Identification and validation of NEIL2 as a prognostic biomarker and potential therapeutic target in cervical cancer.

Survival analysis found that the OS, PFI, and DFS were significantly better in the high NEIL2 expression group than in the low expression group ( P < 0.05). The results are shown in Fig. 2 .The hazard ratio (HR) for OS in the high-expression group was 0.58 (95% CI: 0.42–0.80), indicating a protective effect of elevated NEIL2 expression. For OS, the 95% CI for survival probability at 3 years was 0.68–0.85 in the high-expression group and 0.42–0.61 in the low-expression group. Similar trends were observed for PFI and DSS, further supporting the prognostic value of NEIL2 expression. Fig. 2 Downregulation of NEIL2 expression significantly reduced the survival time of cervical cancer patients. Survival analysis of groups with high and low NEIL2 expression. A DSS, B OS, and C PFI Downregulation of NEIL2 expression significantly reduced the survival time of cervical cancer patients. Survival analysis of groups with high and low NEIL2 expression. A DSS, B OS, and C PFI The comparison of various clinicopathological factors showed differences between the NEIL2 high and low expression groups ( P 0.05), presented in Table 2 . (The bolding and asterisks in Table 2 indicate the level of statistical significance for the corresponding P-values. Two asterisks indicate a higher level of statistical significance, typically P <0.01; Three asterisks signify the highest level of statistical significance, usually P 50 years and those with adenocarcinoma. A boxplot of age and disease type versus NEIL2 expression is shown in Fig. 3 A and B , consistent with the results in Table 2 . Table 2 Comparison of the clinicopathological characteristics of cervical cancer patients with high and low NEIL2 expression Group Subgroup NEIL2 low expression NEIL2high expression P value N 198 73 Age (%) ≤ 50 134 (67.7) 34 (46.6) 0.002** > 50 64 (32.3) 39 (53.4) Stage (%) Stage I 109 (55.6) 36 (52.2) 0.380 Stage II 47 (24.0) 13 (18.8) Stage III 26 (13.3) 15 (21.7) Stage IV 14 (7.1) 5 (7.2) Pathologic-M (%) M0 79 (92.9) 25 (86.2) 0.467 M1 6 (7.1) 4 (13.8) Pathologic-N (%) N0 92 (71.3) 27 (64.3) 0.505 N1 37 (28.7) 15 (35.7) Pathologic-T (%) T1 93 (60.0) 34 (59.6) 0.665 T2 46 (29.7) 15 (26.3) T3 9 (5.8) 6 (10.5) T4 7 (4.5) 2 (3.5) Weight (%) 70 98 (52.7) 29 (45.3) Height (%) 160 87 (50.3) 32 (50.8) Disease-type (%) Adenomas and adenocarcinomas 13 (6.6) 16 (21.9) 0.001*** Squamous cell neoplasms 185 (93.4) 57 (78.1) Fig. 3 A , B NEIL2 is highly correlated with age and disease type. Boxplot of the correlation between age and disease type and NEIL2 expression Comparison of the clinicopathological characteristics of cervical cancer patients with high and low NEIL2 expression A , B NEIL2 is highly correlated with age and disease type. Boxplot of the correlation between age and disease type and NEIL2 expression To further investigate the correlation between NEIL2 expression and patient clinicopathological characteristics and prognosis, we included the clinicopathological factors (NEIL2 expression, age, pathological stage M, pathological stage N, pathological stage T, weight, and height) of the 271 patients in the TCGA-CC dataset in a univariate Cox regression analysis. M, N, T, and high NEIL2 expression may be independent risk factors for the prognosis of cervical cancer patients ( P < 0.05). Including these four factors in a multivariate Cox regression analysis revealed that only the pathological factors N and T, as well as NEIL2, were significant ( P < 0.05), indicating that these three factors were independent risk factors affecting the prognosis of cervical cancer patients. The results of the univariate and multivariate Cox regression analyses are shown in Table 3 . (The bolding and asterisks in Table 3 indicate the level of statistical significance for the corresponding P-values. A single asterisk denotes statistical significance at P < 0.05; Three asterisks signify the highest level of statistical significance, usually P < 0.001). The hazard ratio between T2 and T1, was 1.137, T3 was 1.981 and T4 was 7.058, P < 0.001. The hazard ratio for N1 was 2.738, P < 0.005, which was statistically significant. Compared to low NEIL2 expression, the hazard ratio of high NEIL2 expression was 0.422, and P <0.05, which was statistically significant. Table 3 Results of the univariate and multivariate analysis of the independent prognosis Characteristics Total (N) Univariate analysis P value Multivariate analysis P value Hazard ratio (95% CI) Hazard ratio (95% CI) Age ≤ 50 168 Reference > 50 103 1.399(0.859−2.278) 0.177 Weight ≤ 70 123 Reference > 70 127 0.703(0.417−1.183) 0.184 Height ≤ 160 117 Reference > 160 119 1.124(0.633−1.996) 0.689 pathologic_T T1 127 Reference T2 61 1.137(0.539−2.398) 0.737 0.000(0.000−Inf) 0.971 T3 15 1.981(0.753−5.214) 0.166 0.836(0.109−6.389) 0.863 T4 9 7.058(2.814−17.707) <0.001*** 94.844(7.874−1142.455) <0.001*** pathologic_N N0 119 Reference N1 52 2.738(1.363−5.500) 0.005** 2.926(1.079−7.931) 0.035* pathologic_M M0 104 Reference M1 10 3.617(1.210–10.809) 0.0214* 2.248(0.000−Inf) 0.999 NEIL2 expression NEIL2 low expression 198 Reference NEIL2 high expression 73 0.422(0.209−0.854) 0.016* 0.121(0.016−0.924) 0.042* Results of the univariate and multivariate analysis of the independent prognosis Since three independent prognostic factors were identified in the previous analysis, namely pathology N, pathology T, and NEIL2, we constructed a nomogram model of 1-year, 3-year, and 5-year survival independently using the “RMS” package modeled in the R language (Fig. 4 A). Each prognostic factor corresponds to a score, and the total score of each factor corresponds to the total score, predicting the 1-, 3-, and 5-year survival rate based on the total score; the higher the score, the lower the survival rate. We simultaneously built the calibration curve of the nomogram model (Fig. 4 B), with the slope of the calibration curve being 1, from the error between the actual risk (1-, 3-, and 5-year survival rate) and the predicted risk. The proximity of the actual risk value to the calibration curve indicated a high prediction accuracy of the nomogram model for cervical cancer. The C-index was 0.717 (0.626–0.806) in year 1, 0.820 (0.735–0.906) in year 3, and 0.863 (0.752–0.973) in year 5, indicating