Abstract
Background. Novel treatments such as monotherapy and combined immunotherapy significantly extend overall survival (OS) for hepatocellular carcinoma (HCC) patients, but HCC is susceptible to treatment resistance during long-term therapy. The resistance mechanism to targeted drugs in HCC remains ambiguous, making research on HCC drug resistance targets crucial for the development of precision medicine.
Objectives. To investigate the transcriptional features, biological functions and potential clinical value of the tyrosine kinase inhibitor (TKI)-resistant gene ZNF687 in HCC.
Materials and methods. The TKI-resistant genes of HCC were identified using clustered regularly interspaced short palindromic repeats (CRISPR) in vitro screening. Then, the dependence of HCC cell lines on ZNF687 was investigated in silico. We collected global mRNA datasets of HCC tissue, integrated the mRNA expression characteristics of ZNF687 in HCC and explored the impact of ZNF687 on HCC patient prognoses using the Kaplan–Meier method (in silico). The Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways analyses were then conducted, and a connectivity map and molecular docking technology were applied to find the underlying agent opposing ZNF687.
Results. In vitro, the guide RNA corresponding to ZNF687 was weakly detected in HCC cells, and ZNF687 deficiency was found to inhibit growth in HCC cell lines. ZNF687 mRNA was overexpressed and had a high discriminatory ability for HCC in 2,975 HCC samples, contrasting with 2,340 non-HCC samples. Moreover, an excessive ZNF687 transcript level was related to a worse overall survival (OS) prognosis. Histone modification, spliceosome, transcription coregulator activity, and nucleocytoplasmic transport were the most significant pathways for ZNF687 differential-related gene enrichment. Chaetocin was found to be a candidate compound and presented a strong affinity to ZNF687.
Conclusions. ZNF687 represents a TKI-resistant and growth-dependent gene for HCC, the overexpression of which indicates poor OS for HCC patients. Additionally, ZNF687 is expected to be a druggable target for overcoming TKI resistance, and chaetocin may be a candidate therapeutic compound for ZNF687.
Key words: tyrosine kinase inhibitor, resistance, mRNA, hepatocellular carcinoma, ZNF687
Background
According to statistics from the American Cancer Society (ACC), the mortality rate of liver cancer has fallen compared to previous decades. However, the burden of liver cancer is still heavy, with the estimated number of deaths ranking 5th among all cancer deaths in men and 7th in women.1 Generally, hepatocellular carcinoma (HCC) dominates liver cancer cases (75–85%),2 and the risk factors include hepatitis virus infection, alcohol and aflatoxin, as well as water contamination.3, 4 A lack of specific clinical signs in the early stages of HCC causes many diagnoses to be delayed until an advanced stage. With early diagnosis, liver transplantation and surgical resection are the recommended therapy choices. Because the complete resection of pathological tissue is difficult, patients are at risk of tumor recurrence, metastasis, hemorrhage, infection, and abdominal wall hernia.5, 6 Advanced HCC is common in clinical practice and has a poor prognosis for many complications, such as serious ascites, jaundice, hemorrhage, and hepatic encephalopathy. For instance, the prognosis of advanced HCC combined with obstructive jaundice is poor, and although endoscopic biliary drainage may improve patient outcomes, the risk of cholangitis increases.7, 8 Currently, systematic therapy, especially tyrosine kinase inhibitors (TKI), is suggested for advanced HCC patients who are classed as A class using Child–Pugh score for hepatic function.9 Due to the disorder of protein kinase activity in many malignancies, targeting protein kinases has become a significant anti-cancer strategy. Tyrosine kinase inhibitor is one of the U.S. Food and Drug Administration (FDA)-approved protein kinase inhibitors, which occupy an important position in targeted therapy.10 Monotherapy or combined immunotherapy significantly extends HCC patients’ survival times. Nevertheless, HCC is susceptible to treatment resistance during long-term medication therapy,11 leading to relapse and disease progression.12 The problem is that TKI resistance occurs in the late stage of treatment, which limits long-term therapeutic benefits.13 Regrettably, the resistance mechanisms of targeted drugs in HCC have not been completely elucidated except for epithelial–mesenchymal transition, ATP-binding cassette transporters, hypoxia, autophagy, and angiogenesis.14 Therefore, research on the mechanism and therapeutic target of HCC drug resistance is crucial for improving treatment response, reducing complications, and thus enhancing long-term efficacy.
Clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated (Cas) is an RNA-guided and Cas nuclease-cleaved gene-editing technique that can modify a single gene and reveal its function.15, 16 There are 3 groups of CRISPR/Cas systems, with CRISPR/Cas9 belonging to the 2nd category.17 CRISPR/Cas9 comprises the Cas9 nuclease and CRISPR RNA or gRNA, which can lead Cas9 to specific sites and cut DNA for gene knockout.18 Genome-wide CRISPR knockout screening (GeCKO) technology offers an effective method for observing genomic alterations under certain conditions. During the CRISPR loss-of-function screen, gRNAs are randomly carried into various cells by lentiviral vectors containing Cas9 and puromycin resistance. Cells that are not effectively transfected by the lentivirus are eliminated by puromycin, and a collection of genome-wide mutant HCC cells is created. The genomic expression profile of mutant cells changes under a particular intervention, and high-throughput sequencing is utilized to detect variations between treatment and control groups.19, 20 In this work, we conducted an in vitro CRISPR screen experiment for TKI resistance in HCC cells and gathered genes instead of TKIs. ZNF687 was identified as a TKI-resistant gene and is anticipated to be a therapeutic target for HCC.
Located at chromosome 1q21.3,21 ZNF687 is a recently discovered C2H2-type zinc finger protein that has been reported to be overexpressed in the kidneys, spleen and other hemopoietic organs and to be associated with the proliferation and differentiation of hematopoietic cells.22 Aberrant ZNF687 expression is a driver of some cancers. For instance, ZNF687 mutations are implicated in Paget’s disease of bone, and their overexpression is linked to giant cell tumors associated with Paget’s disease of bone.21, 23, 24 It is now suspected that ZNF687 may induce HCC cells to produce stem cell-like characteristics by upregulating BMI1, NANOG and OCT4, which then contributes to HCC progression. In vitro experiments have demonstrated that ZNF687 knockdown increases the susceptibility of HCC cells to cisplatin; thus, ZNF687 may be engaged in the development of HCC chemoresistance.22 However, no studies have discovered the mechanism of ZNF687 that results in tumor-targeted drug resistance.
Consequently, this is the first comprehensive research to explore the expression status, TKI therapy response, biology, and clinical implications of ZNF687 in HCC. We hypothesized that ZNF687 may be a candidate gene for TKI resistance in HCC, estimated the mean expression level of the ZNF687 gene using global HCC cohorts, and analyzed its capacity to distinguish HCC tissue from controls. Survival curves and clinicopathological characteristics analysis were applied to investigate the association between ZNF687 transcriptional-level expression and patient prognosis. Moreover, we identified the potential molecular mechanism and predicted the underlying therapeutic component based on the biological abnormalities resulting from increased ZNF687 expression.
Objectives
This study was designed to investigate the transcriptional expression features, biological functions and potential clinical value of the TKI-resistant gene ZNF687 in HCC, exploring the probability of overcoming the TKI-resistance problem.
Materials and methods
In vitro genome-wide CRISPR/Cas9 knockout library
Experimental material
The experiment followed the protocol described in the research of Joung et al.25 The human HCC cell line Huh7 was acquired from the cell bank of the Chinese Academy of Sciences (Beijing, China). The genome-wide CRISPR knockout v2 (GeCKO v2) library and gRNA were obtained from the Addgene Corporation (Watertown, USA; https://www.addgene.org/pooled-library/zhang-human-gecko-v2; 1000000048).
Lentivirus transfection
When cell growth reached 70–90% confluence, the adherent cells were separated with trypsin and counted. In total, 1.10×108 cells were transfected with the GeCKO v2 library containing 65,386 specific gRNAs. The multiplicity of infection was controlled to be <0.3.
TKI intervention and DNA extraction
Hepatocellular carcinoma cells successfully infected with lentivirus were dosed with the TKI drug for 21 days. The TKI component we adopted, anlotinib, was sourced from the Jiangsu Zheng Da Tian Qing company (Nanjing, China). Afterward, we used a Quick-DNA Midiprep Plus Kit (D4075) developed by Zymo Research (Orange, USA) to extract the DNA of the surviving HCC cells for library construction and high-throughput sequencing.
