Establishment of liver tumor cell lines from atherogenic and high fat diet fed hepatitis C virus transgenic mice

A syngeneic mouse model bearing a transplanted tumor is indispensable for the evaluation of the efficacy of immune checkpoint inhibitors (ICIs). However, few syngeneic mouse models of liver cancer are available. We established liver tumor cell lines (MHCF1 and MHCF5) from hepatitis C virus transgenic mice fed an atherogenic high-fat diet. MHCF1 and MHCF5 were successfully transplanted into the subcutaneous space of syngeneic C57BL/6 mice, in addition, they efficiently developed orthotopic tumors in the liver of syngeneic C57BL/6 mice. MHCF5 grew rapidly and showed a more malignant phenotype compared with MHCF1. Histologically, MHCF1-derived tumors were a combined type of hepatocellular carcinoma and MHCF5-derived tumors showed a sarcomatous morphology. Interestingly, MHCF1 and MHCF5 showed different sensitivity against an anti-PD1 antibody and MHCF5-derived tumors were resistant to this antibody. CD8 T cells infiltrated the MHCF1-derived tumors, but no CD8 T cells were found within the MHCF5-derived tumors. Gene expression profiling and whole-exon sequencing revealed that MHCF5 displayed the features of an activated cancer stem cell-like signature of sonic hedgehog and Wnt signaling. Therefore, these cell lines could be useful for the identification of new biomarkers and molecular mechanisms of ICI resistance and the development of new drugs against liver cancer.


Allograft and syngeneic tumor models of established mouse liver tumor cell lines. The estab-
lished cell lines were transplanted subcutaneously into immune-deficient NOD-SCID mice. Among the three cell lines (derived from HCV-Tg, HCV-Tg/Ath + HFD, and Pdgfc-Tg tumors), one cell line, an HCV-Tg/ Ath + HFD-derived cell line, developed into tumors in NOD-SCID mice ( Fig. 2A,B). The tumor cells were dissociated and grown on dishes. The cells were grown and passaged several times. We confirmed the cells were positive for albumin (Alb) and alpha-feto protein (Afp) mRNA by RT-PCR after several passages. We named this cell line mouse liver tumor cells derived from HCV full-length Tg No. 1 (MHCF1) (Fig. 2C). MHCF1 cells were then transplanted subcutaneously into syngeneic C57BL/6 mice. MHCF1 cells developed into tumors in C57BL/6 mice (Fig. 2D), whereas Hepa1-6 cells did not (data not shown).
To obtain cell lines that could adapt to the microenvironment of the in vivo mouse model, MHCF1 cells were repeatedly transplanted into NOD-SCID mice, and the developed tumors were dissociated and passaged on culture dishes several times. We performed these procedures three times ( Fig. 2A). After three cycles, the obtained cells were seeded on 96-well plates with limiting dilution and 12 clones were obtained. All clones were negative for Alb and Afp mRNA. Six of the 12 clones were transplanted subcutaneously into syngeneic C57BL/6 mice and all clones developed into tumors. We selected one fast growing clone and named it MHCF5 (Fig. 2D). A trace amount of HCV-RNA was detected in MHCF1 and MHCF5 cells by RT-PCR; however, HCV protein was not detected in either cell line by western blotting (Supplemental Fig. 2). Although the volume of MHCF1-and MHCF5-derived tumors in C57BL/6 mice did not differ significantly within 25 days (Fig. 2E), the overall survival of MHCF5 tumor-bearing mice was significantly shorter than that of MHCF1 tumor-bearing mice (Fig. 2F).

Histopathological features of MHCF1-and MHCF5-derived tumors.
The histopathological features of MHCF1-and MHCF5-derived tumors were analyzed. MHCF1-derived tumors were composed of two cell components, namely, an HCC-like lesion and an intracellular cholangiocarcinoma-like lesion (Fig. 3A). Therefore, MHCF1-derived tumors were considered to be combined HCC and CCC. KRT19 expression was confirmed in the intracellular cholangiocarcinoma-like lesions by immunohistochemical staining (Fig. 3B). In contrast, MHCF5-derived tumors were composed of mesenchymal and fibrotic cells, as reported for sarcomatous HCC after repeated anti-cancer treatment 9 .
