Abstract

Background

   Ovarian cancer is difficult to treat and is, therefore, associated with
   a high fatality rate. Although targeted therapy and immunotherapy have
   been successfully used clinically to improve the diagnosis and
   treatment of ovarian cancer, most tumors become drug resistant, and
   patients experience relapse, meaning that the overall survival rate
   remains low.

Aims

   There is currently a lack of effective biomarkers for predicting the
   prognosis and/or outcomes of patients with ovarian cancer. Therefore,
   we used published transcriptomic data derived from a large ovarian
   cancer sample set to establish a molecular subtyping model of the core
   genes involved in necroptosis in ovarian cancer.

Methods and Results

   Clustering analysis and differential gene expression analyses were
   performed to establish the genomic subtypes related to necroptosis and
   to explore the patterns of regulatory gene expression related to
   necroptosis in ovarian cancer. A necroptosis scoring system (NSS) was
   established using principal component analysis according to different
   regulatory patterns of necroptosis. In addition, this study revealed
   important biological processes with essential roles in the regulation
   of ovarian tumorigenesis, including external encapsulating structure
   organization, leukocyte migration, oxidative phosphorylation, and focal
   adhesion. Patients with high NSS scores had unique immunophenotypes,
   such as more abundant M2 macrophages, monocytes, CD4^+ memory T cells,
   and regulatory T cells. Immune checkpoint CD274 had a greater
   expression in patients with high NSS values.

Conclusion

   This NSS could be used as an independent predictor of prognosis to
   determine the sensitivity of ovarian cancer to various small‐molecule
   inhibitors, immune checkpoint inhibitors, and platinum‐based
   chemotherapy drugs.

   Keywords: biomarkers, clustering analysis, necroptosis, ovarian cancer

1. INTRODUCTION

   Although ovarian cancer accounts for only 2.5% of all malignancies
   among females, the high fatality rate associated with it is responsible
   for 5% of cancer‐related deaths.[42] ^1 Owing to the lack of typical
   symptoms and effective screening methods, the percentage of patients
   with advanced‐stage disease (Federation of Gynecology and Obstetrics
   [FIGO] stage III or IV) exceeds 70%, and the 5‐year survival rate of
   patients with ovarian cancer is approximately 46%.[43] ^2

   Ovarian cancer is also difficult to treat because of the high rates of
   relapse, and relapsed tumor cells gradually become drug resistant.[44]
   ^3 Great advances have been made in the past few decades for the
   treatment of ovarian cancer, including the successful clinical
   application of poly ADP‐ribose polymerase inhibitor (PARPi); however,
   resistance to PARPi remains a challenge.[45] ^4 In addition, there are
   no reliable biomarkers or predictive models to guide the selection and
   application of drugs to treat recurrent ovarian cancer, and second‐ or
   third‐line regimens can only be selected based on evidence‐based
   medical data. Consequently, many patients have no effective therapeutic
   options or die from drug‐related adverse reactions.[46] ^5

   Although cancer immunotherapies using immune checkpoint inhibitors
   (ICIs) have achieved remarkable success in clinical practice,[47] ^6
   only a third of patients with cancer respond to ICIs, and hence, the
   number of patients who can benefit is limited. ICI therapy is also
   limited by the occurrence of immune‐related adverse events caused by
   immune overactivation and the subsequent disturbance of immune
   homeostasis.[48] ^7 Indeed, serious adverse events can lead to
   permanent illness, and, in some cases, death. Therefore, it is
   important to develop predictive and prognostic biomarkers for ICI
   therapies to better understand their benefits and risks and to
   effectively select patients for these therapies.

   Inducing cell death has become a promising novel cancer treatment
   strategy since most tumors display innate resistance to apoptosis.[49]
   ^8 Several recent studies have revealed the interactions between
   different cell death mechanisms and antitumor immunity. For instance,
   the combinatorial application of ICIs and drugs that induce pyroptosis
   and ferroptosis exerts enhanced synergistic antitumor activity, even in
   ICI‐resistant tumors.[50] ^9 Necroptosis is a form of regulated cell
   death that has the same morphological features as necrosis.[51] ^10
   Receptor‐interacting serine–threonine kinase 3 (RIPK3) and mixed
   lineage kinase domain‐like (MLKL) play important roles in necroptosis,
   and recent studies have begun to elucidate the detailed molecular
   mechanisms underlying necroptosis under different pathophysiological
   conditions.[52] ^11 In particular, key components of necroptosis appear
   to be downregulated in tumors,[53] ^12 and accumulating evidence has
   indicated that necroptosis‐deficient cancer cells are poorly
   immunogenic, and hence escape natural and therapy‐elicited
   immunosurveillance. Unfortunately, the role of necroptosis in ovarian
   cancer development remains unclear.

