Advances in hepatocellular carcinoma drug resistance models

Hepatocellular carcinoma (HCC) is the most common primary liver cancer. Surgery has been the major treatment method for HCC owing to HCC’s poor sensitivity to radiotherapy and chemotherapy. However, its effectiveness is limited by postoperative tumour recurrence and metastasis. Systemic therapy is applied to eliminate postoperative residual tumour cells and improve the survival of patients with advanced HCC. Recently, the emergence of various novel targeted and immunotherapeutic drugs has significantly improved the prognosis of advanced HCC. However, targeted and immunological therapies may not always produce complete and long-lasting anti-tumour responses because of tumour heterogeneity and drug resistance. Traditional and patient-derived cell lines or animal models are used to investigate the drug resistance mechanisms of HCC and identify drugs that could reverse the resistance. This study comprehensively reviewed the established methods and applications of in-vivo and in-vitro HCC drug resistance models to further understand the resistance mechanisms in HCC treatment and provide a model basis for possible individualised therapy.


Introduction
Primary liver cancer is a common digestive system malignancy with extremely high rates of incidence and mortality, ranking sixth and fourth, respectively (1).Primary liver cancer includes hepatocellular carcinoma (HCC), intrahepatic cholangiocarcinoma and combined hepatocellular cholangiocarcinoma, with HCC as the most common, accounting for almost 90% of cases (1).Surgery remained the most important treatment method for HCC over the years owing to HCC's poor sensitivity to chemotherapy.However, postoperative patients show a high risk of recurrence or metastasis.Moreover, many HCC patients are diagnosed with advanced tumours and have lost the opportunity for surgery.In recent years, with the gradual popularisation of new treatment techniques, such as radiofrequency, microwave, freezing and TACE, and the development of numerous targeted and immunotherapeutic drugs, the progression-free and overall survival rates of HCC have greatly improved.However, primary and acquired drug resistance to these medications remains the most critical challenge in HCC treatment.This study reviewed original articles about drug resistance of HCC published in the last 5 years.The drug resistance models that were employed are presented, as well as a detailed introduction to some of the major drug resistance mechanisms that were discovered utilising the drug resistance models.Most studies have used traditional commercial HCC cell lines to establish drug resistance models.In-vitro models involve exposing parental HCC cells to medications at either a continuous high or a progressively increasing concentration for 3-6 months or 20-30 generations for the cells to develop drug-specific resistance.For in-vivo drug resistance models, drug-resistant RCC cells can directly be implanted subcutaneously or orthotopically into nude mice.Furthermore, parental RCC cells may be injected subcutaneously or orthotopically into nude mice, followed by prolonged oral drug feeding to acquire drug resistance.Figure 1 presents the establishment methods of traditional drug resistance models.Table 1 presents studies on HCC drug resistance using traditional in-vitro and in-vivo models.Briefly, HepG2, Huh7, SMMC-7721, MHCC97H, MHCC97L, Hep3B, BEL-7402, PLC/PRF/5 and SK-Hep-1 are commonly used to induce acquired drug resistance.Traditional chemotherapy drugs, such as cisplatin, oxaliplatin, 5-FU and doxorubicin, and TKIs, including sorafenib, lenvatinib and regorafenib, are major research topics.Numerous signalling pathways may be involved in resistance to a single agent.A specific drug that could reverse the drug resistanceassociated signalling may present its potential as an alternative or combined treatment.Hepatocellular carcinoma traditional drug resistance models.113) established the Pt-resistant HCC cell line 7404DDP to evaluate the antitumour effect of a cascade targeted and mitochondriondysfunctional nanomedicine (Gal-NP@TPt).When compared to cisplatin, Gal-NP@TPt caused a 9-fold increase in Pt accumulation in 7404DDP cells.Moreover, Gal-NP@TPt caused significant DNA damage in 7404DDP cells.Furthermore, Gal-NP@TPt may mitigate platinum resistance as the ratio of IC50 for 7404DDP to that for 7,404 dropped from 6.34 for cisplatin to 0.71 for Gal-NP@TPt.Kim et al. (133) established secondary sorafenib-resistant HCC cells and cellderived mouse models to investigate the role of SIRT7 in sorafenib resistance.In Huh7SR and SK-Hep1SR cells, hyperactivated pERK1/2 was seen in conjunction with increased SIRT7 expression.In-vivo tumour development was suppressed by inhibiting SIRT7.Mechanically, SIRT7 inhibition eliminates sorafenib resistance by decreasing ERK1/2 phosphorylation via the DDX3X-mediated NLRP3 inflammasome in HCC.

