Abstract

Background

   Cisplatin resistance is among the main reasons for the poor prognosis
   of ovarian cancer (OC) patients. Until now, effective biomarkers for
   predicting cisplatin resistance in OC and specific drugs for reversing
   this resistance are lacking. This study identified the critical gene
   associated with cisplatin resistance in OC and provided a potential
   target for overcoming this resistance.

Methods

   Differentially expressed genes between cisplatin‐resistant and
   ‐sensitive OCs were identified by screening public datasets. Survival
   analysis was conducted to screen prognosis‐related DEGs. CIBERSORT,
   ESTIMATE, and immune checkpoint genes were used to assess the
   association between EMP1 expression and tumor microenvironment
   features. CTRP and GDSC databases were employed to analyze the
   correlation between EMP1 expression and cisplatin resistance.
   Furthermore, immunohistochemistry, qPCR, Western blotting, siRNA
   interference, and the CCK8 assay were performed to verify the role of
   EMP1 in cisplatin resistance in vitro. Finally, xenograft mouse models
   were generated to further confirm the role of EMP1 in cisplatin
   resistance in vivo.

Results

   EMP1 was identified as a critical gene associated with cisplatin
   resistance in OC. According to bioinformatics analyses, increased EMP1
   expression was linked to higher stromal/ESTIMATE scores as well as
   greater ICG expression levels. The in vitro experiments showed that
   EMP1 was highly expressed in cisplatin‐resistant OC tissues and cells,
   and silencing this EMP1 expression enhanced OC cell sensitivity to
   cisplatin. Finally, in vivo experiments confirmed that EMP1 promotes
   tumor growth and cisplatin resistance.

Conclusions

   EMP1 can act as a predictive biomarker for cisplatin resistance in OC
   and as a potential therapeutic target.

   Keywords: cisplatin resistance, differential expression genes,
   functional enrichment analysis, ovarian cancer, survival analysis
     __________________________________________________________________

   In this study, EMP1 was identified as a key gene for cisplatin
   resistance in ovarian cancer by bioinformatics, and it was verified in
   vitro and in vivo.

   graphic file with name CAM4-12-9024-g002.jpg

1. INTRODUCTION

   Ovarian cancer (OC) is the primary cause of gynecological
   cancer‐related mortality.[34] ^1 According to the American Cancer
   Society, 19,880 new cases and 12,810 deaths of OC are estimated in the
   United States in 2022.[35] ^2 Chemotherapy resistance is among the key
   reasons for high mortality and poor prognosis of OC patients.[36] ^3 ,
   [37]^4 Surgical cytoreduction, along with postoperative platinum‐based
   chemotherapy, is the cornerstone of front‐line treatment for OC.[38] ^5
   Cisplatin is one of the most effective chemotherapy drugs for OC that
   can activate apoptosis‐related signaling pathways, DNA damage repair,
   and cell cycle regulation, but resistance is fairly common.[39] ^6 ,
   [40]^7 With initial cisplatin chemotherapy, the chance that the patient
   will react to the treatment is greater than 70%.[41] ^8 However,
   approximately 80% OC patients experience chemotherapy
   resistance‐induced tumor recurrence, which eventually leads to
   death.[42] ^9 Therefore, revealing the molecular mechanism of
   platinum‐based chemotherapy resistance in OC is crucial.

   According to currently available studies, the mechanisms underlying
   cisplatin resistance include apoptosis dysregulation, alteration of the
   tumor microenvironment, improvement of detoxification, enhancement of
   DNA damage repair function, greater immune evasion, and decrease in
   intracellular drug concentration.[43] ^10 , [44]^11 , [45]^12 Based on
   an understanding of the aforementioned resistance mechanisms, some
   molecularly targeted drugs have been developed, including immune
   checkpoint inhibitors, antiangiogenesis inhibitors, and poly‐ADP‐ribose
   polymerase (PARP) inhibitors (listed in Table [46]S1).[47] ^13 However,
   the prognosis of cisplatin‐resistant OC patients has not improved
   radically. Furthermore, because of the complicated mechanisms
   under‐lying cisplatin resistance and the genetic variability of OC
   patients, no effective biomarker is currently available for predicting
   cisplatin resistance in OC in the clinic. Thus, investigating critical
   cisplatin resistance‐linked genes is necessary.

   The transcriptome analysis using gene chips and next‐generation
   sequencing technology has facilitated recently the investigation of
   cisplatin resistance mechanisms.[48] ^14 , [49]^15 , [50]^16 , [51]^17
   However, the outcomes of drug resistance‐related targets found in many
   studies may be inconsistent because of OC heterogeneity, and variations
   in detection technologies and data processing methodologies. Robust
   Rank Aggregation (RRA) is an algorithm that integrates rankings to
   obtain a comprehensive ranking list. This list can combine the results
   of many gene expression datasets to acquire reliable results.[52] ^18 ,
   [53]^19 Therefore, this study identified the critical gene responsible
   for cisplatin resistance in OC by using integrated bioinformatics
   methods and verified the gene in vitro and in vivo.

2. METHODS

2.1. Clinical specimens and datasets

   Clinical samples were collected from the First Affiliated Hospital of
   Yangtze University for the immunohistochemistry (IHC) analyses. A total
   of 87 patient specimens in all were collected, of which 30 had
   cisplatin‐sensitive recurrence, 34 had cisplatin‐resistant recurrence,
   and 23 had benign ovarian tumors. The First Affiliated Hospital of
   Yangtze University's Ethics Committee approved the study (No.
   KY20211206). All patients' families provided their written informed
   permission. Four datasets were obtained from the Gene Expression
   Omnibus (GEO) database ([54]https://www.ncbi.nlm.nih.gov/gds). The
   detailed information is depicted in Table [55]1.

