Acquired Resistance with Epigenetic Alterations Under Long-Term Antiangiogenic Therapy for Hepatocellular Carcinoma

Hepatocellular carcinoma is the second most frequent cause of cancer-related mortality worldwide (1, 2). As the SHARP and Asian-Pacific trials successfully demonstrated that sorafenib prolonged overall survival in patients with advanced hepatocellular carcinoma, it has become the standard of care in this setting (3). A phase I/II trial of lenvatinib for hepatocellular carcinoma patients in Child-Pugh class A recently reported that 14 of 46 patients enrolled have partial response, and a phase III trial was designed to compare the efficacy of lenvatinib and sorafenib (4). In the randomized, double-blind, placebo-controlled international phase III trial (RESORCE), regorafenib achieved an overall survival improvement in patients with hepatocellular carcinoma progressing on sorafenib treatment (5). All the three drugs are classified as multikinase inhibitors mainly targeting angiogenesis through VEGF receptor (VEGFR). We previously reported that a novel VEGFR and Aurora kinase inhibitor JNJ-28841072 remarkably suppresses tumor growth of human hepatocellular carcinoma cell lines by decreasing neovascularization (6), suggesting the angiogenesis dependency as an essential hallmark of hepatocellular carcinoma.

Although the introduction of antiangiogenic therapy leads to tumor regression and in some cases extended survival, the tumors inevitably progress again after a short term of clinical benefit (7). In the phase III C-08 trial of adjuvant therapy prior to surgery for colorectal cancer, the addition of a humanized anti-VEGF mAb bevacizumab transiently improved disease-free survival during the initial year of treatment but not during the remaining period (8). In xenograft mouse models, the tumors showed transient response to VEGFR antagonists, but became more aggressive, invasive, and metastatic following several weeks of administration (9). There is a clear need to reveal the mechanistic basis of this apparent conundrum for the therapeutic targeting of tumor angiogenesis.

Several laboratories have mentioned that epigenetic events are implicated in drug resistance in tumor cells treated with targeted agents such as hormone therapy and tyrosine kinase inhibitors (10–15). Tamoxifen- and fulvestrant-resistant sublines of MCF7 breast cancer cells have diverse gene expression and DNA methylation profiles (12). H3K4 methylation states were altered by the elevated expression of histone demethylase RBP2/KDM5A/JARID1A in gefinitib-tolerant PC9 non–small cell lung cancer cells (13). However, the epigenetic mechanisms underlying drug resistance to antiangiogenic therapy are poorly understood.

Therefore, we derived in vivo drug-tolerant hepatocellular carcinoma cells by repeated treatment of VEGFR inhibitor and serial transplantation in immunodeficient mice for 12 months and identified that aberrant expression of thymosin β 4 (Tβ4) caused by epigenetic alterations could contribute to the achievement of resistance (16).

The human hepatocellular carcinoma cell line HuH7 was obtained from the Human Science Research Resources Bank, authenticated by short tandem repeat DNA fingerprinting (BEX Co. Ltd.). It was cultured in RPMI1640 (Wako) supplemented with 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin (Invitrogen), maintained in a humidified incubator at 37°C in 5% CO₂, and harvested with 0.05% trypsin-0.03% EDTA (Invitrogen). NOD/SCID (NOD.CB17-Prkdc^scid/J) mice were purchased from Charles River Laboratories. All mouse procedures were approved by the Institutional Animal Care and Use Committee of our institute (permission No. 0160074A). JNJ-28841072 (7-[1H-indol-2-yl]-2, 3-dihydro-isoindol-1-ones) was generously provided by Janssen Pharmaceutical Research and Development, and sorafenib tosylate was purchased from Selleck Chemicals.

Three hundred and eight patients underwent curative hepatectomy for hepatocellular carcinoma from 2009 and 2014 at Tokyo Medical and Dental University Hospital. Among them, 30 patients were postoperatively treated with sorafenib (400–800 mg/day) for the recurrence. The initially resected tissues were fixed in 10% formaldehyde solution and embedded in paraffin for histopathologic analysis. Written informed consent was received from all the patients in this study with the approval of Institutional Review Board of our institute (permission no. 1080).

