The identification of differentially expressed genes (DEGs) between pre-invasive and invasive cervical squamous cell carcinomas from GSE7803 was conducted by GEO2R in NCBI under the criteria of corrected adj. p-value <0.05. Gene expression data with clinical information from CESC were collected from The Cancer Genome Atlas (TCGA) Program (https://www.cancer.gov/tcga), which included 3 normal and 306 cervical cancer tissues. The Oncomine database (https://www.oncomine.org) was used to compare the expression level between cervical cancer and normal cervix uteri tissues in several GEO datasets. The UALCAN database (http://ualcan.path.uab.edu/) and GEPIA database (http://gepia.cancer-pku.cn/index.html) were used to assess the expression of PAMR1 in cervical cancer patients (11, 12). Gene expression data of normal cervix tissue were also collected from the Genotype-Tissue Expression Project (GTEx) (13). Kaplan–Meier plotter (http://kmplot.com/analysis/) was used to assess the prognostic value of PAMR1 in cervical cancer patients (14). TIMER platform (https://cistrome.shinyapps.io/timer/) was used to study the differential expression between tumor and adjacent normal tissues for PAMR1 across different types of cancers (15). TNM plot (https://tnmplot.com/analysis/) was used to compare the expression of PAMR1 in normal cervix, cervical cancer, and metastatic tissues (16). Pearson correlation analysis was conducted by R (3.6.3).

Two groups of cervical cancer tissues were collected. The first group included 40 normal cervical tissues, 58 cervical intraepithelial neoplasia (CIN) tissues, and 40 cervical cancer tissues. The second group included 31 paired cervical cancer and matched normal cervix samples. Among these cervical cancer samples, 11 cases had no lymphatic metastasis and 4 cases had lymphatic metastasis. The samples were obtained from the Department of Obstetrics and Gynecology at Shanghai General Hospital and immediately fixed in formalin. The tissues were embedded in paraffin to construct tissue microarray (TMA) for subsequent immunohistochemical (IHC) staining. All clinicopathological diagnoses were confirmed by two pathologists according to the guidelines of the World Health Organization (WHO) classification. The present study was approved by the Ethics Committee of Shanghai General Hospital. Written informed consent was obtained from all subjects before enrollment in this study.

The TMA was boiled in 10 mM sodium-citrate buffer for 5 min to retrieve antigen after dewaxing and rinsing. Then, the TMA was blocked in 4% hydrogen peroxide for 10 min in case of intervention of endogenous peroxidase activity. Thereafter, the TMA was incubated with the anti-PAMR1 rabbit polyclonal antibody (1:250, Proteintech, Wuhan, China) at 4°C overnight. The next day, the TMA was incubated with secondary antibody at room temperature for 30 min and stained with DAB and hematoxylin. Lastly, the TMA was covered with coverslips for microscopic observation. Staining intensity was scored on the following scale: 0 (negative staining), 1 (weak staining), 2 (moderate staining), and 3 (strong staining). The proportion of positively stained areas was evaluated with five levels: 0 (<5%), 1 (5%–25%), 2 (25%–50%), 3 (50%–75%), and 4 (>75%). The final score was the product of the above two indicators.

Human cervical cancer lines HeLa, SiHa, C33A, CaSki, and Me180 and normal cervical epithelial cell line H8 were purchased from the Type Culture Collection of the Chinese Academy of Sciences (Shanghai, China). The cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM medium; GIBCO from Thermo Fisher, USA) with 10% fetal bovine serum (FBS, GIBCO) and 1% penicillin–streptomycin, in a humidified atmosphere of 5% CO₂ at 37°C.