TABLE 1.

   Characteristics of the four chip datasets
   GSE ID[56] ^a Samples Platform Sensitive Resistant Reference
   [57]GSE15372 A2780 [58]GPL570 5 5 Li et al.[59] ^14
   [60]GSE28646 A2780/CP70 [61]GPL570 3 3 Zeller et al.[62] ^15
   [63]GSE58470 IGROV‐1 [64]GPL6947 3 3 Arrighetti et al.[65] ^16
   [66]GSE45553 OVCAR‐8/8C [67]GPL6244 4 4 Chowanadisai et al.[68] ^17
   [69]Open in a new tab
   ^^a

   GSE ID: Gene Expression Omnibus (GEO) series ID from National Center
   for Biotechnology Information (NCBI) GEO database
   ([70]https://www.ncbi.nlm.nih.gov/gds/).

2.2. Identification of cisplatin resistance‐related differentially expression
genes

2.2.1. (DEGs) in OC

   First, the Limma (Bioconductor, [71]https://bioconductor.org) software
   package[72] ^20 was used to normalize the dataset and identify the DEGs
   in cisplatin‐sensitive and ‐resistant cell lines, while the “Robust
   Rank Aggreg” function of the R software package was applied to
   integrate all the ranking gene lists of multiple datasets. Then, the R
   software package cluster profile[73] ^21 was applied to perform
   functional enrichment analysis, including Gene Ontology (GO) enrichment
   analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway
   enrichment analysis on the identified DEGs to predict their potential
   functions, while, gene set enrichment analysis (GSEA) identified the
   potential biological function of EMP1. Furthermore, the PPI network of
   the identified DEGs was constructed with the Search Tool for the
   Retrieval of Interacting Genes/Proteins (STRING,
   [74]https://string‐db.org/).[75] ^22 Finally, the results of PPI
   network was explored and visualized with the Cytoscape (version 3.9.1)
   software, and the hub genes in the PPI network were detected by using
   the cytoHubba plugin, while the modules of PPI network were screened by
   the Molecular Complex Detection (MCODE) plugin of Cytoscape software.

2.3. Survival analysis identifies key genes associated with prognosis

   First, the survival‐related DEGs were obtained through univariate Cox
   analysis of TCGA database survival information. In addition, the Least
   Absolute Shrinkage and Selection Operator (LASSO) regression analysis
   was performed to build a survival prediction model for identifying the
   most important genes for prognoses. Moreover, the Kaplan–Meier survival
   curves were produced with the R Survminer software package to assess
   the prognostic effect of the identified genes on OC survival. Finally,
   the dataset [76]GSE140082 was used to validate the effect of EMP1 on
   the prognosis of OC patients.

2.4. Estimation of tumor microenvironment and tumor infiltrating immune cells

   First, the CIBERSORT algorithm,[77] ^23 the ESTIMATE algorithm,[78] ^24
   and immune checkpoint genes (ICGs) were applied to analyze the
   association between the EMP1 expression with TME features by using
   transcriptome data from TCGA‐OV. Furthermore, genomic mutation data of
   TCGA‐OV were used to calculate tumor mutation burden (TMB),[79] ^25
   which was defined as the total number of altered bases per megabase,
   and only mutations that alter amino acids were included in the
   calculation. In addition, the expression levels of ICGs were calculated
   using TCGA‐OV transcriptome data. The differences in tumor
   microenvironment (TME) score, tumor infiltrating immune cells (TIICs),
   TMB, and ICGs expression between the two EMP1 high‐ and low‐expression
   groups were compared by the Wilcoxon test.

2.5. Correlational of EMP1 and cisplatin drug‐sensitivity score

   We investigated the association between the EMP1 expression level and
   cisplatin resistance with the Genomics of Drug Sensitivity in
   Cancer[80] ^26 (GDSC, [81]https://www.cancerrxgene.org/) and the Cancer
   Therapeutic Response Portal[82] ^27 (CTRP,
   [83]https://portals.broadinstitute.org/ctrp/) databases. The
   cisplatin‐sensitivity response scores were predicted with the
   oncoPredict R package[84] ^28 ([85]https://osf.io/c6tfx/), while
   Spearman correlational analysis was conducted to analyze the
   relationship between the EMP1 expression and the cisplatin
   drug‐sensitivity scores.

2.6. IHC

   The Human Protein Atlas[86] ^29 (HPA,
   [87]https://www.Proteinatlas.org/) was referred to detect the key
   proteins expression in OC and the normal tissues. The EMP1 protein
   expression was assessed by IHC. The specimens were sectioned (4 mm),
   heated at 60°C for 4 h, deparaffinized with xylene, and dehydrated with
   graded alcohol. After antigen extraction with EDTA and a 20 min
   pre‐incubation with 5% normal bovine serum, the sections were incubated
   overnight at 4°C with the primary antibody EMP1 (1:100, ab230445,
   Abcam). After washing, the slides were treated with secondary
   antibodies conjugated with horseradish peroxidase (HRP) against rabbit
   IgG (1:4000, ab205718, Abcam). Then, they were stained with
   diaminobenzidine (DAB, Beyotime, P0203).

2.7. Cell culture

   SKOV3/DDP and SKOV3 cell lines were purchased from the Cancer Biology
   Detection Center, Chinese Academy of Sciences, which were cultivated in
   penicillin‐free RPMI 1640 media with 10% fetal bovine serum and
   streptomycin and incubated at 5% CO[2] and 37°C with saturated
   humidity. The cells could be passaged when the confluence reached >80%.
   The maintenance concentration of cisplatin in the drug‐resistant cell
   line was 1 μg/mL.

