2.1 Clinical samples

Tissue samples from ovarian cancer patients without drug treatment were collected at Fudan University Shanghai Cancer Center from December 25, 2012 to August 31, 2017. Samples were stored at −80℃ until RNA isolation. The patients who recurred within 6 months after chemotherapy were regarded as chemotherapy insensitivity. Instead, the patients who did not recur over 6 months were regarded as chemotherapy sensitivity. A total of 44 patients (range, 30-79 years old) were recruited including 14 chemotherapy insensitive and 18 sensitive patients. The pathological stage was defined according to UICC/AJCC and TNM classification system (https://www.uicc.org/resources/tnm), details were shown in Table 1. The research was authorized by the Ethics Committee of Fudan University Shanghai Cancer Center and informed consent was obtained from all patients.

2.2 RNA isolation and sequencing

RNA was isolated from tumor tissues using TriReagent (Ambion Inc). Agarose gel electrophoresis was performed to determine the extent of RNA degradation and contamination. The purity of the RNA was also measured by Nanodrop (ND-1000). The concentration was precisely quantified by Qubit. The integrity was assayed by Agilent 2100 and samples with a RIN value of 7 or above were used for further analysis. Small RNA sequencing libraries were created according to the IlluminaHTruSeq TM Small RNA Sample Preparation protocol. Segment sizes were selected by AMPure XP beads, and PCR enrichment was conducted to obtain the final cDNA library. HiSeq sequence was conducted after the library passed the inspection.

2.3 Identification of DEGs

The quality of RNAseq data was controlled by Fastp software. RNA-seq data of OC tissues was blasted to the Hg19 human reference genome by STAR software. Quantification and standardization of genes used RSEM software (count FPKM TPM). The read count data were then analyzed to identify DEGs in OC tissues and normal tissues through edgeR v3.12.0 in R v3.2.2 with false positive discovery (FDR) correction. The read count data of normal tissues belong to the Common Fund's Genotype-Tissue Expression (GTEX) (http://commonfund.nih.gov/GTEx/), the largest international normal tissue database. In addition, the read count data were also analyzed to identify DEGs in chemotherapy sensitive and insensitive tissues by the same method. The SVA software was used to eliminate the batch effect. The genes meeting the conditions of |log2 fold change (logFC)| >2 and P < .01 were considered as DEGs.

2.4 Enrichment analysis of DEGs

Gene Ontology (GO) database (http://www.geneontology.org/) and Kyoto Encyclopedia of Genes and Genomes pathway (KEGG) (http://www.kegg.jp) were used to enrich the DEGs in cancer occurring and chemotherapy sensitivity group separately. We utilized r packages “org.Hs.eg.db47” and “clusterProfile48” to identify the proteins belonging to corresponding pathway. P < .05 set as the threshold.

2.5 Real-time PCR

RNA was extracted from tissues with ovarian cancer using DNA/RNA isolation kit (TIANGEN). Concentrations and purity of RNA were analyzed using a NanoDrop ND-1000 Spectrophotometer. Up to 1 μg total RNA from each sample was reverse transcribed to cDNA using a PrimeScriptTM RT reagent kit (Takara). The following primer pairs were used to evaluate the relative expression level of each mRNA: β-actin: F: 5′-AAGGTGACAGCAGTCGGTT-3′, R: 5′-TGTGTGGACTTGGGAGAGG-3′; JUND: F: 5′-CAAGGACGAGCCACAGACG-3′, R: 5′-CCGTGTTCTGACTCTTGAGGG-3′; MUC5B: F: 5′-AACTGCACCGTGTACCTCTG-3′, R: 5′-TCGTGTTGATGCGGACTTGA-3′. All primers were purchased from Sangon Biotech. Quantitative PCR assays were performed using Hieff UNICON Power qPCR SYBR Green Master Mix (YEASEN) and Applied Biosystems QuanStudio Dx Real-Time PCR system. For each sample, 20 μL reactions were set up containing 10 μL 2× SYBR mix, 0.4 μL PCR forward primer (10 μM), 0.4 μL PCR reverse primer (10 μM), 2 μL template cDNA, and 7.2 μL RNase-free water. All PCR reactions were performed in triplicate. The following cycling protocol was used: 95°C for 3 minutes, then 40 cycles with a two-step programme (95°C for 10 seconds, 60°C for 30 seconds), and completed with a product dissociation cycle. The relative expression values for each target gene are shown as 2^−ΔCt.