Drug-resistant mechanisms based on traditional models
3 Patient-derived drug resistance models

Establishment methods of patient-derived drug resistance models
The use of patient-derived HCC cell lines and xenograft mouse models has increased in the study of HCC drug resistance.New drugresistant cell lines may be created using primary cultures of HCC tissues from patients who have developed drug resistance.The resected tumour tissue samples may endure tissue digestion, cell separation and purification and primary cell culture to form a stable cell line.Patient-derived xenografts (PDXs) from patients with drug resistance could be directly implanted in nude mice subcutaneously or orthotopically to establish in-vivo drug resistance models.Figure 2 presents the establishment methods of patient-derived drug resistance models.The patient-derived drug resistance models better reserve individual molecular signature, including DNA copy number alterations, mutations and gene expression levels.The patient-derived drug resistance models may have additional advantages over the traditional models in the investigation of drug resistance mechanisms and individualised treatment as they can more accurately imitate the pathophysiological features of individual patients.

Drug-resistant mechanisms based on patient-derived models
Hu et al. (139) established sorafenib-resistant and sorafenibsensitive PDX mouse models and examined the expression pattern differences between them.KPNA3 was found overexpressed in the sorafenib-resistant PDX models and was further confirmed to induce EMT and sorafenib resistance of HCC cells through KPNA3-AKT-ERK-TWIST signalling.Wang et al. (140) used HCC patient-derived organoid models to investigate the functions of Hedgehog signalling and CD44 in sorafenib resistance.In CD44positive HCC, GANT61 dramatically reduced Hedgehog signalling to reverse sorafenib resistance, demonstrating that sorafenib with Hedgehog signalling inhibitors may be a useful therapeutic strategy for HCC patients with elevated CD44 levels.Hashiba et al. (33) revealed a patient-derived sorafenib-resistant HCC cell line (HCC-SR).The single-cell suspensions obtained from post-sorafenib HCC tissues were subcutaneously injected in nonobese diabetic, severe combined immunodeficient (NOD/SCID) mice, and the subcutaneous tumours were harvested to establish the new cell line.Sorafenib treatment induced higher cell proliferation of HCC-SR than other HCC cell lines, showing the resistance of HCC-SR to sorafenib.CIC S1595P missense mutation was detected in HCC-SR by whole-exome sequence analysis, revealing the potential function of this gene on HCC sorafenib resistance.Prawira et al. (141) established infigratinib-resistant PDX mouse models to explore the function of ribociclib in overcoming infigratinib resistance in HCC.SCID mice were subcutaneously implanted with four HCC PDX tumours with high FGFR1-4 expression, administered infigratinib (15 mg/kg per day) and sacrificed when tumours were 1800 mm 3 .Infigratinib-resistant tumours were re-implanted into SCID mice, and nine treatment cycles were required to maintain infigratinib resistance.Then, the infigratinib-resistant PDX mouse models were treated with either infigratinib alone or in combination with ribociclib and sacrificed for subsequent examination.The results showed that the combined inhibition of FGFR/CDK4/6 pathways is efficient in overcoming infigratinib resistance.Xu et al. (62) established four sorafenibresistant HCC cell lines (HepG2-SR, LM3-SR, Huh7-SR, and SKhep1-SR) and cell-derived and patient-derived sorafenibresistant xenograft mouse models to explore the mechanisms of sorafenib resistance in HCC.They found that circRNA-SORE was overexpressed in sorafenib-resistant HCC cells, induced sorafenib resistance and activated the Wnt/β-catenin pathway by sponging miR-660-3p and miR-103a-2-5p.Moreover, The cytoplasmic binding of circRNA-SORE to the master oncogenic protein YBX1 inhibits PRP19-mediated YBX1 degradation by blocking YBX1's nuclear interaction with the E3 ubiquitin ligase PRP19 (142).Moreover, sorafenib resistance could be spread via exosomal circRNA-SORE transport (142).A local injection of circRNA-SORE shRNA lentivirus could significantly enhance the sensitivity of sorafenib treatment in mouse models.Liao et al. (143) investigated the potential of 17-AAG in overcoming sorafenib resistance using secondary sorafenib-resistant PDX mouse models.The HCC tissues from patients were transplanted into the armpit of severe NSG immunodeficient mice.Sorafenib (80 mg/kg) was administered orally to the mice once a day when the tumour size reached 100 mm 3 .The tumours were significantly resistant to sorafenib in the fourth generation.Additionally, 17-AAG suppressed HSP90α and reversed sorafenib resistance in vivo, indicating the potential of 17-AAG in overcoming sorafenib resistance.Leung et al. (144) created two sorafenib-resistant PDX mouse models, which resembled the emergence of sorafenibinduced acquired resistance in patients with HCC.The tumours from two HCC patients were xenografted into the immunodeficient mice, which were then administered with several cycles of  Hepatocellular carcinoma patient derived drug resistance models.