   This research aimed to elucidate the role of necroptosis in the
   pathogenesis of ovarian cancer, facilitate the development of new
   molecular typing methods, guide treatment options, and effectively
   select patients who may benefit from immunotherapy.

2. METHODS

2.1. Data preparation and preprocessing

   Clinical information and transcriptomic (log2(FPKM+1)) data for ovarian
   cancer (OV) samples were acquired from The Cancer Genome Atlas (TCGA)
   database using the R package TCGAbiolinks. The expression and survival
   information of 352 tumor samples from TCGA‐OV dataset, which included
   no normal samples, was used for subsequent analyses. Because the
   control group for a specific comparative analysis is different, the
   control and experimental groups in each comparative analysis are
   described in detail in the part “Result” or figures. For example, in
   Section [54]3.1 (the experimental group and control group for
   comparison were also marked in Figure [55]1A), the the control group
   refer to patients ≥60 years old. In Section [56]3.2 (the experimental
   group and control group for comparison were also marked in
   Figure [57]2D), the control group are Cluster2.

FIGURE 1.

   FIGURE 1
   [58]Open in a new tab

   Necroptosis affects the development of ovarian cancer. (A) Gene
   enrichment analysis of necroptosis‐related genes in different age
   groups. (B) Top 20 mutation frequencies in the TCGA‐OV cohort.
   Different columns represent individual patients. The bar chart above
   shows the tumor mutation burden (TMB). The numbers on the right
   indicate the mutation frequency of each gene. The stack bar chart below
   shows the conversion rates in each sample. (C) Copy number variation
   (CNV) of necroptosis‐related genes. The ordinate refers to the
   frequency of CNV variation (that is, the proportion of samples with
   copy number amplification or loss). The different genes are listed on
   the horizontal axis. Red dots represent copy number amplification and
   blue dots represent copy number loss. (D) Interaction network of
   necroptosis‐related genes. (E) Survival analysis of core genes in the
   interaction network.

FIGURE 2.

   FIGURE 2
   [59]Open in a new tab

   Identification of different ovarian cancer subtypes based on
   necroptotic genes. (A) Consistent clustering based on the expression of
   15 core necroptosis‐related genes in the TCGA‐OV cohort. (B)
   Distribution of CDF curves in consistent clustering. (C) PCA analysis
   of the two subtypes. (D) Survival status of different subtypes. (E)
   Comparison of HALLMARK pathways in different subtypes. (F) Comparison
   of ESTIMATE score, immune score, stromal score, and tumor purity
   between the two subtypes.

   The prediction model was verified using the [60]GSE132342 dataset from
   the Gene Expression Omnibus (GEO) database. The expression data of this
   dataset were determined using the [61]GPL26748 platform.

   Validation cohort data were processed as follows: (1) deletion of
   no‐load probe; (2) deletion of probes corresponding to multiple genes;
   (3) calculation of median expression values for multiple probes
   corresponding to the same gene.

   Map0421 necroptosis pathways in Homo sapiens were selected from the
   Kyoto Encyclopedia of Genes and Genomes (KEGG, [62]https://www.kegg.jp)
   website. In total, 160 necroptosis‐associated genes were extracted
   after ENTREZID transformation.

2.2. Gene set enrichment analysis (GSEA), copy number variation (CNV),
genetic mutations, and survival analysis of necroptosis‐associated genes

   GSEA was used to calculate whether selected gene sets differed
   significantly between the experimental and control groups. The
   enrichment of necroptosis‐associated genes in patients with different
   clinical features was analyzed using the R “clusterProfiler” package.
   CNVs are copy number amplifications and deletions in genomic regions
   larger than 1 Kilobase (kb). The frequency of CNV refers to the
   proportion of the number of samples in which CNV occurs to the total
   sample. For example, the frequency of GAIN for H2AC18 gene was
   calculated as follow: 76 (samples with copy number amplification in
   genomic regions larger than 1 kb)/258 (samples with CNV data available
   for analysis) = 0.2946. The frequency of LOSS for H2AC18 gene was
   calculated as follow: 8 (samples with copy number deletion in genomic
   regions larger than 1 kb)/258 = 0.031. A CNV was considered common if
   its population frequency is >1%. The “maftools” package was utilized to
   analyze and visualize mutations in necroptosis‐associated genes in
   TCGA‐OV samples.

