
Abrams
SL
,
Montalto
G
,
Cervello
M
, et al
Mutations and deregulation of Ras/Raf/MEK/ERK and PI3K/PTEN/Akt/mTOR cascades which alter therapy response
.
Oncotarget
2012
;
3
:
954
87
.
26.
National Cancer Institute
. 
Sorafenib tosylate with or without everolimus in treating patients with advanced, radioactive iodine refractory Hurthle cell thyroid cancer
.
ClinicalTrials.gov. Available from:
https://clinicaltrials.gov/ct2/show/NCT02143726.
27.
Harwood
FC
,
Klein Geltink
RI
,
O'Hara
BP
,
Cardone
M
,
Janke
L
,
Finkelstein
D
, et al
ETV7 is an essential componenet of a rapamycin insensitive mTOR complex in cancer
.
Sci Adv
2018
;
4
:
eaar3938
.
28.
Nguyen
JT
,
Haidar
FS
,
Fox
AL
,
Ray
C
,
Mendonca
DB
,
Kim
JK
, et al
mEAK-7 forms an alternative mTOR complex with DNA-PKcs in human cancer
.
iScience
2019
;
17
:
190
207
.

Supplementary data