Tumors derived from HuH7 were roughly minced into small pieces and immediately transplanted into NOD/SCID mice. When the tumors reached 100 to 150 mm³ in volume, JNJ-28841072 (100 mg/kg/day) was intraperitoneally administered into the mouse on two consecutive days per week for 2 weeks, but tumors composed of resistant cancer cells developed up to day 28. This procedure was repeated up to 12 times, and cancer cells established from the 6th and 12th generation of the resistant tumors served as HuH7-F6 and HuH7-F12, respectively.

Cells were suspended in 100-μL Matrigel (BD Biosciences) and subcutaneously injected into NOD/SCID mice. The volume of the growing tumors was monitored every 2 days, and calculated by the formula; volume = length × width² × 0.5. JNJ-28841072 (100 mg/kg/day) and sorafenib (30 mg/kg/day) were administered after tumors became palpable (approximately 250 mm³).

Total RNA was extracted from cells by using RNeasy Protect Mini Kit (Qiagen). The integrity of the obtained RNA was confirmed by using 2100 Bioanalyzer (Agilent Technologies). Contaminating DNA was removed by digestion with RNase-Free DNase Set (Qiagen). Complementary RNA was prepared from 100 ng of total RNA from each sample with 3′ IVT Express Kit (Affymetrix). The hybridization and signal detection of GeneChip Human Genome U133 Plus 2.0 Array (Affymetrix) were performed in accordance with the manufacturer's instructions. Gene expression data have been deposited in the Gene Expression Omnibus (GEO) under accession GSE93595.

The four microarray datasets of two pairs of the parental (HuH7-F0) and filial hepatocellular carcinoma cell lines (HuH7-F6 and HuH7-F12) were normalized using the robust multiarray average method in the R statistical software (version 3.0.3) and the Affy Bioconductor package. To investigate how the biological functions changed during the acquisition of resistance to antiangiogenic therapy, the gene set enrichment analysis (GSEA) was performed with the MSigDB gene sets (C2: chemical and genetic perturbations; version 5.0).

For single-stranded complementary DNA synthesis, 1 μg of total RNA was reverse-transcribed by SuperScript III Reverse Transcriptase (Invitrogen). The primer sets and amplification conditions for PCR are listed in Supplementary Table S1. GAPDH RNA was used as an endogenous control.

Genomic DNA was obtained from cells by phenol–chloroform extraction. Bisulfite treatment of DNA was performed with EZ DNA Methylation-Gold (Zymo Research), and then methylation-specific PCR (MSP) and bisulfite sequencing (BSS) were conducted. Briefly, the PCR reaction was performed for 35 cycles in a 25-μL mixture comprising bisulfite-modified DNA, 2.5 μL of 10× PCR buffer, 1.25 μL of 25 mmol/L dNTPs, 25 pmol/L of each primer, and 1 U of JumpStart RedTaq polymerase (Sigma-Aldrich). The conditions for PCR were listed in Supplementary Table S1.

Chromatin immunoprecipitation (ChIP) assay was performed by using ChIP-IT Express Kit (Active Motif) according to the manufacturer's protocol. Immunoprecipitated DNA enrichment was normalized to the input. The antibodies used in this study were anti-H3K4me3, anti-H3K9me3, anti-H3K27me3, and anti-H3K27ac, all of which were purchased from Active Motif. Normal rabbit IgG (Cell Signaling Technology) was used as a negative control for each assay.

Cells were lysed by using RIPA Buffer (Thermo Fisher Scientific) with a Protease Inhibitor Cocktail Kit (Sigma Aldrich). Aliquots containing 30 μg of cell lysates were denatured in 5× Sample Buffer (Wako), electrophoretically resolved on SDS-polyacrylamide gels (Wako), and then transferred onto Immobilon polyvinyldifluoride membranes (Millipore). The membrane blots were blocked with 2% skimmed milk (Cell Signaling Technology) for an hour at room temperature, and then incubated with primary antibodies at 4°C overnight. After the appropriate secondary antibodies were added for an hour, the signals were developed with Immun-Star AP Substrate (Bio-Rad) and observed by using LAS-3000 (Fujifilm).

