Detailed Molecular Mechanism and Potential Drugs for COL1A1 in Carboplatin-Resistant Ovarian Cancer

Carboplatin resistance in ovarian cancer (OV) is a major medical problem. Thus, there is an urgent need to find novel therapeutic targets to improve the prognosis of patients with carboplatin-resistant OV. Accumulating evidence indicates that the gene COL1A1 (collagen type I alpha 1 chain) has an important role in chemoresistance and could be a therapeutic target. However, there have been no reports about the role of COL1A1 in carboplatin-resistant OV. This study aimed to establish the detailed molecular mechanism of COL1A1 and predict potential drugs for its treatment. We found that COL1A1 had a pivotal role in carboplatin resistance in OV by weighted gene correlation network analysis and survival analysis. Moreover, we constructed a competing endogenous RNA network (LINC00052/SMCR5-miR-98-COL1A1) based on multi-omics data and experiments to explore the upstream regulatory mechanisms of COL1A1. Two key pathways involving COL1A1 in carboplatin resistance were identified by co-expression analysis and pathway enrichment: the “ECM-receptor interaction” and “focal adhesion” Kyoto Encyclopedia of Genes and Genomes pathways. Furthermore, combining these results with those of cell viability assays, we proposed that ZINC000085537017 and quercetin were potential drugs for COL1A1 based on virtual screening and the TCMSP database, respectively. These results might help to improve the outcome of OV in the future.


INTRODUCTION
Ovarian cancer (OV) is the leading cause of death among women with gynecological malignancies and is characterized by high recurrence and mortality rates (1). Each year, 225,500 new cases of ovarian cancer are diagnosed, with 140,200 cancer-specific deaths worldwide (2). Owing to the use of chemotherapy, which is the mainstay of OV treatment, the mortality rate has decreased in recent decades (1). Chemotherapy, especially carboplatin, is the primary treatment for OV and can improve patients' overall survival and quality of life (3). However, most OV patients receiving carboplatin chemotherapy develop chemoresistance, which leads to treatment failure (4).
Recently, many studies have demonstrated that the COL1A1 (collagen type I alpha 1 chain) gene is a potential therapeutic target with an important role in chemoresistance (5,6). Most of these studies focused on the relationship between the expression of COL1A1 and chemoresistance. Some researchers found that the expression of COL1A1 was associated with resistance of OV to taxol (7), cisplatin (8), paclitaxel, doxorubicin, topotecan, vincristine, and methotrexate (5). However, the molecular mechanism by which COL1A1 participates in carboplatinresistant OV has remained unclear; thus, the development of potential targeted therapeutic drugs is challenging.
In the present study, we performed data mining based on largescale multi-omics data to explore the detailed molecular mechanism of COL1A1 in carboplatin-resistant OV and to identify potential drugs to target COL1A1. Our study provides new insight into the chemoresistance of OV at the molecular level and explores potential therapeutic drugs to overcome carboplatin resistance in OV and improve the outcomes of OV patients.

Data Collection
The gene expression count profile (transcriptome sequencing and microRNA [miRNA] profiles) and corresponding clinical data of OV patients were collected from The Cancer Genome Atlas (TCGA) database (accessed on September 21, 2019). Proteincoding genes and long non-coding RNAs (lncRNAs) were isolated from the transcriptome sequencing data. Based on their clinical data and previous study (9), 98 OV patients were categorized into a carboplatin-nonresistant group (complete response and partial response; n = 84) and a carboplatinresistant group (stable disease and progressive disease; n = 14). Patients' clinical characteristics are shown in Table S1. The use of TCGA data in the present study was in accordance with TCGA publication guidelines. As the patient data originated from the TCGA database, no further ethical approval was required.

Study Design
The workflow of this study is shown in Figure 1. In order to confirm the function of COL1A1 in carboplatin-resistant OV, we used weighted correlation network analysis (WGCNA), an unsupervised analysis method, to identify carboplatinresistance-related genes. Then, we performed hub gene analysis and survival analysis to further validate the key role of COL1A1 in carboplatin-resistant OV. Subsequently, we explored the upstream regulatory mechanisms of COL1A1 by constructing a competing endogenous RNA (ceRNA) network. Moreover, we performed co-expression analysis and pathway enrichment to identify the downstream regulatory mechanisms of COL1A1. Finally, we identified candidates for drug-repurposing by virtual screening based on the structure of COL1A1 and the traditional Chinese medicines in the TCMSP database. Furthermore, we performed experiments to evaluate the results of the analysis.

