Human epididymis protein 4 and Lewis y enhance chemotherapeutic resistance in epithelial ovarian cancer through the p38MAPK pathway

Ovarian cancer has a high mortality rate due to diculties in early detection and chemotherapy resistance. Human epididymal protein 4 (HE4) has been adopted as a novel serum biomarker for early ovarian cancer diagnosis, we have previously detected the presence of Lewis y antigen modications on HE4 in ovarian cancer cell lines. In this study, we aimed to analyze the expression of HE4 and Lewis y antigen in human ovarian cancer in order to analyze their correlation with each other, as well as with the clinical pathological parameters of ovarian cancer patients. We rst used immunohistochemistry to detect the respective expressions of these compounds in two patient groups (chemotherapy-resistant and sensitive) containing a total of 95 patients. Then, we adopted a bioinformatic approach and used online large-sample databases (TCGA, CCLE and GTEx; Metascape, Cytoscape) to explore the potential mechanisms of action of these compounds. Our results demonstrate that high HE4 and Lewis y expressions could be used as markers for chemotherapy-resistance and poor prognosis in ovarian cancer patients. These two expressions were widely correlated in various cancer tissues and are thought to act by activating the p38 MAPK pathway and inducing VEGFA, PTGS2,EGR1,andHIFI1A, thereby promoting malignant biological behavior and resistance in ovarian cancer. reveals possible affect could be very useful in ovarian cancer detection and/or the development of new targeted therapies. was rst discovered in the distal epithelial cells of the epididymis and named accordingly. In recent years, HE4 has stirred some interest in the ovarian cancer research eld and has been adopted as a novel serum biomarker for early ovarian cancer diagnosis 2 .The ROMA index, an ovarian cancer risk prediction model that takes into account the levels of HE4 and CA125, can be used to assess the risk of ovarian cancer in premenopausal women 3, 4 . Lewis y antigen is a tumor-associated carbohydrate antigen that is part of multiple receptor structures on cellular surfaces. Our previous in vivo and in vitro studies demonstrated that Lewis y antigen is associated with the biological behaviors of tumors, including proliferation, adhesion, and chemotherapy resistance 5–7 . Our recent study showed the presence of Lewis y antigen modications on the HE4 structure; such modications were detected in ovarian cancer cell lines, cell culture media, and malignant ovarian tumor tissues 8 . Although research on the effect of HE4 on the biological behavior of ovarian cancer has been on the rise as of late, the role of HE4 drug resistance and the underlying mechanisms still remain unclear. Besides, the effect of fucosylation on the biological function of HE4 and the relationship between HE4 and Lewis y have not been comprehensively explored yet. Therefore, the present study employed immunohistochemical techniques to determine the expression of HE4 and Lewis y antigen in chemotherapy-resistant and chemotherapy-sensitive epithelial ovarian cancer tissues. We further evaluated the correlation between the expression of these proteins and their association with chemotherapy resistance and poor prognosis in ovarian cancer. Moreover, we used different bioinformatic techniques to explore the possible mechanisms of action of HE4 and Lewis y. that modied ovarian cancer the p38MAPK signaling PPI the and MCODE identied four hub genes among the overlapped genes (VEGFA, EGR1, HIF1A, and PTGS2); the known functions of these and breakthroughs for prospective studies on HE4 and

were Stage II, 62 cases were Stage III, and 2 cases were Stage IV. Furthermore, resistance to chemotherapy was evaluated according to the guidelines of the National Comprehensive Cancer Network (NCCN) 9 ; the ovarian cancer patients were divided into a chemotherapy-resistant group, a partially-sensitive group (partial sensitive group), and a sensitive group. Patients who received PC-based chemotherapy for the rst time and obtained clinical remission along with patients suffering from treatment-induced recurrence during or within six months of the chemotherapy period were included in the drug resistance group; patients with recurrence 6-12 months after chemotherapy were included in the partially sensitive group; while patients with no recurrence or with recurrence 12 months or more after chemotherapy were included in the sensitive group. The main clinical features of ovarian cancer recurrence include: 1) a continuous increase in CA125; 2) a mass is noted during gynecological examination; 3) a mass is observed during imaging examination; 4) ascites is evident; 5) intestinal obstruction of unknown cause. The 95 patients in this study were grouped accordingly into the following: 35 patients were included in the chemotherapy-resistant group; whereas, the remaining 60 were included in the sensitive group, including six partially-sensitive patients. The overall mean age of patients was 54.36 ± 9.41 years (range: 24 to 78 years, median age: 53.0 years); the mean age of drug-resistant patients was 56.24 ± 9.50 years (range: 34 to 76 years, median age is 56 years); while that of the patients in the sensitive group was 53.31 ± 9.27 years (range: median age of 24 to 78 years is 53 years).There was no signi cant difference in age between the two groups (F = 0.104, P = 0.147).