The Tβ4 construct was amplified by using the primer pair shown in Supplementary Table S1, and subcloned into HindIII and BamHI sites in pEGFP-N1 (Clontech Laboratories). HuH7 cells were transfected with the plasmid for Tβ4–EGFP fusion protein expression by using Neon Transfection System (Invitrogen). The transfected cells were selected in culture media containing 400 μg/mL G418 (Invitrogen), and utilized as HuH7-Tβ4 cells in this study.

Cells were seeded onto small coverslips in 6-well plates, and incubated for 24 hours to allow cell attachment. The cells were fixed with 4% paraformaldehyde at 4°C for 15 minutes, and permeabilized with 0.1% Triton X-100 for 5 minutes prior to incubation in 3% BSA for 30 minutes at room temperature. The blocking buffer was removed and the cells were incubated with primary antibodies at 4 °C for an hour. After washing with PBS, they were additionally incubated with fluorescence-conjugated secondary antibodies (Invitrogen) for an hour, and the cellular DNA was subsequently covered and labeled with ProLong Gold antifade reagent with DAPI (Invitrogen). The slides were viewed with a fluorescent microscope (Carl Zeiss).

HuH7-F0, -F12, and -Tβ4 cells were plated separately at several densities from 1,000 to 2,000 cells in low-attachment plates (24-well Ultra Low Cluster Plate; Corning), and incubated in serum-free DMEM/F12 media (Wako), and the total number of spheres was counted a week after incubation.

The double-chamber migration assay was performed by using a transwell chamber (24-well plate, 8-μm pores; BD Biosciences). For the invasion assay, Matrigel-coated (BD Biosciences) transwells (0.1 mg/mL) were prepared by incubation in serum-free media for 2 hours at 37°C in 24-well plates. The bottom chambers were filled with 0.8 mL culture media without antibiotics. Then, cancer cells (8 × 10⁴ in 0.3 mL serum-free media) were seeded onto the top chambers and incubated at 37°C for 24 hours. The cells on the top surface of the filters were removed by using cotton wool swabs. The remaining cells were then fixed with 100% methanol and stained with Giemsa solution, and the number of cells migrating or infiltrating into the bottom surface was counted in three randomly selected high-magnification fields (100×) for each sample.

Transplanted tumor tissues were fixed overnight in 4% paraformaldehyde, embedded in paraffin, and sectioned (4 μm thick). Tumor specimens of the 30 hepatocellular carcinoma patients described above were also sectioned (4 μm thick). They were stained with an automated immunostainer (DISCOVERY XT; Ventana Medical Systems) by using heat-induced epitope retrieval and a Standard Diaminobenzidine Detection Kit. Incubation times for the primary and secondary antibodies were 2 and 1 hours, respectively. The primary antibodies used were anti-Tβ4 (sc-67114, 1:200; Santa Cruz Biotechnology) and anti-CD31 (ab28364, 1:50; Abcam). Secondary antibodies included universal secondary antibody (Ventana Medical Systems). All tissue sections were counterstained with hematoxylin.

Statistical analysis was performed by using SPSS statistics Version 23.0 (IBM). Two-sided Student t tests and multiple comparison tests were used to analyze for differences between continuous values of two independent groups. The Fisher exact test was applied to analyze categorical variables. Survival curves were constructed by using the Kaplan–Meier method and compared with the log-rank test. P value less than 0.05 was considered statistically significant.

To clarify the mechanism of in vivo resistance to long-term antiangiogenic therapy in hepatocellular carcinoma, we first established a xenograft mouse model as schematically represented in Fig. 1A. Briefly, after the VEGFR and Aurora kinase inhibitor JNJ-28841072 was intraperitoneally injected into NOD/SCID mice bearing tumors of HuH7 cells, the tumors eliciting resistance to this inhibitor were excised into small aliquots and directly engrafted in a second group of NOD/SCID mice. By repeating this process 6 and 12 times, drug-refractory HuH7-F6 and -F12 cells were derived from the parental cancer cells (HuH7-F0), respectively. We validated that JNJ-28841072 reduced the size of transplanted tumors of HuH7-F0, but not those of HuH7-F12 (Fig. 1B). Sorafenib, an antiangiogenic drug approved for advanced hepatocellular carcinoma, could not extend survival time in the mice with HuH7-F6 or -F12 tumors because of failure to suppress the tumor growth (Fig. 1C and D). Thus, HuH7-F6 and -F12 exhibited enhanced in vivo tolerance to antiangiogenic therapy.