2.8. CCK8 assays

   The cells were collected and injected at 7 × 103 cells/well on a
   96‐well cell culture plate. They were left overnight at 37°C under a 5%
   CO2 atmosphere. The cells were grouped according to the cisplatin
   concentration (0, 0.195, 0.780, 3.125, 12.500, 50, and 100 mg/L). After
   the required time for cell culture, 10 μl of CCK8 (MedChemExpress,
   HY‐K0301) was added to each well, and the cells were cultured at 37°C
   for 2 h. The OD450 value of each well was measured by a
   Spectrophotometer (BioTek, C22.2NO.1010.1).

2.9. Real‐Time qPCR (RT‐qPCR)

   The primer sequences for glyceraldehyde 3‐phosphate dehydrogenase
   (GAPDH) and EMP1 were as follows: GAPDH forward
   5′‐TCAAGAAGGTGGTGAAGCAGG; reverse 5′‐TCAAAGGTGGAGGAGTGGGT‐3′; EMP1
   forward 5′‐TGAAGAT GCCCTCAAGACAGT‐3′; reverse
   5′‐AGCCCAGGATGTAGGAATAGC‐3′. The discovered genes' Ct value were
   contrasted with the Ct values of GAPDH, an internal control gene. The
   2^−ΔΔCt equation gained the relative expression of EMP1.

2.10. Western blotting assay

   The treated cells were incubated in the radioimmunoprecipitation assay
   (RIPA) lysis buffer (Beyotime), and the lysis buffer was collected
   after centrifugation at 12,000 rpm for 5 min. The bicinchoninic acid
   (BCA) kit was used to assess the protein content (Beyotime, P0010).
   After loading 30 μg of protein per lane, the proteins were separated
   using 10% SDS‐PAGE. Then, the proteins were electrotransferred onto a
   polyvinylidene fluoride (PVDF) membrane (Millipore). The membranes were
   incubated overnight after blocking for 2 h with 5% skimmed milk. EMP1
   antibody (1:1000, ab230445, Abcam) was diluted with a blocking
   solution. The PVDF membranes were dipped in the primary antibody
   incubation solution and left to incubate overnight at 4°C. The PVDF
   membranes were treated with an HRP‐conjugated secondary antibody
   (1:5000, ab205718, Abcam) for 2 h at room temperature. The BeyoECL Plus
   Kit (Beyotime, P0018) was used to probe the membranes.

2.11. Colony formation

   The cell suspension was diluted and seeded in a 6‐well plate at the
   density of 500 cells per well. Then, the culture dish was placed in a
   37°C under 5% CO[2] incubator for 2 weeks. When the clones appeared,
   they were rinsed twice with PBS, fixed with 4% paraformaldehyde for
   10 min, stained with 0.5% crystal violet for 30 min, and counted and
   photographed under an optical microscope.

2.12. siRNA transfections

   siRNA‐291, siRNA‐335, and siRNA‐567 of EMP1 were designed and
   synthesized by Jtsbio Co. Ltd. Following incubation for 48 h after
   transfection of SKOV3/DDP cells, the expression level of EMP1 was
   assessed by qPCR and Western blotting. The following sequences were
   used: siRNA‐291 guide 5′‐GGUCCACAU CGCUACUGUUTT‐3′, passenger
   5′‐AACAGUAGCGAUGU GGACCTT‐3′; siRNA‐335 guide
   5′‐GCAGUGACAGCCUGUCAUATT‐3′, passenger 5′‐UAUGACAGGC UGUCACUGCTT‐3′;
   siRNA‐567 guide 5′‐GCAGUAUCACCACGGCUAUTT‐3′, and passenger
   5′‐AUAGCCGUGGUGAUACUGCTT‐3′.

2.13. Xenograft tumors in nude mice

   SKOV3 cells were transfected with an EMP1 lentiviral vector from
   GenePharma Inc., while the SKOV3/DDP cells were transfected with a
   shEMP1 lentiviral vector from the same source. Female BALB/c nude mice
   of 4–5 weeks of age were purchased from Yaokang Biotechnology Co. Ltd.
   and assigned to eight groups of three mice in each group. The cell
   suspension was injected into the left axilla subcutaneous of the nude
   mouse at a cell concentration of 1 × 10^7/mL. After 5 mm tumor
   formation, the DDP group was injected with the concentration of 1 mg/mL
   and the dose of cisplatin 10 mg/kg via the tail vein once every 4 days
   for a total of four times. At the end of drug treatment, the mice were
   sacrificed. All animal experiments were carried out in accordance with
   IACUC of Hubei Provincial Center for Safety Evaluation of Food and Drug
   regulations and were given their approval (IACUC no.202210174).

2.14. Statistical analyses

   The means and standard deviations of descriptive statistics were
   calculated.[88] ^30 One‐way ANOVA and the Tukey's post‐hoc test were
   conducted to compare multiple groups. The integral optical density
   (IOD) was used to quantify the expression of EMP1.[89] ^31 IOD and the
   density of Western blotting bands were measured with the Image‐Pro Plus
   v6.0 (Media Cybernetics, Inc.) software. IC50 (50% inhibitory
   concentration) values were calculated using the GraphPad Prism 8
   (GraphPad Software). The Student's t‐test was applied to compare two
   sets of data. The differences were considered to be statistically
   significant at p < 0.05.

