A total of three pairs of cervical cancer and adjacent tissues were collected from three cervical cancer patients who underwent surgical operation from July 2019 to August 2012 in Department of Gynecology and Obstetrics, Affiliated Jiangning Hospital of Nanjing Medical University. After sequencing, we collected another 20 pairs of samples for data verification ( Supplementary Table 1 ). Among them, 15 pairs (15 cervical cancer tissues and 15 controls) were collected from June 2020 to December 2020 for qPCR assay. Another five pairs of samples (five cervical cancer tissues and five controls) were collected between January 2021 and February 2021 to indirectly measure the number of immune cells in cervical cancer samples.

Specimens were frozen in liquid nitrogen immediately after operation and stored at −80°C until extraction. All samples were confirmed by histopathological examination. This study was approved by the ethics committee of the hospital. Informed consent to collection and use of the biological samples was obtained from each patient.

Total RNA was isolated using a RNeasy mini kit (Qiagen, Germany) and analyzed by 1% agarose gel electrophoresis (Bio-Rad, USA) to ensure that no degradation occurred. The RNA libraries were constructed using the TruSeq RNA Sample Preparation Kit (Illumina, USA). After purification, libraries were quantified using a Qubit 8000 (Life Technologies, USA) and validated with an Agilent 2100 (Agilent Technologies, USA) to confirm the insert size. Then, clusters were sequenced on an Illumina HiSeq 2500 instrument (Illumina, USA). Library construction and sequencing were performed at Shanghai Yuanshen Biomedical Technology Co., Ltd.

Differentially expressed lncRNAs and mRNAs were identified through fold-change filtering (|log2FC|>1 and p<0.05) using the “edgeR” package in R. The differentially expressed RNA profiles were normalized by log2 transformation.

The Pearson correlation coefficient was calculated, and the R value (cutoff >0.95) was used for each pair of lncRNA-mRNA interactions. The lncRNA-mRNA correlation network was constructed by Cytoscape software.

Functional enrichment analysis of lncRNA-target mRNAs was performed using Metascape (https://metascape.org/). All statistically enriched terms (Gene Ontology and Kyoto Encyclopedia of Genes and Genomes) were identified based on accumulative hypergeometric p values.

To investigate the predictive value of the expression levels of lncRNAs and mRNAs for the survival of cervical cancer patients, Kaplan–Meier survival analysis was performed using GEPIA (http://gepia2.cancer-pku.cn/). The statistical significance was set at p < 0.05. Then, the data obtained from the analysis were verified by StarBase (http://starbase.sysu.edu.cn/).

To analyze the independent effects of individual miRNAs on the overall survival of patients with colon cancer, we performed univariate and multivariate Cox proportional regression analysis with an online tool (SangerBox tools, http://sangerbox.com/Tool). We constructed a Cox proportional hazards regression model and calculated the risk value of each patient through the formula (risk score = b × exp (RNA1) + b × exp (RNA2) +… + b × exp (RNAn), where b represents the multivariate Cox regression coefficient and exp () represents the expression level of prognostic RNAs. Next, we calculated the survival rates of the high-risk and low-risk groups and plotted the 1-year, 3-year, and 5-year survival receiver operating characteristic (ROC) curves to test the feasibility of the prediction ability of the model.

Total RNA was extracted from cervical cancer samples using an RNeasy mini kit (Qiagen, Germany) and analyzed by 1% agarose gel electrophoresis (Bio-Rad, USA) to ensure that no degradation had occurred. A Qubit 8000 (Life Technologies, USA) was used to measure the RNA concentration. Then, the RNA was reverse transcribed into cDNA. qRT-PCR was performed using SYBR Premix Ex Taq (Takara, China) on an ABI7500 system (Applied Biosystems, CA). The following cycling parameters were used: initial denaturation at 95°C for 30 s, followed by 35 cycles of 95°C for 5 s, 58°C for 30 s and 95°C for 60 s, and 60°C for 30 s. The primer sequences for PCR were as follows: PCBP1-AS1, forward: 5′-CCAACCTGATACATTGCCT-3′ and reverse 5′-TGGAAGAAATTCCCTGCTG-3′, GAPDH: forward 5′-CTCCTCCACCTTTGACGCTG-3′ and reverse 5′-TCCTCTTGTGCTCTTGCTGG-3′. Primers were synthesized by Sangon Biotech (China). GAPDH was used as a control. The mean value of triplicate experiments was used to calculate relative lncRNA expression using the formula ΔCt = Ct^mean lncRNAs − Ct^mean GAPDH. Expression fold changes were calculated using the 2-^ΔΔCt method.