Immunohistochemistry
The ovarian tissues of each group were sliced into 5 µm-thick serial sections. The expression of HE4 and Lewis y was detected by the streptavidinperoxidase linked (SP) immunohistochemical method per the manufacturer's instructions [12] 10 . Each batch of sections was compared with positive and negative controls; normal human epididymis tissue was used as HE4 positive control; human gastric cancer tissue was used as Lewis y antigen positive control; while phosphate buffered saline (PBS) was added instead of primary antibodies in both negative controls (blanks).Blank buffered photographs were replaced with phosphate buffered saline (PBS) as the primary antibody. The working concentration of rabbit anti-human HE4 antibody (Abcam, UK) was 1:50 and that of mouse anti-human Lewis y antibody (Abcam, UK) was 1:200.

Analyzing the immunohistochemical results
After immunohistochemical staining, the appearance of brownish-yellow particles in the cell membranes and cytoplasm of the tissue samples was used to determine positivity. Moreover, samples were scored according to the intensity of the obtained color; no color, light yellow, brownish-yellow, and tan, were scored as 0, 1, 2, and 3, respectively. Additionally, for each slice, the average percentage ( ve consecutive observations) of colored cells in the visual eld of a high-power eld (HPF) microscope (400x) was calculated and the corresponding samples were assigned a score as follows: 0 points if the positive cell rate was < 5%; 1 point if the rate is between 5-25%, 2 points if the rate is between 26-50%, 3 points if the rate is between 51-75%, and 4 points if the rate is between > 75%. After that, the two obtained scores were multiplied to get the nal score; 0 to 2 points represented negative expression (-), 3 to 4 points re ected weak positivity (+), 5 to 8 points represented moderate positivity (++), and 9 to 12 points re ected strong positivity (+++); (-) and (+) were de ned as low expression, while (++) and (+++) were de ned as high expression. Two different observers performed the visual inspections to avoid any errors.

Prognostic analysis
After referring to patient hospitalization data, we performed follow-up phone calls with all the patients; the last follow-up was performed on May 2019. Next, the overall survival (OS) time was analyzed, which refers to the time from the follow-up date to the death of the patient due to disease, or the follow-up deadline.
Correlation and large-sample-database analysis The correlation between the expression of the two compounds in ovarian cancer tissues was analyzed. Simultaneously, in order to explore the universality of this correlation, we downloaded the pan-cancer mRNA expression data (https://tcga-data.nci.nih.gov) and the expression data of normal tissues from the Genotype-Tissue Expression project portal (GTEx, https://www.gtexportal.org/home/) 11 ; moreover, cancer cell line data was also acquired from the Cancer Cell Line Encyclopedia (CCLE) 12 . We then analyzed the correlation between the mRNA expression of the associated genes (WFDC2 and FUT1) in R software by Pearson correlation analysis.