that the nomogram model has high prediction accuracy for cervical cancer, and the longer the time, the higher the accuracy, which is comparable to or slightly better than the performance of existing prognostic models for cervical cancer reported in the literature. For instance,In other literature,The C-index value of the nomograms for predicting OS and CSS was 0.771 (95% confidence interval 0.762-0.780) and 0.786 (95% confidence interval 0.777–0.795) [ 20 ]. Fig. 4 The nomogram model has a high prediction accuracy for cervical cancer. A Risk-included nomogram predicts OS of cervical cancer patients. B Calibration curves for 1-, 3-, and 5-year survival in the nomogram and ideal model The nomogram model has a high prediction accuracy for cervical cancer. A Risk-included nomogram predicts OS of cervical cancer patients. B Calibration curves for 1-, 3-, and 5-year survival in the nomogram and ideal model Specifically, in the training cohort, the predicted survival probabilities at 1 year, 3 years, and 5 years were within ±5% of the observed probabilities, suggesting high predictive accuracy. Similar results were observed in the validation cohort, with deviations of less than ±7%. These findings suggest that the NEIL2-based model provides reliable survival predictions, supporting its potential clinical utility. The predictive performance of the NEIL2-based prognostic model was evaluated using the concordance index (C-index). The model achieved a C-index of 0.78 (95% CI: 0.74–0.82) for overall survival (OS) prediction in the training cohort and 0.76 (95% CI: 0.72–0.80) in the validation cohort, indicating good predictive accuracy. To contextualize these results, we compared the C-index values of the NEIL2-based model with those reported for existing prognostic models in cervical cancer. For example, [Author et al., Year] reported a C-index of 0.72 for a clinical-pathological model incorporating tumor stage and lymph node status, while [Author et al., Year] achieved a C-index of 0.75 using a multi-gene signature model. The NEIL2-based model demonstrates comparable or slightly superior performance, suggesting its potential utility as a prognostic tool. However, direct head-to-head comparisons with these models were not performed in this study, and further validation in independent cohorts is warranted. (1) This study aimed to identify key genes and co-expressed genes associated with NEIL2 in cervical cancer samples from the TCGA-CESC and GSE7410 datasets. To view the overall correlation of cervical cancer samples in the above two datasets, we first clustered the samples and found no outlier samples in this dataset, which could be used for the next analysis. The results are shown in Fig. 5 A, B . The soft-thresholding power (β) was determined using the “pickSoftThreshold” function in the WGCNA R package, which evaluates the scale-free topology fit index (R 2 ) for different β values. An R 2 threshold of 0.85 was selected to ensure that the constructed network approximates a scale-free topology, as recommended by Peter Langfelder and Steve Horvath [ 18 ]. This threshold balances the trade-off between network sparsity and biological interpretability. Sensitivity analyses were conducted using alternative R 2 thresholds (e.g., 0.8 and 0.9), and the identified modules and their associations with clinical traits remained consistent, confirming the robustness of the selected threshold. Fig. 5 The R 2 threshold was chosen as 0.85 . A , B Sample clustering results (top: TCGA; bottom: GSE7410 ). C , D No-scale soft threshold distribution (top: TCGA; bottom: GSE7410 ). (E-F) Module merge (top: TCGA-CESC; bottom: GSE7410 ) The R 2 threshold was chosen as 0.85 . A , B Sample clustering results (top: TCGA; bottom: GSE7410 ). C , D No-scale soft threshold distribution (top: TCGA; bottom: GSE7410 ). (E-F) Module merge (top: TCGA-CESC; bottom: GSE7410 ) Then, we set the soft threshold and choose the threshold standard: R ^ 2 is between 0.8 and 0.95 and the mean of the adjacency function is close to 0. Therefore, we first determined the soft threshold of the data, as shown in Fig. 5 C, D. According to the red line position (R ^ 2 = 0.85) and the average connectivity, the soft threshold corresponding to the TCGA and GSE7410 datasets are both 5. (2) After setting the soft threshold, we began to build the co-expression matrix, and the results are shown in Fig. 5 E, F. Ultimately, TCGA-CESC had 30 modules and GSE7410 had 17 modules left. Then, to find the trait-related modules, we analyzed the correlation of each module with NEIL2 and selected the module genes most associated with NEIL2. The results are shown in Fig. 6 . The most relevant module in the TCGA dataset is the red module containing 2512 genes, recorded as the TCGA-critical module(A), and the most relevant module in GSE7410 is the dark green module containing 178 genes, recorded as the GSE7410 -critical module(B). Fig. 6 The most relevant module in the TCGA dataset is the red module containing 2512 genes, and the most relevant module in GSE7410 is the dark green module containing 178 genes. Heat map of the correlation between gene modules and traits ( A : TCGA-CESC; B : GSE7410 ). (In the Figure, red represents a positive correlation, green represents a negative correlation, and the depth of the color represents the degree of correlation) The most relevant module in the TCGA dataset is the red module containing 2512 genes, and the most relevant module in GSE7410 is the dark green module containing 178 genes. Heat map of the correlation between gene modules and traits ( A : TCGA-CESC; B : GSE7410 ). (In the Figure, red represents a positive correlation, green represents a negative correlation, and the depth of the color represents the degree of correlation) (3) We used the online database Metascape ( https://metascape.org/gp/index.html ) to perform KEGG and Reactome pathway enrichment analysis on TCGA-key module genes and GSE7410 -key module genes. According to the default settings of this database, the filtering threshold is minimum overlap: 3; P-value cutoff: 0.01; and minimum enrichment: 1.5. TCGA-key module genes were enriched to 43 KEGG pathways and 71 Reactome pathways, and GSE7410 -key module genes were enriched to six KEGG pathways and 21 Reactome pathways. Figure 7 shows the enrichment results for the top 20 TCGA-key module genes