The TIMER (https://cistrome.shinyapps.io/timer/) correlation module was used to evaluate potential relationships between PCBP1-AS1 expression and immune infiltrates.

GSEA was performed using normalized RNA-Seq data obtained from TCGA-cervical cancer. The number of permutations was set to 100. Using GSEA, we further analyzed GO terms and KEGG pathways to investigate possible biological functions of PCBP1-AS1 (p-value <0.05).

HeLa cells were obtained from the American Type Culture Collection (ATCC). The cell lines were cultured as suggested by ATCC. The cells were cultured in Dulbecco’s modified Eagle’s medium (Invitrogen, USA) supplemented with 10% fetal bovine serum (Invitrogen, USA), 100 U/ml penicillin (Sigma, USA), and 100 µg/ml streptomycin (Sigma, USA) under a humidified atmosphere of 5% CO₂ at 37°C.

PCBP1-AS1 small interfering RNA (si-PCBP1-AS1) and the corresponding control (si-NC) were purchased from RiboBio (Guangzhou, China). The PCBP1-AS1 overexpression plasmid (pCDH-GFP-PCBP1-AS1) and corresponding control plasmid (NC) were also purchased from RiboBio (Guangzhou, China). All oligomers and plasmids were transfected into HeLa cells using Lipofectamine 3000 reagents (Invitrogen, USA) based on the manufacturer’s protocol. Briefly, when HeLa cell densities were approximately 60% in 12-well plates (Corning, USA), 50 nM siRNA oligos or 2 µg overexpression plasmids were introduced into cells using Lipofectamine 3000 reagents (Invitrogen, USA). Untreated cells were set as blank groups and transfected with empty vectors, and NC-siRNA was used as a negative control. At 24 h post transfection, the efficiency of knockdown and overexpression was determined by qRT-PCR and fluorescence microscopy. Subsequent experiments were performed at 48 h after transfection.

The proliferation of HeLa cells was measured by cell proliferation using Cell Counting Kit-8 (Sigma, USA) in 96-well plates. Then, 3,000 cells/well were incubated for 12, 24, and 48 h. All cells were then incubated with CCK-8 reagent (10 μl per well) for 3 h, and a microplate reader (Thermo, USA) was utilized to detect the absorbance of each well at 450 nm. Each experiment was carried out three times.

HeLa cells were seeded in plates (96 wells, Corning, USA) at 5 × 10⁵ cells/well with culture medium at 37°C with 5% CO₂. Then, the confluent cell monolayer was scratched with a sterile 200 μl pipette tip, and Opti-MEM™-reduced serum medium (Gibco, USA) was added. Microscope photos were taken after 0, 12, 24, and 48 h to record the scratched areas. ImageJ software was used to evaluate the percentage of closure.

At 48 h after transfection, HeLa cells were collected to prepare a single-cell suspension. The HeLa cell suspension (3 × 10³ cells/well) was added to the Transwell upper chamber (Corning, USA), and DMEM (20% FBS) was added to 24-well plates in the lower chamber. The upper chamber was coated with Matrigel. After 24 h, 4% paraformaldehyde (Sigma, USA) was applied to fix the cells, and the cells were stained with 1% crystal violet (Sigma, USA). Cells were observed and counted under an optical microscope (Olympus, Japan).