Analyzing the microarray expression pro le of HE4 and Lewis y gene transfection In our previous studies, we analyzed changes in the genome-wide gene expression pro le of epithelial ovarian cancer cell lines after transfection of HE4 and Lewis y genes (WFDC2 and FUT1) and noted changes in many differentially expressed genes (DEGs) [13][14][15] . In this study, we analyzed DEG pro les, up-and down-regulation, using the Venny-based online software (http://jvenn.toulouse.inra.fr/app/example.html).Moreover, we used the "VennDiagram" R package (RStudio, Version 1.2.5019) to illustrate our results.

Gene Ontology (GO) and cluster network plot analyses
The overlapped genes were uploaded to the online gene annotation and analysis resource Metascape 16 to perform a GO enrichment analysis; this enrichment analysis tested all genes in the human genome in the following databases: KEGG pathway, GO Biological process, Reactome gene sets, Canonical pathways, and CORUM. The inclusion criteria for this analysis included a P value < 0.01, a minimum count of 3, and an enrichment factor > 1.5 and DEGs were grouped according to their similarity; the most statistically signi cant term in each cluster was used to represent said cluster. Further, in order to further study the relationship between terms, a subset of enriched terms (similarity > 0.3) was selected and presented as a network graph, by connecting the edges; Cytoscape (https://cytoscape.org) was used to visualize the enriched terms, where each point represents an enriched term, rst colored by cluster ID, and then colored according to its P value.
Pathway analysis, integration of protein-protein interaction network, and module analysis KEGG Mapper (https://www.kegg.jp/ kegg / mapper.html) was used to visualize speci c genes in the enriched pathways. Next, the list was uploaded to the STRING database (http://string-db.org) in order to form a protein interaction (PPI) network analysis; STRING is an online network that strictly evaluates the direct (physical) or indirect (functional) interactions between proteins. The results were then imported, via stringApp, to Cytoscape (https://cytoscape.org) to visualize molecular interaction networks and biological pathways; stringApp is a Cytoscape application that can easily import STRING networks into Cytoscape; it retains the many STRING features and integrates data from associated databases. In this study, two Cytoscape plugins (MCC plugin and MCODE plugin) were used to screen hub genes and cytoHubba, a Java plugin for Cytoscape, was utilized for identifying hub objects and sub-networks from complex interaction groups; it provides 11 topological analysis methods, of which MCC (Maximal Clique Centrality) has better accuracy in predicting key protein performance 17 . Whereas, MCODE clusters a given network based on topology to generate highly interconnected subgraphs of molecular complexes and partial paths.

Statistical analysis
Statistical analyses were performed using SPSS software (Version 26.0, SPSS, IBM, USA). Count data was tested by χ2 test and measurement data was tested by t test. Spearman rank correlation analysis was used to analyze the correlation between the expression of these two proteins, while Pearson correlation analysis was performed to analyze the expression of the two proteins in the large sample databases and form a gene mRNA expression correlation. Moreover, Logistic Regression was used to analyze various clinical pathological parameters and the relationship between HE4 and Lewis y expression and chemotherapy resistance. The Kaplan-Meier estimator (Log-rank test) was used to perform a survival analysis. Meanwhile, the Cox model was used to analyze the relationship between various clinical pathological parameters and the expression of HE4 and Lewis y with prognosis (Methods: Enter); "Ggplot2," "forestplot" packages were used in R software (Version 3.6.1, RStudio Version 1.2.1335); P 0.05 was considered statistically signi cant.

Results
HE4 and Lewis y antigen expression are correlated with ovarian cancer clinicopathological parameters, especially in chemotherapy-resistant tissues HE4 was mainly expressed in the cell membranes and cytoplasm of tissue samples but was rarely expressed in the perinuclear region. The positive expression rates of HE4 and Lewis y in the ovarian cancer chemotherapy-resistant group were 97.1% and 88.6%, respectively, signi cantly higher than the expression rates of 76.7% and 68.3% in the sensitive group (P = 0.008 and 0.026, respectively) ( Figure 1, Table 1). We further explored the highest expression rates of the two groups and found that the HE4 and Lewis y antigen rates in the chemotherapy-resistant group (80.0%, 80.0%) were signi cantly higher than those in the chemotherapy-sensitive group (30.0%, 30.0%), as shown in Figure 1A and Table 1(P<0.001 and<0.001, respectively).