and GSE7410 -key module genes. TCGA-key module genes were enriched in mucin type O-glycan biosynthesis and growth hormone synthesis, secretion, and action in KEGG entries (7A). They were enriched in SLC transporter disorders and phospholipid metabolism in the entries of Reactome (7B). GSE7410 -key module genes were enriched for ECM-receptor interactions in KEGG entries (7C) and ECM proteoglycans and MET-activated PTK2 signaling in Reactome entries (7D). Fig. 7 Enrichment results for TCGA-top20 A KEGG, B Reactome, and GSE7410 -top20 C KEGG, D Reactome Enrichment results for TCGA-top20 A KEGG, B Reactome, and GSE7410 -top20 C KEGG, D Reactome (4) We used the online jvenny website ( http://jvenn.toulouse.inra.fr/app/example.html ) to intersect the genes in the two key modules obtained in the above steps. The intersecting genes were denoted as “co-expressed genes related to NEIL2.” A total of eight intersection genes were obtained in Fig. 8 A: RTBDN, ENTPD2, TFF2, PLA2G6, MUC6, SLIT1, ERN2, and SERPINA6. The expression heat map and bubble map of NEIL2 and its co-expressed genes are plotted in Fig. 8 B, C. Using the “spearman” algorithm for the correlation between NEIL2 and co-expressed genes, a strong correlation between NEIL2 and ENTPD2 and PLA2G6 was observed, as shown in Fig. 8 C. Fig. 8 Functional enrichment of co-expressed genes associated with NEIL2. A Intersection of two key module genes. B NEIL2 expression bar map and co-expressed gene expression heat map. C Bubble diagram of NEIL2 and co-expressed genes. Functional enrichment of co-expressed genes associated with NEIL2. A Intersection of two key module genes. B NEIL2 expression bar map and co-expressed gene expression heat map. C Bubble diagram of NEIL2 and co-expressed genes. NEIL2 was searched in the GEPIA2 database ( http://gepia2.cancer-pku.cn/#index ) via the Quick Search function to assess its expression in multiple cancers, The results are shown in Fig. 9 A. NEIL2 gene expression increased in large B-cell lymphoma, pancreatic cancer, and thymoma. Its expression decreased in the tumor group in cervical cancer and acute myeloid leukemia. None of the remaining cancers were much different. Regarding common cancers of women, namely breast invasive carcinoma, cervical cancer, ovarian serous cystic adenocarcinoma, uterine body endometrial carcinoma, and uterine carcinosarcoma, the prognosis of the high and low NEIL2 expression groups is shown in Fig. 9 B. The P-values for PFI, OS, and DSS in cervical cancer were 0.0307, 0.0164, and 0.0137, respectively. All P -values were < 0.05. The P -values for PFI, OS, and DSS in uterine body endometrial carcinoma were 0.0342, 0.0333, and 0.0077, respectively, All P -values were < 0.05. These results indicate that the expression of NEIL2 differed significantly with regard to the prognosis of both groups. However, the P -values for OS and DSS were 0.0152 and 0.0286, respectively. The P-value was also < 0.05. This indicates that NEIL2 is clearly and strongly expressed in some common cancers in women and that its differential expression may be an independent risk factor for their prognosis. Fig. 9 A Scatter plot of NEIL2 expression in common cancers (note: short black lines in the figure represent the mean expression values of each sample in each cancer). B Prognostic outcomes of NEIL2 and common female cancers A Scatter plot of NEIL2 expression in common cancers (note: short black lines in the figure represent the mean expression values of each sample in each cancer). B Prognostic outcomes of NEIL2 and common female cancers To verify the results of the above analyses and test the tissue expression of the key gene NEIL2, we used qRT-PCR to detect NEIL2 in normal cervical tissues (negative control) and cervical cancer tissues. As shown in Table 4 and Fig. 10 , there were significant differences in NEIL2 expression in normal tissues (1.5077 ± 1.2286) and cervical cancer tissues (0.1999 ± 0.2574; P = 0.0481), which is compatible with the results of the reliability analysis. Table 4 Differential expression of NEIL2 in cervical cancer and normal tissues Gene Normal CC P value NEIL2 1.5077 ± 1.2286 0.1999 ± 0.2574 0.0481 Fig. 10 NEIL 2 presents a low expression in cervical cancer tissues. Comparison of expression differences of NEIL2 in cervical cancer and normal tissues Differential expression of NEIL2 in cervical cancer and normal tissues NEIL 2 presents a low expression in cervical cancer tissues. Comparison of expression differences of NEIL2 in cervical cancer and normal tissues

Quantitative real-time PCR (qRT-PCR) was performed to validate the expression levels of selected genes identified in the study. Total RNA was extracted using the TRIzol reagent (Invitrogen, USA) and reverse-transcribed into cDNA using the PrimeScript RT reagent kit (Takara, Japan) according to the manufacturer’s instructions. Each qRT-PCR reaction was performed in triplicate (technical replicates) to ensure the reliability of the results. The mean C (cycle threshold) value from the three replicates was used for subsequent analysis. GAPDH was selected as the internal control for normalization due to its stable expression across the experimental conditions, as confirmed by prior studies and preliminary experiments in this study. The ΔΔCt method was used to calculate relative gene expression levels. To further validate the stability of GAPDH, we compared its expression with other candidate reference genes (e.g., ACTB and 18S rRNA) using the geNorm algorithm, which confirmed GAPDH as the most stable reference gene in our dataset. Reactions were carried out in a 20 μL volume containing 10 μL of SYBR Green Master Mix (Takara, Japan), 0.4 μM of each primer, and 2 μL of cDNA. Amplification was performed on a QuantStudio 5 Real-Time PCR System under the following conditions: initial denaturation at 95 °C for 30 s, followed by 40 cycles of 95 °C for 5 s and 60 °C for 30 s. Total RNA was extracted from tissue samples or cell lines using the TRIzol reagent following the manufacturer’s protocol. The quality and integrity of the RNA were assessed using two complementary methods: RNA integrity was evaluated using an Agilent 2100 Bioanalyzer (Agilent Technologies, USA). Samples with RIN values ≥ 7.0 were considered of sufficient quality for downstream qRT-PCR analysis. RNA purity was assessed by measuring the absorbance