Total protein was extracted using RIPA lysis buffer (Beyotime, China), and separated using 10% SDS-PAGE (Beyotime, China). Next, proteins were transferred onto PVDF membranes (Beyotime, China). Then, the target protein membrane was blocked with 5% nonfat milk for 24 h. Subsequently, the membranes were incubated with specific primary antibodies against CD4 (Beyotime, China, 1:1,000), CD19 (Beyotime, China, 1:1,000), CD56 (Beyotime, China, 1:1,000), and GAPDH (Beyotime, China, 1:1,000) for 24 h at 4°C. GAPDH was used as a control. Afterwards, the membranes were incubated with secondary antibodies (Beyotime, China, 1:1,000) for 4 h at room temperature. TMB color liquid (Beyotime, China) was used to detect protein bands. ImageJ software was used to analyze the gray value of the bands. Protein levels were calculated using the ratio of target protein/GAPDH.

GraphPad Prism 8.0 software (California, USA) was utilized to perform statistical analysis. The discrepancies between two groups were compared by t-test. The differences were deemed statistically significant at p < 0.05.

To understand the lncRNAs and mRNAs involved in cervical cancer pathogenesis, we performed lncRNA microarray detection in three pairs of cervical cancer tissues and matched adjacent tissues. On average, 82.58 million reads were obtained for each sample ( Supplementary Table 2 ). Among the 10,675 detected expressed lncRNAs, we identified 130 differentially expressed lncRNAs between cervical cancer tissues and adjacent tissues (|log2FC|>1 and p < 0.05), of which 48 were upregulated and 82 were downregulated in the cancer tissue, as shown in Figure 1A ; the red points represent statistically significant upregulated differentially expressed lncRNAs, and the blue points represent downregulated differentially expressed lncRNAs. With this same criterion, we identified 656 significantly differentially expressed mRNAs, of which 293 were upregulated and 363 were downregulated ( Figure 1A ). The top 10 lncRNAs (up- and down) and mRNAs (up- and down) are shown in Tables 1 and 2 , respectively. Among the annotated differentially expressed lncRNAs, antisense lncRNAs accounted for 37.50% of upregulated lncRNAs, whereas the majority of downregulated lncRNAs were lincRNAs and antisense lncRNAs (accounting for 40.24% each) ( Figure 1D ). The distribution of differentially expressed lncRNAs across chromosomes was also analyzed. Among the downregulated differentially expressed lncRNAs, chromosome 1 had the most differentially expressed lncRNAs (n = 9), followed by chromosome 5 (n = 8) ( Figure 1C ). In the same analysis, chromosome 2 had the most DElncRNAs (n = 7) among the upregulated differentially expressed lncRNAs ( Figure 1C ). The heat map results in Figure 1B show that lncRNA and mRNA expression was distinct between cervical cancer tissues and adjacent tissues. In summary, the results from the lncRNA microarray analysis indicated that aberrantly expressed genes, including mRNAs and lncRNAs, may play important roles in the development and progression of cervical cancer.

lncRNAs can regulate the transcription, translation, and splicing of downstream target mRNAs. To understand the correlation between the expression of differentially expressed lncRNAs and differentially expressed mRNAs, a lncRNA-RNA interaction network was constructed. The lncRNA-mRNA coexpression pairs in the network were selected with a threshold of correlation ≥0.95, resulting in a network consisting of 514 nodes and 7,102 significant coexpression relationships, including 127 differentially expressed lncRNAs and 387 differentially expressed mRNAs ( Figure 2A ), suggesting that these differentially expressed lncRNAs might regulate downstream target mRNAs mainly through induction mechanisms. As shown in Figure 2A , the coexpression regulatory network was divided into two parts (cis and trans), which are the two regulatory mechanisms by which lncRNAs regulate downstream genes. Then, using the jActive module, we identified a highly active subnetwork module (ActivePath Score = 7.20; Figure 2B ) from the network, including 66 nodes and 192 edges with 19 lncRNAs and 47 mRNAs. GO and KEGG enrichment analyses were performed to analyze the functions of the differentially expressed mRNAs in each subnetwork module of the network. The results showed that these differentially expressed mRNAs were significantly enriched in epidermal development, regulation of hormone levels, cell cycle, epidermal cell differentiation, phosphorylation, and cell resistance. The main pathways were the HNF3A pathway and regulation of the intracellular estrogen receptor signaling pathway ( Figure 3 ), which indicated that these differentially expressed mRNAs may be related to EMT. EMT causes dissociated epithelial cells to acquire migration and invasive capacities and confers cancer cells with the ability to migrate to distant tissues.

