Prediction of Cervical Cancer Outcome by Identifying and Validating a NAD+ Metabolism-Derived Gene Signature

Cervical cancer (CC) is the second most common female cancer. Excellent clinical outcomes have been achieved with current screening tests and medical treatments in the early stages, while the advanced stage has a poor prognosis. Nicotinamide adenine dinucleotide (NAD+) metabolism is implicated in cancer development and has been enhanced as a new therapeutic concept for cancer treatment. This study set out to identify an NAD+ metabolic-related gene signature for the prospect of cervical cancer survival and prognosis. Tissue profiles and clinical characteristics of 293 cervical cancer patients and normal tissues were downloaded from The Cancer Genome Atlas database to obtain NAD+ metabolic-related genes. Based on the differentially expressed NAD+ metabolic-related genes, cervical cancer patients were divided into two subgroups (Clusters 1 and 2) using consensus clustering. In total, 1404 differential genes were acquired from the clinical data of these two subgroups. From the NAD+ metabolic-related genes, 21 candidate NAD+ metabolic-related genes (ADAMTS10, ANGPTL5, APCDD1L, CCDC85A, CGREF1, CHRDL2, CRP, DENND5B, EFS, FGF8, P4HA3, PCDH20, PCDHAC2, RASGRF2, S100P, SLC19A3, SLC6A14, TESC, TFPI, TNMD, ZNF229) were considered independent indicators of cervical cancer prognosis through univariate and multivariate Cox regression analyses. The 21-gene signature was significantly different between the low- and high-risk groups in the training and validation datasets. Our work revealed the promising clinical prediction value of NAD+ metabolic-related genes in cervical cancer.


Introduction
Cervical cancer is a huge burden for women around the world and a global health concern, as well as becoming one of most frequent cancers that threaten the health and lives of women. CC has contributed significant cancer-associated deaths, especially in developed countries [1,2]. Patients are treated based on the established treatment standards by the international Federation of Gynecology and Obstetrics (FIGO) and the National Comprehensive Cancer Network (NCCN). As treatment measures have matured, most cases of early disease achieve a satisfactory level of survival and prognosis. However, for the advanced stage, the unoptimistic prognosis is ascribed to a lack of timely and accurate measures depending on the clinical condition of patients [3][4][5]. Metastasis and recurrence are significant hurdles to the prognosis of cervical cancer. Hence, there is an urgent need to identify biomarkers of progression to reverse the outcome of CC.
Cervical cancer cells are involved in energy metabolism reprogramming, which promotes their rapid cell proliferation [6,7]. Malignant tumor cells display increased lactate production, which preferentially requires aerobic glycolysis over oxidative phosphorylation. NAD+ serves as a vital coenzyme for energy transduction and aerobic glycolysis, which mediates redox reactions in metabolic pathways [8], as well as being a substrate for NAD+-dependent enzymes, such as sirtuins and poly ADPribose polymerases (PARPs) [9,10]. NAD-mediated signaling events regulate transcription, cell cycle progression, DNA repair, and metabolic regulation through these enzymes [11]. NAD+ is essential for various cancer-related processes including the citric acid (TCA) cycle, oxidative phosphorylation, fatty acid metabolism, and antioxidant defense metabolism [12].
Accumulating evidence has uncovered that cancer cells exhibit a dependency on metabolic pathways regulated by NAD+. Of note, decreasing the level of NAD+ suppresses glycolytic activity and causes a metabolic collapse in numerous cancers [13][14][15]. Nicotinamide (NAM) plays an essential role in the metabolic pathway that produces NAD+ [16]. As a precursor to the coenzyme NAD+, nicotinamide participates in cellular energy metabolism in the mitochondrial electron transport chain [17]. Researchers have found that nicotinamide inhibited the proliferation of HeLa cells and significantly increased ROS accumulation and GSH depletion at relatively high concentrations [18]. Moreover, nicotinamide phosphoribosyltransferase (Nampt) catalyzes the rate-limiting step of the mammalian NAD salvage pathway. More notably, nicotinamide phosphoribosyltransferase (Nampt) levels increased with the SCC grade [19].
This study constructed and validated an NAD+ metabolic-related gene prognostic signature and explored the molecular signatures in cervical cancer. Through integrative analysis of the relationship between NAD+ metabolism and CC, this study may help in providing potential predictors and treatments of CC.