Identification of the Hub Genes in Carboplatin-Resistant OV
To obtain modules related to carboplatin resistance in OV, WGCNA was performed using R package WGCNA (10). A total of 8,791 genes with the top 70% median absolute deviations were screened from the database (n = 98) based on carboplatin response. Gene expression modules with similar patterns were identified by the dynamic tree cut method (minModuleSize = 50, mergeCutHeight = 0.25, and deepSplit = 1). An unsigned network type was used to retain relationships between modules and sample type (nonresistant or resistant). Modules with the highest absolute correlation values and p < 0.05 were considered to be significantly related to carboplatin resistance. Then, the identified carboplatin resistance genes were used to construct a network based on WGCNA. Subsequently, we employed two different methods to determine the hub gene in the carboplatin resistance module obtained by WGCNA. We assumed that a biological network G = (V, E) is an undirected network, where V is the collection of nodes within the network and E is the edge set. We used another notation, G = (V[G], E[G]), to represent a network, where V(G) is the collection of nodes in a network G, and E(G) is the collection of edges in a network G. For a set S, we used |S| to denote its cardinality (i.e., the number of elements in the set). Given a node v, N(v) denotes the collections of its neighbors. The two methods were as follows. The gene ranked in first place by both methods was considered to be the hub gene in carboplatin-resistant OV. Finally, survival analysis was performed using KmPlot with an auto-selected best cutoff (11).

Identification of Differentially Expressed mRNAs, miRNAs, and lncRNAs
We used the R package edgeR (12) to normalize and analyze significantly differentially expressed lncRNAs (DElncRNAs), miRNAs (DEmiRNAs), and mRNAs (DEmRNAs) between the carboplatin-resistant OV group (n = 14) and the carboplatinnonresistant OV group (n = 84). According to the previous study (13), we wanted to obtain more candidates. So, the cutoff values were |log2 fold change| ≥ 0.4 and p < 0.05. The DElncRNAs, DEmiRNAs, and DEmRNAs were identified based on these thresholds.

Co-Expression Analysis of COL1A1 and Pathway Enrichment Analysis
To explore the downstream regulatory mechanism of COL1A1, we performed co-expression analysis between COL1A1 and genes related to carboplatin resistance according to WGCNA, with cutoff values of Pearson correlation coefficient > 0.9 and p < 0.005. After obtaining the genes co-expressed with COL1A1, a KEGG overrepresentation test was performed using R package clusterProfiler (17) with a cutoff of p < 0.01. Gene set enrichment analysis (GSEA) (18) was also used to explore the potential molecular mechanisms in the carboplatin-resistant (n = 14) and carboplatin-nonresistant (n = 84) groups based on the expression profiles of all protein-coding genes, with cutoff values of false discovery rate (FDR) < 0.25 and p < 0.05. These results were combined to obtain the key pathways involving COL1A1 in carboplatin-resistant OV.

Potential Drug-Repurposing and Traditional Chinese Medicine
The three-dimensional (3D) structure of COL1A1 was downloaded from the Protein Data Bank (PDB; 5CVB, https:// www.rcsb.org/), and its binding sites were identified by Schrodinger Maestro (19). Then, we built a library of 2,106 US Food and Drug Administration (FDA)-approved drugs obtained from the ZINC15 database (20). Finally, we performed virtual screening and molecular docking with Schrodinger Maestro to identify potential drug-repurposing. We also used TCMSP (http://www.tcmspw.com/tcmsp.php) (21) to find traditional Chinese medicines that might target COL1A1.

Real-Time Quantitative PCR
Total RNA was extracted using TRIzol reagent (Invitrogen). cDNA was synthesized using a cDNA synthesis kit (TIANGEN).
Quantitative real-time PCR analysis was performed in triplicate with SYBR Green (TIANGEN) and specific primers (Table S2) using a CFX Connect Real-Time PCR Detection System (Bio-Rad, USA). U6 was used as an internal control for miRNAs. The relative expression levels of mRNAs or lncRNAs were evaluated relative to glyceraldehyde 3-phosphate dehydrogenase (GAPDH). Relative expression values were calculated using the 2 -△△Ct method.