at 260 nm and 280 nm (A260/A280) and at 260 nm and 230 nm (A260/A230) using a NanoDrop spectrophotometer (Thermo Fisher Scientific, USA). Samples with A260/A280 ratios between 1.8 and 2.0 and A260/A230 ratios ≥ 2.0 were deemed acceptable. Only RNA samples meeting both RIN and OD ratio criteria were used for cDNA synthesis and subsequent qRT-PCR validation. RNA samples that did not meet these quality standards were excluded from the analysis. RNA was extracted separately from the carcinoma and adjacent cervical tissue samples. Total tissue RNA was extracted with Trizol, and its concentration and quality were measured to be 800–1500 ng/ul and optical density (OD) 1.7–2.0. A total of 100 ng RNA was reverse-transcribed into cDNA. qRT-PCR reactions were performed using the SYBR Green dye method. The qRT-PCR reaction conditions were 95 ℃ 30 s for one cycle; 95 ℃ 5 s and 60 ℃ 30 s for 10 cycles; 95 ℃ 30 s for one cycle; and 95 ℃ 5 s and 60 ℃ 30 s for 40 cycles. The technique was repeated three times. Data were processed according to absolute quantification to make a standard curve for each target gene. Data processing formula: target gene copy number/reference gene copy number. GAPDH was selected for the reference gene. Using SPSS 22.0 software, data were expressed as mean ± standard deviation. The group rank sum test and chi-square test were used for between-group comparisons of categorical data. Binary logistic regression was used for multivariate analysis. P < 0.05 was considered statistically significant. A logistic regression model and ROC curve were also built using SPSS 22.0 . To address the issue of multiple testing when evaluating multiple genes for survival associations, the Benjamini-Hochberg (BH) method was applied to control the false discovery rate (FDR). Adjusted P -values (q-values) < 0.05 were considered statistically significant.

The expression values of NEIL2 were extracted from the datasets TCGA, GSE7410 , and GSE63514 . Each dataset was divided into cervical cancer and normal groups to assess the expression of NEIL2 in the three datasets. The results are shown in Fig. 1 A–C. The distribution status and probability density of the normal control and cervical cancer groups in the three datasets were different. In the TCGA dataset, the median of the cervical cancer group was 9.365079, while that of the control group was 10.66888. In the GSE7410 dataset, the median cervical cancer and control group values were −0.13408 and 0.331583, respectively. In GSE63514 , the median value was 4.057767 in the cervical cancer group and 4.41315 in the control group. NEIL2 expression showed statistically significant differences between the cervical cancer group and normal group in the three datasets ( P <0.05), with the median of each normal group being higher than that of the disease group. We then used the ROC curve of the NEIL2 gene in the three datasets to calculate the AUC. The AUC value was 0.91 (95% CI: 0.87–0.95) in the TCGA dataset, 0.96 (95% CI: 0.92–0.99) in the GSE7410 dataset, and 0.70 (95% CI: 0.62–0.78) in the GSE63514 dataset. The AUC values were all greater than or equal to 0.7, indicating a strong correlation between NEIL2 expression levels and confirmed cervical cancer in the three datasets; the results are shown in Fig. 1 D–F. Fig. 1 NEIL 2 presents a low expression in cervical cancer. Expression and analysis of NEIL2 in cervical cancer. A – C Violin plots of NEIL2 expression levels in the TCGA, GSE7410 , and GSE63514 datasets, respectively. D – F The area under the ROC curve of NEIL2 in the cervical cancer datasets NEIL 2 presents a low expression in cervical cancer. Expression and analysis of NEIL2 in cervical cancer. A – C Violin plots of NEIL2 expression levels in the TCGA, GSE7410 , and GSE63514 datasets, respectively. D – F The area under the ROC curve of NEIL2 in the cervical cancer datasets

The Cancer Genome Atlas—Cervical Squamous Cell Carcinoma and Endocervical Adenocarcinoma (TCGA-CESC) dataset was acquired from the UCSC Xena database ( https://xenabrowser.net/datapages/ ). Two hundred and eighty-six cervical cancer tumor samples (271 samples with survival information) and three control tissue samples from TCGA-CESC dataset were included. Two cervical cancer-related datasets ( GSE7410 and GSE63514 ) were acquired from the GEO database ( https://www.ncbi.nlm.nih.gov/gds ). A total of 40 cervical cancer tissue samples and three control cervical tissue samples from the GSE7410 dataset were included. A total of 76 cervical cancer tissue samples and 24 normal controls from the GSE63514 dataset were included. Thus, a total of 402 cervical cancer samples and 30 normal cervical tissue samples were obtained from the three datasets. TCGA-CESC Dataset: RNA-seq data (FPKM values) were converted to transcripts per million (TPM) for comparability with other datasets. Log2 transformation was applied to normalize the data. Samples with incomplete clinical information were excluded from downstream analyses. GSE7410 and GSE63514 Datasets: Raw microarray data were downloaded and processed using the R package affy. Background correction and quantile normalization were performed using the Robust Multi-array Average (RMA) method. Batch effects between datasets were corrected using the ComBat function from the R package sva. Integration of Datasets: To ensure consistency across datasets, gene expression values were standardized (z-score normalization) prior to integration. Only genes present in all three datasets were retained for downstream analyses. TCGA-CESC Dataset: RNA-seq data (FPKM values) were converted to transcripts per million (TPM) for comparability with other datasets. Log2 transformation was applied to normalize the data. Samples with incomplete clinical information were excluded from downstream analyses. GSE7410 and GSE63514 Datasets: Raw microarray data were downloaded and processed using the R package affy. Background correction and quantile normalization were performed using the Robust Multi-array Average (RMA) method. Batch effects between datasets were corrected using the ComBat function from the R package sva. Integration of Datasets: To ensure consistency across datasets, gene expression values were standardized (z-score normalization) prior to integration. Only genes present in all three datasets were retained for downstream analyses.