2.3. Protein–protein interaction (PPI) network analysis

   The interactions among necroptosis‐associated genes were investigated
   in the STRING database. A PPI network was generated, and a network
   diagram was generated using Cytoscape (v.3.9.1).

2.4. Clustering of ovarian cancer subtypes

   Univariate Cox regression analysis was performed on the 160
   necroptosis‐associated genes using the R “survival” package. Core genes
   were screened using a threshold of p < .05. TCGA‐OV tumor samples were
   classified based on these core genes. These samples were subjected to
   consistent clustering analysis using the “ConsensusClusterPlus” package
   with Euclidean distance and the pam clustering method (reps = 100,
   pItem = 0.8). The Kaplan–Meier (KM) method was used to generate the
   prognostic survival curves for different ovarian cancer subtypes. The
   log‐rank test was utilized to evaluate the significance of differences
   between these curves.

   Gene set variation analysis (GSVA) for different ovarian cancer
   subtypes was performed using the “msigdb.v7.4. symbols.gmt” gene set,
   which was extracted from the MSigDB database. Results with p < .05 were
   considered statistically significant.

2.5. Gene ontology (GO) enrichment and KEGG pathway analysis of
differentially expressed genes (DEGs)

   Differential patterns of expression between the two ovarian cancer
   subtypes were analyzed using the R “limma” package. DEGs were screened
   according to these parameters: log2(Fold Change, FC) > 0.585
   (or < −0.585) and p < 0.05. Specially, log2(FC) > 0.585 means
   up‐regulation, while log2(FC) < −0.585 suggests down‐regulation. The
   “ClusterProfiler” package was used to conduct KEGG and GO analyses of
   the DEGs with the following parameters: pvalueCutoff = 0.05,
   pAdjustMethod = “BH,” and qvalueCutoff = 0.5. The top 10 pathways were
   visualized.

2.6. DEG re‐clustering analysis

   TCGA‐OV samples were re‐clustered based on the identified DEGs using
   the R “ConsensusClusterPlus” package with Euclidean distances and the
   k‐means clustering algorithm (reps = 100, pItem = 0.8). The R
   “survival” package was used to analyze survival in the re‐clustered
   subtypes, and “ggplot2” was used to draw box plots of differences in
   the expression of necroptosis‐related genes.

2.7. Construction of the necroptosis scoring system (NSS)

   To quantify necroptosis for individual tumors, we constructed a scoring
   system named “NSS.” First, univariate Cox regression analysis was
   performed, aiming to discover the hazard ratio (HR) and prognostic
   significance of DEGs. Prognosis‐related genes were defined as those
   with p < .01. Next, the NSS was constructed using principal component
   analysis (PCA) based on the prognosis‐related genes. Principal
   components 1 and 2 were feature scores. NSS was developed using the
   genes with the highest correlation (or anti‐correlation) in the gene
   set, thus reducing the contribution weight of genes with other
   unrelated members of the gene set. The following formula was used:
   [MATH: <mrow><mi>NSS</mi><mo linebreak="goodbreak">=</mo><mi
   mathvariant="normal">∑</mi><mspace width="0.25em"></mspace><mfenced
   open="(" close=")"><mrow><mi
   mathvariant="italic">PC</mi><mn>1</mn><mi>i</mi><mo
   linebreak="goodbreak">+</mo><mi
   mathvariant="italic">PC</mi><mn>2</mn><mi>i</mi></mrow></mfenced><mo>,<
   /mo></mrow> :MATH]
   (1)

   where i represents individual tumor samples.

   Ovarian cancer samples were grouped into different groups based on the
   median values by NSS, and the survival status of patients was analyzed.
   Univariate and multivariate Cox regression analyses were performed to
   explore the independent prognostic value of the NSS. The same formula
   was used to calculate the NSS in the validation dataset.

2.8. Immuno‐infiltration analysis

   The CIBERSORT algorithm of the “IOBR” package was utilized to calculate
   immune cell fractions in the tumor microenvironment, with perm = 200.
   The matrix score, immune score, Estimation of Stromal and Immune cells
   in Malignant Tumor tissues using Expression (ESTIMATE) score, and tumor
   purity were calculated using the R “ESTIMATE” package.