Cell Viability
Cell viability was detected by cell counting kit-8 (CCK-8; TongRen) assay following the manufacturer's instructions. The CCK-8 test solution was added 30 min before the end of treatment, and the absorbance was measured at 450 nm using a microplate reader. Carboplatin-resistant cells were exposed to a concentration gradient (0, 0.01, 0.1, 1, 10, and 100 mM) of ZINC000085537017 or quercetin for 24 h. To understand the influence of ZINC000085537017 and quercetin on sensitivity to carboplatin, resistant cells were pretreated with 1 mM ZINC000085537017 or 10 mM quercetin for 24 h, followed by incubation with 20% maximal inhibitory concentration (IC 20 ) or IC 50 doses of carboplatin for 48 h, and then subjected to cell viability assays.

Statistical Analysis
Statistical analysis was performed using R 3.6.3 (R Foundation for Statistical Computing, Vienna, Austria). Normal distribution and homogeneity of variance tests were performed before the statistical analysis. The Wilcoxon test was used to evaluate the expression of COL1A1 between the carboplatin-resistant (n = 14) and carboplatin-nonresistant groups (n = 84), and t-tests were used to compare data between the two groups; p < 0.05 was considered statistically significant.

COL1A1 Has an Important Role in Carboplatin-Resistant OV
A total of 98 samples (84 nonresistant to carboplatin and 14 resistant to carboplatin) were included in WGCNA. We selected b = 6 as the appropriate soft-thresholding value to ensure a scalefree network, and 16 modules were identified. These modules are shown in distinct colors in Figure 2A. Then, the correlations between module eigengenes and the clinical trait of interest (resistance to carboplatin) were determined ( Figure 2B, Table  S3). The modules with the highest absolute correlation values and p < 0.05 were considered to be significant carboplatin-resistancerelated modules. Based on the cutoffs used, the yellow module was screened as significantly related to carboplatin resistance in OV ( Figure 2B). A total of 412 genes of the yellow module (Table S4) were found to be significantly related to carboplatin resistance by WGCNA. Subsequently, we constructed a network based on these 412 genes, then used the Deg and MNC methods to identify the hub gene involved in carboplatin-resistant OV. COL1A1 was ranked first place in the hub gene analysis by both methods (Table S5). We also found that COL1A1 mRNA was significantly overexpressed in the carboplatin-resistant OV group (p < 0.05, Figure 2C). We used overall survival analysis with KmPlot to further validate the role of COL1A1 in carboplatin resistance. The results showed that high mRNA expression of COL1A1 was associated with poor prognosis in OV (p < 0.05, Figure 2D). Taken together, these results showed that COL1A1 plays an important part in carboplatin resistance in OV.

Identification of the Downstream Regulatory Mechanism of COL1A1
To explore the downstream mechanism of COL1A1 in carboplatin-resistant OV, we analyzed the genes co-expressed with COL1A1 among the 412 carboplatin-resistance-related genes. There were 14 genes co-expressed with COL1A1 (Table S11) according to the cutoff values of absolute Pearson correlation coefficient > 0.9 and p < 0.005. In addition, 11 KEGG pathways were enriched in the KEGG over-representation test based on the co-expressed genes of COL1A1 ( Figure 4A). Another method for pathway analysis, GSEA targets the expression across the whole genome. GSEA analysis produced a total of five pathways (Table S12) in the carboplatin-resistant group and 22 pathways (Table S13) in the carboplatin-nonresistant group, using cutoff values of FDR < 25% and p < 0.05. Based on the KEGG overrepresentation test and GSEA results ( Figure 4B), two overlapping pathways, "ECM receptor interaction" and "focal adhesion," were identified.