3. RESULTS

3.1. Identification of EDGs in cisplatin‐resistant and ‐sensitive OC cell
lines

   The results of normalizing gene expression levels from the four
   microarray datasets were shown in Figure [90]S1. The Spearman
   correlation analysis revealed that all correlation coefficients between
   different samples were >0.92 (Figure [91]S2). The principal component
   analysis revealed that cisplatin‐sensitive and ‐resistant cell lines
   could be clearly separated in each dataset (Figure [92]S3). DEGs
   identified using the Limma software package in each dataset were shown
   in a volcano plot (Figure [93]1A) and in Table [94]S2. Cluster heatmaps
   displayed the top 100 DEGs in each dataset (Figure [95]S4). In total,
   231 DEGs, including 124 upregulated and 107 downregulated, were
   obtained using the RRA algorithm (Table [96]S3). The heatmap
   illustrated the top 25 upregulated DEGs and the top 25 downregulated
   DEGs (Figure [97]1B).

FIGURE 1.

   FIGURE 1
   [98]Open in a new tab

   The differential expression genes (EDGs) in each datasets. (A)
   differential expression genes (DEGs) for cisplatin‐sensitive and
   cisplatin‐resistant cell lines in [99]GSE15372, [100]GSE28646,
   [101]GSE45553, and [102]GSE58470. (B) Heatmap depicting the log2 fold
   change (FC) value of the top 50 differential expression genes (DEGs) in
   each dataset. Orange dots represent genes whose expression levels are
   significantly upregulated, green dots represent genes whose expression
   levels are significantly downregulated, and gray dots represent genes
   whose expression levels have not changed significantly. The abscissa
   represents the name of the dataset. The ordinate represents the name of
   the gene. Orange shading indicates that the gene expression is
   upregulated (log2 FC >0). Green shading indicates that the gene
   expression is downregulated (log2 FC <0). GSE ID: Gene Expression
   Omnibus (GEO) series ID from the National Center for Biotechnology
   Information (NCBI) GEO database.

3.2. Functional enrichment analysis and construction of protein–protein
interaction network of DEGs

   According to the GO enrichment analysis, 86 GO items were obtained,
   including 73, four, and nine items related to biological process (BP),
   cellular components (CC), and molecular functions (MF), respectively
   (Table [103]S4). The top 20 enriched GO items are listed in
   Figure [104]2A. According to the KEGG enrichment analysis, DEGs were
   primarily enriched in immune‐related pathways, and the top 10 enriched
   pathways are listed in Figure [105]2B. The protein–protein interaction
   (PPI) network of DEGs is displayed in Figure [106]2C. Two modules were
   found using the MCODE plugin in Cytoscape software. Module 1
   (Figure [107]2D) contained 16 proteins, while Module 2 contained 12
   proteins (Figure [108]2E). The top 20 genes with the highest degree of
   connection were identified as hub genes by using the cytohubba plugin
   (Table [109]S5).

FIGURE 2.

   FIGURE 2
   [110]Open in a new tab

   Function enrichment analysis and protein–protein interaction (PPI)
   network analysis of differential expression genes (DEGs). (A) The top
   10 GO enrichment analysis items. (B) The top 10 enriched KEGG pathways.
   PPI network of (C) all DEGs, (D) Module 1, and (E) Module 2. AGE:
   advanced glycation endproduct; BP: biological process; CC: cellular
   component; GO: Gene Ontology; IL‐17: interleukin‐17; KEGG: Kyoto
   Encyclopedia of Genes and Genomes; MAPK: mitogen‐activated protein
   kinase; MF: molecular function; RAGE: receptor for AGE. Each oval is a
   protein, and the gray lines indicate interactions between the proteins.
   The orange ovals represent genes whose expression levels are
   significantly upregulated. The green ovals represent genes whose
   expression levels are significantly downregulated.

3.3. EMP1 associated with prognosis and its expression increased with stage

   Univariate Cox regression analysis identified 23 prognostic genes in
   the TCGA‐OV dataset (Table [111]S5), and LASSO regression analysis
   screened out 15 most prognostic genes, namely PDE4B, LMO4, GBP1,
   HOXD11, SLITRK3, TGFBI, PSMB9, UST, TSPAN7, ANKRD22, EMP1, STARD13,
   UBB, TGM2, and AXL. The selection of regression coefficients and
   optimal model parameters is presented in Figure [112]S5. The
   Kaplan–Meier survival analysis revealed that six of the aforementioned
   15 genes were significantly associated with survival. Among them, the
   high expression of EMP1, SLITRK3, and STARD13 was associated with poor
   OC prognosis, while that of GBP1, LMO4, and PSMB9 was associated with
   favorable prognosis (Figure [113]3A). The correlation analysis between
   key genes and disease stages discovered that EMP1 expression increased
   gradually as tumor stages progressed, although no statistically
   significant difference was noted (Figure [114]3B). The expression of
   GBP1, LMO4, PSMB9, STARD13, SLITRK3 proteins in OC, and normal tissues
   in the HPA database is displayed in Figure [115]S6.

FIGURE 3.

   FIGURE 3
   [116]Open in a new tab

   Kaplan–Meier survival and tumor staging analyses. (A) Kaplan–Meier
   survival analysis of EMP1, GBP1, LMO4, PSMB9, SLITRK3, and STARD13. (B)
   Correlational analysis between the key genes and tumor stages.

3.4. Identification of biological functions of EMP1

   In total, 379 samples in the TCGA‐OV cohort were classified into high
   (n = 190) and low (n = 189) expression groups by the median of EMP1
   gene expression. Overall, 96 DEGs were identified, of which 92 were
   significantly upregulated and four were significantly downregulated
   (Figure [117]4A, Table [118]S6). In addition, a heatmap was used to
   display 96 DEGs and revealed that the significantly upregulated genes
   in the high EMP1 expression group mainly encoded extracellular matrix
   proteins such as collagen fibers (COL1A1, COL1A2, COL3A1, COL5A1,
   COL5A2, COL6A1, COL6A2, COL6A3, COL8A1, COL10A1, COL11A1, and COL12A1)
   and matrix proteases (MMP2, MMP7, MMP11, MMP13, and TIMP3)
   (Figure [119]4B). Further, the GSEA GO enrichment analysis showed that
   EMP1 was involved in collagen fibril organization, extracellular matrix
   organization, integrin‐mediated cell adhesion, and angiogenesis
   (Figure [120]4C, Table [121]S7). The GSEA KEGG analysis found that the
   genes enriched in the high EMP1 expression group were mainly associated
   with focal adhesion, the PI3K‐Akt signaling pathway, and proteoglycans
   in cancer, as well as cell adhesion molecules. The low expression genes
   were mainly enriched in ribosomes (Figure [122]4D, Table [123]S7).