2.9. Predictive value of NSS for chemotherapy drug sensitivity

   The genomics study of cancer drug sensitivity (GDSC) database data were
   analyzed using the R‐encapsulated oncoPredict to assess the IC[50]
   value of a drug in different NSS groups, which can be used to evaluate
   the sensitivity of an antitumor drug. Spearman's correlation was
   calculated between the NSS values and the IC[50] of drugs.

2.10. Statistical analysis

   All statistical analyses were performed using R (v.4.1.2). To determine
   the significance of differences in expression levels, infiltration
   ratios, and various eigenvalues, the Wilcoxon rank‐sum test was used to
   compare two groups of samples, and the Kruskal–Wallis test was used to
   compare multiple groups of samples. In all figures, significant
   differences are indicated as follows: ns = p > .05, * = p ≤ .05,
   ** = p ≤ .01, *** = p ≤ .001, and ^**** = p ≤ .0001. Survival curves
   were generated using the KM method. The significance of differences was
   evaluated using the log‐rank test.

3. RESULTS

3.1. Necroptosis affects the pathophysiology of ovarian cancer

   Based on the ovarian cancer transcriptomic data in the TCGA‐OV dataset,
   we found that necroptosis‐related genes were significantly
   downregulated in patients with ovarian cancer aged ≥60 years
   (Figure [63]1A) and that 27.52% (71/258) of patients harbored genetic
   mutations, including JAK3, NLRP3, DNM1L, RNF31, PARP4, PLA2G4C, PYGL,
   and so on (Figure [64]1B). In addition, CNV was common, especially in
   the H2AC18 and H2AC19 genes (Figure [65]1C). Further exploration of the
   interactions among the necroptosis‐associated genes using String data
   revealed that TNF, IL1B, STAT1, CASP8, and IFNB1 had high network
   connectivity (Figure [66]1D). In addition, patients with high STAT1,
   CASP8, and IFNB1 expression tended to have a better prognosis, as
   determined using KM analysis (Figure [67]1E). Taken together, these
   findings suggest that necroptosis participates in the pathophysiology
   of ovarian cancer.

3.2. Identification of ovarian cancer subtypes based on
necroptosis‐associated genes

   Next, we performed univariate Cox regression analysis on the
   necroptosis‐associated genes. Among the 160 necroptotic genes, 15 core
   genes were selected with p < .05. Based on the expression data for
   these 15 core genes in TCGA‐OV, consensus clustering was conducted
   using the R “ConsensusClusterPlus” package (Figure [68]2A). A
   cumulative distribution function (CDF) curve was generated, based on
   which samples were divided into the two subtypes (Figure [69]2B). The
   differences between Cluster1 and Cluster2 were determined using PCA
   (Figure [70]2C). KM analysis revealed that Cluster1 was associated with
   a better survival than Cluster2 (Figure [71]2D). Therefore, we explored
   the biological behaviors of the different subtypes by using GSVA to
   evaluate differences in HALLMARK pathways (Figure [72]2E). Common
   signaling pathways that were activated in tumors differed significantly
   between different subtypes, including those related to DNA repair,
   Hedgehog, TGF‐β, TNF‐α via NF‐κB, Notch, Wnt‐β‐catenin, angiogenesis,
   Il2‐stat5, and the inflammatory response (Figure [73]2E). Immune cell
   infiltration analysis using the CIBERSORT algorithm revealed
   significant differences between the two subtypes, including M1
   macrophages, follicular helper T cells, γδT cells and CD4 memory
   resting T cells (Figure [74]S2). ESTIMATE data analysis further
   revealed that patients with a Cluster1 subtype had significantly lower
   ESTIMATE and matrix scores than those with a Cluster2 subtype
   (Figure [75]2F). In general, ovarian cancer could be divided into two
   subtypes based on differential expression of necroptotic genes.