Potential Drug-Repurposing and Traditional Chinese Medicine
We used virtual screening with Schrodinger Maestro 2019-1 to identify potential drugs that could be repurposed to target COL1A1. The 3D protein structure of COL1A1 was downloaded from the PDB (5CVB, Figure S2A), and the active site was found by Schrodinger Maestro ( Figure S2B). According to the glide scores (Table S14), ZINC000085537017 (Cangrelor) was the top hit from the structure-based virtual screening process. The 3D structure of ZINC000085537017 is shown in Figure 5A. There were six H-bonds and one pi-pi interaction in the ligand-protein complex ( Figure 5B). The docking results for ZINC000085537017 and COL1A1 are shown in Figure 5C. Furthermore, we found that COL1A1 was targeted by quercetin based on the TCMSP database ( Table S15). The docking results for quercetin and COL1A1 are shown in Figure S2C. To identify whether ZINC000085537017 or quercetin affected carboplatin sensitization in OV, first, carboplatin-resistant cell lines (A2780-carboplatin and SKOV3carboplatin) were assayed for cell viability after treatment with a concentration gradient of ZINC000085537017 or quercetin (0, 0.01, 0.1, 1, 10, 100 mM) for 24 h. As shown in Figures S3A-B and S3C-D, 0.01-1.00 mM ZINC000085537017 and 0.01-10.0 mM quercetin did not significantly inhibit cell growth, but 10-100 mM ZINC000085537017 and 100 mM quercetin significantly  As shown in Figure 5D, 1 mM ZINC000085537017 significantly enhanced the carboplatin sensitivity (p < 0.05) of carboplatinresistant cells treated with IC 20 carboplatin, although it did not affect cell viability responses to IC 50 carboplatin ( Figure 5D). Similarly, 10 mM quercetin significantly enhanced the sensitivity of carboplatin-resistant cells to carboplatin (IC 20 and IC 50 ) after 48 h of treatment (p < 0.05, Figure 5E).