After exploring the relationship between HE4 and Lewis y expression and the clinicopathological parameters, we found that HE4 expression was related to FIGO stage, differentiation, and drug resistance (P =0.014 0.048, and 0.008 respectively); whereas, Lewis y expression was associated with lymph node metastasis and drug resistance (P =0.026 and 0.026, respectively). In terms of high expression rate, we found that high expression of both HE4 (P<0.001 and 0.001, respectively) and Lewis y (P<0.001 and<0.001, respectively) were correlated with lymph node metastasis and drug resistance (Table 1).
HE4 and Lewis y antigen are independent risk factors of drug resistance in ovarian cancer Age, FIGO stage, pathological grading and type, and HE4 and Lewis y antigen expression of ovarian cancer patients were included in the multivariate logistic regression analysis. As shown in Table 2, the FIGO stage, degree of differentiation, lymph node metastasis, and HE4 and Lewis y expression are independent risk factors for drug resistance (HR=2.882, 2.241, 6.118, 3.421, and 3.836; P=0.003, 0.015, <0.001, <0.001, and <0.001, respectively); as for the univariate analysis, factors with a P value less than 0.05 were included in the multivariate analysis. Our results suggest that poor differentiation and positive HE4 and Lewis y antigen expression are independent risk factors for chemotherapy resistance in ovarian cancer (HR=2.953, 2.972, and 3.254, respectively; P=0.022, 0.004, and 0.004, respectively) ( Table 2).

HE4 expression is an independent risk factor that affects prognosis in ovarian cancer
To further analyze the overall survival (OS) rate, we followed-up with the two ovarian cancer patient groups and saw that, among the 95 patients,56 patients died, 35 survived, and 4 failed to follow-up; the overall median survival time of the patients was 48.00 ± 1.55 months (95% CI, 44.97-51.03 months).
Although the median survival time of patients with positive Lewis y expression was lower than that of patients with negative expression, the difference was not statistically signi cant (45.00 ± 2.76 months vs. 54.00 ± 5.18months; P = 0.016) ( Table 3); moreover, there was no correlation between pathological differentiation, pathological type, lymph node metastasis and patient OS (P>0.05) (Table 3).
Further, only factors with a P value less than 0.05 after KM analysis were included in the Cox multivariate model analysis; results showed that HE4 positive expression and chemotherapy resistance were independent risk factors related to the prognosis of ovarian cancer patients (HR = 2.894 and 3.512, respectively; P<0.001 and 0.045, respectively) ( Table 4).
Correlation between HE4 and Lewis y at protein and mRNA levels cases+, 11 cases++, 5 cases +++) (Figure 2A); the correlation between the expression of these two proteins was further con rmed by Spearman correlation (R = 0.327, P = 0.001).
To explore the general signi cance of the correlation between HE4 and Lewis y, we investigated the correlation between the mRNA expression of their respective genes (WFDC2, FUT1) using the large sample databases TCGA, CCLE, and GTEx. As shown Figure 2B, the correlation was con rmed via the TCGA pan-cancer (R=0.351, P<0.001) and the TCGA ovarian cancer tissue (R=0.268, P<0.001) databases; moreover, we noted similar results using the CCLE pan-cancer (R=0.450, P<0.001) and the CCLE ovarian cancer cell line (R=0.478, P <0.001) databases ( Figure 2C), along with the GTEx database(R= 0.410, P<0.001) ( Figure 2D). Therefore, the expression of HE4 and Lewis y were shown to have a positive correlation via all the aforementioned databases. This correlation has universal signi cance in various types of cancer tissues and cells, especially in ovarian cancer cells and tissues, and the potential interaction mechanism between them deserves further exploration.