FIGURE 4.

   FIGURE 4
   [124]Open in a new tab

   Biological functions of EMP1. (A) Volcano plot depicting differential
   expression genes (DEGs) between the EMP1 high‐ and low‐expression
   groups. (B) Heatmap illustrating 96 DEGs. (C) Gene Set Enrichment
   Analysis (GSEA) showing five GO entries that were significantly
   enriched, and (D) GSEA revealed five KEGG pathways that were enriched.

3.5. EMP1 associated with tumor immunity microenvironment

   First, the ESTIMATE algorithm was used to generate TME scores (stromal
   score, immunological score, and ESTIMATE score) for the two groups. The
   stromal scores (p < 0.001) and ESTIMATE scores (p = 0.0016) were
   significantly higher in the EMP1 high‐expression group than in the
   low‐expression group (Figure [125]5A). Then, the CIBERSORT algorithm
   compared the immune landscapes of the two groups and discovered that
   T‐cells follicular helper, plasma cells, and resting mast cells had
   significantly lower infiltration in the EMP1 high‐expression group than
   in the low‐expression group, while resting CD4 memory T‐cells,
   neutrophils, and activated mast cells were significantly more
   infiltrated in the EMP1 high‐expression group than in the
   low‐expression group (Figure [126]5B). Additionally, the results showed
   that no significant difference in the TMB score between the EMP1 high‐
   and low‐expression groups (p = 0.45) (Figure [127]5C). Furthermore,
   several ICGs were found significantly upregulated in EMP1 high
   expression group (p < 0.05), including CD244, CD48, CD86, CTLA4,
   HAVCR2, IDO1, LAIR1, TNFRSF14, but only BTNL2 gene expression was
   slightly downregulated (Figure [128]5D).

FIGURE 5.

   FIGURE 5
   [129]Open in a new tab

   Correlation between EMP1 gene and tumor immune microenvironment of
   ovarian cancer. (A) Differences in tumor microenvironment (TME) scores
   between EMP1 high‐ and low‐expression groups. (B) Radar chart
   displaying the correlation between the EMP1 expression and
   tumor‐infiltrating immune cells (TIICs). (C) Correlational between the
   EMP1 expression and tumor mutation burden (TMB). (D) Differences in
   immune checkpoint genes (ICGs) expression between EMP1 high‐ and
   low‐expression groups. *p < 0.05; **p < 0.01; ****p < 0.0001.

3.6. EMP1 overexpression in cisplatin‐resistant OC and correlation with poor
prognosis and cisplatin resistance

   EMP1 expression was detected in all groups (Figure [130]6A), but the
   expression level was different. The IOD values of the benign,
   cisplatin‐sensitive, and cisplatin‐resistant groups were
   11.228 ± 2.251, 9.571 ± 3.153, and 17.712 ± 2.251, respectively. The
   IOD levels in the cisplatin‐ and cisplatin‐resistant groups had
   statistically significant differences (p < 0.001), whereas those in
   benign ovarian tumors and the cisplatin‐sensitive group exhibited no
   statistically significant difference (p = 0.088) (Figure [131]6B).
   Table [132]2 summarizes the association between EMP1 expression levels
   and the clinicopathological characteristics of OC patients. The EMP1
   expression level was not significantly correlated with patient age,
   pathological type, pathological grade, or the International Federation
   of Gynecology and Obstetrics stage. The cutoff value of high and low
   EMP1 expression levels was determined by the ROC curve, while that of
   IOD was 11.42 (Figure [133]6C). The area under the ROC curve (AUC) was
   8.822 (95% CI: 0.711–0.932). In [134]GSE140082, the OS of EMP1
   high‐expression group was shorter than that of EMP1 low‐expression
   group (log‐rank p = 0.034). Among all patients, the overall survival
   (OS) was shorter in patients with high EMP1 expression than in those
   with low EMP1 expression [median: 14.263 (95% CI: 12.298–16.228) vs.
   22.000 (95% CI: 13.143–30.857) months, log‐rank p < 0.001]. No
   significant difference in OS was observed between patients with high
   and low EMP1 expression in the cisplatin‐sensitive group [median:
   21.921 (95% CI: 10.723–33.119) vs. 18.555 (95% CI: 10.186–26.924)
   months, log‐rank p = 0.35]. In the cisplatin‐resistant group, patients
   with high EMP1 expression had significantly shorter OS than those with
   low EMP1 expression [median: 14.000 [95% CI 9.363–18.637] vs. 24.378
   (95% CI: 18.275–30.481) months, log‐rank p = 0.0017] (Figure [135]6D).
   Moreover, the drug sensitivity scores of cisplatin in GDSC1, GDSC2, and
   CTRP datasets were predicted using the OncoPredict software package,
   and EMP1 expression was linked to cisplatin resistance
   (Figure [136]6E).

FIGURE 6.