DISCUSSION
Carboplatin is the cornerstone of chemotherapy for OV. However, drug resistance to this agent continues to present challenges, leading to a poor prognosis for OV patients with a 5-year survival rate of only 25-30% (4,24). Therefore, the mechanism of resistance to carboplatin in OV has become a focus of research in recent years. Increasing evidence has shown that COL1A1 has an important role in chemoresistance and could represent a potential therapeutic target (5, 6), but the mechanism of COL1A1 in carboplatin-resistant OV has remained unclear. In the present study, we determined the detailed molecular mechanism involving COL1A1 in carboplatin resistance and identified potential targeted drugs (both traditional Chinese medicine and FDA-approved drugs). These results provide new information and supporting data that could help to improve the outcomes of OV patients. Many hub genes involved in carboplatin resistance have been identified by screening of differentially expressed genes (25)(26)(27)(28). However, they were selected by an artificially set threshold, potentially excluding some important genes. In the present study, carboplatin resistance genes were screened by WGCNA, an unsupervised analysis method, making our results potentially more realistic and objective. We first screened carboplatinresistance-related genes by WGCNA, then used hub gene analysis to identify COL1A1 as the hub gene. Furthermore, the expression of COL1A1 mRNA was found to be higher in carboplatin-resistant OV (p < 0.05), and survival analysis showed that high expression of COL1A1 mRNA was correlated with poor prognosis (p < 0.05), further demonstrating the pivotal role of COL1A1 in carboplatin resistance. The mRNA expression of COL1A1 was also found to be increased significantly (p < 0.05) in two carboplatin-resistant cells by real-time PCR. Although many studies have reported that the expression of COL1A1 was related to chemoresistance in OV (5)(6)(7)(8), this was the first time that COL1A1 had been shown to play an important part in carboplatin-resistant OV. Furthermore, as type I collagen is composed of COL1A1 and COL1A2 (29), we speculated that COL1A2 might also have an important role in carboplatinresistant OV, although there have been no reports about the role of COL1A2 in carboplatin resistance. According to the WGCNA and hub gene analysis results, COL1A2 was found in the 412 carboplatin resistance gene sets and was ranked 12th and 20th by Deg and MNC, respectively. Previously, Januchowski et al. reported that mRNA levels of COL1A2 and COL1A1 were significantly increased in OV cell lines resistant to cisplatin, paclitaxel, doxorubicin, topotecan, vincristine, and methotrexate (5). Taken together, these results suggest that COL1A1 and COL1A2 could be used as molecular targets for new antitumor drugs against carboplatin-resistant OV.
Emerging evidence indicates that ceRNA networks have an important role in chemoresistance to cancers (30,31) and can provide therapeutic targets. In the present study, we used bioinformatics analysis to identify the ceRNA network of COL1A1. We also performed real-time PCR and KmPlot analysis to further confirm that the ceRNA network was LINC00052/ SMCR5-miR-98-COL1A1 ( Figure 6). Moreover, COL1A1 was previously found to be regulated by miR-98 in hypertrophic scarring (32) and muscular dystrophies (33); overexpression of miR-98 could increase cell apoptosis and enhance sensitivity to cisplatin in lung adenocarcinoma (34); and low miR-98 expression was correlated with temozolomide resistance of glioma (35). Although there have been no reports on the relationship between LINC00052/SMCR5 and chemoresistance, high expression of LINC00052 was found to promote gastric cancer cell proliferation and metastasis (36) and progression of head and neck squamous cell carcinoma (37). However, some of the experimental results in this study were inconsistent with our expectations. We propose two possible reasons for this. First, the associations observed between dysregulated expression of some candidates and prognosis of OV might not have been causal. Second, false positive results may have been generated in our analysis. Overall, in this study, we identified a ceRNA network (LINC00052/SMCR5-miR-98-COL1A1), which expanded our understanding of the upstream regulatory mechanism of COL1A1 in carboplatin-resistant OV and could provide therapeutic targets to improve the prognosis of OV. Furthermore, the "ECM-receptor interaction" and "focal adhesion" KEGG pathways were identified as downstream pathways of COL1A1 involved in carboplatin-resistant OV. We identified these two key pathways using the KEGG overrepresentation test based on the co-expressed genes of COL1A1 and GSEA, suggesting that COL1A1 promoted carboplatin resistance in OV through these pathways. According to previous studies, the "ECM-receptor interaction" pathway is involved in platinum- (38), paclitaxel-, and topotecan-resistant OV (39), trastuzumab-resistant gastric cancer (40), and temozolomide-resistant glioblastoma (41). Moreover, to date, many drug resistance mechanisms involving the extracellular (C) 3D structure of ZINC000085537017's binding mode in COL1A1. Yellow represents hydrogen bonds, green represents pi-pi interactions, and purple represents salt bridges. ZINC000085537017 (D) and quercetin (E) enhanced the cytotoxicity of carboplatin in resistant cells. Cells were pretreated with ZINC000085537017 or quercetin for 24 h, followed by incubation with carboplatin for 48 h, and were then subjected to cell viability assays. The results are presented as mean ± SD (n = 6) and were normalized to the control (*p < 0.05).
matrix have been identified across cancer types; these mechanisms have been classified into a range of categories including physical barriers to treatment (hypoxia, pH, and interstitial fluid pressure) and cell-adhesion-associated drug resistance (42). The "focal adhesion" KEGG pathway has been shown to be associated with taxol-(43) and cisplatin-resistant OV (44,45). Taken together, these results indicated that COL1A1 was involved in carboplatin resistance in OV through the "ECM-receptor interaction" and "focal adhesion" KEGG pathways ( Figure 6).
In the present study, we also explored potential drugrepurposing by virtual screening of FDA approved drugs and traditional Chinese medicines targeting COL1A1, which might expand potential therapeutic strategies for carboplatin-resistant OV treatment. We found that ZINC000085537017 and quercetin were potential drugs for treatment of COL1A1. Although 1 mM ZINC000085537017 did not affect cell viability in response to carboplatin, we expected that ZINC000085537017 and quercetin would enhance the sensitivity of carboplatin-resistant cells based on the cell viability assays. We speculated that the leaching toxicity of IC 50 carboplatin in resistant cells might have exceeded the influence of 1 mM ZINC000085537017, resulting in no significant differences being found when IC 50 carboplatin was combined with 1 mM ZINC000085537017 in these cells. Previous study showed that quercetin could increase the sensitivity of OV to cisplatin (46); however, in contrast to these studies, which focused on the relationship between the expression of COL1A1 and chemoresistance, we not only showed that COL1A1 was a therapeutic target but also identified some potential drugs. These results could help accelerate the development of drugs to improve the outcomes of carboplatin-resistant OV patients.

CONCLUSION
In summary, we identified that COL1A1 has an important role in carboplatin-resistant OV by WGCNA; this result was further validated by survival analysis. Then, we constructed a ceRNA network for COL1A1 by bioinformatics analysis and experiments to expand understanding of the upstream regulatory mechanism of COL1A1 in carboplatin-resistant OV and identify potential therapeutic targets that could be used to improve the prognosis of OV. Moreover, we found that COL1A1 participated in carboplatin resistance in OV through the "ECMreceptor interaction" and "focal adhesion" KEGG pathways by co-expression analysis and pathway enrichment. Furthermore, combining these results with those of experiments, we found that ZINC000085537017 and quercetin were potential drugs for COL1A1 by virtual screening based on the structure of COL1A1 and the TCMSP database. These findings could accelerate drug development to improve the outcomes of carboplatin-resistant OV patients.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding authors.