The PAMR1 small interfering RNA (siRNA), negative control siRNA (siNC), and overexpression vector pEX-3-PAMR1 were purchased from Shanghai GenePharma Co. Ltd. The siRNA sequences were as follows: siPAMR1-2: sense GCCUGGAGUUUGACUACAUTT, antisense AUGUAGUCAAACUCCAGGCTT and siPAMR1-3: sense CGGUUUCCAUGCCAUUUAUTT, antisense AUAAAUGGCAUGGAAACCGTT. Lipofectamine 3000 (Invitrogen, Carlsbad, CA, USA) was used to transfect with siRNA or plasmids according to the instructions of the manufacturer. The efficiency of transfection was identified by using quantitative real-time PCR (qRT-PCR) analysis.

Total RNA was extracted from cervical cell lines by using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) according to the instructions of the manufacturer. The cDNA was synthesized with random primers using a HyperScript III RT SuperMix (NovaBio, Shanghai, China); 2× S6 Universal SYBR qPCR Mix (NovaBio, Shanghai, China) was used to perform a qRT-PCR reaction according to the guidelines of the manufacturer. GAPDH was used as internal controls. The 2^−ΔΔCt method was used to analyze relative expression levels. The primers were all synthesized by Sangon (Sangon Biotech, Shanghai, China). Primers used in the present study are listed in Supplementary Table S1 .

Cervical cancer cells were lysed in 1× SDS loading buffer to extract cellular protein. Equal amounts of protein were separated on SDS-polyacrylamide gels and transferred onto a PVDF membrane (Millipore, Billerica, MA, USA) at 300 mA for 100 min. The membranes were blocked with 5% BSA (Roche, Mannheim, Germany) for 1 h, and then incubated with the primary antibody overnight at 4°C. Antibodies against human vimentin (#5741), E-cadherin (#3195), N-cadherin (#13116), LC3A/B (#12741), SQSTM1/p62 (#8025), and GAPDH (#5174) were purchased from Cell Signaling Technology (USA). Antibody against PAMR1 (55310-1-AP) was purchased from Proteintech (China). Antibodies against MAPKAP-1 (F-3) and ULK1 (F-4) were purchased from Santa Cruz Biotechnology (USA). Antibody against human ATF4 (WL02330) was purchased from Wanleibio (China). After washing with Tris-buffered saline with Tween-20 (TBST) three times, the membranes were incubated with secondary antibody for 1 h at room temperature. The proteins were visualized using ECL chemiluminescence (NCM Biotech) and quantified with ImageJ (National Institutes of Health, Bethesda, MD, USA).

Cell counting kit-8 (CCK8, Yeasen, China) was used according to the instructions of the manufacturer. Cells were seeded at an appropriate density in 96-well plates and cultured at 37°C. Ten microliters of CCK8 solution was added to each well after 24, 48, 72, and 96 h, respectively, and incubated for 1 h in an incubator. Relative cell density was determined at 450 nm using the Varioskan Flash Multi-Mode Microplate Reader (Thermo Scientific, USA).

For cell wound-healing assay, after the cells were cultured to 90% confluence in six-well plates, they were scratched with a 10-μl micropipette tip in the center of the well. Then, the cells were incubated in 1% FBS medium and images were captured at 0 and 24 h after injury to evaluate the migration rate. All experiments were repeated independently in triplicate.

For Transwell assay, 1 × 10⁵ cells of each group in 200 μl serum-free medium were seeded in the upper chamber (8.0 μm pore, Corning, USA) without (migration) or with (invasion) Matrigel (BD Bioscience, USA). Five hundred microliters of culture-medium with 10% FBS was added to the lower chamber. After incubating for 24 h, the upper chambers were fixed with 4% paraformaldehyde for 15 min and then stained with 0.1% crystal violet for 15 min. The cells were then photographed on the reverse side of the upper chambers. Five random fields were selected to calculate cells of migration and invasion.