Identi cation of hub genes by gene expression and enrichment analyses During our previous analysis of gene expression sequences in ovarian cancer cells, we found several genes that were affected by HE4 and Lewis y expression. In the case of HE4, 1,113 genes were up-regulated with the increase in HE4 expression while 1,531 genes were down-regulated with its downregulation 14 . As for Lewis y, 177 genes were up-regulated while124 were down-regulated 13 . Here, we found through Venny analysis, 23 up-regulated and 9 down-regulated genes, that were commonly affected by the two proteins. The overlapped genes were then uploaded to Metascape for an enrichment analysis; we found extensive interactions between these overlapped genes ( Figure 3B). Moreover, GO enrichment analysis revealed that the two proteins were involved in the positive regulation of different elements, including the p38MAPK cascade, cell cycle, and tumor necrosis factor, in addition to biological behaviors such as response and autophagy ( Figure 3C, D).
Moreover, pathway analysis suggested that the overlapped DEGs are involved in pathways related to MAPK signaling, HIF-1 signaling, TNF signaling, and cell cycle regulation; the involved genes and associated P values are shown in Figure 4A. Interestingly, the MAPK pathway was the most signi cantlyenriched pathway and had the highest number of involved genes ( Figure 4B), which suggests this pathway plays a big role in the malignant biological behavior of ovarian cancer.
Furthermore, Cytoscape analysis was performed to identify the hub genes among the overlapped genes. We used the MCC algorithm in the cytoHubba plugin to analyze the PPI and ranked the genes according to their obtained scores; the top 10 genes with the highest scores were considered hub genes ( Figure  4C). The MCODE plug-in was used to further analyze the PPI; from the module analysis, we obtained a module containing four nodes and six edges. The obtained module was composed of four up-regulated genes, which were among the 10 hub genes obtained by cytoHubbaanalysis( Figure 4D). Therefore, we con rmed that VEGFA, EGR1, HIFIA, and PTGS2 were hub genes that play a core role in the network diagram.

Discussion
Ovarian cancer is a common malignant tumor with the third-highest morbidity rate and the highest mortality rate among female reproductive system cancers; what is even more unsettling is the low ve-year survival rate, which does not exceed 45%. Further, more than 75% of ovarian cancer patients are discovered at advanced stages of the disease 1 , which is one of the main reasons behind the poor prognosis of ovarian cancer, along with chemotherapy resistance. Therefore, developing e cient methods for early screening and drug resistance prediction is crucial; of late, many studies have been aimed at exploring the mechanism of drug resistance in ovarian cancer and developing treatment strategies to overcome this menacing phenomenon.
Human epididymal protein 4 (HE4), also known as WFDC2, is a member of the Whey acidic protein family that was rst discovered in human distal epididymal epithelial cells. In 2008, the US Food and Drug Administration (FDA) identi ed HE4 as a serum marker for monitoring recurrence and progression in epithelial ovarian cancer (EOC) patients. Thereafter, HE4 started gradually stirring more and more attention; several studies have demonstrated the usefulness of HE4 in the early diagnosis of EOC and the identi cation of pelvic masses, especially when used in combination with the risk of ovarian malignancy algorithm (ROMA); moreover, HE4 exhibited higher sensitivity and speci city than its predecessors, including CA125 3,18 . In terms of function and mechanism, HE4 can promote the proliferation, invasion, metastasis, and drug resistance of ovarian cancer cells 10,19,20 . This protein also forms the HE4-ANXA2 complex by combining with membrane-associated protein A2 (ANXA2) and promotes malignant behaviors such as cell proliferation, adhesion, invasion, and metastasis through MAPK and focal adhesion signaling pathways 20,21 . Notably, HE4 knockdown can inhibit the invasive, cell growth, and tumor progression of ovarian cancer by inhibiting the JAK / STAT3 pathway 22 .