2.6 Immunohistochemical staining and evaluation

Formalin-fixed and paraffin-embedded sections were dewaxed in xylene and hydrated in grade alcohol, followed by inhibition of endogenous peroxidase activities with methanol containing 0.3% H₂O₂. After boiling in 10 mmol/L of citrate buffer (pH 6.0) for antigen retrieval and cooling down, the sections were blocked with 1% BSA and incubated overnight at 4°C with rabbit polyalonal antibody to JunD (Abcam, ab28837, 1:200) and MUC5B (Abcam, ab87376, 1:100). On the second day, these sections were incubated for another 45 minutes at 37°C. After washing with PBS, the sections were incubated with HRP-conjugated secondary antibody (Shanghai Long Island Biotech) for 1 hour at room temperature, followed by reaction with diaminobenzidine, and counterstaining with hematoxylin.

3.1 OC-related biological processes and pathways

A total of 1588 DEGs were identified between 44 OC and 133 healthy controls (HCs) (belonging to GTEX), including 945 up-regulated genes and 643 down-regulated genes in OC compared with HCs as shown in volcano plot (Figure 1A). The heat map successfully clustered the two groups of samples separately (Figure 1B).

GO analysis showed that several biological processes, such as cellular component movement, cornification, mitotic sister chromatid segregation, blood circulation, locomotion, protein localization, and chromosome segregation were associated with OC occurring (Figure 1C). KEGG results demonstrated that cell cycle pathway was related to OC occurring (Table 2).

Taken together, the mechanism of OC occurring is complex involving multiple genes and biological processes.

3.2 Chemotherapy sensitivity-related biological processes and pathways

A total of 249 DEGs were identified between 18 chemotherapy sensitive OC patients and 14 chemotherapy insensitive OC patients, 108 genes were up-regulated and 141 genes were down-regulated in chemotherapy sensitive OC patients compared with chemotherapy insensitive OC patients as shown in volcano plot (Figure 2A).

Furthermore, GO analysis demonstrated that DEGs mainly participated in the following biological processes: responding to calcium ion, organophosphorus, purine-containing compound, cAMP, lipopolysaccharide, regulating of synapse organization, structure or activity, and positive regulating of cell projection organization (Figure 2B).

KEGG analysis showed that DEGs were mainly enriched in four cancer-related signaling pathways including MAPK signaling pathway (Figure 3A), ERBB signaling pathway (Figure 3B), IL-17 signaling pathway (Figure 3C) and TNF signaling pathway (Figure 3D). MAPK and ERBB pathway were reported to be associated with cisplatin resistance in OC patients. MAPK inhibitors combined with miR-139-5p or metformin could improve cisplatin sensitivity in cisplatin-resistant ovarian cancer.^{24, 25} NRF2 could affect the sensitivity of ovarian cancer cells to rapatinib and erotinib by regulating the ERBB signaling pathway.²⁶ IL-17 and TNF signaling pathway were identified to be related with chemotherapy sensitivity in ovarian cancer patients for the first time.

3.3 Key genes associated with chemotherapy sensitivity in OC patients

To identify the crucial genes that were associated with chemotherapy sensitivity in OC, we picked out 35 DEGs which were shared in both chemotherapy sensitivity group and OC occurring group (Table 3). Among them, five common DEGs (MUC5B, NR4A1, JUND, NRG1, and JUNB) participating in the four chemotherapy sensitivity-related signaling pathways were regarded as key chemotherapy sensitive genes (Figure 3). We found that JUND, JUNB, NR4A1, and NRG1 were down-regulated in chemotherapy sensitive OC patients while MUC5B was up-regulated in chemotherapy sensitive OC patients.

3.4 Gene expression verified by qPCR and IHC

JUND and MUC5B are involved in IL-17 signaling pathway, and JUND is in the upstream of MUC5B. So we think these two genes are worth to be studied. Their expression levels were further verified by qPCR and IHC. The results of NGS and qPCR existed significant correlation for JUND (n = 40, R = .33, P < .05) and MUC5B (n = 38, R = .65, P < .0001) (Figure 4A,B). The relative expression level of JUND was decreased significantly in 13 sensitive tissues compared with 14 insensitive tissues (P < .05) (Figure 4C). While the relative level of MUC5B was increased in 13 sensitive tissues but without statistically significant (P = .29) (Figure 4D). IHC was performed in eight sensitive tissues and eight insensitive tissues to verify the expression level of JUND and MUC5B. The results showed that, the positive rate of JUND expression in the chemotherapy sensitive group was 37.5%, which was lower than chemotherapy insensitive group (87.5%) (Figure 4E). Fifty percent of samples were positive expression of MUC5B in chemotherapy sensitive group, while the positive rate was 11.1% in chemotherapy insensitive group (Figure 4F).