3.3. Differential expression analysis and regulatory patterns of different
subtypes based on necroptosis

   To further investigate the different biological behaviors of each
   subtype, DEGs between subtypes based on the core genes were extracted
   using the “limma” package (Figure [76]3A), and the enrichment analysis
   of these genes was performed using the R “clusterprofiler” package. GO
   analysis revealed that the enriched biological processes included DNA
   packaging, protein‐DNA complex subunit organization or assembly, DNA
   conformational changes, nucleosome organization, chromatin assembly or
   disassembly, and chromatin remodeling (Figure [77]3B). Meanwhile,
   enriched cellular components included the protein‐DNA complex,
   nucleosome, DNA packaging complex, chromosomal region, condensed
   chromosome, and kinetochore (Figure [78]3C). In addition, the molecular
   function annotations included protein heterodimerization activity,
   tubulin binding, sulfur compound binding, G protein‐coupled receptor
   binding, and nucleosomal DNA binding (Figure [79]3D). KEGG pathway
   analysis revealed 10 distinct pathways, including systemic lupus
   erythematosus, alcoholism, neutrophil extracellular trap formation,
   viral carcinogenesis, cell cycle, necroptosis, oocyte meiosis,
   progesterone‐mediated oocyte maturation, transcriptional dysregulation
   in cancer, and p53 (Figure [80]3E). To further explore the patterns of
   necroptosis regulation, unsupervised clustering analysis was conducted,
   and then we classified patients into different genomic subtypes
   (Figure [81]3F). Consistent with the clustering performed using core
   necroptosis genes, unsupervised clustering analysis revealed two
   distinct genomic phenotypes, namely geneCluster1/2 (Figure [82]3G).
   Notably, geneCluster2 was also associated with a poor prognosis
   (Figure [83]3H), and most of the 15 genes showing prognostic values
   were differentially expressed between the two genomic subtypes
   (Figure [84]3I). Overall, the two subtypes of ovarian cancer had
   different expression patterns.

FIGURE 3.

   FIGURE 3
   [85]Open in a new tab

   Differential expression analysis and regulatory pattern analysis of
   different subtypes based on necroptosis. (A) Heat map of DEG expression
   distribution in different subtypes. (B) Biological processes (BP)
   enriched in GO analysis. (C) Cellular components (CC) enriched in GO
   analysis. (D) Molecular functions (MF) enriched in GO analysis. (E)
   KEGG pathway enrichment analysis. (F) Results of unsupervised
   clustering analysis of DEGs (classification number κ = 2). (G) Heat map
   of DEGs in different genomic subtypes. (H) Survival status of different
   genomic subtypes. (I) Expression of 15 core necroptosis‐related genes
   associated with prognosis in different genomic subtypes.

3.4. Construction and validation of a scoring system associated with
necroptosis

   To quantify necroptosis for individual patients, a scoring system
   associated with necroptosis, namely NSS, was established. Univariate
   Cox regression analysis identified 51 prognosis‐related genes.
   Principal components 1 and 2 of the PCA method were chosen as the
   feature scores to construct the NSS. According to the median NSS
   values, samples were divided into different NSS groups with a high or
   low score (Figure [86]4A, Figure [87]S1). The low NSS group tended to
   have a better survival prognosis than the high NSS group
   (Figure [88]4B,C). To verify the stability of the prognostic model, the
   [89]GSE132342 cohort was analyzed using the NSS. The survival status
   was also significantly different between different groups, consistent
   with the trend observed in the training cohort. Low‐score patients
   tended to have better survival than those with a high NSS score
   (Figure [90]4D–F). Altogether, the NSS was established based on the
   DEGs between different subtypes of ovarian cancer.

FIGURE 4.

   FIGURE 4
   [91]Open in a new tab

   Construction and validation of a scoring system associated with
   necroptosis. (A) NSS distribution for each sample in the TCGA‐OV
   cohort. (B) Survival status of each sample. (C) Survival analysis of
   high‐ or low‐NSS patients. (D) NSS distribution for each sample in the
   [92]GSE132342 cohort. (E) Survival status of each sample. (F) Survival
   analysis of high‐ or low‐NSS patients.

3.5. NSS was closely associated with the clinical characteristics of ovarian
cancer patients

   Univariate Cox regression analysis indicated that the NSS had a close
   correlation with overall survival (OS; Figure [93]5A). Next,
   multivariate Cox regression analysis revealed that the NSS could be an
   independent predictor of OS (Figure [94]5B). An alluvial map was used
   to visualize the variation in individual patient attributes for
   different groupings based on clusters, gene clusters, NSS values, and
   survival status (Figure [95]5C). The NSS was significantly higher for
   the Cluster2 and geneCluster2 subtypes than for the Cluster1 and
   geneCluster1 subtypes (Figure [96]5D,E); however, there was no
   significant difference in the NSS among different clinical groups
   (Figure [97]S3). In general, the scoring system was closely associated
   with the clinical characteristics of patients with ovarian cancer.

FIGURE 5.

   FIGURE 5
   [98]Open in a new tab

   Association of NSS with clinical characteristics of patients with
   ovarian cancer. (A) Univariate Cox regression analysis of independent
   risk factors for overall survival in patients with ovarian cancer. (B)
   Multivariate Cox regression analysis of independent risk factors. (C)
   Alluvial map showing changes among cluster, geneCluster, survival
   status, and NSS. (D) NSS comparison in different clusters. (E) NSS
   comparison in different geneClusters.