Library construction
The library construction was completed by the Beijing Nuo He Zhi Yuan Technology Company (Beijing, China), according to the following process. The extracted DNA was randomly digested into 350 bp fragments using a Covaris breaker (Covaris, Woburn, USA). The library was then prepared after conducting terminal repair and adding a poly-A-tail, and index connectors sequentially to the DNA fragments.
Analysis of sequencing results
and negative screening strategies
Two treated samples and 2 control samples were subjected to high-throughput sequencing by the Beijing Nuo He Zhi Yuan Technology Company. After connector data, unknown data and low-quality test data were removed and quality control was conducted using FastQC software, high-quality gene counts were included. Model-based analysis of genome-wide CRISPR/Cas9 knockout software was employed for positive and negative screening. Both analysis tools were employed with default parameters, and genes corresponding to significantly reduced gRNAs were considered potential TKI-resistant genes.19
Gene effect
The gene effect can reflect the necessity of a specific gene in various cell lines. We downloaded the data for “CRISPR (DepMap 22Q2 Public + Score Chronos)” from the Cancer Dependency Map website (DepMap; https://depmap.org/portal/). The “gene effect” values of each cancer cell line after ZNF687 knockout or inhibition were extracted, and a scatterplot was drawn using the ggplot2 package. The effects of ZNF687 deficiency on the growth vitality of 20 HCC cell lines were then investigated.
Collection strategy of mRNA datasets
Microarray and RNA sequencing datasets were retrieved and screened from multiple databases, including the Gene Expression Omnibus (GEO) (https://www.ncbi.nlm.nih.gov/geo/), ArrayExpress (https://www.ebi.ac.uk/arrayexpress/), Oncomine (https://www.oncomine.com/), The Cancer Genome Atlas (TCGA) (https://portal.gdc.cancer.gov/), and Genotype-Tissue Expression (https://www.gtexportal.org/) on March 1, 2023, using the following search terms: (malignancy OR cancer OR tumor OR neoplasia OR carcinoma OR carcinomatosis) AND (hepatocellular OR liver OR hepatic OR HCC). The datasets were included if they met the following criteria, namely: 1) the species was homo sapiens; 2) there were tissue samples; and 3) there were at least n = 3 HCC and noncancerous tissue samples. Simultaneously, samples that had been genetically modified or treated with drugs were excluded. The screened data were then further processed, and datasets from the same platform were merged and eliminated using batch effect, and unstandardized datasets were normalized with the log2(x + 1) method.
Combined analysis in silico
We integrated the 35 public datasets and computed the standardized mean difference (SMD) to compare the ZNF687 mRNA expression differences between HCC tissue and noncancerous tissue. For each dataset, the sample number, average expression level and standard deviation (SD) of ZNF687 in HCC tissue and noncancerous tissue were listed. The I2 index and Q test were used to examine the overall heterogeneity of the data, with I2 > 50% or p < 0.10 illustrating obvious heterogeneity, for which the random-effect model was chosen. Otherwise, the fixed-effect model was used to combine the SMDs. To appraise the differential diagnostic significance of ZNF687 for HCC, a summary receiver operating characteristic (SROC) curve was created. Sensitivity, specificity and likelihood ratios were used to assess the effectiveness of the diagnostic test, and publication bias was detected using the Egger’s test.