Dataset Processing
We downloaded the clinical characteristics and NAD+ metabolic-related genes of 309 CESC patients, including cervical squamous cell carcinoma and endocervical adenocarcinoma patients and non-cervical cancer tissues refer to adjacent normal tissues, from the TCGA portal (https://portal.gdc.cancer.gov/) in October 2021. The inclusion criteria were as follows: (a) samples diagnosed as CC; (b) samples with mapped clinical information and gene expression matrix; (c) samples with complete clinical information including age, FIGO stage, tumor size status, lymph node status and metastasis status. Samples lacking follow-up data were excluded. Patients were randomly split into a training set for identification and a testing set for validation in Table 1.

NAD+ Metabolic-Related Genes
NAD+ metabolic-related genes were identified through chi-square test analysis, and the p value < 0.05 was set up for infiltration data. All cervical cancer patients were divided into Cluster 1 or Cluster 2 via the Consensus Cluster Plus R package [20]. Survival analysis was conducted, and the expression levels between the two subgroups were assessed.

Construction and Validation of the NAD+ Metabolic-Related Gene Prognosis Model
Multivariate Cox regression analysis was used to identify 21 significant NAD+ metabolic-related genes for construction of the prognostic signature. The risk score for each cervical cancer patient was calculated as per the following formula: Risk score = Σ Expn *βn, where Expn stands for the expression value of each gene, and βn represents the coefficient of each target gene. We then divided cervical cancer patients into the low-risk or high-risk group based on the cut-off risk score.

Application and Assessment of Prognosis Model
In order to evaluate the use of the signature, Kaplan-Meier survival curves were applied to calculate the difference in the progression-free interval (PFI) and overall survival (OS) between the different risk groups in the training and testing sets. Time-dependent receiver operating characteristic (ROC) curves were applied to verify the prediction accuracy of this constructed prognostic model. Univariate and multivariate Cox regression analyses further validated independent clinical prognoses. Nomograms that included age, FIGO stage, tumor size, lymph node status, metastasis status, and risk score were used to predict the 5-and 10-year survival probabilities through Kendall's tau-b correlation analysis.

Function Enrichment Analysis
The Gene Ontology (GO) analysis method explores the possible KEGG pathways that might implicate the gene-predicted signature. The reference gene set was retrieved from Cluster Profiler [21], and the pathways were considered significant when p ≤ 0.05 or FDR ≤ 0.05.

Statistical Analysis
All statistical analyses were carried out using R version 4.0.2. Differentially expressed genes were identified using a chi-square test. Kendall's tau-b correlation analysis was used to estimate the correlation analysis of clinical factors and risk scores. Survival differences were assessed through Kaplan-Meier analyses and log-rank tests. A p value or FDR ≤ 0.05 was regarded as statistically significant.

Clinical Data and Identification of NAD+ Metabolic-Related Genes in Cervical Cancer
The clinical characteristics and 39 NAD+ metabolic-related genes of 293 cervical cancer patients and normal tissues were confirmed. There were 10 significantly differentially expressed genes: NADSYN1, PARP9, PARP14, SIRT1, SIRT3, BST1, PTGIS, QPRT, AOX1, NT5E (Wilcoxon test p < 0.05; Figure 1A). To explain the influences of NAD+ metabolicrelated genes in the development of CC, we divided CC samples into Clusters 1 (n = 148) and 2 (n = 145) via the Consensus Cluster Plus R package and found that k = 2 realized the optimal clustering stability from k = 2 to k = 9 ( Figure 1B). Demographic, clinical, and pathologic characteristics of the patients with CC are shown in Table 1. Principal components analysis (PCA) showed significant differences between Cluster 1 and Cluster 2 subtypes ( Figure 1C). The PFI and OS of Cluster 2 were shorter than those of Cluster 1 ( Figure 1D,E).