3.6. Exploration of molecular background based on the NSS

   To understand differences in the molecular background of ovarian cancer
   based on the NSS, single sample GSEA (ssGSEA) was performed to analyze
   changes in HALLMARK pathways in the TCGA‐OV cohort. Forty‐two of the
   fifty pathways showed significant differences between groups with high
   or low NSS values (Figure [99]6A). The NSS correlated positively with
   the “apical junction” and “UV response DN” pathways and negatively with
   the “DNA repair” and “E2F targets” pathways (Figure [100]S4). GSEA
   based on gene sets from biological processes in the GO dataset (GO‐BP)
   revealed that the five most significant pathways were involved in
   external encapsulating structure organization, leukocyte migration,
   mitochondrial gene expression, mitochondrial translation, and oxidative
   phosphorylation (Figure [101]6B). Focal adhesion, oxidative
   phosphorylation, Parkinson's disease, ribosomes, and spliceosomes were
   identified through GSEA of KEGG pathways between different NSS groups
   (Figure [102]6C). Using CIBERSORT, we found that the high NSS group
   showed a higher abundance of immune cell subsets, such as M2
   macrophages, monocytes, and CD4 memory resting T cells, than the low
   NSS group (Figure [103]6D). Further analysis revealed that the
   ESTIMATE, immune, and matrix scores of patients with a high NSS were
   significantly higher than those of patients with a low NSS
   (Figure [104]6E). In addition, the expression level of the immune
   checkpoint molecule CD274 was higher in patients with a high NSS than
   in those with a low NSS (Figure [105]6F). Overall, the NSS represents
   the activation of specific oncogenic pathways and immunomodulatory
   molecules.

FIGURE 6.

   FIGURE 6
   [106]Open in a new tab

   Exploration of molecular background based on the NSS system. (A)
   Pathways between groups with high or low NSS values. (B) Biological
   processes with significant changes between high‐ and low‐NSS groups.
   (C) KEGG pathways with significant changes between high‐ and low‐NSS
   groups. (D) Immune infiltration scores (cibersort) of high‐ and low‐NSS
   groups. (E) Comparison of ESTIMATE score, immunoscore, stromal score,
   and tumor purity between the two subtypes. (F) Expression of CD274
   between the two subtypes.

3.7. Potential influence of the NSS as a therapeutic strategy

   Using the R “oncoPredict” package and anticancer drug responses
   available in the Genomics of Drug Sensitivity in Cancer (GDSC), we
   combined transcriptomic data with predicted IC[50] values for various
   drugs in the TCGA‐OV samples. Spearman's correlation analysis between
   the NSS and log2(IC[50]) of drugs suggested that drugs, such as
   sepatronium bromide and selumetinib, had a negative correlation with
   the NSS, whereas drugs, such as nilotinib and oxaliplatin, showed a
   positive correlation with the NSS (Figure [107]7A,B). These drugs
   included various small molecule inhibitors, including MEK/ERK pathway
   inhibitors (selumetinib, trametinib, and VX‐11e), a survivin inhibitor
   (sepantronium bromide), a CDK1 inhibitor (RO‐3306), a BCR‐ABL inhibitor
   (nilotinib), a USP1‐UAF1 inhibitor (ML 323), an estrogen receptor
   antagonist (fulvestrant), a PORCN/Wnt inhibitor (Wnt‐C59), an IDH
   inhibitor (AGI‐6780), and cisplatin sensitivity (oxaliplatin;
   Figure [108]7A,B). We then explored whether the NSS could predict
   patients who could benefit from immunotherapy using the
   Gide_CancerCell_ipiPD1_Pre cohort. Although no significant differences
   in OS were observed between these two groups following KM analysis, the
   low NSS group presented a trend to have a better survival
   (Figure [109]7C). No significant differences in NSS were noted between
   the nonresponsive (PD/CD) and responsive (CR/PR) immunotherapy groups
   (Figure [110]7D); however, the responsive immunotherapy group had a
   lower NSS than did the nonresponsive immunotherapy group
   (Figure [111]7E). Besides, the NSS values correlated with the IC[50] of
   olaparib (Figure [112]7F), and samples with a high NSS tended to have a
   high log (IC[50]) value (Figure [113]7G). In the VanAllen_Science
   cohort with ipilimumab treatment for melanoma, patients with a high NSS
   were reported to benefit from treatment and had an increased survival
   time (Figure [114]7H). Therefore, the NSS has guiding significance for
   the prediction of therapeutic drug efficacy and is conducive for
   decision‐making for treatment programs.