Recent studies have found that HE4 is associated with chemotherapy resistance in tumor cells [23][24][25] ; HE4 could induce drug resistance in ovarian cancer cells by activating the AKT and ERK pathways and was associated with poor prognosis in ovarian cancer patients 24 ; HE4 could also interact with EGFR, IGFR, and, the transcription factor, HIF1α, thereby promoting ovarian cancer cell invasiveness and drug resistance and, consequently, worsening the prognosis of ovarian patients 26 ; the over expression of HE4 was also noted to promote cisplatin-and paclitaxel-resistance in ovarian cancer, where the down-regulation of HE4 reversed resistance to many drugs.
In this study, we used para n sections of 95 ovarian cancer patient tissues and evaluated the predictive value of HE4 in chemotherapy-resistance by immunohistochemistry. We found that HE4 positive expression was signi cantly higher in tissues from drug-resistant patients than in tissues from the sensitive group. Notably, HE4 was an independent risk factor for chemotherapy resistance whose high expression signi cantly lowered patient survival; interestingly, HE4 expression and chemotherapy resistance were both independent risk factors for patient prognosis. These results indicate that HE4 may be a predictor of chemotherapy resistance and poor prognosis in ovarian cancer and may provide a reference for future studies aiming to further explore the role of HE4 in chemotherapy-resistant ovarian cancer.
Sugar complexes are an important part of the cell membrane; exposed sugar chains are involved in important processes such as cell growth and differentiation 27 . Lewis y antigen, a tumor-associated carbohydrate antigen (TACA), is a double fucosylated oligosaccharide that is over-expressed in epithelial malignancies including ovarian, pancreatic, prostate, rectal, and non-small cell lung cancers. Because the Lewis y antigen is located on the cell surface, it can modify the structure of cell surface membrane proteins, thereby affecting cellular functions. Previous studies by our group have proven that Lewis y is an important component of TGF-β, EGFR, ANXA2, CD44, CD147, MUC1 and other cell membrane proteins [28][29][30] ; moreover, Lewis y glycosylation further promoted malignant biological behaviors in ovarian cancer such as cell proliferation, invasion, metastasis, drug resistance, autophagy, and apoptosis 6, 10, 31 . Furthermore, we have also identi ed Lewis y antigen modi cations on the HE4 structuure 8 . In the current study, we found that Lewis y antigen is an independent risk factor of chemotherapy resistance whose expression in the resistant group was signi cantly higher than that in the sensitive group, similar to what we saw with HE4; the two proteins, HE4 and Lewis y, were positively correlated. Next, after analyzing the correlation between the mRNA expression of HE4 and Lewis y using large-sample databases (TCGA, CCLE, GTEx), we con rmed that these compounds were positively correlated at the mRNA level. These ndings provide some evidence regarding the relationship between HE4 and Lewis y; however, further pre-clinical studies are needed to prove the potential correlation between HE4 and Lewis y. Nonetheless, HE4 and Lewis y play an important role in chemotherapy-resistant ovarian cancer and may have potential applications in clinical practice.
In order to further explore the potential underlying mechanisms behind the effect of HE4 and Lewis y on the biological behavior of ovarian cancer, we conducted bioinformatic analyses of these two indicators. Although we have previously explored gene expression alterations in ovarian cancer cell lines after transfection with WFDC2 and FUT1, we did not explore how the transfected genes altered cellular genetic pro les; however, commonly-altered genes can provide essential information regarding the mechanism by which HE4 and Lewis y affect the malignant biological behavior of ovarian cancer. Therefore, we performed Venny analysis and found 32 overlapped genes, 23 of which were up-regulated while nine were down-regulated. We went further and analyzed these genes using Metascape in order to identify the most commonly-enriched GO terms; among the different pathways highlighted, several were related to oncogenetic processes, such as: "positive regulation of p38MAPK cascade," "positive regulation of cell cycle," "response to tumor necrosis factor," and "autophagy." These results con rmed that HE4 and Lewis y participate in tumorigenesis and drug resistance in epithelial ovarian carcinomas. In addition, through online pathway analysis, we found that these overlapped DEGs are involved in important signaling pathways such as MAPK, VEGF, HIF-1, and TNF, among which MAPK was the most signi cantly-activated pathway.