Direct detection of clinical drug-resistant samples from HCC patients
Peripheral blood and tumour tissues from HCC patients may be directly detected for gene expression levels using quantitative or semi-quantitative techniques such as qPCR, western blot or immunohistochemistry.In drug-resistant HCC samples, highly expressed genes are generally more likely to perform as drug resistance genes, whereas suppressed genes may be able to withstand drug resistance.Table 2 presents research on the direct detection of clinical drug-resistant samples from HCC patients.
Circulating tumour DNA (ctDNA) sequencing is a minimally invasive method that enables the collection of repeat samples.Hatlen et al. ( 19) sequenced the ctDNA of acquired fisogatinib-resistant HCC patients in a fisogatinib phase I trial.Two patients with disease progression had mutations in the gatekeeper and hinge-1 residues in the FGFR4 kinase domain.Further, subsequent experiments using in-vivo and in-vitro secondary fisogatinib-resistant models demonstrated that acquired fisogatinib resistance was related to FGFR4 kinase domain mutations.Yu et al. (67) examined the HCC tissues of patients who experienced recurrence after primary HCC resection and sequential sorafenib treatment and found an inverse expression between CYP1A2 and NF-κB p65 in the sorafenib-naïve primary HCC compared with its paired sorafenib-experienced recurrence.Moreover, Weng et al. (73) divided patients who received two cycles of sorafenib treatment into the sorafenib-sensitive and sorafenib-resistant groups.RNA sequencing showed higher circFOXM1 levels in sorafenib-resistant HCC tissues.Functional experiments revealed circFOXM1-induced sorafenib resistance by upregulating MECP2 expression via sponging miR-1324.Ma et al. (87) discovered that transcriptional factor myocyte enhancer factor 2D (MEF2D) was overexpressed in sorafenib-resistant HCC cell lines and HCC specimens, indicating a poor prognosis for sorafenib-treated HCC patients.Mechanically, coupling HDAC4 with MEF2D may activate ERK by inhibiting SPRY4, causing sorafenib resistance of HCC.Wang et al. (160) examined the compared tumour tissues (pretreatment and post-progression samples) of one HCC patient with acquired resistance to a combined treatment of atezolizumab and bevacizumab who subsequently underwent surgical resection of the tumour.The number of CD8+ T cells in the tumour area and PD-L1 level in tumour-infiltrating immune cells were decreased in the drugresistant HCC tissue.Additionally, the drug-resistant tissue presented more progenitor/hepatoblast features in the gene expression profile.The abovementioned results show that the acquired resistance to the combined treatment may be caused by the immune-excluded tumour microenvironment and tumour dedifferentiation.Lu et al. (161) found that circTMEM181 expression was upregulated in puncture biopsies of HCC tissues from anti-PD1 antibody-resistant patients compared to those from anti-PD1-sensitive patients.Furthermore, a high exosomal circTMEM181 level was associated with immunosuppression of microenvironment and anti-PD1 resistance in HCC.Mechanistically, exosomal circTMEM181 promoted CD39 expression by sponging miR-488-3p in macrophages.They further created macrophagespecific CD39 knockout mice and discovered that CD73 expression in HCC cells and CD39 expression in macrophages could impair the function of CD8 + T cells and induce anti-PD1 resistance by activating the eATP-adenosine pathway.Meng et al. (170) discovered that CD10 + ALPL+ neutrophils were more abundant in the tumour tissues of anti-PD-1-resistant patients than in those of anti-PD-1-sensitive patients.Mechanically, tumour cells secreted NAMPT, which reprogrammed CD10 + ALPL+neutrophils via NTRK1, keeping them immature and preventing their maturation and activation.This was how CD10 + ALPL+neutrophils were generated.Immunosuppressive CD10 + ALPL+ neutrophils further mediated ongoing T-cell exhaustion, which increased resistance to anti-PD-1 treatment in HCC.
5 Transgenic drug resistance models