Prognostic Model Correlated with Clinicopathological Characteristics
To examine the prognostic value of our model, we used both the univariate and multivariate Cox regression analyses on CC clinical characteristics including age, FIGO stage, tumor size, lymph node status, metastasis status, and risk group, which revealed the risk group could be the independent predictive factor of CC (p < 0.001). All the details are shown in Table 3. Observations revealed that the PFI and OS in the low-risk group were significantly longer than those in the high-risk group in patients aged both >45 and ≤45 years; in FIGO stages I and II-IV; in tumor sizes ≤ 4, >4 cm, and Tx; and in the N0, N1, Nx, M0, M1, and Mx classifications. (Figure 3A-D,I-L, Figure 4A-C,G-I, Figure 5A-C,G-I and Figure 6A Table 4.

Prognostic Nomogram and Functional Enrichment Analysis
Prognostic nomograms incorporating clinicopathological characteristics and the prognostic signature of the 21 NAD+ metabolic-related genes were constructed to quantify the probabilities for predicting the 5-and 10-year PFI and OS of cervical cancer patients ( Figure 7A,B). We further performed functional enrichment analysis of GO and KEGG and selected the top 30 enriched GO terms and KEGG pathways, which are shown in Figure 7C,D. Biological process (BP), cellular component (CC), and molecular function (MF) revealed NAD+ metabolic-related genes involved in extracellular structure organization, the extracellular matrix, the PI3K-Akt signaling pathway, and the MAPK signaling pathway, which participate in the process and development of CC.

