Derivation and Validation of a Necroptosis-Related lncRNA Signature in Patients with Ovarian Cancer

Background Ovarian cancer (OC) is the leading cause of gynecologic malignant tumors. The role of necroptosis-related lncRNAs (NRLs) in OC remains unclear. This study aims to explore the association between NRLs and prognosis in OC patients. Methods The Cancer Genome Atlas (TCGA) and GTEx datasets were used to obtain OC's data. A NRLs signature associated with overall survival (OS) was constructed by Cox-LASSO regression analysis in training cohort for calculating risk score and then validated in testing cohort. Subsequently, the area under the curve (AUC) and Kaplan–Meier survival analysis were used to evaluate the predictive accuracy of the risk score. Finally, the immune infiltration and functional enrichment were compared between different risk groups. Results A 8-NRLs signature including AC245128.3, AL355488.1, AC092794.1, AC068888.2, AL590652.1, AC008982.2, FOXP4-AS1, and Z94721.1 was identified to assess the OS of OC. Kaplan–Meier survival analysis, AUC value, and Cox regression analysis confirmed its predictive value and showed that the clinical outcomes were worse for high-risk patients. There were also differences in immunological functioning and immune pathways between the high-risk and low-risk groups. Conclusions The signature based on eight NRLs has significant values in predicting prognostic prediction in OC, as well as providing a new sight for targeted therapies.


Introduction
Ovarian cancer (OC) is a common cancer which is the leading cause of cancer-related mortality in the reproductive system worldwide. In Asian countries, its incidence showed an upgraded trend in recent years [1]. As a heterogenous group of malignancies, its prognosis is obviously different in various pathological stages and classifications. Despite advances in early detection and drug development, clinical outcomes still remain unsatisfactory. When diagnosed, 60%-75% of patients are already in the middle and advanced stages (stage III-IV) and their 5-year overall survival (OS) rate is less than half [2].
ere is an urgent need to search for accurate predictors and prospective prognostic markers in order to contribute to a better prognosis of OC and give trustworthy information to guide suitable personalized treatment options.
Necroptosis is a type of planned cell death that combines the characteristics of both necrosis and apoptosis [3]. Recent researches have revealed the important role of necroptosis in tumorigenesis and metastasis [4]. Meanwhile, as a member of the noncoding RNA family, long noncoding RNA (lncRNA) lacks the capability of encoding a protein [5]. Along with the development of related researches, accumulating studies suggest that lncRNAs act as a vital role in development, progression, cell survival, and genesis of tumor [6,7]. Many lncRNAs are associated with drug response in various cancers, including OC [8]. To date, emerging evidence has shown the potential of lncRNAs in regulating necroptosis for cancer biology [9]. However, the clinical significance of necroptosis-related lncRNAs (NRLs) remains largely unknown. Remarkably, it is crucial to figure relevant lncRNAs closely linked to necroptosis and prognosis in OC.
is research was conducted to analyze the lncRNAs expression dataset in OC from GTEx database and e Cancer Genome Atlas (TCGA), and screened for NRLs with prognostic value. An eight autophagy-related lncRNA signature of predicting the OC prognosis was identified.

Datasets and Data Preprocessing.
e TCGA and GTEx databases were used to gather the clinical data (n � 587), RNA sequencing profiles (n � 379), and normal ovarian epithelial tissue RNA sequencing profiles (n � 88) for OC. Only 374 individuals were kept for further study after excluding patients lacking RNA sequencing and survival time.
e patients were split into two groups (training cohort and testing cohort) at a ratio of 3 : 7 using the R software "caret" package. With the usage of annotation documents of the GENCODE database, lncRNAs and protein-coding genes were identified. Moreover, based on earlier research, 67 necroptosis-related genes (NRGs) were retrieved. Meanwhile, the "limma" program was used to investigate the differences in NRG expression between OC and normal samples. Subsequently, using the 13,832 lncRNAs and differential expression NRGs discovered, Pearson correlation analysis was performed (p < 0.001, correlation coefficient >0.4). A total of 161 NRLs were eventually chosen for further bioinformatics investigation.