   FIGURE 6
   [137]Open in a new tab

   EMP1 expression associated with prognosis and cisplatin resistance. (A)
   Representative immunohistochemistry images of EMP1 protein in the ovary
   tissues, benign group (scale bar = 100 μm), cisplatin‐sensitive group
   and cisplatin‐resistant group (scale bar = 200 μm). (B) Quantification
   of the IHC results of EMP1 is depicted in the scatter plot (one‐way
   ANOVA and Tukey's post‐hoc test). (C) The ROC curve of EMP1. (D)
   Kaplan–Meier survival analysis of EMP1(log‐rank test) in
   [138]GSE140082, all group, cisplatin‐sensitive group, and
   cisplatin‐resistant group. (E) Correlation analysis of drug sensitivity
   scores in the GDSC1 dataset, GDSC2 dataset, and (C) CTRP dataset.
   *p < 0.05; **p < 0.01; ****p < 0.0001.

TABLE 2.

   Clinicopathological characteristics and their associations with the
   expression level of EMP1 in the OC tissues
   Clinicopathological features Case No. EMP1 expression level χ ^2
   p‐value
   Low (n = 31) High (n = 33)
   Age
   <50 20 9 (45.0%) 11(55.0%) 0.318 0.711
   ≥50 44 22 (50.0%) 22(50.0%)
   Histological type
   Serous 37 15 (40.5%) 22 (59.5%) 4.052 0.230
   Mucinous 19 10 (52.6%) 9 (47.4%)
   Clear cell 6 5 (83.3%) 1 (16.7%)
   Endometrioid 2 1 (50%) 1 (50%)
   Pathological grade
   1/2 25 12 (48.0%) 13 (52.0%) 0.003 0.955
   3 39 19 (48.7%) 20 (51.3%)
   FIGO stage
   III 27 13 (48.1%) 14 (51.9%) 0.002 0.968
   IV 37 18 (48.6%) 19 (51.4%)
   Cisplatin response
   Sensitive 30 24 (80.0%) 6 (20.0%) 22.524 <0.001 [139]^***
   Resistant 34 7 (20.6%) 27 (79.4%)
   [140]Open in a new tab

   Bold value indicate statistical significance. ***p < 0.001.

3.7. In vitro validation of the role of EMP1 in cisplatin resistance

   The IC50 value of cisplatin for the SKOV3 cell line was 4.970 mg/L,
   whereas that for the SKOV3/DDP cell line was 9.650 mg/L
   (Figure [141]7A). EMP1 mRNA expression was significantly higher in
   SKOV3/DDP cells than in SKOV3 cells (t = 13.97, p < 0.0001)
   (Figure [142]7B). Accordingly, EMP1 protein expression was higher in
   SKOV3/DDP cells than in SKOV3 cells (t = 7.555, p = 0.0016)
   (Figure [143]7C). Colony formation rates were 58.8% for SKOV3 and
   87.26% for SKOV3/DDP, and a significant difference was observed between
   the two groups (t = 20.38, p < 0.0001) (Figure [144]7D). The qPCR
   result revealed that siRNA‐219 had the highest efficiency of silencing
   EMP1 mRNA (Figure [145]7E). Therefore, siRNA‐219 was selected for the
   follow‐up experiments. EMP1 mRNA was significantly downregulated in the
   EMP1‐siRNA group compared with the NC (negative control)‐siRNA group
   (p < 0.0001) (Figure [146]7F). Western blotting further confirmed the
   silencing effect of siRNA on EMP1 protein (Figure [147]7G). At 24 h
   after siRNA transfection, the CCK8 assay revealed that the IC50 value
   of cisplatin in the NC‐siRNA group was 11.000 mg/L, while that in the
   EMP1‐siRNA group was 6.255 mg/L (Figure [148]7H). At 48 h after siRNA
   transfection, the IC50 value in the NC‐siRNA group was 5.734 mg/L,
   while that in the EMP1‐siRNA group was 2.966 mg/L (Figure [149]7H).

FIGURE 7.

   FIGURE 7
   [150]Open in a new tab

   In vitro and in vivo validation of EMP1. (A) CCK8 of SKOV3 and
   SKOV3/DDP cell lines. (B) Bar graphs depicting the EMP1 mRNA expression
   measured by qRT‐PCR. (C) The expression of EMP1 protein detected by
   Western blotting. (D) Colony formation of SKOV3 cells and SKOV3/DDP
   cells. (E) The inhibitory efficiency of different siRNAs. (F) qPCR
   detection of the inhibition efficiency of siRNA‐219. (G) The inhibitory
   effect of siRNA‐219 was verified by Western blotting and illustrated in
   the bar graph. (H) At 24 h and 48 h after siRNA transfection, the CCK8
   assay was performed to show the IC50 value of cisplatin in different
   groups. (I) The gross image of SKOV3‐ and SKOV3/DDP‐derived tumors. (J)
   Bar graphs showing the weights of SKOV3 and SKOV3/DDP cell tumors. (K)
   The effects of EMP1 on tumor size (volume) in SKOV3 and SKOV3/DDP cell
   xenograft mice. *p < 0.05; **p < 0.01; ***p < 0.001, ****p < 0.0001.

3.8. EMP1 promoted tumor growth and cisplatin resistance in vivo

   In SKOV3 tumors, upregulation of EMP1 expression accelerated tumor
   growth and enhanced cisplatin resistance (Figure [151]7I). The tumor
   weight and volume were significantly higher in the EMP1 group than in
   the NC group (p = 0.0003, p = 0.0043). However, no statistically
   significant differences were observed between the EMP1 + DDP and
   NC + DDP groups in tumor weight and volume (p = 0.2503, p = 0.0703)
   (Figure [152]7J, K). In SKOV3/DDP tumors, downregulation of EMP1
   expression inhibited tumor growth and reduced cisplatin resistance of
   tumor cells (Figure [153]7I). Tumor weight and volume were considerably
   reduced in the shEMP1 group compared with the shNC group (p = 0.0004,
   p = 0.0074), and in the shEMP1 + DDP group compared with the shNC+DDP
   group (p = 0.0013, p = 0.0193) (Figure [154]7J, K).