Prognostic analysis in silico
The clinicopathological information of 333 HCC patients was downloaded from TCGA, and the clinical value of ZNF687 mRNA expression levels on HCC patient outcomes was investigated using univariate Cox regression with clinicopathological features. A Schoenfeld residual was analyzed for proportional hazards (PH) assumption, and a Martingale residual was analyzed to detect whether the log-hazard function was linearly related to the continuous variable ZNF687 expression. McFadden’s pseudo-R2 was employed to determine the goodnes-of-fit, with a McFadden’s pseudo-R2 closer to 1 indicating better goodness-of-fit. During data cleaning, clinical parameters with more uncertain information (e.g., “Unknown,” “unreport,” “Tx,” “Nx,” and “Mx”) were eliminated. For the identification of overall survival (OS), if a patient’s survival status was “alive”, the survival time was selected as the last follow-up time. In the event that a patient’s survival status was “dead”, the OS was selected as the time of death. Finally, the impact of each clinicopathological parameter on HCC patient outcomes was assessed by independently calculating the hazard ratios (HRs) of ZNF687 expression, age, sex, race, primary T (pT) stage, and American Joint Committee on Cancer (AJCC) stage. Additionally, the survminer R package was utilized to identify the optimum cutoff value, which divided the 333 samples into a high-expression ZNF687 group and a low-expression ZNF687 group, and the relationship between ZNF687 mRNA status and HCC patient prognosis was explored through the Kaplan–Meier method. The computational formula was as follows:
McFadden’s R2 =
= [(likelihood (null) – likelihood (model))/
/likelihood (null)]
Biological pathway exploration
An expression profile of 371 HCC tissue samples and 50 control samples (data source: TCGA) was executed using the limma package in R 4.1.1 (R Foundation for Statistical Computing, Vienna, Austria). The genes with a log fold change (logFC) of >1 were defined as differentially high expression genes of HCC. Genes with a strongly positive relationship to ZNF687 (Pearson’s correlation coefficient >0.75) were acquired using Pearson’s analysis.26 ZNF687-related differential genes were obtained by intersecting differential highly expressed genes and ZNF687 strongly positively correlated genes, which were used for enrichment analysis of the Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways. The STRING tool v. 11.5 (https://cn.string-db.org/) was employed to calculate the connection score between ZNF687-related differential genes, and the potential protein complex subnetworks were analyzed using the molecular complex detection (MCODE) algorithm with Cytoscape software (https://cytoscape.org).
Candidate drugs prediction
and evaluation in silico
A connectivity map (CMAP) is a query tool for predicting candidate drugs by comparing similarities or dissimilarities between the reference perturbation signatures and the input gene set.27 In this study, due to the limitation of no more than 150 genes in the CMAP tool for drug prediction, the ZNF687 genes positively related (Pearson’s coefficient >0.80) and simultaneously differentially upregulated in HCC were inputted to CMAP (https://clue.io/query) to predict the compounds opposing ZNF687. The two-dimensional (2D) structure of each molecular drug was downloaded from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/), and energy was minimized via Chem3D software (PerkinElmer Informatics, Inc, Waltham, USA). The AlphaFold structure of ZNF687 (ID: Q8N1G0) was downloaded from the UniProt database (https://www.uniprot.org/). Subsequently, PYMOL 2.5.2 (https://pymol.org) and Autodock Vina 1.1.2 (https://autodock.scripps.edu) were used to simulate the docking of each candidate compound, and the molecular structure of ZNF687.28 Discovery Studio v. 4.5 software (https://www.3ds.com/products/biovia/discovery-studio) was used to visualize the docking results.
Statistical analyses
R 4.1.1, Stata v. 17 (StataCorp LLC., College Station, USA) and IBM SPSS v. 26 (IBM Corp., Armonk, USA) were used for statistical analyses. Standardized mean difference was calculated as an effect indicator to reveal the mRNA expression status of ZNF687 in HCC tissue. The I2 index and Q test were used to examine the overall heterogeneity of the data, and sensitivity analysis and meta-regression analysis were used to explore the sources of heterogeneity. The summary receiver operating characteristic (SROC) curve, sensitivity, specificity, and positive and negative likelihood ratios were analyzed to evaluate the discrimination efficiency of ZNF687 for HCC. For the area under the SROC curve (AUC), the criteria for verifying its efficacy were as follows: 0.50−0.70 indicated low estimated capacity, 0.70−0.80 indicated moderate estimated capacity, 0.80−0.90 indicated good, estimated capacity, and an AUC > 0.90 indicated strong estimated capacity. The Egger’s test was utilized for detecting publication bias. The univariate Cox regression and Kaplan–Meier methods were applied to explore the prognostic risk factors. The analysis was considered statistically significant at a value of p < 0.05.
Results
CRISPR positive and negative screening in vitro
Under the positive and negative screening results (Figure 1), 904 lowly enriched gRNAs (logFC < 0) and 949 gRNAs highly enriched gRNAs (logFC > 0) were identified (t-test with a p < 0.05). Compared with the control group, gRNA corresponding to ZNF687 was weakly detected in the whole-genome mutant HCC cells treated with TKI (logFC = –0.56, t-test with a p = 0.048), which suggested that for ZNF687-defective HCC cells, they were more sensitive to TKI intervention and more likely to die.