The MAPK signaling pathway regulates almost all cellular processes, including gene expression, cell cycle, cell survival, cell death, and cell movement 32 . There are multiple mechanisms underlying HE4-mediated chemoresistance, including dysregulation of MAPK signaling (inhibition of EGR1 and p38) and changes in the level or stability of tubulin; notably, recombinant HE4 up-regulates the expression levels of α-tubulin, β-tubulin, and microtubule-associated protein tau (MAPT) 33 . Moreover, TGF-β1 regulates the expression of Lewis y by activating the MAPK/c-Fos pathway, thereby further affecting the proliferation, invasion, and prognosis of malignant epithelial tumors 34 . In malignant melanoma, siRNA inhibition of Lewis y expression can prevent the activation of the EGFR/MAPK pathway and inhibit cell proliferation 35 . The above-mentioned studies support our ndings regarding the relationship of HE4 and Lewis y and the MAPK signaling pathway. Therefore, we speculate that Lewis y modi ed HE4 promotes the chemotherapy-resistant biological behaviors of ovarian cancer through the p38MAPK signaling pathway. We then constructed a PPI network, using the MCC and MCODE algorithms, and identi ed four hub genes among the overlapped genes (VEGFA, EGR1, HIF1A, and PTGS2); analyzing the known functions of these genes can provide ideas and breakthroughs for prospective studies on HE4 and Lewis y.
VEGF-mediated MAPK/ERK and PI3K/AKT signaling pathways play an important role in regulating the growth, proliferation, migration, and angiogenesis of tumor cells and vascular endothelial cells 36 ; moreover, VEGFA has a known role in regulating tumor cell chemotherapeutic drug sensitivity, In addition to its important function in reproductive health 37,38 . EGR1 is bidirectional regulator of pathological process in malignant tumors; in breast cancer and lung cancer tissues, the EGR1 down-regulation plays a tumor-suppressive role and inhibits the proliferation and migration of tumor cells 39,40 ; whereas, in colorectal and prostate cancers, EGR1 plays an oncogenic role where its signi cant up-regulation promotes the proliferation and migration of tumors 41,42 . Recent studies have found that EGR1 is associated with drug resistance 43 and metastasis 44 in breast cancer, as well as autophagy 45 . Numerous studies have suggested a correlation between HIF1α and drug resistance in tumors, such as ovarian cancer 46,47 . Besides, HIF1α has been also shown to affect drug resistance in tumors by regulating apoptosis and autophagy 47,48 . PTGS2, also known as COX-2, plays an important role in in ammation and cancer development; its expression plays a vital role during various stages of a large number of cancer types [49][50][51] . Recent studies have found that COX-2 is associated with drug resistance in breast cancer 50 and bladder cancer 51 . Studies on a variety of malignant tumors have shown that autophagy is related to tumor drug resistance [52][53][54] ; however, this effect is bidirectional. Moreover, autophagy can also promote the death of drug-resistant cells 55

Conclusions
In conclusion, we speculate that the modi cation of Lewis y residues on HE4 can promote chemotherapy resistance in ovarian cancer by activating p38MAPK signaling. Furthermore, we believe that the four hub genes VEGFA, EGR1, PTGS2, and HIF1A play an important role in the pathogenesis of ovarian cancer and may open the door for better screening methods and/or therapeutic targets after further investigations.  Table4. COX regression analysis of prognosis. Figure 3 Venny analysis and GO enrichment analysis. Through Venny analysis, we found 23 overlapped genes that increased together with HE4 and Lewis y, and 124 overlapped genes that decreased together with HE4 and Lewis y (panel A). We performed an online Metascape analysis of these differentially expressed gene lists andnoted the presence of an interaction between them (panelB); GO analysis suggested that these differentially expressed genes are involved in important pathways including MAPK, cell cycle, etc. (panel C). The obtained terms and their respective P values were used to construct a network graph (panel D), where the left side is sorted by enriched terms and the right side is sorted by P value.