4. DISCUSSION

   Bioinformatics analysis is extensively used to identify treatment
   targets for diverse diseases and establish mechanisms underlying cancer
   initiation, invasion, and metastasis. For example, by using
   bioinformatics methods in their study on the mediating neuroprotective
   effect, Jiang F et al. predicted that PTEN gene was the direct target
   of miR‐374a‐5p.[155] ^32 Yang et al. used the GEO database to find
   essential genes and therapeutic medicines for OC.[156] ^33

   To discover the possible therapeutic targets for cisplatin resistance
   in OC, we combined and evaluated data from the four GEO datasets,
   including 15 cisplatin‐sensitive and 15 cisplatin‐resistant cell line
   samples. In total, 231 DEGs were found, of which 124 genes were
   upregulated and 107 were downregulated. Then, bioinformatics approaches
   such as GO and KEGG enrichment, PPI network construction, hub gene
   selection, and survival analysis were used to further investigate these
   DEGs. The bioinformatics analysis showed that high EMP1, SLITRK3, and
   STARD13 expression was associated with poor prognosis, whereas
   increased GBP1, LMO4, and PSMB9 expression improved prognosis.

   EMP1 expression increased as the cancer stage progressed, and
   therefore, we focused on the EMP1 gene. EMP1 gene expression was
   positively correlated with the drug sensitivity score of cisplatin in
   GDSC1, GDSC2, and CTRP datasets. IHC verified that cisplatin‐resistant
   OC tissues had higher EMP1 expression levels than other groups.
   Additionally, no discernible variations in EMP1 expression were
   observed between the group with cisplatin‐sensitive OC and the group
   with benign ovarian tumors. According to RT‐qPCR, cisplatin‐resistant
   SKOV3/DDP cells exhibited considerably higher EMP1 mRNA expression than
   cisplatin‐sensitive SKOV3 cells. Similarly, according to Western
   blotting results, EMP1 protein expression was significantly upregulated
   in SKOV3/DDP cells. Thus, we demonstrated that EMP1 is highly expressed
   in cisplatin‐resistant tissues and cell lines and is involved in
   cisplatin resistance.

   The GSEA allows detecting if a set of preset genes displays
   statistically significant expression variations among different
   groups.[157] ^34 By identifying ARID1A‐linked signal pathways, Yuanyuan
   Feng et al. employed GSEA to seek physiological processes activated
   differently in hepatocellular carcinoma.[158] ^35 In this study, GSEA
   KEGG analysis found that the genes enriched in the high EMP1 expression
   group were mainly involved in focal adhesion, the PI3K‐Akt signaling
   pathway, proteoglycans in cancer, and cell adhesion molecules. GSEA
   enrichment analysis identified GO BP entries related to the EMP1 gene
   function, including extracellular matrix organization,
   integrin‐mediated cell adhesion, and angiogenesis. EMP1 is a
   membrane‐bound glycoprotein having four conserved hydrophobic
   transmembrane domains.[159] ^36 It plays an essential function in cell
   apoptosis, proliferation, adhesion, differentiation, tumor development,
   and metastasis.[160] ^37 , [161]^38 , [162]^39 According to Liu et al.,
   EMP1 is upregulated in OC, leading to invasiveness, migration,
   proliferation, and epithelial–mesenchymal transition(EMT).[163] ^40
   Furthermore, EMP1 is a biomarker in gefitinib‐resistant cancers,
   suggesting its role in the regulation of drug resistance
   development.[164] ^41 By downregulating EMP1 and boosting PI3K/AKT
   phosphorylation in gastric cancer, Ni et al. concluded that miR‐95‐3p
   helps in the establishment of cisplatin resistance and promotes cell
   proliferation, migration, and invasion.[165] ^42 Moreover, EMP1 is
   linked to nodal metastases in oral SCC[166] ^43 and metastatic
   recurrence in colorectal cancer.[167] ^44 Moreover, increased EMP1
   expression was associated with a poor 5‐year event‐free survival in
   patients with precursor‐B ALL.[168] ^45 The extracellular matrix,
   mesothelial cells, endothelial cells, cells originating from the bone
   marrow, adipose cells, immunological and inflammatory cells,
   fibroblasts, and myofibroblasts form the OC TME.[169] ^46 Neutrophils
   exhibit antitumor effects in the early stages of tumor progression, but
   as the tumor progresses, they become more protumor based on
   microenvironmental cues.[170] ^47 Our study showed that the ESTIMATE
   score and Stromal score of the EMP1 high‐expression group were
   significantly higher than those in the EMP1 low‐expression group
   (p < 0.05), indicating that EMP1 gene was associated with matrix
   infiltration, which was consistent with the finding of GSEA GO
   enrichment analysis. Besides, the proportions of resting memory CD4+
   T‐cells, neutrophils, and activated mast cells in the EMP1
   high‐expression group were significantly higher than those in the EMP1
   low‐expression group, suggesting that EMP1 may be involved in the
   regulation of innate and adaptive immunity. Moreover, we found that
   eight ICGs were significantly upregulated in EMP1 high‐expression group
   (p < 0.05), and only one gene was slightly downregulated, suggesting
   that EMP1 gene has a potential immunomodulatory function. Therefore,
   the aforementioned results indicated that EMP1 is associated with the
   tumor immune microenvironment and is a biomarker for drug resistance
   and prognosis.