We performed RNA-seq analysis of MHCF1 and MHCF5 cells together with healthy liver and Hep55.1C cells, which were derived from carcinogen-induced liver tumors developed in C57BL/6 mice 10 . MHCF1 cells expressed hepatocyte markers such as Alb, Afp, hepatocyte nuclear factor 4 alpha (Hnf4a) and apolipoprotein e (Apoe) as well as cholangiocyte markers, such as integrin beta-4 (Itgb4), aquaporin 1 (Aqp1), keratin 7 (Krt7), and Krt19. In contrast, MHCF5 cells lost the expression of hepatocyte markers, but maintained the expression of cholangiocyte markers (Itgb4, Aqp1, and Krt7) (Fig. 3C). Hepatocyte markers were expressed at a higher level in MHCF1 cells than in Hep55.1C cells, while Hep55.1C cells expressed cholangiocyte markers to some degree. These results were confirmed by RTD-PCR (data not shown). www.nature.com/scientificreports/  www.nature.com/scientificreports/  www.nature.com/scientificreports/ Establishment of orthotopic liver tumor models of MHCF1 and MHCF5 cells by splenic tumor cell injection. To examine whether MHCF1 and MHCF5 cells could develop into tumors in the liver, they were injected into the spleen of C57BL/6 mice. At 30 days after injection, the mice were sacrificed and the livers were examined. We found that the whole liver was almost completely occupied with a tumor (white colored lesion) (Fig. 4A). The infiltration of immune cells was investigated by immunohistochemical staining of CD4-, CD8-, and CD68-positive cells. Interestingly, CD4 T cells preferentially accumulated in the marginal area of MHCF1-derived tumors, while CD8 T cells infiltrated the MHCF1-derived tumors (Fig. 4B). In contrast, no CD4 or CD8 T cell accumulation was observed in MHCF5-derived tumors (Fig. 4C). For CD68-positive monocytes/macrophages, there was no difference in the degree of infiltration between MHCF1-and MHCF5-derived tumors.
Gene expression profiling of MHCF1 and MHCF5 cells. To reveal the molecular signatures of MHCF1-and MHCF5-derived tumors, gene expression profiling was performed using a gene chip. The gene expression profiles of MHCF1 cells, MHCF5 cells, and healthy mouse liver were compared. First, we examined the expression of highly expressed genes (57 representative genes that were within the top 1.5% of all transcripts) in primary human hepatocytes (deduced from the personal RNA-sequencing data of PXB cells; PhoenixBio, Hiroshima, Japan). In addition to Alb and Hnf4a, other hepatocyte-specific genes such as apolipoproteins (Apoe, Apob, and Apoa1), coagulation factors (F5 and Fgb), bile acid transporter (Abcc3), and complement proteins (C2 and C3) were expressed in MHCF1 cells, but not in MHCF5 cells (Supplemental Fig. 3). We next examined the expression of HCC-specific genes in MHCF1 and MHCF5 cells according to Hoshida's HCC classification (S1, S2, and S3) 11 . In the GeneChip Mouse Gene 1.0 ST Array (Affymetrix, Santa Clara, CA) used in this study, 199 S1 genes, 101 S2 genes, and 208 S3 genes could be analyzed. We found that 79% of S1 and S2 genes (236 out of 300) were upregulated in MHCF1 and/or MHCF5 cells (Fig. 5A), while 23% of S3 genes (48 out of 208) were upregulated in MHCF1 and/or MHCF5 cells. Thus, MHCF1 and MHCF5 cells shared the features of S1 and S2 subclasses rather than the S3 subclass. Among the 236 upregulated genes in MHCF1 and/ or MHCF5 cells, 149 genes were commonly upregulated in both cell lines, including Akt3, Ctnnb1, Smad2, and Hif1a, indicating the activation of Wnt, Tgf-β, Akt, and Myc signaling. Sixty-seven genes were upregulated only in MHCF5 cells, including Col4, Rho, and Tcf4, indicating the activation of Wnt and Tgf-β signaling. Twenty-one genes were upregulated only in MHCF1 cells, including the tumor markers Afp and Cpc3. The results indicated that MHCF1 cells more resembled the S2 subclass, while MHCF5 cells more resembled the S1 subclass.