Construction of a Prognostic
Signature. Univariate Cox regression analysis was conducted to identify prognostic lncRNAs (p value <0.05). ese predictive lncRNAs were then used to identify lncRNAs implicated in signature creation with multivariate Cox, least absolute shrinkage and selection operator (LASSO) regression analysis. To develop the model and manage the complexity of LASSO regression, we employed the suitable λ. e following formula was used to determine the risk score: risk score for In addition, patients with different prognostic lncRNAs expressions were classified by using ConsensusClusterPlus package. By locating the inflection point of the sum of squared errors, the ideal k value was determined (SSE). e decline slowed down after k � i, and k � i was selected.

Exploring Clinical Benefit.
e risk score for each OC patient was calculated using the algorithm above. Risk signatures for predicting survival were assessed by area under the curve (AUC) and Kaplan-Meier survival analysis. e median value in the ROC curve, which is used to choose "high-risk" and "low-risk" groups, was determined by calculating the risk score of each patient. In training and testing cohorts, we pooled clinical variables and ran univariate and multivariate Cox regression analyses, respectively. Finally, based on the coefficients of the above multivariate Cox regression, to create a nomogram, we utilized R software "regplot" package.

Immune Infiltration Analysis.
We employed 6 methods (TIMER, CIBERSORT, QUANTISEQ, MCP-counter, XCELL, and EPIC) to estimate the abundances of immune cells classified by risk groups in order to investigate variations in immune cell infiltration. SsGSEA and ESTIMATE algorithms were also performed to compare immunological functions and pathways in low-and high-risk groups. More crucially, we looked at how immunological checkpoint and human leukocyte antigen-associated genes were expressed in various risk groups. e "ggplot2" and "clusterProfiler" programs in R software were then used to conduct gene GSEA enrichment analysis using differently risk groupings.

Drug Sensitivity Analysis.
e IC50 was computed in R using the pRRophetic package, and the medicines were found in the Genomics of Drug Sensitivity in Cancer database.

Statistical Analysis.
e R programming language conducted all statistical analyses (v.4.0.1). In the preceding section, detailed statistical approaches for transcriptome data processing were discussed. e difference is statistically significant when P < 0.05.

Landscape of Necroptosis-Related Genes.
According to the expression of 67 genes associated with necroptosis between normal and tumor samples, we finally got 21 specific necroptosis-related genes NRGs in OC (|Log2FC|>1 and P < 0.05), as shown in Figure 1(a). Of them, 8 were upregulated, and 13 were downregulated (Figure 1(b)). As shown in Figure 1(c), the correlation analysis of 21 NRGs showed that DDX58 had the strongest positive correlation with CYLD (r � 0.39).

Identification of NRLs.
For TCGA-OC cohort, the aforementioned 21 NRGs, as well as all annotated lncRNAs, were used and analyzed through Pearson correlation analysis (correlation coefficients>0.4 and p < 0.001). 161 NRLs were identified (Figure 2(a)). ese NRLs were adopted in univariate Cox analysis to find survival-related NRLs. Finally, for the following analyses, 11 NRLs were examined ( Figure 2(b)). Interestingly, the Wilcox test revealed that all survival-related NRLs were highly expressed in the tumors except MYCNOS and AC245128.3, which was highly expressed in the normal samples (Figures 2(c)-2(d)).