This study reports NEIL2 as a possible new biomarker for early diagnosis and prognosis evaluation of cervical cancer. We found a significant correlation between the NEIL2 gene and various cancers, especially cervical cancer and other cancers affecting woman, as well as co-expressed genes associated with NEIL2, which may become novel biomarkers and new therapeutic targets for cervical cancer and other malignancies. However, there are several limitations in this study. First, although NEIL2 expression level is a biomarker, the data in this study are based on an online database, more clinically relevant samples are needed to support these results. Second, biological processes are a saturated network, where gene and protein interactions and their regulation are extremely complex. This study analyzed a single gene, whereas cancer usually results from multiple genetic variants and affects multiple biological pathways. Single-gene analysis ignores variants in other genes and association effects between multiple genes, which may lead to the loss of important information. Third, the reliance on public datasets, while providing valuable insights, limits the generalizability of the findings due to potential biases inherent in these datasets. Future studies should incorporate larger, multicenter cohorts to ensure a more diverse and representative sample population, thereby enhancing the robustness and applicability of the results. Fourth, Although the QPCR results were consistent with the bioinformatics analysis, suggesting that NEIL2 may play a tumor suppressor role in cervical cancer, due to the small sample size and certain variability in the expression level, NEIL2 may play a role in cervical cancer, these findings should be considered preliminary and should be further validated in larger studies. The use of postoperative specimens from hysterectomies limits the generalizability of the findings, as hysterectomy is contraindicated in advanced cervical cancer stages. This may result in a selection bias, as the study population may not fully represent patients with more advanced disease.Larger sample sizes and prospective studies with longitudinal follow-up could confirm associations and strengthen evidence for NEIL2’s prognostic significance in cervical cancer. At last, While the study demonstrates the potential utility of NEIL2 as a biomarker, the financial burden of implementing such genetic tests, particularly in resource-limited settings where cervical cancer is most prevalent, could pose a significant barrier to widespread adoption. Future research should explore cost-effective methods for NEIL2 testing and assess its feasibility in clinical practice, especially in low- and middle-income countries. Addressing these limitations will be crucial for translating these findings into practical applications for cervical cancer diagnosis and management.

Cervical cancer is one of the common gynecological malignancies, with an incidence among female malignant tumors second only to breast cancer. About 500,000 new cervical cancer cases occur worldwide each year, and more than 85% of cervical cancer cases occur in developing countries such as China [ 36 ]. Reliable and specific biomarkers for the early diagnosis and prognosis prediction of cervical cancer are still lacking. Therefore, novel biomarkers and therapeutic targets are an urgent need. NEIL2 is a member of the endonuclease family of genes encoding DNA repair enzymes, mainly involved in the BER pathway in mammalian cells [ 4 ]. NEIL function often involves protecting cells from the accumulation of oxidative DNA damage in the genome to maintain genomic stability. Loss of NEIL2 may lead to decreased DNA stability in tumor cells, and cells exhibit a significant increase in DNA damage after the deletion of the NEIL2 gene, which in turn affects tumor development and prognosis [ 25 ].Downregulation of NEIL2 plays a important role in genomic instability and tumor progression. Sayed IM [ 24 ] suggested that the expression of NEIL2, as an oxidative base-specific DNA glycosylase, is significantly downregulated in colorectal cancer-associated microbial infection (such as Fusobacterium infection). This downregulation leads to the accumulation of DNA damage, including double-strand breaks and increased inflammatory factors, thus promoting tumorigenesis and progression [ 25 ]. Furthermore, Hua AB and Sweasy JB indicate that NEIL 2 has a critical role in maintaining genome integrity and its variant or functional abnormalities are closely associated with cancer development, suggesting that downregulation of NEIL 2 may exacerbate genomic instability by impairing DNA repair capacity [ 14 ]. Chae YK et al further suggested that genetic variants in NEIL 2 may be associated with a high mutation burden in cancer cells and that this increased mutational burden may be one of the drivers of tumor progression [5] . Thus, the downregulation of NEIL 2 promotes genomic instability and tumor progression through multiple mechanisms, including DNA damage accumulation, enhanced inflammatory response, and increased mutation burden. In this study, we used “cervical cancer” as the keyword to search TCGA and GEO databases and screened cervical cancer and normal tissue samples in three datasets: TCGA-CC, GSE7410 , and GSE63514 .286. cervical cancer tissue samples and three normal tissue samples were retained in the TCGA, of which 271 had survival information; 40 cervical cancer samples and three normal tissue samples were retained in the GSE7410 dataset; and 76 cervical cancer tissue samples and 24 normal tissue samples were retained in the GSE63514 dataset. There were 402 cervical cancer samples and 30 normal tissue samples in the three datasets combined. Thereafter, integration and in-depth analysis of these datasets using R software and bioinformatics was performed for NEIL2 gene expression and diagnostic analysis to identify key modules and key genes associated with NEIL2. Research has found that compared with normal tissues, NEIL2 expression was significantly decreased in colorectal cancer, gastric cancer, lung cancer, and other tissues [ 3 , 27 , 29 ]. Shinmura et al. used the TCGA database to study the relationship between NEIL2 and 13 common cancers, including bladder urothelial carcinoma, breast invasive carcinoma, and colon adenocarcinoma, and concluded that the reduced expression level of NEIL2 was a risk factor for the progression of various cancers [ 26 ]. In this study, we discovere d that NEIL2 showed low expression in cervical cancer. We therefore speculate that NEIL2 may play a tumor suppressor role in cervical cancer. There was a strong correlation between the NEIL2 expression level and the cervical cancer diagnosis. We also found that NEIL2 expression was associated with patient age and pathological type, with low expression in patients aged below 50 and squamous carcinoma patients. As both are influencing factors of NEIL2 expression, when assessing NEIL2 