Next, on the basis of the previously reported gene expression data of healthy biliary epithelial cells and CCC cells (GSE22633) 12 , we selected 32 CCC-specific genes and examined their expression in MHCF1 and MHCF5 cells. These CCC-specific genes were expressed in MHCF1 and MHCF5 cells (Fig. 5B). Similarly, we selected 32 sarcomatous-specific genes deduced from CCC with sarcomatous changes 12 and examined their expression in MHCF1 and MHCF5 cells. Sarcomatous-specific genes were expressed at a higher level in MHCF5 cells than in MHCF1 cells (Fig. 5C). These results supported the histological findings of MHCF1-and MHCF5-derived tumors; MHCF1-derived tumors consisted of a combined type of HCC sharing the features of HCC and CCC, whereas MHCF5-derived tumors were poorly differentiated CCC with sarcomatous features.
Finally, we performed pathway comparisons among MHCF1 cells, MHCF5 cells, and healthy liver as previously described 13 . Compared with healthy liver, the expression of cell cycle (Cdk, etc.), DNA damage response (Brca, Rb, p53, etc.), Erad pathway (endoplasmic reticulum-associated), and oncogene (sonic hedgehog [Shh] and Src) genes was upregulated in MHCF1 and MHCF5 cells, whereas the expression of coagulation, complement, lectin, and nuclear receptor genes was downregulated (Table 1) (Supplemental Fig. 4A). Interestingly, when comparing MHCF1 and MHCF5 cells, the expression of Tgf-β and inflammation pathway genes was upregulated in MHCF5 cells compared with MHCF1 cells, and the expression of coagulation genes was downregulated in MHCF5 cells compared with MHCF1 cells (Table 1) (Supplemental Fig. 3B). Therefore, MHCF5 cells expressed genes that are characteristic of epithelial to mesenchymal transition.

Different sensitivity of MHCF1-and MHCF5-derived tumor against an anti-PD1 antibody.
The sensitivity of MHCF1-and MHCF5-derived tumors against an anti-PD1 antibody was evaluated. The expression of immune checkpoint molecules (PDL-1, PDL-2, and Galectin-9) in MHCF1 and MHCF5 cells is shown in Fig. 6A. The expression of these immune checkpoint molecules was upregulated in MHCF1 cells compared  www.nature.com/scientificreports/ to MHCF5 cells and healthy liver. For MHC class I molecules, the expression of H2-D1, H2-K1, and H2-K2 was downregulated in MHCF1 and MHCF5 cells compared to healthy liver. An anti-PD1 antibody (100 μg/mouse for 5 times) was administered intraperitoneally to MHCF1-or MHCF5derived tumor-bearing mice and tumor volumes were evaluated (Fig. 5). Interestingly, the anti-PD1 antibody reduced the volume of MHCF1-derived tumors by approximately 70%, while MHCF5-derived tumors showed no response to the anti-PD1 antibody. Thus, MHCF1-and MHCF5-derived tumors showed different sensitivity to the anti-PD1 antibody.

Discussion
Treatment using ICIs for liver cancer has yielded some encouraging results, but the percentage of patients responding to single-agent therapies remains low. Therefore, the development of combination therapies for liver cancer is crucial and urgent. A syngeneic mouse model bearing an orthotopic or subcutaneous transplanted tumor is indispensable for the evaluation of the efficacy of ICIs or combinations of ICIs and other anti-tumor drugs. However, few syngeneic models of liver cancer are available. BNL-1MEA is a BALB/c-derived and chemically (methylcholanthrene epoxide) transformed cell line 14 . The HCA-1 cell line is derived from a spontaneously developed liver cancer in C3H mice 15,16 . However, genetically engineered mice such as Tg or knockout mice are made mainly from C57BL/6 mice; therefore, a syngeneic tumor model based on C57BL/6 mice is more useful to reveal the molecular mechanisms of drug resistance and immune-evading mechanisms of tumors in the setting of tumor microenvironments. The responsible genes for immune evasion by tumors can be analyzed by transplanting a tumor into responsible gene knockout or Tg mice. The RIL-175 cell line was established from C57BL/6 mice; however, this cell line has already been genetically modified by knocking out p53 (p53 −/− ) and transduced with H-RasV12 17 . The Hepa1-6 cell line is derived from C57L mice; however, we observed that subcutaneous Hepa1-6 HCC cells grew transiently in C57BL/6 mice and subsequently diminished 18 . The Hep55.1C cell line is derived from carcinogen-induced liver tumors of C57BL/6 mice 10 and utilized in a syngeneic orthotopic murine model of HCC 19 . However, the etiology of liver cancer in this model is different from that of the clinical setting.