To identify the differentially expressed lncRNAs and mRNAs with potential prognostic value, the expression levels of 19 differentially expressed lncRNAs and 47 differentially expressed mRNAs in the network of the subnetwork module were analyzed using a univariate Cox proportional hazards regression model. Only one lncRNA (PCBP1-AS1) and four mRNAs (FAM222A, FHAD1, WDR62, and SBK1) were identified as prognostic factors (p < 0.05; Figure 4B ). Kaplan–Meier curve analysis showed that PCBP1-AS1 was negatively correlated with OS (p < 0.05), and all mRNAs were positively correlated with OS (p < 0.05) ( Figure 4A ). Meanwhile, we evaluated the relationship among PCBP1-AS1 and four mRNA expression levels and various clinicopathological parameters of cervical cancer patients. Expression data and clinical characteristics were obtained from TCGA-cervical cancer database. The results showed that the expression of PCBP1-AS1, FAM222A, FHAD1, WDR62, and SBK1 was significantly correlated with tumor clinical stage, pathologic TNM, and lymphatic invasion (p < 0.05) ( Figure 4C ). From the multivariate Cox regression analysis, PCBP1-AS1 (p = 0.046; HR = 0.407, 95% CI, 0.156–1.06) and SBK1 (p = 0.047; HR = 0.804, 95% CI, 0.641–1.007) were independent prognostic factors ( Table 3 , Supplementary Figure 1 ).

This project focuses on the regulation of lncRNAs; thus, PCBP1-AS1 was selected as a candidate for further determination of its role in cervical cancer pathogenesis. LncLocator prediction results revealed that PCBP1-AS1 was localized to the cytosol ( Figure 5A ). qRT-PCR was performed to detect PCBP1-AS1 expression levels in 15 cervical cancer tissues and 15 controls. The clinical characteristics of the cervical cancer samples are summarized in Supplementary Table 1 . As shown in Figure 5B , PCBP1-AS1 had significantly higher expression levels in cervical cancer tissues than in their normal counterparts (p < 0.001). This result was consistent with the microarray analysis and TCGA data ( Figure 5C ). In addition, the expression of PCBP1-AS1 was positively correlated with FAM222A (p = 0.035), FHAD1 (p = 0.027), and SBK1 (p < 0.001) ( Figure 5D ). Moreover, univariate analysis revealed that PCBP1-AS1 expression, tumor stage, pathologic T stage, and lymph vascular invasion were significantly correlated with the OS of cervical cancer patients ( Table 4-a ). Our multivariate analysis revealed that PCBP1-AS1 expression might be an independent factor for the prognosis of cervical cancer ( Table 4-b , Figure 5E ). Meanwhile, the ROC curve AUC of PCBP1-AS1 expression for predicting survival was 0.603 ( Figure 5F ), which indicated that PCBP1-AS1 possessed the potential prognostic ability of cervical cancer. Furthermore, as shown in Figure 5G , we uncovered a correlation between PCBP1-AS1 expression and clinicopathologic characteristics. Increased PCBP1-AS1 expression levels in cervical cancer were significantly correlated with tumor stage, pathologic TNM (p < 0.05), and lymph invasion (p = 0.0402). These results indicated that cervical cancer patients with high levels of PCBP1-AS1 expression are more likely to promote the initiation and growth of cervical cancer than patients with low levels of PCBP1-AS1 expression due to the effect of tumor stage, pathologic TNM, and lymph invasion.