Construction of a Risk Signature.
A total of 11 NRLs were subjected to LASSO regression analysis in order to reduce the number of genes in the signature (Figures 3(a)-3(b)). After that, 10 NRLs were recovered using LASSO and submitted to multivariate Cox regression analysis (stepwise method) to develop a risk stratification system (Figure 3(c)). Eventually, the risk score for OC patients was determined by multiplying the

Clinical Benefits of Risk Signature.
e OC patients of TCGA cohort were separated into two risk categories based on the median value of risk scores: high risk and low risk.
All of these showed that the high-risk group had a poor prognosis (

Construction of Nomogram.
We depicted the risk signature based on the aforementioned risk formula since the risk signature's formula is hard and the nomogram may intuitively relate to clinical practice. In multivariate Cox regression, we merged statistically significant indications to create a visual prognostic model ( Figure 6(a)). Furthermore, the nomogram's calibration curve revealed that the prediction curves in two cohorts are almost identical to the standard curve. e projected survival rate is closely similar   3.6. Immunity Analysis. e main enrichment pathways of various risk categories were investigated using the GSEA algorithm. In the high-risk group, CHEMOKINE SIG-NALING PATHWAY, CYTOKINE RECEPTOR IN-TERACTION, and HEMATOPOIETIC CELL LINEAGE were dominant (Figure 7(a)). In the low-risk group, important pathways were BASAL CELL CARCINOMA, HEDGEHOG SIGNALING PATHWAY, and RIBOSOME ( Figure 7(b)). A heatmap of immune infiltration was created, based on six algorithms (TIMER, CIBERSORT, QUANTISEQ, MCP-counter, XCELL, and EPIC) (Figure 7(c)) to investigate the association between various risk groups and immune cell infiltration. Interestingly, in TIMER algorithm, T cell CD8 + , neutrophil, macrophage, and myeloid dendritic cell were positively correlated with risk score (Figure 7(d)). Meanwhile, results of ESTIMATE algorithm revealed high-risk group had a higher score in stromal, immune, and estimate ( Figure 8(a)). ose who are at high risk are more likely to develop hot tumors and react to immunotherapy. e "ssGSEA" R package was used to quantify the enrichment scores of immune cell subpopulations and their related activities in order to further investigate the relationships between risk scores and immune cells and functions (Figure 8(b)). Given the significance of checkpoint immunotherapy, it is worth emphasizing that there are large variances in immune checkpoint expression across various risk categories (Figure 8(c)).

Drug Effectiveness Analysis.
e drug sensitivity of OC chemotherapeutic drugs, which are often used in clinics, was examined. Six chemotherapeutic drugs had their IC50 values measured in OC patients, and four of them had statistically significant differences across risk subgroups (Figure 9). e high-risk group had considerably higher IC50 values of etoposide, docetaxel, doxorubicin, and cisplatin (P < 0.05). It was discovered that OC patients in the low-risk group, as determined by their risk profile, were more responsive to the chemotherapeutics mentioned above.

Molecular Subtypes Based on NRLs.
e OC patients from TCGA were divided into different subtypes based on consensus algorithm and 8 NRLs expression in risk signature (Figure 10(a)). Interestingly, the majority of C2 subtypes were classified as high-risk group, and the majority of C1 subtypes belonged to the low-risk group (Figure 10(b)). C2 showed the worst prognostic efficacy, while C1 had a higher probability of survival (Figure 10(c)). Moreover, results of ESTIMATE algorithm revealed C2 group had a higher score in stromal, immune, and estimate. Heatmap also showed that the molecular subtypes were statistically significant with risk groups and FIGO staging (Figure 10(d)). Taken together, our molecular subtype results provided another insight into the ability of 8 NRLs to differentiate patients (Figure 10(e)).

