The data were acquired from the TCGA database (https://portal.gdc.cancer.gov/) and comprised normalized whole-genome mRNA expression data of 379 HGSOC samples, somatic mutation data, and corresponding clinical data. After removing 1 patient without complete survival information, 378 HGSOC patients with simultaneously accessible OS and mRNA expression profile data were included in the study. GSE149940 and GSE32062 were extracted from GPL4133 and GPL6480 using the GEO database (http://www.ncbi.nlm.nih.gov/geo). The validation cohort (GSE32062) were external cohorts. The platform annotation files downloaded from the database were adopted to convert the probe data in the matrix files into gene symbols. To eliminate the batch effect, the “sva” R package (https://bioconductor.org/packages/release/bioc/html/sva.html) was utilized [15]. These data were obtained from publicly available web sources and we conducted this research in accordance with applicable protocols regarding the use of databases.

The human ovarian carcinoma SKOV3 cell lines and normal ovarian cell line IOSE80 were obtained from (ATCC, Manassas, VA, USA). IOSE80 cells were cultured in RPMI-1640 medium (BasalMedia, L210KJ, Shanghai, China) with 10% fetal bovine serum (Gibco, cat. no. 10270-106, Thermo Fisher Scientific, Waltham, MA, USA), and SKOV3 cells were cultured in McCoy’s 5a Medium (cat. no. 30-2007, American Type Culture Collection (ATCC)) with 15% fetal bovine serum. Cells were cultivated at 37 °C in a humidified atmosphere containing with 5% CO${}_2$. All cell lines were authenticated shortly before use by the short tandem repeat (STR) profile, carried out by Genewiz (Genewiz, Suzhou, China) and Genetic Testing Biotechnology Corporation (Genetic Testing Biotechnology Corporation, Suzhou, China). Cells were routinely tested for mycoplasma by PCR.

The immune cells marker gene sets were acquired from another publication [16]. Furthermore, for each immune-related cell, the R package “gsva” (https://bioconductor.org/packages/release/bioc/html/GSVA.html) was utilized to perform ssGSEA to calculate the enrichment score [17]. The “ConsensusClusterPlus” algorithm (50 iterations, resample rate of 80%) was used to cluster the HGSOC samples into three distinct groups (low-, medium-, and high-immunity) according to immune cell enrichment scores in ssGSEA [18]. To confirm that the three subgroups had distinct immunological profiles, we employed the R package “estimate” (https://bioinformatics.mdanderson.org/estimate/rpackage.html) to determine the immune, ESTIMATE, and stromal scores of each tumor sample [19]. Differential expression analysis of DEGs between low- and high-immunity groups was realized using the “limma” R package (https://bioconductor.org/packages/release/bioc/html/limma.html) [20]. p 0.05 and $|$log2 fold change (FC)$|$$>$1.0 were adjusted as thresholds. Additionally, The Molecular Signatures Database (MSigDB) (https://www.gsea-msigdb.org/gsea/msigdb/index.jsp) was used to retrieve leukocyte migration sets. Then, the Venn diagrams were plotted to detect leukocyte migration-related differentially genes (LMRGs) using the R package “VennDiagram” (http://cran.r-project.org/web/packages/VennDiagram/index.html) [21].

In order to generate a prognostic multigene signature in the training set, univariate Cox analysis and LASSO regression analysis were conducted. The “glmnet” R package (https://cran.r-project.org/web/packages/glmnet/index.html) was used to implement the abovementioned processes [22]. In order to calculate the risk score, the following equations were used: risk score = $\sum$ (expression level of each gene $\times$ corresponding regression coefficient). We categorized the patients into two groups according to the median value of risk score as the cutoff, namely: low-risk and high-risk groups. In addition, the Kaplan-Meier curve and the “survival” R package (https://cran.r-project.org/web/packages/survival/index.html) were used to examine variations in survival rates. Time-dependent ROC curves for 1-, 3-, and 5-year survival were plotted to appraise the prediction performance of the model. Furthermore, in the TCGA cohort, we utilized univariate and multivariate Cox regression analyses for OS to evaluate the prognostic association between the risk signature and other clinical parameters (such as age, stage, residual tumor, and TCGA subtype). Finally, using the GEO cohort (GSE32062) as external validation set, a similar approach was used to test the reliability and generality of the risk signature.