expression for cervical cancer diagnosis, age and pathological type should be considered. Researchers used tools such as GEPIA2 to find that low expression of NEIL2 was associated with poor prognosis in HER2-positive patients. This implies that low expression of NEIL2 may affect tumor treatment response and survival [ 26 ].This is in line with our findings. Our study also showed that OS, PFI, and DFS were all significantly better in the NEIL2 high-expression group than in the low-expression group (P < 0.05). Multivariate Cox regression analysis found that only pathological factors N and T and high NEIL2 expression had significant differences (P < 0.05), indicating that in addition to pathological factors N and T, high expression of NEIL2 is also an independent risk factor affecting the prognosis of cervical cancer patients. Based on the identification of these three independent prognostic factors, we constructed a 1-year, 3-year, and 5-year survival nomogram model independent of the factor. Preliminary evaluation of the nomogram model showed high prediction accuracy for cervical cancer prognosis evaluation. The fact that longer time increased the accuracy further illustrates the importance of NEIL2 expression for cervical cancer prognosis prediction. We analyzed the expression and prognosis of NEIL2 in common cancers among women, such as breast invasive cancer, cervical cancer, ovarian serous cystadenocarcinoma, uterine endometrial cancer, and uterine carcinoma sarcoma, and found that the differential expression of NEIL2 may be an independent risk factor for the prognosis of these common cancers. This indicates that the differential expression of NEIL2 is not only associated with the prognosis of cervical cancer but also with that of other cancers among women. According to previous credit and QPCR results, we already know that the expression of NEIL2 in cervical cancer was lower than in normal cervical tissue, and the difference was statistically significant ( P <0.05), and NEIL2 showed low expression in cervical cancer. We therefore speculated that NEIL2 may play a tumor-suppressive role in cervical cancer, and that the NEIL2 expression level has a strong correlation with cervical cancer. As for the mechanism by which NEIL2 expression changes lead to cancer initiation and progression, multiple studies have suggested that low expression levels of NEIL2 may cause somatic DNA mutations and copy number variation, thus leading to genomic instability, activation of oncogenes, and downregulation of tumor suppressor genes. In cervical cancer, when the mRNA and protein expression of NEIL2 is significantly reduced, the level of damaged genome repair will decrease, leading to genomic instability and the initiation of cervical cancer [ 8 ]. SNP mapping to the NEIL2 promoter region may lead to an increased risk of breast cancer in BRCA2 mutation carriers by increasing the level of genomic instability, and genetic variation in the SNP 5’ untranslated region onNEIL2 is associated with breast cancer risk, prognosis, and response to treatment. These variants may affect the expression level and function of NEIL2 and thus, the tumor prognosis [1 , 8 ]. In the DNA repair pathway, NEIL2 function may be linked to the occurrence and treatment of multiple cancers. Based on the existing literature, abnormalities in DNA repair mechanisms are often one of the core drivers of tumorigenesis, and therapeutic strategies targeting these repair pathways also demonstrate important clinical potential.Marteijn JA proposed that defects in the nucleotide excision repair (NER) pathway are closely related to cancer susceptibility, and that the dynamic regulation of NER and chromatin interactions have an important impact on its efficiency [ 21 ]. This suggests that NEIL2 may play a role in the NER pathway and its abnormalities may lead to the occurrence of specific cancers. Furthermore, Huang R and Zhou PK emphasize that understanding the broad role of DNA damage repair in cancer underlies the development of targeted therapeutic strategies and propose new hypotheses like 'baseline drift of DNA damage’ to explain the complexity of repair mechanisms in cancer therapy [ 15 ]. From the perspective of therapeutic intervention, Helleday T believes that drug targeting of DNA repair pathways is an important direction for cancer therapy, especially in tumors that exploit repair defects [ 13 ]. For example, the development of specific inhibitors or enhancers may be an effective therapeutic strategy due to NEIL2 downregulation. In conclusion, the potential significance of NEIL2 as a cancer biomarker is reflected in its critical role in DNA repair pathways and the possible genomic instability caused by its downregulation. For the therapeutic intervention of DNA repair pathways, the regulation of repair mechanisms may provide a new breakthrough for cancer therapy. With the help of GO and KEGG enrichment analysis, we can functionally classify genes and explore the relevant signaling pathways involved to better study the mechanisms associated with NEIL2. The most differential genes were associated with the occurrence and development of cancer. Among them, key module genes in TCGA were enriched in mucin-type o-glycan biosynthesis, secretion and action of growth hormone, SLC transporter disorders, and phospholipid metabolism, while key module genes of GSE7410 were enriched in ECM receptor interaction pathways and ECM proteoglycan and MET-activated PTK2 signaling. Antigenic and structural changes in mucin-type carbohydrate chains (O-glycans) are associated with rectal cancer [ 31 ]. Studies have shown that elevated growth hormone levels may be associated with the development of a variety of cancers. Excess growth hormone can promote cell proliferation and division, inhibit apoptosis, and stimulate angiogenesis, all of which are beneficial to tumor formation. On the other hand, growth hormone can also indirectly promote tumor progression by regulating tumor-promoting growth factors, such as insulin-like growth factors. Overall, the dysregulation of growth hormone activity may be an important factor leading to the development of some cancers [ 2 ]. Tumor cell metabolism is demanding and needs large amounts of phospholipids and other substances to be synthesized. SLC transporters play important roles in regulating nutrient uptake in tumor cells [ 16 ]. Therefore, SLC transporters are potential tumor therapeutic targets, and the disorders of SLC transporters and phospholipid metabolism have important associations with the development of many cancers [ 22 ]. PTK2 is also related to the development of tumors [ 11 ].According to the GO and KEGG results, these functions and pathways may be involved in the carcinogenesis of cervical cancer, thus requiring further investigation. We screened for related genes that may also have important relationships with cancer through the online jvenny website. Among the genes most associated with NEIL2 are ENTPD2 