Carcinogenic effect of ZNF687
The gene effect of ZNF687 was less than 0 in 19 HCC cell lines (SNU182, PLCPRF5, JHH2, JHH4, JHH6, SNU387, HLF, SKHEP1, SNU423, JHH5, SNU761, SNU398, JHH7, SNU475, JHH1, HUH7, SNU886, HUH1, and SNU449) and greater than 0 only in the HEPG2 cell line (Figure 2, Supplementary Table 1). Thus, the high expression of ZNF687 likely promoted HCC cell growth.
Overexpression of ZNF687 and its discriminatory efficacy against HCC
In total, 35 mRNA datasets were collected (Figure 3), including 2,975 HCC tissue samples and 2,340 noncancerous tissue samples. The ZNF687 transcriptional expression in each dataset, displayed as mean and SD, is shown in Supplementary Table 2. The SMD was greater than 0 in most separate datasets, and the combined SMD was 1.10 (95% confidence interval (95% CI): 0.87–1.33), indicating that ZNF687 was upregulated at the transcriptional level in large HCC samples (Figure 4A). The results of the I2 index and Q test suggested high heterogeneity (I2 = 90.4%, Q test with a p < 0.01), so the random-effect model was applied. For the investigation of heterogeneity, sensitivity analysis indicated that the pooled SMD was greater than 0 and stable after removing any of the datasets (Supplementary Fig. 1). Based on the result of the meta-regression analysis, the sequencing technique contributed to the heterogeneity (p < 0.01), but neither the merged datasets nor the sample size demonstrated any relationship to the heterogeneity (Supplementary Table 3). The AUC was 0.87 (95% CI: 0.83–0.89) (Figure 4B), the sensitivity was 0.71 (95% CI: 0.64–0.77) (Figure 4B, Figure 5) and the specificity was 0.88 (95% CI: 0.84–0.89) (Figure 4B, Figure 5). The positive and negative likelihood ratios were respectively 6.09 (95% CI: 4.25–8.71) (Figure 6) and 0.33 (95% CI: 0.27–0.41) (Figure 6), indicating a high discrimination capacity of ZNF687 for HCC. Finally, no compelling publication bias was found according to the Egger’s test (p = 0.12) (Figure 4C).
Clinical significance
Based on the Kaplan–Meier curves, OS time appeared more depressed in the ZNF687-overexpression group (288 samples) than in the ZNF687 underexpression group (45 samples), with an HR of 2.00 (95% CI: 1.09–3.69, log-rank test with a p < 0.05) (Figure 7). This illustrates that elevated mRNA expression of ZNF687 may be connected with poor prognosis in HCC. According to the Supplementary Fig. 2, the Schoenfeld residuals of “ZNF687 (continuous),” “Age,” “Gender,” “pT stage,” and “AJCC stage” show no relationship with time and comply with the PH assumption (PH test, p > 0.05), but the covariate “Race” does not comply with the PH assumption (PH test, p < 0.05). The predictor ZNF687 (continuous) shows no linearity to the log hazard (Supplementary Fig. 3). In line with univariate Cox analysis (Table 1), individuals with a high ZNF687 expression level had a higher risk of HCC death (HR = 1.01, 95% CI: 1.00–1.02, Wald test with a p = 0.04, McFadden’s pseudo R2 < 0.01). The probability of death in “T3 + T4” and “III + IV” HCC patients was higher than that in “T1 + T2” and “I + II” patients, with HRs of 2.47 (95% CI: 1.68–3.61, Wald test with a p < 0.01, McFadden’s pseudo R2 = 0.02) and 2.45 (95% CI: 1.67–3.58, Wald test with a p < 0.01, McFadden’s pseudo R2 = 0.02), respectively. Age and gender were not statistically significant regarding the HCC outcome (Wald test with a p > 0.05, McFadden’s pseudo R2 < 0.01).
GO and KEGG functional pathways
We intersected 2,705 differentially expressed HCC genes and 401 strongly positive ZNF687-related genes, of which 214 ZNF687-related differential genes were obtained and considered to have an oncogenic effect in HCC cooperating with ZNF687. The GO analysis showed that the 214 ZNF687-related differentially expressed genes were significantly enriched for “histone modification” of the biological process (BP) and “nuclear speak” of the cellular component (CC), and had “transcription coregulator activity” of molecular function (MF). For the KEGG pathway analysis, “nucleocytoplasmic transport” and “spliceosome” predominated (Figure 8). Six protein-protein interaction networks were listed through the MCODE algorithm (Table 2). The GO-BP analysis was performed on the network with the highest score, and it was noted that the gene within MCODE-Cluster1 was significantly related to the regulation of mRNA processing (Figure 9).