Microarray analysis was performed to compare gene expression patterns between HuH7-F0, -F6, and -F12. The GSEA demonstrated that gene signatures involved in stem cell features; embryonic, neural, and hematopoietic stem cell-like (17) and hepatoblastoma-like (18), as well as proliferative phenotypes; subclass S2 of Hoshida's classification (19) and subclass G2 of Boyault's classification (20) were extremely augmented in the two resistant cancer cells (Fig. 2A), consistent with the recent findings that sorafenib-resistant hepatocellular carcinoma xenografts exhibit the similar molecular characteristics (21). Next, the expression levels of 119 genes were more than 4-fold upregulated in HuH7-F6 cells while those of 2,199 genes in HuH7-F12 (Fig. 2B). Among the genes differentially expressed between the acquired resistant and sensitive cancer cells (Supplementary Table S2), we highlighted Tβ4, X-linked, which is a G-actin monomer binding protein involved in actin–cytoskeleton organization. In the microarray data, Tβ4 was one of the most overexpressed genes in both HuH7-F6 and -F12 (27.7- and 30.7-fold in the microarray data as shown in Supplementary Table S2), which was confirmed by qPCR (Fig. 2C). We also found that the expression level of CD133, a liver progenitor marker, was upregulated in these two sublines (Fig. 2C).

Several previous findings have implicated a distinct epigenetic status in the maintenance of drug resistance (10–15). To determine the transcriptional regulation of Tβ4 by epigenetic mechanism, we initially exposed HuH7 cells to two epigenetic drugs, a demethylating agent 5-aza-2′-deoxycytidine (5-aza-dC) and a histone deacetylase (HDAC) inhibitor trichostatin A (TSA). The Tβ4 expression levels were markedly elevated not only in the cells treated with 5-aza-dC treatment but also cells treated with TSA (Fig. 3A), implying that epigenetic changes could be important to Tβ4 expression in HuH7 cells.

As the human Tβ4 gene contains an extensive CpG island from the promoter region to the exon 2 according to the UCSC database and the Methprimer program (Fig. 3B), we investigated methylation states at the promoter in each HuH7 subclone. In MSP, HuH7-F0 cells harbored dense methylation at the region (Fig. 3C), and the PCR band displayed unmethylated DNA in HuH7-F0 cells with 5-aza-dC treatment (data not shown). However, both methylated and unmethylated patterns were detected at the same locus in HuH7-F6 and -F12 cells (Fig. 3C). Bisulfite sequencing illustrated that the methylation ratio of HuH7-F0, HuH7-F6, and HuH7-F12 were 94.0%, 55.8%, and 45.3%, respectively (Fig. 3D). These findings indicated demethylation at the Tβ4 promoter during the formation of sorafenib resistance.

As histone modification may be associated with Tβ4 expression as well as DNA methylation (Fig. 3A), we examined the methylation status of histone H3, that is, the levels of H3K4 trimethylation (me3), H3K9me3, and H3K27me3 in HuH7-F0 and its clonal cells by using semiquantitative ChIP assay. Among the three regions (CP1, CP2, and CP3) of the Tβ4 gene (Fig. 4A), the levels of H3K4me3 at the CP2 region in HuH7-12 cells were higher than those in HuH7-F0 (Fig. 4B), but the H3K9me3 and H3K27me3 levels at the three regions were not distinct between the two cell lines. The patterns of the H3K27ac levels, which are known to be restored by inhibiting HDAC activity, were very similar to those of the H3K4me3 (Fig. 4B). Quantitative ChIP analysis demonstrated that the H3K4me3 and H3K27 acetylation (ac) levels at the CP2 region in HuH6-F6 and HuH6-F12 cells were significantly increased compared with those in HuH7-F0 cells (Fig. 4C). Thus, active marks of histone modification, H3K4me3 and H3K27ac, were enriched at the promoter region in the drug-refractory subclones with high Tβ4 expression. In addition, the protein levels of H3K4me3 and H3K27ac were more strongly upregulated in HuH-F6 and HuH7-F12 than those in HuH7-F0 (Fig. 4D), implying genome-wide alterations of histone modification.