Discussion
NAD+ is an essential coenzyme that mediates glycolysis, where increased NAD+ levels facilitate the rate of aerobic glycolysis and initiate cancer cells via lactate dehydrogenase (LDH) and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) [22]. NAD+ can be synthesized via the de novo pathway, Preiss-Handler pathway, and salvage pathway from tryptophan, nicotinic acid (NA), and nicotinamide (NAM), respectively. An increase in the levels of NAD+ enhanced the malignant phenotype in the cells and led to poorer outcomes, accompanied by NAMPT overexpression, which is mainly produced by the NAD+ salvage pathway in cancer cells and involved in several pro-cancer pathways mediated by NAMPT [23][24][25]. NAMPT was observed to be overexpressed in a broad range of malignant tumors and to modulate cancer cell proliferation, metastasis, and induction of angiogenesis [13,14,26]. Consumption of NAD+ sensitizes cancer cells to oxidative damage by disrupting antioxidant defense systems, reducing cell proliferation, and triggering the cell death of manipulating cell signaling pathways [27,28]. As major NAD+ consumers, SIRTs, PARPs, and CD38 function in many critical processes in cancers. Different SIRT isoforms are essential for DNA repair, energy metabolism, and transcription in cancer cells. Several recent studies found that SIRT modulators exhibit potent biological effects in ovarian, endometrial, and breast cancer cells [29][30][31]. PARPs play pivotal roles in NAD+-dependent DNA damage repair and transcriptional regulation involved in cancer progression. PARP inhibitors have been comprehensively used for gynecologic cancers [32]. Studies demonstrate that CD38 promotes cervical cancer cell growth by inhibiting apoptosis and reducing ROS levels [33]. This necessity supports the idea that NAD+ metabolism is involved in cancer development and progression, and that targeting NAD+ metabolism is likely to be a very intriguing therapeutic concept in repressing underlying tumor progression.
In our study, we first explored clinical characteristics, confirmed 39 NAD+ metabolicrelated genes among 293 cervical cancer patients and para-cancerous tissues, and identified 10 differentially expressed genes. By dividing CC samples into clusters and further investigating NAD+ metabolic-related genes implicated in CC progression, we pinpointed 21 prognostic NAD+ metabolic-related genes (ADAMTS10, ANGPTL5, APCDD1L, CCDC85A, CGREF1, CHRDL2, CRP, DENND5B, EFS, FGF8, P4HA3, PCDH20, PCDHAC2, RASGRF2, S100P, SLC19A3, SLC6A14, TESC, TFPI, TNMD, ZNF229) using univariate and multivariate regression analyses. To explore the clinical value of these 21 genes, the signature was validated in the testing and training cohorts, indicating its reliability and efficacy. The signature could effectively classify CC patients into high-risk and low-risk groups in both sets. The predicated model showed a strong predictive performance, and the survival rates of the low-risk group were significantly higher than those of the high-risk group.
The AUC values of the ROC curve were 0.830 and 0.772, and 0.803 and 0.787, respectively, in the training and testing sets, which proved that this signature has reliable accuracy as a prognostic factor in CC patients. Moreover, the results of the multivariate analysis in different datasets clarified that the NAD+ metabolic-related signature is an independent factor after adjusting for other clinical factors (including age, FIGO stage, tumor size, lymph node status, metastasis status, and risk group).
Among the 21 candidate genes, ADAMTS10, ANGPTL5, APCDD1L, CCDC85A, CGREF1, CHRDL2, DENND5B, EFS, P4HA3PCD, H20, PCDHAC2, RASGRF2, SLC6A14, TESC, TNMD, and ZNF229 have not been reported in CC; our study is the first to characterize these genes as a prognostic signature. SLC19A3, CRP, and FGF8 have been reported to have clinical prognostic value in CC progression [34]; our study is the first to characterize SLC19A3 as a prognostic signature through NAD+ metabolism. TFPI has been reported to be related to survival in breast cancer [35]. S100P has been reported as a prognostic signature in cholangiocarcinoma [36] and hepatocellular carcinomas [37]. Moreover, our gene functional enrichment analysis revealed that the related genes were enriched in the process of the extracellular matrix and the PI3K-Akt and MAPK signaling pathways. Research has highlighted that the ECM initiates cell proliferation, migration, and invasion in CC [38,39]. Previous studies illustrated that the PI3K-Akt and MAPK signaling pathways are involved in the progression and metastasis of CC as well as being therapeutic targets [40,41].
In this study, we investigated the functional implication of NAD+ and genes related to NAD+ metabolism for the outcome of CC. NAD+ metabolism is involved in pro-cancer pathogenesis and is considered a promising therapeutic for oncotherapy. This NAD+ metabolic-related signature is based on population databases; further in vitro or in vivo experimental data as well as clinical experiments are needed to confirm our findings.
Recent research shows that NAD+ metabolism-derived gene signatures could predict the prognosis of ovarian cancer and immunotherapy [42]. In some ways, NAD+ metabolism-related gene signatures (models) could be associated with the prognosis in some types of carcinomas. Additionally, we also explored the characterization of 22 types of immunocyte infiltration and the proportion of immunocyte infiltration in the two NAD+ metabolic-related clusters in cervical cancer, but there were no statistically significant differences. Perhaps this difference could be attributed to the different types of tumors; thus, more experiments are needed for further exploration.
In conclusion, we established a NAD+ metabolic-related gene signature for CC. The external validation of this signature was significantly correlated to the risk value and OS of CC patients, and this signature represents a promising biomarker to predict cervical cancer prognosis. This study demonstrates a new NAD+ metabolism-related prognostic model and therapeutic target for cervical cancer.  Institutional Review Board Statement: Not applicable. Ethical approval from the committee/ institutional review board was not required because all information acquired from the TCGA is publicly accessible and de-identified (https://portal.gdc.cancer.gov/).

Informed Consent Statement: Not applicable.
Data Availability Statement: The data presented in this study are available on request from the corresponding author.