GSEA was utilized to determine if a previously established gene set exhibited obvious differential expression between the low- and high-risk groups in the enrichment of MSigDB Database (c2.cp.kegg and c5.go.bp. v7.2. symbols.gmt). Then, it was assumed that the phenotypic labels were represented by the low- and high-risk and gene set permutations were undertaken 1000 times for each analysis. Classification of the gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways that were enriched in differential phenotype was done using the normalized enrichment score (NES) and the false discovery rate (FDR).

The TME in HGSOC exhibits immunological properties such as activation of the anticancer immunity cycle, tumor infiltrating immune cells (TIICs) infiltration, and expression of inhibitory immune checkpoints and immunomodulators. From the study conducted by Charoentong et al. [23], we initially compiled data on 92 immunomodulators, such as receptors, chemokines, and MHC. In the anticancer immune response, there are seven steps inside this cancer immunity cycle [24]. These steps and their influence on the ultimate fate of tumor cells were examined by Xu et al. [25] using ssGSEA, based on the gene expression levels of individual samples. Thereafter, in order to reduce the possibility of computation mistakes, we computed the infiltration levels of TIICs using seven separate algorithms: xCell, Cibersort-ABS, TIMER, quanTIseq, MCP-counter, Cibersort, and EPIC [25, 26, 27, 28, 29, 30]. Using earlier research, we were able to identify the effector genes of TIICs.

Then, type I interferon (IFN) response and its marker genes were obtained from Akul’s study [31]. From Auslander’s investigation [32], we also obtained ten inhibitory immune checkpoints with treatment significance for further evaluation. The other gene sets, which present T cell-inflamed gene expression profile (GEP) or immune cytolytic activity (CYT), were collected from earlier studies [33, 34]. Finally, we obtained the T cell receptor (TCR) and B cell receptor (BCR) Shannon Entropy data from Vésteinn Thorsson et al. [35].

In order to identify single nucleotide variants (SNVs), single nucleotide polymorphisms (SNPs), and insertion-deletions (INDELs), we utilized the WES somatic mutations data from both the low-risk (n = 136) and high-risk (n = 135) groups with the aid of VarScan2.39 program. The Fisher’s exact test was utilized to assess the differential mutation genes that had a p-value 0.05. The somatic mutations were visualized utilizing the “maftools” R package (https://www.bioconductor.org/packages/release/bioc/html/maftools.html). Furthermore, aneuploidy scores and homologous recombination deficiency (HRD) score was obtained from Vésteinn Thorsson et al. and Taylor et al. [35, 36]. Mutant-allele tumor heterogeneity (MATH) and Tumor mutation burden (TMB) was obtained from the somatic mutation data of 271 tumor samples, using “maftools” R package [37, 38].

The Genomics of Drug Sensitivity in Cancer (GDSC) (https://www.cancerrxgene.org/) was used to estimate each patient’s chemotherapeutic response. Ridge regression was employed to evaluate the half-maximal inhibitory concentration (IC50) and the “pRRophetic” R package (https://github.com/paulgeeleher/pRRophetic) was used to conduct 10-fold cross-validation [39].

We used two computational approaches to anticipate the immunotherapy response in HGSOC patients at low- and high-risk groups in order to investigate the association between the immunological signature and immunotherapy effectiveness. To begin with, Tumor Immune Dysfunction and Exclusion (TIDE) (http://tide.dfci.harvard.edu) was utilized to anticipate each sample’s anti-CTLA4 and anti-PD1 immunotherapy response on the basis of the transcriptome patterns [40]. Second, we retrieved patient immunophenoscore (IPS) from The Cancer Immunome Atlas [23].