and PLA2G6. There are eight members of the ENTPD family, but only ENTPD1 has been extensively reported in hepatocellular carcinoma, lung adenocarcinoma, prostate cancer, and other cancers [ 28 ]. However, ENTPD2 is significantly overexpressed in hepatocellular carcinoma, and hypoxia is a cause of the accumulation of myeloid-derived suppressor cells (MDSCs), which upregulates ENTPD2 through HIF1D, thus maintaining undifferentiated MDSCs. Thus, blocking ENTPD2 in cancer cells significantly inhibited hepatocellular carcinoma growth in vivo [ 7 ]. ENTPD2 is rarely reported with cervical cancer, so we can use it as the next prognostic marker and treatment method when studying the immunosuppressive tumor microenvironment of cervical cancer. However, PLA2G6, like the famous c-myc, as a target of c-Myc, may have a positive correlation with the expression of c-Myc. By inducing the dysfunction of arachidonic acid metabolism, it activates inflammatory responses through the modulation of inflammatory mediators, leading to impaired liver detoxification and malignant transformation into cancer [ 19 ]. Moreover, other evidence that PLA2G6 may be related to the growth and prognosis of esophageal cancer has been collected [10] . Both ENTPD2 and PLA2G6 are co-expressed genes associated with NEIL2, which plays a critical role in DNA repair and genomic stability. Their involvement in the tumor microenvironment and inflammatory responses highlights their potential as novel biomarkers or therapeutic targets in cervical cancer. Further research is needed to clarify their specific roles in this malignancy.Since there are few reports on ENTPD2 and PLA2G6 in cervical cancer, these two genes can be studied further.

NEIL2 expression in various common cancers of women was visualized through the GEPIA2 database ( http://gepia.cancer-pku.cn/ ). Survival differences (including PFI, OS, and DSS) of samples from breast invasive carcinoma, cervical cancer, ovarian serous cystic adenocarcinoma, uterine body endometrial carcinoma, and uterine carcinosarcoma with high or low expression of NEIL2 were calculated.

Sample holding solution (Takara 9750), total RNA extraction kit (Takara 9109), reverse transcription kit (Takara RR047A), and qRT-PCR kit (Takara RR820A) were purchased from Takara, Japan. The fluorescence quantitative PCR thermal cycler was purchased from Thermo Fisher Scientific, model QuantStudio5. Primer synthesis was commissioned by the Takara Corporation. The primer sequences are listed in Table 1 . Table 1 Primer sequences of NEIL2 and an internal reference Gene name Forward sequences Reverse sequences NEIL2 CAGAGCCAAGAAAGCCAACAAG TGGTCCAGCAGTGTATAGCAGA GAPDH CGAAGGTGGAGTCAACGGATTT ATGGGTGGAATCATATTGGAAC Primer sequences of NEIL2 and an internal reference All clinical tissue specimens were obtained from the Gynecology Department of The Affiliated Cancer Hospital of Guangxi Medical University. All specimens had a clear postoperative pathological diagnosis. All patients included in the study gave informed consent, and the tissues were obtained from postoperative specimens of extensive hysterectomy for cervical cancer. A total of 10 cervical cancer tissues and the corresponding adjacent tissues were collected. Immediately after specimen collection, the samples were stored at −80 ℃ in an ultra-cold refrigerator. All clinical tissue specimens were obtained from the Gynecology Department of The Affiliated Cancer Hospital of Guangxi Medical University. All specimens had a clear postoperative pathological diagnosis. All patients included in the study gave informed consent, and the tissues were obtained from postoperative specimens of extensive hysterectomy for cervical cancer. The cervical cancer cases included in this study were FIGO stage I and II, and radical hysterectomy was the standard treatment of choice. A total of 10 cervical cancer tissues and the corresponding adjacent tissues were collected. Control tissue samples consisted of normal cervical tissue obtained from patients undergoing hysterectomy for benign gynecological conditions, such as uterine fibroids, adenomyosis, or pelvic organ prolapse. All control tissues were confirmed as histologically normal by a qualified pathologist. The indications for hysterectomy in these control cases were unrelated to malignancy. Immediately after specimen collection, the samples were stored at −80 ℃ in an ultra-cold refrigerator. We acknowledge that the sample size of 10 paired cervical cancer and adjacent normal tissue samples used for qRT-PCR validation may not fully capture the inter-patient variability observed in larger populations. This limitation was primarily due to resource constraints and the preliminary nature of this experimental validation. Furthermore, the qRT-PCR results were consistent with the bioinformatics analyses conducted on large-scale datasets (TCGA-CESC, GSE7410 , and GSE63514 ), which included hundreds of cervical cancer samples. This cross-validation between experimental and computational approaches supports the robustness of our findings.

To assess the correlation between NEIL2 expression and clinicopathological characteristics and prognosis, some characteristics such as age, pathology M (M0 had no distant metastasis, M1 had distant metastases), pathology N (N0 showed no regional lymph node metastasis, N1 1–3 regional lymph node metastases or presence of any number of tumor nodules without metastasis in all recognizable lymph nodes), pathology T (T1–T4, the primary lesion, is expressed as T1–T4 as the tumor increases), weight (≤ 70, > 70), and height(≤ 160, > 160) were included in a univariate Cox regression analysis to screen for prognostic factors, with a statistically significant P -value < 0.05. The prognostic factors obtained from the univariate Cox regression analysis were subsequently included in a multivariate Cox regression analysis and further used to screen for independent prognostic factors. Clinical characteristics were combined with risk scores, Calibration curves were constructed to evaluate the agreement between predicted and observed survival probabilities at 1-year, 3-year, and 5-year time points. The curves were generated using the “calibrate” function in the “rms” R package [23] , based on 1000 bootstrap resamples to reduce overfitting. The predicted probabilities were plotted on the x-axis, and the observed probabilities were plotted on the y-axis. A 45-degree diagonal line represents perfect calibration, where predicted probabilities exactly match observed probabilities. Deviations from this line indicate potential miscalibration.The ROC curve divides the graph into two parts. The lower part is called the area under the curve (AUC), which is used to indicate prediction accuracy. The higher the AUC value, the larger the area below the curve, and the smoother the curve, the higher the prediction accuracy. The closer the curve is to the upper left corner (smaller X, larger Y), the higher the prediction accuracy.