FIGURE 7.

   FIGURE 7
   [115]Open in a new tab

   Potential influence of the NSS as a therapeutic strategy. (A) Left: NSS
   correlated negatively with drug log2(IC50). The top six with the
   highest absolute spearman correlation coefficients are shown. Right:
   Differences in log2(IC50) between high‐ and low‐NSS groups for the six
   drugs. (B) Left: NSS correlated positively with drug log2(IC50). The
   top six with the highest absolute spearman correlation coefficients are
   shown. Right: Differences in log2(IC50) between high‐ and low‐NSS
   groups for the six drugs. (C) Survival of high‐ and low‐NSS groups in
   the immunotherapy cohort. (D) Proportion of patients who did and did
   not respond to immunotherapy in the high‐ and low‐NSS groups. (E) Score
   comparison of patients with and without treatment response. (F)
   Correlation between olaparib IC50 and NSS. (G) Difference in olaparib
   IC50 between high‐ and low‐NSS groups. (H) Survival of high and low‐NSS
   groups in the immunotherapy cohort.

4. DISCUSSION

   Ovarian cancer is a common gynecological cancer that is difficult to
   treat. The overall survival rate remains low as there is currently a
   lack of effective biomarkers for predicting the prognosis or outcomes
   of patients. In this study, we used published transcriptomic data
   derived from a large ovarian cancer sample set to establish a molecular
   subtyping model of the core genes involved in necroptosis in ovarian
   cancer.

   Previous evidence has suggested that necroptosis is essential in
   tumorigenesis, tumor progression, and tumor immunity; however, the
   overall effect of necroptosis in cancer has not yet been determined and
   may depend on the cancer type and disease stage. For instance,
   necroptosis was detected in areas of tumor necrosis and is thought to
   play a critical role in tumor growth and metastasis.[116] ^13 ,
   [117]^14 In contrast, inducing necroptosis can delay tumor progression
   by killing cancer stem cells or ameliorating inflammation.[118] ^15 ,
   [119]^16 , [120]^17 In ovarian cancer, the specific pathophysiological
   function of necroptosis is still unclear, and its relationship with the
   development of ovarian cancer is poorly studied. In this study, we
   performed a systematic analysis of necroptosis in ovarian cancer and
   found that genetic variations in necroptotic genes were frequent. We
   identified a series of necroptosis‐related genes involved in the
   pathogenesis of ovarian cancer, including STAT1, CASP8, and IFNB1, and
   provided a preliminary analysis of their interactions. GSEA
   demonstrated that some essential pathways are involved in necroptosis
   in ovarian cancer, such as DNA packaging, protein‐DNA complex assembly,
   chromatin remodeling and organization, functional changes in
   organelles, cellular adhesion, the recruitment of immune cells, and
   oxidative phosphorylation. These abnormally activated pathways may
   therefore be promising targets for drug intervention research.

   Analyzing the expression of necroptotic genes in large samples to
   establish prognosis‐related gene signatures is important for the
   identification of regulatory patterns, molecular typing of tumors,
   guiding therapeutic options, and predicting therapeutic effects. For
   example, a novel necroptosis‐related gene signature established in
   gastric cancer was able to predict early diagnosis, prognosis, and the
   response to immunotherapy.[121] ^18 Moreover, a necroptosis‐related
   signature was established in hepatocellular carcinoma to evaluate
   prognosis and profile the tumor immune microenvironment.[122] ^19 In
   osteosarcoma, a necroptosis‐related gene signature was constructed and
   found to be strongly associated with immunity.[123] ^20 Further, a risk
   model consisting of five genes (UBD, ISG20, CXCL11, HLA‐DOB, and
   ATP1A3) was constructed using least absolute shrinkage and selection
   operator regression and could predict the survival and prognosis of
   patients with ovarian cancer, while also revealing the immune
   microenvironment to a certain extent.[124] ^21 In this study, we
   established models using two consistent clustering methods. We
   established molecular subtypes based on core necroptosis‐related genes
   and then developed genomic phenotypes of necroptosis based on
   unsupervised clustering analysis of 160 DEGs. The two clustering
   analyses yielded consistent results, suggesting that there are two
   different modes of necroptosis in ovarian cancer. PCA was used for
   dimension reduction to establish prognosis‐related gene signatures to
   preserve the information in the original data to the greatest extent.
   The different subtypes showed striking differences in survival status,
   immune infiltration, stromal abundance, and activated pathways.
   Finally, we constructed a novel scoring system named “NSS,” and
   validated the system using the [125]GSE132342 cohort. The NSS had a
   close correlation with clinical characteristics, composition of immune
   cells or stroma, and activation of essential pathways, such as the
   apical junction, UV response, DNA repair, and E2F target pathways. In
   addition, a high NSS was closely correlated with higher CD274
   expression, dissatisfactory response to PAPRi, poorer survival status,
   and poor immunotherapy sensitivity.