To investigate the oncogenic role of Tβ4 in hepatocellular carcinoma cells, we used a gain-of-function strategy by using HuH7 cells stably overexpressing Tβ4 (HuH7-Tβ4), which was observed by immunocytochemistry (Fig. 5A). Considering the enrichment of stem cell properties and aggressive phenotypes in the subclones of HuH7 cells refractory to antiangiogenic therapy (Fig. 2A), we characterized biological traits of HuH7-F12 and HuH7-Tβ4 in vitro. HuH7-Tβ4 cells strikingly generated spheres at higher frequency than HuH7-F0, similarly to HuH7-F12 (Fig. 5B; ref. 21). Compared with HuH7-F0, HuH7-F12 and HuH7-Tβ4 showed increased cellular migration and invasion capacities (Fig. 5C and D).

Overexpression of Tβ4 promoted tumor growth in vivo (Fig. 6A), which was immunohistochemically validated in the xenograft of HuH7-F12 and HuH7-Tβ4 (Fig. 6B). We next evaluated the in vivo effects of Tβ4 on resistance to antiangiogenic therapy. Similarly to the sorafenib-resistant F12, HuH7-Tβ4 tumors were refractory to the VEGFR inhibitor, resulting in short survival periods and large tumor volumes (Fig. 6C and D). Sorafenib targeting VEGF pathway could exhibit antiangiogenic activity in the HuH7-F0 tumors, but not in the HuH7-F12 or HuH7-Tβ4 tumors (Fig. 6E and F). Taken together, ectopic expression of Tβ4 in hepatocellular carcinoma cells could mimic malignant transformation undergone in the course of acquired resistance to antiangiogenic drugs.

The clinicopathologic significance of Tβ4 expression was surveyed in 30 patients receiving sorafenib administration in our institute (22). Seven hepatocellular carcinoma samples displayed strong positive staining for Tβ4, whereas 23 did weak or no staining (Fig. 7A), and there was no difference in any clinical factors between the Tβ4-high and Tβ4-low groups (Supplementary Table S3). It is noteworthy that the progression-free survival time of the Tβ4-high was markedly shortened compared with that of the Tβ4-low/negative group (Fig. 7B), suggesting that Tβ4 expression could predict the prognosis of patients treated with antiangiogenic therapy.

We elucidated the mechanism underlying acquired resistance to antiangiogenic therapy by using a xenograft mouse model in this research. Although the study of in vivo effects was necessarily more difficult than in vitro, an understanding of the factors may be critical to the development of strategies to overcome or prevent the emergence of drug tolerance (16). Moreover, because sorafenib and other kinase inhibitors approved for hepatocellular carcinoma mainly target the tumor microenvironment by deregulating VEGF-dependent angiogenesis, only in vivo models recapitulate the process of acquired resistance in the clinic. Two groups have elegantly demonstrated sorafenib resistance in hepatocellular carcinoma by using in vivo models. Rudalska and colleagues performed screening assays with hydrodynamic tail-vein injection of Nras^G12V and a focused shRNA library into p19^Arf-deficient mice, and then identified that silencing of Mapk14 could sensitize mouse hepatocellular carcinoma to sorafenib therapy (23). Establishing hepatocellular carcinoma xenograft mouse models treated with a cycle of sorafenib, Tovar and colleagues compared the biological and molecular properties between resistant and sensitive tumors, and argued that acquisition of tumor-initiating cell-like features and activation of IGF/FGF signaling could cause drug tolerance (12). We here presented a potent strategy of repeated administration, which is more closely similar to therapeutic use for hepatocellular carcinoma than the two prior studies.