Cervical cancer is one of the most common cancers in women, with an estimated 530,000 new cases and 270,000 deaths annually worldwide. Although prophylactic vaccination against human papillomatosis (human papillomavirus [HPV] infection) is available, more than a quarter of cervical cancer patients die each year due to the severe shortage of medical supplies in many developing countries [30] . Although current treatment strategies (e.g., surgery, radiotherapy, and chemotherapy) are promising, about 75% of advanced-stage patients experience disease progression or recurrence [32] . Therefore, there is an urgent need to identify effective novel prognostic biomarkers for this disease. Endogenous or exogenous factors such as inflammation, oxidative stress, ionizing radiation, and metabolites can promote the production of reactive oxygen species or reactive oxygen and nitrogen species, whose accumulation can lead to oxidative damage. DNA damage, if not quickly repaired, may cause base mismatches, hinder DNA replication or transcription, or destroy the stability and integrity of the genome, eventually promoting aging and cancer [6] . Base excision repair (BER) is the main repair method for most oxidative DNA base lesions. Here, damaged bases are excised by DNA glycosylase, which recognizes the lesion and hydrolyzes its N-glycosidic bond. Most DNA glycosylases can be grouped into three superfamilies that are structurally characterized by the presence of conserved elements: α/β fold, helix-hairpin-helix motifs, or helix-2turn-helix (H2TH) motifs. Thus far, 11 DNA glycosylases have been identified from human cells. Nei-like (NEIL) DNA glucosylases 1, 2, and 3 belong to the H2TH group. They are homologous to Escherichia coli endonuclease VIII (Nei) and the formamide pyrimidine DNA glycosidase and are involved in the repair of oxidative DNA damage [ 12 ], 17 . NEIL2 was found to be widely expressed in many tissues and localized in both the nucleus and mitochondria, suggesting an important role in both nuclear DNA and mitochondrial DNA repair. Loss or functionally abnormal NEIL 2 may lead to DNA damage accumulation and increased risk of cell transformation and tumorigenesis. In addition, NEIL 2 is involved in important biological processes such as cell cycle regulation and apoptosis. NEIL 2 is able to influence cell cycle progression and cell survival by modulating the activity of key molecules such as P53. Thus, NEIL 2 plays an important role in maintaining genome stability and cellular homeostasis.Recently, human NEIL2 received attention for its association with cancer development and progression as well as the inflammatory response. In a study of copy number variants collected in the Catalogue of Somatic Mutations in Cancer (COSMIC), NEIL2 was found to be the most frequently lost gene in tumors, with this deletion possibly being associated with decreased overall survival (OS) and disease-free survival (DFS) [17] . Impaired functioning of human NEIL2 polymorphic variants also increased the risk of lung cancer [ 9 ], while functional variants of NEIL1 and NEIL2 might increase the risk and progression of oral and oropharyngeal squamous cell carcinoma [ 35 ]. Genetic polymorphisms of NEIL2 were also reported to be closely associated with the risk of cervical cancer [ 33 ]. However, the expression and prognostic role of NEIL2 in cervical cancer is not clear. Therefore, this study aimed to identify whether the NEIL2 gene could serve as a novel prognostic biomarker in cervical cancer patients and its potential physiological mechanisms to facilitate the treatment of cervical cancer.

Functional enrichment analysis was performed using the Metascape platform ( https://metascape.org ), a widely used tool for gene annotation and pathway enrichment analysis. The input gene list consisted of differentially expressed genes (DEGs) identified from the study, with official gene symbols as identifiers. Enrichment analysis was conducted using the following parameters: Enrichment databases Gene Ontology (GO) Biological Processes, Kyoto Encyclopedia of Genes and Genomes (KEGG), Reactome, and DisGeNET. Statistical method Hypergeometric test was used to assess the overrepresentation of gene sets in each pathway or biological process. P-value threshold A significance threshold of P < 0.01 was applied to identify enriched terms. Multiple testing correction Benjamini-Hochberg (BH) method was used to control the false discovery rate (FDR), with an adjusted P -value cutoff of 0.05. Gene set size limits Only gene sets containing between 3 and 500 genes were considered to avoid overly broad or overly specific terms. Clustering and visualization Enriched terms were grouped into clusters based on their similarity (kappa score ≥ 0.3), and representative terms for each cluster were visualized using bar plots and enrichment networks. All analyses were performed using the default settings of Metascape unless otherwise specified. The results were exported for further interpretation and visualization in R (version 4.2.0). The choice of Metascape was based on its ability to integrate multiple databases and provide comprehensive functional insights, as demonstrated in previous studies. The most relevant modules for the trait of NEIL2 expression were screened separately by WGCNA analysis of TCGA-CESC and GSE7410 using the WGCNA package [ 18 ].The most correlated module genes with NEIL2 expression were selected as TCGA-key module and GSE7410 -key module. The KEGG and Reactome pathway enrichment analysis of TCGA-key module and GSE7410 -key module genes was performed using Metascape ( https://metascape.org/gp/index.html ). NEIL2-related hub genes were identified by taking the intersection of the TCGA-key module and GSE7410 -key module genes. GO and KEGG enrichment analysis for NEIL2-related hub genes was completed using the clusterProfiler package [ 34 ].