   The close interaction between necroptosis and immune regulation has
   made necroptosis‐mediated anti‐tumor immune response the focus of
   research in recent years, although its specific regulatory mechanisms
   remain unclear. Necroptosis can induce CD8^+ T cell cross‐priming,
   dendritic cell maturation, cytotoxic T cell cross‐priming, and IFN‐γ
   production in response to tumor antigen stimulation, which play
   essential roles in antitumor immunity.[126] ^22 , [127]^23 In
   pancreatic ductal adenocarcinoma, the necroptotic pathway can remodel
   the immunosuppressive tumor microenvironment via CXCL1 and Mincle
   signaling, or switch tumor‐associated macrophages toward an M2‐like
   phenotype, which promotes immune tolerance and immunotherapeutic
   resistance and induces disease progression.[128] ^24 , [129]^25
   Overall, it was probable that chronic spontaneous necroptosis promoted
   tumor development by suppressing antitumor immunity, and massive
   induction of necroptosis could substantially undermine tumor growth and
   trigger immunogenic responses. The NSS established in this study
   correlated significantly with prognosis and could be utilized as an
   independent predictor of prognosis. The NSS correlated significantly
   with different tumor cell regulatory patterns, immune cell composition,
   and immune checkpoint molecule expression. Furthermore, the NSS was
   closely related to the sensitivity of various small molecule
   inhibitors. In addition, the NSS also correlated significantly with
   prognosis in cohorts treated with anti‐PD1 antibodies and was closely
   related to therapeutic responses. Furthermore, the NSS was
   significantly associated with drug sensitivity to olaparib and clinical
   outcomes in the cohort treated with anti‐CTLA4 antibodies. The NSS
   score can indicate the different biological composition and immune
   infiltration patterns in the pathogenesis of the tumor. It is closely
   related to small molecule inhibitors and immunotherapy, suggesting that
   the NSS has important application value in treatment selection,
   efficacy monitoring, and prognosis prediction.

   Despite these notable findings, our study has some limitations. First,
   we were unable to validate the key necroptosis‐related genes in
   clinical samples of ovarian cancer. Second, the specific biological
   functions associated with necroptosis require further exploration and
   verification using in vivo and in vitro experiments. Third, the sample
   size of the cohort used to judge the efficacy of immunotherapy and
   sensitivity of PARP inhibitors was small. Therefore, it is necessary to
   conduct validation studies using a larger sample size to verify the
   utility of the NSS for selecting novel therapies or combination
   therapies.

AUTHOR CONTRIBUTIONS

   Conceptualization: Shuqin Chen and Huiling Lai; Methodology: Guofen
   Yang; Software: Yunyun Guo and Linxiang Wu.; Validation: Aligu Yusufu;
   Formal analysis: Qiyu Zhong, Zhouzhou Liao and Jianyu Ma;
   Writing—original draft preparation, Huiling Lai and Wen Shi;
   Writing—review and editing: Huiling Lai and Shuqin Chen. All authors
   have read and agreed to the published version of the manuscript.

CONFLICT OF INTEREST STATEMENT

   The authors declare no conflicts of interest.

ETHICS STATEMENT

   This research does not include any human or animal studies or studies
   that require the approval of an institutional review board.

Supporting information

   FIGURE S1. ROC curves for NSS in the training and validation sets.
   [130]Click here for additional data file.^ (274.9KB, jpg)

   FIGURE S2. Comparison of immune cell infiltration scores in the two
   subtypes.
   [131]Click here for additional data file.^ (387.4KB, jpg)

   FIGURE S3. NSS differences for different clinical characteristics.
   [132]Click here for additional data file.^ (114.9KB, jpg)

   FIGURE S4. Correlation between NSS and different HALLMARK pathways.
   [133]Click here for additional data file.^ (2MB, jpg)

ACKNOWLEDGMENTS