Molecular analysis extracted Tβ4 as a gene overexpressed in the hepatocellular carcinoma subcloned cells refractory to multiple VEGFR inhibitors (Fig. 2), and the clinical expression of Tβ4 could predict the prognosis of hepatocellular carcinoma patients treated with sorafenib (Fig. 7). Several previous studies have hinted that Tβ4 might contribute to the development of resistance to these types of anticancer drugs; Tβ4 overexpression accelerated malignant progression including tumor xenograft growth in colorectal cancer cells (24), and knockdown of Tβ4 attenuated sphere-forming capacities in vitro and tumorigenicity in vivo by promoting differentiation in glioma cells (25). Thus, Tβ4 could cause gain of cancer stem cell–like phenotypes, consistent with the current data.

Epigenetic alterations have increasingly been recognized as integral for their possible roles in treatment resistance (11). However, all the anticancer drugs used in these studies target cancer cells, not tumor microenvironment pathways such as angiogenesis, and are administered to cancer cells for the generation of resistant subclones in vitro, not in vivo (12–15). In contrast, we have identified that H3K4me3 and H3K27ac levels were globally elevated in the hepatocellular carcinoma cells surviving under the inhibition of angiogenesis, providing the first evidence that dynamic epigenetic states of cancer cells could be influenced by modulating the tumor microenvironment in vivo. In this work, we particularly focused on epigenetic regulation of Tβ4, and observed that the Tβ4 promoter region in the drug-resistant HuH7 cells was demethylated and enriched with active histone marks H3K4me3 and H3K27ac, that is, exhibited open chromatin states. Cumulative findings indicate that an open chromatin state contributes to maintenance of pluripotency in stem cells and also dedifferentiation with aberrant activation of oncogenes in cancer cells (26), and then were consistent with the current observation of the similar epigenetic alterations during acquisition of stemness and drug resistance in this study. Knoechel and colleagues have recently discovered that a chromatin regulator BRD4 binds acetylated histones linked with open chromatin structures at the enhancers of critical genes like MYC and BCL2 to facilitate the gene transcription in the drug-tolerant persisters of T-cell acute lymphoblastic leukemia (14). They have also reported the intriguing data that a BRD4 inhibitor JQ1 triggered growth arrest and apoptosis specifically in the resistant cells by decreasing core gene expression, which implies that epigenetic drugs antagonizing chromatin regulators recruited at the Tβ4 promoter locus can sensitize the resistant hepatocellular carcinoma cells to antiangiogenic therapy (14).

In summary, we here established a xenograft model of antiangiogenic therapy resistance by long-term exposure with serial transplantation, which closely mimics clinical conditions. This process was accompanied with the gain of cancer stem cell–like features and epigenetic alterations inducing the aberrant expression of Tβ4. Tβ4 was responsible for the drug resistance, and then could be the molecular target as well as the surrogate marker for sorafenib antiangiogenic treatment in patients with advanced hepatocellular carcinoma.

No potential conflicts of interest were disclosed.

Conception and design: Y. Ohata, S. Tanaka

Development of methodology: Y. Ohata, K. Nakao, S. Matsumura

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Ohata, S. Shimada, Y. Akiyama, K. Nakao, M. Tanabe, T. Ochiai, A. Kudo

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Ohata, S. Shimada, Y. Akiyama, K. Mogushi, K. Nakao, S. Tanaka

Writing, review, and/or revision of the manuscript: Y. Ohata, S. Shimada, Y. Akiyama, S. Tanaka

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Shimada, S. Matsumura

Study supervision: S. Tanaka

Other (collection of clinical samples): A. Aihara, Y. Mitsunori, D. Ban, S. Arii

We thank Ms. Hiromi Nagasaki for technical assistance.

This work was supported by Grant-in-Aid for Scientific Research (A), Scientific Research on Innovative Areas, Challenging Exploratory Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan; Research Grant from the Princess Takamatsu Cancer Research Fund; P-DIRECT, P-CREATE, and Research Program on Hepatitis from AMED (Japan Agency for Medical Research and Development).

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.