Therapeutic compound opposing ZNF687
In total, 84 genes were input into CMAP, and the top 3 compounds (amiloride, chaetocin and phloretin) were identified (Table 3). As depicted in Figure 10, the minimum binding energies of amiloride (PubChem CID: 16231), chaetocin (PubChem CID: 11657687) and phloretin (PubChem CID: 4788) docking ZNF687 protein were correspondingly –5.9 kcal/mol, –8.9 kcal/mol and –6.5 kcal/mol, of which chaetocin exhibited the highest affinity to the ZNF687 protein. The enlarged three-dimensional (3D) structure and the 2D interactions of the binding site are presented in Figure 10B. Based on the docking result of chaetocin and the ZNF687 protein, the small compound was predicted to form 8 hydrogen bonds with the amino acid residues SER704, ASN705, ALA702, ALA699, GLY701, LEU714, and PRO715, 3 hydrophobic bonds with amino acid residues MET713, PRO698 and LEU695, and an unfavorable acceptor–acceptor interaction with the ANA699 residue.
Discussion
Since 2007, TKI has dramatically improved the treatment of HCC,29 yet according to a study from 2008, acquired resistance to TKI occurs within 6 months after using TKI drug.30, 31 It is reported that patients with sorafenib resistance had worse OS.32 Recently, 2nd- and 3rd-generation TKI have been developed to treat TKI-resistant patients,33, 34 but they are inadequate for overcoming the difficulty of TKI resistance. The CRISPR screening is a genome-wide editing technology extensively applied in tumor drug-resistance research. CRISPR knockout, CRISPR inhibition and CRISPR activation screens are 3 common methods to explore the drug-resistance mechanism and identify responsible genes.35 In this study, we conducted an in vitro CRISPR knockout screen and confirmed a potential TKI-resistant gene, ZNF687. Because ZNF687 has been reported as overexpressed in HCC, we also collected global cohorts to demonstrate it is an oncogene and is expected to be a druggable target opposing TKI resistance.
In the in vitro study, the gRNA of ZNF687 was enriched only to a low level, indicating that ZNF687 knockout likely diminished TKI resistance for HCC cells. This is not the first time a zinc finger protein has been revealed as participating in TKI resistance. In 2020, zinc finger protein 703 was found to induce sorafenib resistance via transactivating CLDN4 expression.36 However, the underlying mechanism of TKI resistance is unclear, except for presently known processes, including epithelial–mesenchymal transition, ATP-binding cassette transporters, hypoxia, autophagy, and angiogenesis.14 Considering that no directly relevant evidence followed, we further explored the potential molecular mechanism of ZNF687 in HCC. ZNF687-related differentially expressed genes were enriched for histone modification, nucleocytoplasmic transport, spliceosome, and transcription coregulator activity pathways. Histone modification is essential to resistance, with histone modification inhibition, such as histone methylase inhibitors, being demonstrated to reverse tumor drug resistance.37 Additionally, for nucleocytoplasmic transport, it is well known that cancer cells can escape antitumor attacks through the normal nuclear-cytoplasmic transport process or nuclear pore complex. For example, the transport receptor protein CRM1 can mediate drug-target proteins exported from the nucleus. Consequently, antitumor pharmaceuticals cannot take effect in the nucleus, facilitating resistance.38 Transcriptional coregulator activity and spliceosome pathways are also implicated in the gene transcription and transcript modification processes. Using the MCODE algorithm and GO enrichment analysis, we showed that ZNF687-related differential genes were significantly enriched for the regulation of mRNA processing and splicing. In recent decades, it has been shown that, apart from gene mutations, mRNA alterations are crucial for the occurrence and progression of tumors. Aberrant splicing and polyadenylation of mRNAs are connected with resistance to antitumor therapy, and certain tumors are highly sensitive to components that inhibit splicing.39 To overcome tumorigenesis and drug resistance caused by abnormal RNA splicing, research on and improvement of splice variant-specific siRNAs, splice-switching antisense oligonucleotides, and small molecule inhibitors aimed at splicing factors, splicing factor kinases, and aberrant carcinogenic protein isoforms recently been proposed.40 Among the 5 genes found to be participating in the regulation of the mRNA processing pathway in this study, PTBP1, RBMX, HNRNPU, and CPSF6 have been previously shown to enhance HCC development.41, 42, 43, 44 In particular, RBMX benefits sorafenib resistance in HCC cells. An obvious reduction in cell viability with increasing sorafenib concentration was observed in RBMX-deficient Huh7/Hch7-SR cells.42 Moreover, HNRNPU can encourage cisplatin resistance in bladder cancer.45