We used gene set enrichment analysis (GSEA) v3.0 (http://www.broadinstitute.org/gsea/) to investigate significantly enriched genes of PAMR1 (17, 18). The expression level of PAMR1 was divided into high and low phenotype by median expression in TCGA-CESC data. c2.cp.kegg.v7.2.symbols.gmt (curated), c6.all.v7.2.symbols.gmt (oncogenic signatures), c5.go.v7.2.symbols.gmt (Gene Ontology), and h.all.v7.2.symbols.gmt gene sets (Hallmarks) from MSigDB were used as the reference gene sets. Gene set permutation was performed 1,000 times for each analysis. Normalized p-value <0.05 and FDR (false discovery rate) q-value <0.05 were considered threshold values to estimate statistical significance. The complete GSEA analysis results are presented in Supplementary Table S2 . All the diagram was drawn by R (3.6.3).

The SPSS 22.0 software was conducted for statistical analyses. Paired and unpaired continuous variables were compared by Student’s t-test. One‐way ANOVA was used for comparison among multiple groups. Kaplan–Meier analysis was performed to compare the overall survival (OS) rate between the high and low PAMR1 gene expression groups using the p-value determined in the log-rank test. All data are presented as the means ± SD. p-value <0.05 was considered statistically significant in all tests.

To identify DEGs related to the invasion of cervical cancer, we first analyzed the DEGs between pre-invasive and invasive cervical cancer samples using the GEO dataset GSE7803. This dataset contained seven high-grade squamous intraepithelial lesions and 21 invasive squamous cell carcinomas of the cervix samples. As a result, a total of 1,067 DEGs were identified with adjusted p-value <0.05, which included 578 upregulated genes and 489 downregulated genes. We found that PAMR1 was one of most downregulated DEGs in invasive cervical cancer (p = 3.13e-6, logFC = −3.635, Figure 1A ).

We assessed the expression of PAMR1 in cervical cancer patients by analyzing a public database. As shown in Figures 1B–E and Supplementary Figure S1A , the expression of PAMR1 was lower in cervical cancer compared with normal cervix uteri samples. We found that PAMR1 was negatively correlated to individual cancer stage ( Figure 1F ), but not tumor grade of cervical cancer ( Supplemental Figure S1B ). However, PAMR1 had lower expression in metastatic cervical cancer samples than in primary cervical cancer samples by using TNM plotter ( Supplementary Figure S1C ). Additionally, the expression of PAMR1 was also decreased in a variety of cancer types, such as colon adenocarcinoma (COAD), head and neck squamous cell carcinoma (HNSC), kidney chromophobe (KICH), liver hepatocellular carcinoma (LIHC), lung adenocarcinoma (LUAD), and rectum adenocarcinoma (READ) ( Supplementary Figure S1D ).

Next, we detected the protein level of PAMR1 via IHC in a group of 138 samples which included 40 normal cervical epithelial, 16 CIN I, 20 CIN II, 22 CIN III, and 40 cervical cancer tissues. Normal cervical epithelial samples exhibited stronger staining of PAMR1 than both CIN and cervical cancer tissues ( Figures 2A, B ). However, the difference among the three CIN groups and between the CIN and cervical cancer groups was not statistically significant ( Figure 2B ). We also analyzed the expression of PAMR1 in 31 pairs of cervical cancer and adjacent normal epithelial tissues. The score for PAMR1 was significantly lower in cervical cancer tissues than in adjacent normal tissues ( Figures 2C, D ), and the protein level of PARM1 was decreased in patients who had positive lymph node metastasis ( Figure 2E ). Moreover, Kaplan–Meier analysis of overall survival (OS) showed that low PAMR1 expression is not only associated with poor prognosis in cervical cancer patients (log rank p = 0.011, Figure 2F ) but also predicted worse OS in patients with endometrial carcinoma, lung adenocarcinoma, and liver hepatocellular carcinoma ( Supplementary Figures S2A–C ).