To understand the influence of PCBP1-AS1 in tumor-infiltrating lymphocytes, we analyzed the possible correlations between PCBP1-AS1 expression and levels of immune infiltration in cervical cancer. As shown in Figure 6 , PCBP1-AS1 expression showed a positive correlation with the levels of CD8⁺ T cells (p < 0.05), CD4⁺ T cells (p < 0.05), B cells (p < 0.05), cancer-associated fibroblasts (p < 0.05), myeloid dendritic cells (p < 0.05), eosinophils (p < 0.05), mast cells (p < 0.05), neutrophils (p < 0.05), and regulatory T cells (p < 0.05). In contrast, the presence of macrophages and monocytes was negatively correlated with the levels of PCBP1-AS1 expression. The results indicated that PCBP1-AS1 played an important role in immune infiltration in cervical cancer. Meanwhile, to study whether the cervical cancer immune microenvironment was different in cases with high PCBP1-AS1 levels compared those with to low levels, we downloaded an RNA expression profile obtained from TCGA. The cervical cancer samples were divided into two groups with the median value of PCBP1-AS1 expression as a cutoff. Then, we explored the expression profiles to obtain a fraction of 18 immune cell subtypes and assessed the differences in their expression levels in the two PCBP1-AS1 expression groups ( Figure 7A ). B cells, CD4+ T cells, M0 macrophages, M2 macrophages, and activated NK cells were significantly affected by PCBP1-AS1 expression. M2 macrophages and activated NK cells were increased (p < 0.05) in the low expression group compared to the high expression group. In contrast, CD4+ T cells, B cells, and M0 macrophages were increased in the high expression group (p < 0.05). We collected five pairs of samples (five cervical cancer tissues and five adjacent tissues) and extracted total protein from them. Differences in CD4+ T cells (CD4), B cells (CD19), and NK cells (CD56) in normal samples and cervical cancer samples were detected by western blotting. The results showed that CD56 protein was present at low levels in the tumor samples. In contrast, CD4 and CD19 protein levels were higher in tumor samples than in normal samples, which indirectly confirmed the results of the above analysis of immune infiltration ( Figures 7B, C ). In addition, we analyzed the correlations between 18 types of immune cells ( Figure 7D ), which revealed that the different infiltrating immune cell subpopulations of cervical cancer were moderately correlated.

To further analyze the function of PCBP1-AS1, GSEA was performed, and the most differentially (FDR q-val < 0.250, NOM p-val < 0.050) enriched signaling pathways and functions were selected based on the normalized enrichment score (NES). As shown in Figure 8B , the GO sets of molecular functions and biological processes significantly associated with PCBP1-AS1 expression were cell adhesion, cell migration, cell proliferation, regulation of apoptosis, cell resistance, and chromatin regulation. KEGG pathway analysis showed that the four pathways with the strongest positive correlations with PCBP1-AS1 expression were protein export, proteasome, p53 signaling pathway, and glycolysis gluconeogenesis; the four pathways with the strongest negative correlations were phosphatidylinositol signaling, basal cell carcinoma, bladder cancer, and Notch signaling, as shown in Figure 8A . The above GO and KEGG pathway annotations are shown in Table 5 . These results revealed that the expression level of PCBP1-AS1 was strongly associated with GO functions and pathways regulating cell function (cell adhesion, migration, proliferation, apoptosis, and resistance) and chromosome and protein activity.

To further address the biological function of PCBP1-AS1, we used siRNA and overexpression plasmids to alter the expression level of PCBP1-AS1 in HeLa cells and analyzed the effect of PCBP1-AS1 on HeLa cell proliferation, migration, and invasion. Fluorescence microscopy analysis of PCBP1-AS1 revealed successful transfection of the overexpression plasmid ( Figure 9A ), and siRNA‐PCBP1-AS1 led to a significant decrease in PCBP1-AS1 expression in HeLa cells ( Figure 9B ). The effect of PCBP1-AS1 on cell proliferation was detected by the CCK-8 assay. Compared with the negative control (NC) group, overexpression of PCBP1-AS1 significantly promoted HeLa cell proliferation, whereas HeLa cell proliferation was significantly impaired by PCBP1-AS1 knockdown (p < 0.001) ( Figure 9E ). Meanwhile, the wound-healing assay revealed that overexpression of PCBP1-AS1 significantly enhanced HeLa cell migration, and siRNA-PCBP1-AS1 showed a notably slower scratch closure rate than control cells ( Figures 9C, D ), which revealed that silencing PCBP1-AS1 inhibited HeLa cell migration (p < 0.001). Furthermore, the Transwell assay demonstrated that PCBP1-AS1 knockdown HeLa cells displayed significantly lower invasion potential than the control cells (p < 0.001) ( Figures 9F, G ). Collectively, these results suggest that the expression level of PCBP1-AS1 affected the proliferation, migration, and invasion of cervical cancer cells.