We performed in silico analysis from tissue and at the cell level and revealed the carcinogenicity of ZNF687. Regarding biological functions, zinc finger proteins are involved in transcriptional regulation, protein interactions and post-transcriptional regulation.46 Nevertheless, aberrant expression or dysfunction of zinc finger proteins causes hepatocarcinogenesis.46, 47 For instance, ZNF687 was previously discovered to promote downstream target gene transcription through binding to the enhancer, thus contributing to HCC.22 Compared with former research, we adopted an increased number of samples and public CRISPR/Cas9 gene-editing data, concluding that ZNF687 is a prognostic factor for HCC patients and is significant for HCC cell growth.
We determined that ZNF687 is TKI-resistant and oncogenic, and we subsequently supposed it to be a therapeutic target for overcoming TKI resistance. After further exploration with drug prediction and molecular docking, a small compound chaetocin exhibited a potential to resist ZNF687 in this study. Chaetocin is a natural metabolite from Chaetomium and has been reported to have an antitumor effect on various malignancies, including HCC.48, 49 Notably, chaetocin was previously claimed to overcome TKI resistance for chronic myelogenous leukemia.50 Consequently, we propose that chaetocin can oppose TKI resistance to HCC through targeting ZNF687.
The highlights of the present study include the execution of in vitro CRISPR screening to identify the TKI-resistant gene ZNF687, integrated global sequencing datasets and public CRISPR/Cas9 knockout data to attest its carcinogenicity, and implemented a molecular simulation to declare ZNF687 a druggable target. We comprehensively illuminated the possible functions of ZNF687 and the biological mechanism. The underlying molecular mechanism of ZNF687 may be related to histone modification, spliceosome, transcription coregulator activity, and nucleocytoplasmic transport.
Limitations
During the collection of mRNA datasets of HCC tissue and calculating the demonstrated transcriptional expression status of the ZNF687 gene, the authors detected high heterogeneity. However, random-effects analysis was adopted to make up for the deficiency. Conversely, the biological pathways that ZNF687 may be involved in were predicted in this study, but they have not been verified through in vivo or in vitro experiments. Thus, robust experimental verification should be conducted for an intensive understanding of ZNF687.
Conclusions
ZNF687 was shown to be a TKI-resistant and growth-dependent gene for HCC, and overexpression of ZNF687 indicates poor OS for HCC patients. Additionally, ZNF687 is expected to be a druggable target for overcoming TKI resistance, and chaetocin may be a candidate therapeutic compound for ZNF687.
Supplementary data
The Supplementary materials are available at https://doi.org/10.5281/zenodo.11075935. The package includes the following files:
Supplementary Table 1. Gene effects of zinc finger protein 687 deletion on the growth of hepatocellular carcinoma cell lines.
Supplementary Table 2. Statistical description and the true positive, false positive, true negative and false negative values of zinc finger protein 687 mRNA in datasets.
Supplementary Table 3. Exploring the sources of high heterogeneity using univariable meta-regression analysis
Supplementary Fig. 1. Sensitivity analysis using the one-by-one removal method.
Supplementary Fig. 2. Proportional hazards assumption test by Schoenfeld residual method. A. ZNF687-continuous (p = 0.16); B. Age (p = 0.97); C. Gender (p = 0.15); D. Race (p < 0.05); E. pT stage (p = 0.66); F. AJCC stage (p = 0.53).
Supplementary Fig. 3. Log-linear hypothesis test by Martingale residual method.
Data availability
The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.
Consent for publication
Not applicable.
.png)
.png)
.png)
.png)
.png)
.jpg)
.png)
.jpg)
.jpg)