In cervical cancer cell lines, the mRNA expression of PAMR1 was lower compared with that in normal cervical epithelial cell line H8 ( Supplementary Figure S3A ). To explore the function of PAMR1 on cell migration and invasion, two siRNA-targeted PAMR1 and overexpression plasmids of PAMR1 were designed and transfected into HeLa and Me180 cells, respectively. The efficiencies of transfection were validated by qRT-PCR ( Supplementary Figures S3B, C ). We first determined the role of PAMR1 on cell proliferation using the CCK8 assay. The results showed that overexpression of PAMR1 significantly inhibited cervical cancer cell growth ( Supplementary Figures S3D, E ). The wound-healing and Transwell assays were then employed to confirm the role of PAMR1 on cell migration and invasion. The results showed that PAMR1 knockdown markedly promoted migration and invasive activities in HeLa and Me180 cells ( Figures 3A, B ), whereas overexpression of PAMR1 obtained opposite results ( Figures 3C, D ). Since epithelial-to-mesenchymal transition (EMT) is an important biological process for the migration and invasion of cancer cells, we then analyzed the influence of PAMR1 on EMT markers. As a result, qRT-PCR and Western blot both confirmed the inhibitory effect of PAMR1 on EMT by inhibiting N-cadherin expression and promoting E-cadherin expression ( Supplementary Figures S4A–C ). Taken together, these findings suggested that PAMR1 inhibited the proliferation, migration, and invasion of cervical cancer cells.

To identify PAMR1-related signaling pathways in cervical cancer, GSEA analysis between PAMR1 high and low expression datasets from TCGA-CESC was conducted to reveal enriched signaling pathways which were activated or suppressed by PAMR1. The 15 most significantly enriched signaling pathways of the KEGG or Hallmark pathway database based on their normalized enrichment score were selected to be plotted ( Figures 4A–E ). For example, high expression of PAMR1 was positively associated with UV response, myogenesis, and KRAS signaling, whereas low expression of PAMR1 was positively correlated with MYC targets, mTORC1 signaling, G2M checkpoint, hypoxia, and E2F targets ( Figures 4A–E ). Additionally, KEGG analysis also revealed PAMR1 function in vascular smooth muscle contraction, cell adhesion, and glucose metabolism. We then conducted C6: oncogenic signature gene sets for analysis and got consistent results. PAMR1 was correlated with inhibited expression of target genes which were upregulated by MYC and mTOR ( Supplementary Figures S5A, B ). Moreover, GO analysis showed the biological process, cellular component, and molecular function of PAMR1 in cervical cancer ( Supplementary Figure S5C ).

As mentioned above, we found that the MYC target and mTORC1 signaling pathways were remarkably enriched in PAMR1 low expression phenotype, suggesting that PAMR1 might inhibit these two signaling pathways. To confirm our hypothesis, we analyzed the expressions of the core genes in these two pathways with upregulation or downregulation of PAMR1. In MYC target signaling, ATF4 and MXD3 were significantly negatively regulated by PAMR1 ( Figures 5A, B ). In the mTOR signaling pathway, the expressions of most genes were regulated by PAMR1, among which SIN1, ATG1, and MLST8 expressions showed significant differences ( Figures 5C, D ). In order to confirm the relevance between PAMR1 and these target genes in cervical cancer tissues, we analyzed TCGA-CESC data and GEO microarray data (GSE44001). In TCGA-CESC data, the expressions of MYC and MAPKAP1 were negatively related to PAMR1 expression ( Supplementary Figure S6A ). In the GEO dataset, MYC and MXD3 were also found negatively related to PAMR1 expression and ULK1 was found to have a positive correlation with PAMR1 ( Supplementary Figure S6B ). Moreover, the protein level of ULK1 was upregulated and that of MAPKAP1 was downregulated in PAMR1-overexpressed HeLa cells, whereas opposite results were observed in PAMR1 knockdown cells ( Supplementary Figure S7A ). As an autophagy initiation factor, ULK1 has been proven to promote autophagic flux and inhibit cancer progress and metastasis (19, 20). We then examined the effect of PAMR1 on autophagy using Western blot. Overexpression of PAMR1 increased LC3 expression and decreased p62 expression, whereas LC3 was reduced and p62 was raised when silencing PAMR1 ( Supplementary Figure S7B ).