Discussion
OC is a heterogenous disease. Although overall survival rates for OC patients have improved greatly, metastasis and recurrence are the leading causes of death [10]. Age, pathological stage, lymph node metastasis, and distant metastasis have all been employed in clinical studies to predict the prognosis of       Journal of Oncology OC, but the accuracy is inadequate. e different clinical outcomes in patients with the same staging indicate that the traditional staging system is not fully functional to predict the prognosis. Relevant biomarker tumor of prognosis needs to be investigated urgently. Many aspects of ovarian cancer biology are regulated by lncRNAs [11]. Necroptosis is a sort of cell death which is regulated, acting as a double-edged sword in the cancer development. Key mediators of its pathway alone or in combination can promote tumor metastasis and progression [12]; however, necroptosis, on the other hand, has also been reported as a fail-safe mechanism that protects against tumor development when apoptosis is impaired [13]. Most certainly, necroptosis plays a big part in cancer. e lncRNAs related to necroptosis may be novel molecular biomarkers and therapeutic targets for OC. Owing to the important role of necroptosis in cancer, its related lncRNA has aroused more attention. With unprecedented accumulation of tumor data in international public databases, now is the new era of data technology for OC research [14,15]. lncRNAs were involved in development, progression, and metastasis of cancer through various signal pathways [16,17]. Recent report introduced a prognostic risk model of NRLs in gastric cancer [18], while the significance of NRLs in OC prognosis was yet unknown. Twenty-one specific NRGs were differentially expressed between normal and OC samples. Based on it, we identified the 8 lncRNA prognostic signatures related to necroptosis (AC245128.3, AL355488.1, AC092794.1, AC068888.2, AL590652.1, AC008982.2, FOXP4-AS1, and Z94721.1) as an adjunct to established clinical prognostic factors for OC. Until now, a number of studies of FOXP4-AS1 (lncRNA forkhead box P4 antisense RNA 1) in cancer had been reported. An experimental study demonstrated that by sequestering miR-3184-5p to upregulate FOXP4-AS1, high expression of FOXP4-AS1 boosted cell proliferation and inhibited apoptosis, demonstrating an oncogenic effect in prostate cancer [19]. Zhao et al. [20] found FOXP4-AS1 had a significant influence in cervical cancer development. Additionally, FOXP4-AS1 acted as an unfavorable prognostic factor in neoplastic disease, such as nasopharyngeal carcinoma, mantle cell lymphoma, and colorectal cancer [21][22][23]. Also, AC092794.1 was used to construct a risk signature for lung adenocarcinoma [24]. So far, little is known regarding the effect of other lncRNAs. In consequence, it is necessary to conduct experiments to further clarify their biological functions in OC.
To explore the signature feasibility, univariate and multivariate COX analysis, and ROC analysis were applied to compare this prognostic feature with clinical indexes. e samples were separated into high-risk and low-risk subgroups based on the median risk score. High-risk individuals had a worse clinical result. e findings revealed that the NRL signature might be a predictive factor for OC patients, which was reliable and stable. It was also suggested that the risk model of the 8 NRLs is superior to other clinicopathological factors. e relevance of the NRL signature in OC was further shown by correlative immune algorithms. e immune cell enrichment analysis revealed a close connection between necroptosis and tumor immunity. e high-risk and lowrisk groups had significantly different immune scores, stromal scores, and ESTIMATE scores. High-risk groups might be more prone to be hot tumors and responding to immunotherapy.

Conclusions
In summary, a novel necroptosis-related prognostic risk model consisting of 8 lncRNAs (AC245128.3, AL355488.1, AC092794.1, AC068888.2, AL590652.1, AC008982.2, FOXP4-AS1, and Z94721.1) was identified for OC patients. Furthermore, the lncRNA signature may be used to guide personalized treatment and enhance the prognosis of OC patients. Given the scarcity of studies on the mechanism and interactions among various NRLs in OC, further research is needed to confirm the clinical utility and reveal the underlying pathways.

Conflicts of Interest
e authors have no conflicts of interest to declare.

Authors' Contributions
Linling Zhu and Hao Chen contributed to the conception and design of the study. Jiaoyan He and Xinyun Yang contributed to acquisition of data. Jianfeng Zheng and Wenhua Liu provided administrative, technical, and material support. Linling Zhu and Jiaoyan He contributed to the writing of the manuscript. Wenhua Liu and Hao Chen contributed to revision of the manuscript and graphic design. All authors reviewed the manuscript and approved the version to be published. Linling Zhu and Jiaoyan He have contributed equally to this work and shared first authorship.