In this study, we developed new mouse liver tumor cell lines (MHCF1 and MHCF5 cells) from HCV-Tg mice fed an Ath + HFD. We previously reported an Ath + HFD mouse model that develops hepatic steatosis, inflammation, fibrosis, and insulin resistance, resembling human non-alcoholic steatohepatitis 20 . Non-alcoholic fatty liver disease and non-alcoholic steatohepatitis are frequently associated with chronic hepatitis C and they are accelerated in HCV-related liver disease 21 . Therefore, tumor cell lines derived from HCV-Tg mice fed an Ath + HFD would have clinical significance.
Our cell lines were successfully transplanted into the subcutaneous space of syngeneic C57BL/6 mice and tumor growth was evaluated. In addition, these cell lines efficiently developed orthotopic tumors in the liver www.nature.com/scientificreports/ www.nature.com/scientificreports/ of syngeneic C57BL/6 mice. These cell lines are unique, because no host tumor suppressor genes or oncogenes were manipulated. A trace mount of HCV-RNA was detected in both cell lines; however, no HCV protein was detected. Thus, the effect of HCV proteins on the tumor phenotypes of these cell lines would be minimal. Histologically, MHCF1 tumors showed the combined type of HCC sharing the features of epithelial-like HCC lesions and ductal cholangiocyte-like lesions. In contrast, MHCF5-derived tumors showed the features of poorly differentiated sarcomatous HCC.
According to Hoshida's HCC classification (S1, S2, and S3) 11 , MHCF1 and MHCF5 cells were classified as S1 and/or S2 subclasses, which are associated with poor clinical survival. MHCF1 was more like the S2 subclass with the activation of Akt and Myc and positivity for the tumor marker Afp. In contrast, MHCF5 was more like the S1 subclass with the activation of Wnt and Tgf-β (Fig. 5A). It could be speculated that MHCF1 cells originated from progenitor cells that can develop into hepatocytes and cholangiocytes. We performed 3 cycles of transplantation of MHCF1 cells into NOD-SCID mice. During these repeated transplantation steps, rapidly growing and aggressive tumors could be selected in vivo. Interestingly, the in vivo selection of MHCF1 cells enabled us to isolate tumor cells (MHCF5 cells) with a different phenotype and genotype compared with the original MHCF1 cells.  www.nature.com/scientificreports/ Gene expression profiling and whole-exon sequencing supported the histological findings of the MHCF1and MHCF5-derived tumors. Interestingly, MHCF5 cells lost the hepatocyte gene signature, but retained the expression of representative CCC marker genes together with a sarcomatous gene signature (Fig. 5B,C). It is characteristic that MHCF5 cells obtained an epithelial to mesenchymal transition signature and activated the cancer stem cell-like signature of Shh and Wnt signaling.
Interestingly, the MHCF1-and MHCF5-derived tumors showed different sensitivity to the anti-PD1 antibody. The anti-PD1 antibody reduced the volume of MHCF1-derived tumors by 70%, while MHCF5-derived tumors were resistant to this antibody (Fig. 6B). Interestingly, CD8 T cells infiltrated the MHCF1-derived tumors, but no CD8 T cells were found within the MHCF5-derived tumors (Fig. 4). Although we could not see any difference in the infiltration of CD68-positive cells between these tumors, further studies evaluating myeloid-derived suppressor cells should be performed. In this study, the expression of immune checkpoint molecules, such as PD-L1, PD-L2, and Galectin-9, was higher in MHCF1 cells than in MHCF5 cells (Fig. 6A), which might be related to the favorable response of MHCF1 cells to the anti-PD1 antibody to some degree. However, the expression of these molecules is regulated by multiple factors including MAPK, PI3K, AKT HIF1, STAT3, and NF-κB 22 . Moreover, a recent report showed that the activation of Wnt signaling could be related to resistance against anti-PD1 therapy 23 . Shh and Wnt signaling are co-operating signaling pathways that are essential for embryonic development 24 . The detailed underlying mechanisms of the different responses of MHCF1 and MHCF5 cells to the anti-PD1 antibody should be studied further.