Establishment methods of transgenic drug resistance models
Gene-editing techniques could be used in studies on mechanisms of drug resistance by direct insertion or knockout of drug-resistant genes in cells or animal models.An RNA-guided DNA endonuclease derived from the type II CRISPR bacterial immune system, namely clustered regularly interspaced short palindromic repeats (CRISPR)-associated protein 9 (Cas9), has been extensively employed as a highly effective gene-editing method because of its capacity to target novel genes through the simple modification of single-guide RNA (sgRNA) sequences (176).The precise Watson-Crick base pairing between Cas9's guide RNA and the target DNA region and a direct contact between Cas9 and a short DNA protospacer adjacent  (179,180).A shift in the reading frame caused by random insertions or deletions may result in mutations at the targeted locations (176).Homologous recombination with an introduced homologous donor DNA may be used to complete homology-directed repair (181, 182).

Drug-resistant mechanisms based on transgenic models
The transcription factor called hepatic leukaemia factor (HLF) is a member of the proline and acidic amino acid-rich family, which are known to regulate circadian rhythms (183).Xiang et al. (184) constructed an HLF-knockout mouse model (HLFPB/PB mouse) and found that sorafenib resistance requires HLF-mediated c-Jun activation, demonstrating that the HLF-c-Jun axis may control the haepatoma's response to sorafenib.A piggyBac transposonencoding CAG-RFP (a red fluorescent protein sequence under the CAG promoter) was inserted into the HLF gene to create the HLFPB/PB mice.Diethylnitrosamine was injected into the abdomen of mice to induce haepatoma development.The mice were then euthanised 7, 8 or 9 months later.Haepatoma cells that overexpressed HLF became resistant to growth inhibition and cell death caused by sorafenib.The c-Jun interference eliminated sorafenib resistance in haepatoma cells overexpressing HLF.Furthermore, HLF interference weakened c-Jun activation due to sorafenib and made haepatoma cells more susceptible to sorafenib.Wei et al. (185) found phosphoglycerate dehydrogenase (PHGDH) to be a critical sorafenib-resistance gene by genomewide CRISPR/Cas9 library screening.The Human GeCKOv2A CRISPR knockout pooled library contains 65,386 sgRNAs targeting 19,052 human genes and 1864 miRNAs.They constructed a stable Cas9-expressing HCC cell line (MHCC97L-Cas9) by lentiviral transfection and transfected it with the GeCKOv2A library.Mutant cells were selected by puromycin and treated with sorafenib and DMSO for a week and were collected.The sgRNAs were amplified by PCR for subsequent high-throughput sequencing and bioinformatic analysis.PHGDH targeting sgRNAs were negatively selected after sorafenib treatment.They further established CRISPR/Cas9 knockout and RNAi knockdown cell models and found that inhibition of PHGDH promotes HCC apoptosis by suppressing the synthesis pathway upon sorafenib treatment.Sueangoen et al. ( 186) developed transgenic cell models to evaluate the functions of seven HCC-derived EGFR mutations on erlotinib resistance.Seven missense mutations (K757E, N808S, R831C, V897A, P937L, T940A and M947T) in the kinase domain of EGFR were recognised and retrovirally transducted into NIH-3 T3 cells.T790M and L858R were used as erlotinib-resistant and erlotinibsensitive mutant controls, respectively.In vitro experiments showed that the seven EGFR mutants are dependent on EGF and resistant to erlotinib.Sofer et al. (187) performed a genome-wide CRISPR/ Cas9 activation screen in Huh7 and recognised hexokinase 1 as a critical factor in promoting regorafenib resistance in HCC.Moreover, Chen et al. recognised FGF21 as a sorafenibresistant gene in HCC by CRISPR/CAS9 genome library screening.Mechanically, FGF21 induces sorafenib resistance by directly combining with NRF2 to prevent NRF2 ubiquitination degradation.Huang et al. (188) used CRISPR/CAS9 genome library screening and discovered that lenvatinib resistance was mediated by DUSP4 deficiency through the activation of MAPK/ERK signalling.Another gene associated with lenvatinib resistance, LAPTM5, was identified by Pan et al. (189).MiR-3689a-3p was found to be the most upregulated miRNA in sorafenib-sensitive HCC by CRISPR/ Cas9 screens (190).Mechanically, miR-3689a-3p may suppress CCS/SOD1-dependent mitochondrial oxidative stress to regulate sorafenib resistance.Zhu et al. (191) used a transgenic HCC mouse model driven by MYC overexpression and beta-catenin (encoded by CTNNB1) activation (termed MYC-lucOS; CTNNB1) (192), which shows resistance to anti-PD-1 immunotherapy, to evaluate the effect of the combination of anti-PD-L1 immunotherapy and anti-VEGF.After 4 weeks of the combined treatment, a significant survival improvement and a decline in the proportion of mice with tumours were observed in this model, indicating the potential of the anti-PD-L1/anti-VEGF dual treatment in overcoming resistance to either of a single agent.Martin et al. (193) used CRISPR/Cas9 to directionally knock out tumour suppressor genes in HCC cells.The p53-knockout HepG2 cells performed increased malignant properties and multidrug resistance to cisplatin, regorafenib, sorafenib and doxorubicin.Alb-R26Met mice carry a conditional mouse-human chimeric Met transgene into the Rosa26 locus.The Alb-R26Met HCC is resistant to sorafenib.ADAMTSL5 overexpression was noted at the early stages of liver carcinogenesis, and its upregulation was reproduced in the Alb-R26Met HCC model.Various oncogenic inputs associated with HCC decreased as a result of ADAMTSL5 abrogation, such as MET, EGFR, PDGFRβ, IGF1Rβ and FGFR4 receptor tyrosine kinases, which were all expressed and/or phosphorylated to a lesser extent, showing the potential role of ADAMTSL5 in HCC drug resistance (194).