   Cisplatin resistance can result from any factor that affects
   cisplatin's capacity to bind to DNA and prevent apoptosis, including
   pre‐target resistance (cell modification before cisplatin binds to
   tumor cell targets), on‐target resistance (changes in DNA‐cisplatin
   conjugates), post‐target resistance (changes in downstream signaling
   pathways that induce apoptosis), and off‐target resistance (changes in
   cellular pathways that are not directly related to cisplatin).[171] ^13
   , [172]^48 , [173]^49 The aforementioned discussion clarifies that EMP1
   can affect tumor cell apoptosis, proliferation, differentiation, and
   adhesion, and EMT and PI3K/AKT pathways, and therefore, EMP1 may be
   involved in cisplatin post‐target resistance and off‐target resistance
   in OC.

   In this study, the survival analysis results of our follow‐up data
   showed that high EMP1 expression in OC tissues was associated with poor
   prognosis. The median OS was only 14.263 months in the high EMP1
   expression group compared with 22.000 months in the low‐expression
   group. The difference was more pronounced in the cisplatin‐resistant
   group. Clone formation assays also verified that the SKOV3/DDP cell
   group with high EMP1 expression exhibited stronger proliferation.
   Subsequently, siRNA interference was used to reduce EMP1 expression in
   the SKOV3/DDP cell line. Consequently, we observed a shift in SKOV3/DDP
   cell line's cisplatin resistance. The IC50 value of the NC‐siRNA group
   was 5.734 mg/L, whereas that of the EMP1‐siRNA group was 2.966 mg/L.
   The results revealed that silencing EMP1 enhanced the sensitivity of
   the SKOV3/DDP cell line to cisplatin. In vivo experiments showed that
   EMP1 overexpression promoted tumor growth and enhanced cisplatin
   resistance of tumor cells, whereas silencing EMP1 expression inhibited
   tumor growth and reduced tumor cell resistance to cisplatin. Therefore,
   our study results demonstrated that EMP1 is a tumor promoter and is
   involved in cisplatin resistance.

   The study strengths are as follows. First, our study used multiple
   databases such as GEO, TCGA, GDSC, and CTRP to screen for critical
   genes related to cisplatin resistance in OC, and therefore, the
   screening results exhibited a good robustness. Second, the study used
   tissue samples from patients with recurrent OC, and therefore, the
   sensitivity of each patient to cisplatin could be determined. Moreover,
   the long‐term follow‐up data from our center confirmed the adverse
   effect of EMP1 on OC patient prognosis. However, this study has several
   limitations. On the one hand, the sample size for IHC was small because
   of the small percentage of patients with recurrent OC who could undergo
   reoperation. This implies that multicenter studies can be conducted to
   expand the sample size of patients undergoing reoperation for recurrent
   OC in the future, thereby making the results more reliable. On the
   other hand, the specific mechanism underlying EMP1's involvement in
   cisplatin resistance in vitro and in vivo was not explored here;
   therefore, future studies could clarify the mechanism of EMP1‐mediated
   cisplatin resistance and develop new EMP1‐targeting drugs.

5. CONCLUSIONS

   In conclusion, this was the first study exploring the role of EMP1 in
   cisplatin resistance in OC. The results indicate that EMP1 could be
   used as a predictive biomarker of cisplatin resistance and poor
   prognosis for OC patients. This finding offers a potential target for
   developing drugs to overcome cisplatin resistance in OC.

AUTHOR CONTRIBUTIONS

   Qingsong Zeng: Conceptualization (lead); data curation (lead); formal
   analysis (equal); investigation (equal); methodology (equal); resources
   (equal); software (lead); validation (lead); visualization (equal);
   writing – original draft (lead); writing – review and editing (equal).
   Cunjian Yi: Funding acquisition (lead); methodology (equal); project
   administration (equal); resources (lead). Jingzhi Lu: Funding
   acquisition (equal); methodology (equal); resources (equal); software
   (equal). Xiaowen Wang: Methodology (equal); resources (equal);
   validation (equal); visualization (equal). Keming Chen: Funding
   acquisition (equal); methodology (equal); resources (equal);
   supervision (equal). Li Hong: Formal analysis (lead); methodology
   (lead); project administration (lead); resources (lead); supervision
   (lead); validation (lead); writing – review and editing (equal).

CONFLICT OF INTEREST

   The authors declare no potential conflicts of interest.

ETHICS APPROVAL

   The studies involving human participants were reviewed and approved by
   the Ethics Committee of the First Affiliated Hospital of Yangtze
   University (Ethics No.KY20211206). The patients/participants provided
   their written informed consent to participate in this study.

Supporting information

   Figure S1.
   [174]Click here for additional data file.^ (7.7MB, tif)

   Figure S2.
   [175]Click here for additional data file.^ (5.1MB, tif)

   Figure S3.
   [176]Click here for additional data file.^ (4.1MB, tif)

   Figure S4.
   [177]Click here for additional data file.^ (7MB, tif)

   Figure S5.
   [178]Click here for additional data file.^ (2.8MB, tif)

   Figure S6.
   [179]Click here for additional data file.^ (8.4MB, jpg)

   Figure S7.
   [180]Click here for additional data file.^ (1.9MB, jpg)

   Table S1.
   [181]Click here for additional data file.^ (10.1KB, xlsx)

   Table S2.
   [182]Click here for additional data file.^ (647KB, xlsx)

   Table S3.
   [183]Click here for additional data file.^ (19.6KB, xlsx)

   Table S4.
   [184]Click here for additional data file.^ (59.1KB, xlsx)

   Table S5.
   [185]Click here for additional data file.^ (13.8KB, xlsx)

   Table S6.
   [186]Click here for additional data file.^ (19.9KB, xlsx)

   Table S7.
   [187]Click here for additional data file.^ (173.1KB, xlsx)

ACKNOWLEDGMENTS