In summary, we newly developed two mouse liver tumor cell lines, MHCF1 and MHCF5 cells, derived from HCV-Tg/Ath + HFD mice. MHCF1-derived tumors consisted of a combined type of HCC, whereas MHCF5derived tumors were CCC with a malignant phenotype and resistant to anti-PD1 therapy. These cell lines could be useful for the generation of syngeneic mouse tumor transplant models and the identification of new biomarkers and molecular mechanisms of ICI resistance.

Materials and methods
Mice. The generation and characterization of Ath + HFD mice were performed as previously described 20 .
HCV-Tg mice were generated as previously described 8 . Pdgfc-Tg mice were kindly provided by Dr. Jean S. Campbell (University of Washington, Seattle, WA) 7 . C57BL/6 J and NOD.CB17-Prkdcscid/J mice were purchased from Charles River Laboratories Japan (Yokohama, Japan). Colonies of HCV-Tg and Pdgfc-Tg mice were maintained by crossing with C57BL/6 J mice for at least 5 generations. All animal experiments were approved by the Ethics Committee for the Care and Use of Laboratory Animals at the Takara-Machi Campus of Kanazawa University, Japan and were carried out in compliance with the ARRIVE guidelines 2.0. All experiments were performed in accordance with relevant guidelines and regulations.

Tumor cell dissociation from transplanted mice. A Tumor Dissociation Kit, Mouse (Miltenyi Biotec)
was used according to the manufacturer's instructions. Briefly, tumors were cut into small pieces in Petri dishes and transferred into a gentleMACS C Tube containing enzyme mix. The tissue was dissociated using a gen-tleMACS Dissociator (Miltenyi Biotec). The cell suspensions were washed three times with phosphate-buffered saline and passed through a 100-μm strainer with 10 mL Dulbecco's modified Eagle's medium. These cells were cultured as described above.
For immunofluorescence staining, the slides were incubated with an Alexa Fluor 488-labeled anti-cytokeratin 19 antibody (1:200; Abcam) diluted in 1% bovine serum albumin/phosphate-buffered saline for 2 h at room temperature. The slides were mounted using DAPI, and the cells were viewed using an image analysis system (BIOREVO BZ-9000; Keyence, Osaka, Japan).
Quantitative RTD-PCR. Total RNA was isolated using a High Pure RNA Isolation Kit (Roche Applied Science, Pleasanton, CA), and cDNA was synthesized with a High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA) according to the manufacturers' instructions. RTD-PCR was performed using the 7900HT Real-Time PCR System (Applied Biosystems). The primer pairs and probes for mouse genes (Alb, Afp, Hnf4a, Aqp1, Krt7, Krt19, and Gapdh) were obtained from the TaqMan assay reagents library.
RNA-seq and gene chip analysis. Total  www.nature.com/scientificreports/ sample were aligned to the mouse genome (mm10) using Subread 25 , and transcript abundance was shown by the count data by using HTSeq 26 . Expression data were adjusted by a total of 10 million counts. Gene chip analysis of MHCF1 cells, MHCF5 cells, and mouse liver was performed using a GeneChip Mouse Gene 1.0 ST Array (Affymetrix) as described previously 27 . Functional ontology enrichment analysis was conducted to compare the BioCarta Pathway process distribution of the differentially expressed genes 13 . LS/KS permutation tests were performed for pathway comparison (P < 0.05) (BRB-ArrayTools; https:// brb. nci. nih. gov/ BRB-Array Tools).
Whole-exon sequencing. A whole-exon sequencing was performed as previously described 28 . A Sure-Select Human All Exon V4 Kit (Agilent Technologies, Santa Clara, CA) was used for whole-exon capture, and the HiSeq 2000 Sequencing System (Illumina) was used for massive parallel sequencing. Sequence reads were mapped against the UCSC Genome Browser mm10. Sequence variations, including single nucleotide polymorphisms and insertions/deletions, were detected using Genome Analysis Toolkit software (Broad Institute, Cambridge, MA). Whole-exon sequencing and analysis was performed at Riken Genesis (Tokyo, Japan). Pathway analysis was conducted using MetaCore (Thomson Reuters, New York, NY).

Statistical analysis.
For the animal experiments, a sample size of 5 was chosen for each experimental group. Data represent the mean ± standard error of the mean from two or three independent experiments. The significance of between-group comparisons was tested by one-way or two-way analysis of variance with Bonferroni's multiple comparisons test using Prism 8.0 software (GraphPad Software, La Jolla, CA).