Summary
To further understand drug resistance mechanisms in HCC, useful models must be developed.Different kinds of models have their own advantages and disadvantages.Traditional drug resistance models are easier to establish.Meanwhile, the corresponding experimental results have higher stability and reproducibility.However, individual differences in HCC gene expression could not be reflected in a single cell line.Patientderived models maintain more individual traits, which are essential for studying the various drug resistance pathways related to distinct clinical subtypes of cancer.Despite the difficulty of the construction process, patient-derived models may be a better choice to investigate the drug resistance mechanisms of various malignancies.Sometimes a direct detection of clinical drugresistant samples may be a simpler method to screen for drugresistant genes.Gene-editing methodologies can be employed to generate genetically engineered cell lines or animal models that exhibit resistance to particular pharmaceutical agents.Theoretically, a gene-editing cell line may be an ideal model to investigate the function of a specific gene on drug resistance.It has to be admitted that none of the aforementioned techniques is flawless.In the actual research, we should choose the appropriate FIGURE 1 sorafenib treatment to acquire sorafenib resistance.Two sorafenibresistant HCC cell lines were further developed from the above two PDX models, with stronger self-renewal and tumourigenicity.RNA-sequencing recognised EPH receptor B2 (EPHB2) as the most significantly upregulated kinase in sorafenib-resistant PDXs.Functional experiments demonstrated that EPHB2 may increase tumour stemness and induce sorafenib resistance through the EPHB2/β-catenin/TCF1 positive feedback loop.Prawira et al. (145) established seven acquired infigratinib-resistant PDX models and developed a new infigratinib-resistant HCC cell line from one of these PDXs.Infigratinib-resistant tumours presented higher p-ErbB2, p-ErbB3 and EZH2 levels.Mechanically, EZH2 may promote infigratinib resistance by upregulating the ErbB family.Gao et al. (84) found high USP29, HIF1α and GLUT1 levels in sorafenib-resistant PDX tumours, and follow-up research revealed that USP29-induced sorafenib resistance by mediating HIF1α stabilisation and upregulated glycolysis.Mok et al. (146) used two drug-resistant tumour xenografts derived from HCC patients to investigate driving resistance and CSC repopulation in HCC.The xenografts mimicked the development of acquired resistance to sorafenib or lenvatinib treatment observed in HCC patients.RNA sequencing showed that cholesterol production was most

TABLE 1
Researches on HCC drug resistance using traditional in-vitro and in-vivo models.

TABLE 1 (
Continued) (112)to investigate the effect of ribociclib on the treatment of sorafenibresistant HCC.HepG2 cells that were continuously incubated with sorafenib at escalating concentrations for 8 months, up to a final sorafenib concentration of 4 μM, were used as sorafenib-resistant cells and maintained in the medium with 4 μM sorafenib.They discovered that ribociclib reduced Rb expression and induced G1 cell cycle arrest in SR-HepG2 cells with Rb-high/p16-low protein expression profiles, indicating that ribociclib may be a good choice for the treatment of certain sorafenib-resistant HCC.Qiu et al.(13)established sorafenibresistant HuH7 cells and xenograft mouse models and examined a group of patients with advanced, recurrent HCC to evaluate the clinical significance of sorafenib therapy.They found that high 14-3-3η expression and low miR-16 expression were related to sorafenib resistance and poor prognosis of HCC.Moreover, Chen et al.(36)initiated secondary sorafenib-resistant HuH-7 cells and xenograft mouse models.EphA2 was recognised as a crucial molecule in sorafenib resistance by quantitative phosphoproteomic analysis.It has been confirmed in in vivo animal models that sorafenib resistance can be successfully treated by concurrently inhibiting EphA2.Yang et al.(63)established sorafenib-resistant HepG2 cells and analysed the expression differences of circRNAs between sorafenib-sensitive and sorafenib-resistant HepG2 cells.CircFN1 was upregulated in sorafenib-resistant HepG2 cells and induced sorafenib resistance through the miR-1205/E2F1 signalling pathway.Zhao et al.(70)created sorafenib-resistant in-vitro and in-vivo models to explore the role of CBX4 in sorafenib resistance of HCC.CBX4 was upregulated in sorafenib-resistant Huh7 and PLC cells.Tumour growth could be suppressed by CA3-and UNC3866-mediated YAP1 and CBX4 inhibition in vivo.Younis et al.(71)designed an ultra-small lipid nanoparticle encapsulating sorafenib and midkine-siRNA and examined the effect of the new nanoparticles on treating sorafenibresistant HCC with sorafenib-resistant cell-derived (HepG2) xenograft mouse models.The tumours in the xenograft models were eradicated by 70% using the nanoparticles, demonstrating the potential of this new approach in HCC treatment with sorafenib resistance.Xu et al.(79)constructed sorafenib-resistant in-vitro and in-vivo models to explore the role of UBQLN1 in HCC sorafenib resistance.For the sorafenib-resistant HCC mouse model, 100 million HCCLM3 cells were initially implanted into the flank of a BALB/c mouse.The formed tumours were re-implanted into 4-week-old BALB/c nude mice and fed with sorafenib (30 mg/kg/day).The mice that survived after 8 weeks of treatment were regarded as sorafenib-resistant mice.The ROS levels decreased in sorafenib-resistant HCC cells.However, sorafenib-resistant cells have better mitochondrial function and integrity with less mitochondrial content and respiratory capacity.Huh7-R cells.Mechanically, the knockdown of ZFAS1 caused the downregulation of various mRNAs related to stemness and notch signalling pathways, indicating the critical role of this noncoding RNA in HCC sorafenib resistance.Huang et al.(112)created lenvatinibresistant HCC cell lines and cell-derived xenograft mouse models.The two essential parts of the tRNA N7-methylguanosine (m7G)

TABLE 2
Researches on the direct detection of clinical drug-resistant samples from HCC patients.