Real time PCR was performed as previously described [41]. The expression levels of CD2 were measured by normal ovarian cell (IOSE80) and ovarian cancer cell (SKOV3) on ABI 7700 system using the following primers: forward CD2-F: 5′-TCAAGAGAGGGTCTCAAAACCA-3′, reverse CD2-R 5′-CCATTCATTACCTCACAGGTCAG-3′; GAPDH-F: 5′-TGACTTCAACAGCGACACCCA-3′, reverse GAPDH-R 5′-CACCCTGTTGCTGTAGCCAAA-3′. Total RNA from cultured cells was isolated at 80% confluence with TRIzol reagent (T9424, Sigma Aldrich, STL, USA). Total RNA (1 µg) was reverse transcribed into cDNA using a reverse transcription kit (Takara, Japan). Real-time PCR was performed using SYBR Premix Ex Taq (Takara, Otsu, Japan) according to the manufacturer’s instructions. PCR reaction conditions: denaturation at 95 °C for 10 s, annealing at 60 °C for 15 s, and extension at 72 °C for 30 s. This cycle is amplified for 45 times, and the melting curve is analyzed after the cycle. 23 HGSOC samples (OC-1601) were procured from Servicebio Company (Wuhan, China) and utilized to IF with anti-human CD2 (ab4055, Abcam, MA, USA), anti-PD-1 antibody (ab213524, Abcam, MA, USA) and CD8-specific antibody (ab4055, Abcam, MA, USA). We evaluated only the proportion of cells that had a high membrane staining intensity (brown staining) for CD2 and CD8. The corresponding secondary antibodies included CY5-TSA (G1224, Servicebio), which was utilized to detect PD1, FITC-TSA (G1223, Servicebio), which was utilized to detect CD8, and CY3-TSA (G1222, Servicebio), which was utilized to detect CD2. DAPI was used to highlight the nuclei. Lastly, we calculated the percentages of positive cells across the field.

In order to categorize tumor, stromal, and immune cells, we implemented hierarchical clustering based on the ovarian Cancer (OV) single-cell sequencing data (GSE118828) collected from the electronic website of the Tumor Immune Single-cell Hub (TISCH) (http://tisch.comp-genomics.org/) [42, 43]. After that, the expression of CD2 in these cells was determined, and the findings were shown graphically using scatter plot.

R software version 4.0.3 (https://www.r-project.org/) was utilized to conduct all statistical analyses and plot drawings. The unpaired Student’s t-test and the Whitney U-test were utilized to compare variations between two groups with normal distributed and non-normal distributed variables, respectively [44]. Categorical variables were compared utilizing a Chi-square test [45]. With regard to non-parametric data, the Wilcoxon rank-sum test was employed when two groups were involved, whereas the Kruskal-Wallis test was employed when more than two groups were involved. Using Pearson’s chi-square test, correlations between qualitative variables were examined. A p-value less than 0.05 was interpreted as having statistical significance if it was not indicated above.

With the application of the ssGSEA method, 48 infiltrating immune cells were integrated into HGSOC tissues in order to measure their immunological capability (Fig. 1A). Consensus clustering analysis was used to categorize the overall TCGA cohort into three different groups (high-immunity: 217 samples; median-immunity: 120 samples, and low-immunity: 42 samples). For the purpose of validating the immunity of the three immune groups, we included the ESTIMATE, the immune, and the stromal scores in the heatmap (Fig. 1B). We discovered that when compared to the group with low immunity, the group with high immunity exhibited stronger immune components as well as lesser tumor purity.

Fig. 1.

Immune microenvironment clustering of HGSOC and Construction of Prognosis Prediction Model. (A) There are 48 immune-related gene sets in ssGSEA that have been enriched in HGSOC cancer The gene sets are comprised of immune processes and immune cells. This heatmap also displays the tumor purity, stromal scores, immune scores, and ESTIMATE scores. (B) Comparison of leukocyte migration-related differential expression genes (DEGs) of low- and high-immune infiltration groups in this heatmap. (C) Nine genes were found to be correlated with the prognosis according to the LASSO Cox analysis. (D) 1000-round cross-validation. was used to determine the best values for the penalty parameter.

To examine the function of leukocyte migration in modulating microenvironment immunity of HGSOC, we performed limma analysis to identify 39 differentially expressed leukocyte migration-related genes in the low- and high-immunity group (Fig. 1B, Supplementary Table 1). Then, Univariate Cox and LASSO regression analyses were carried out in order to detect significant prognostic biomarkers, and the results were used to create a risk signature (Fig. 1C,D). The formula for the risk signature was determined using corresponding coefficients: risk score = 0.6054 × CXCR2 - 0.2539 × CCR7 - 0.2659 × SELL - 0.0251 × CD2 + 0.0127 × TREM1 - 0.8400 × TBX21-0.1818 × CCL13 + 0.3648 × ITGAM + 0.4273 × SIRPG.

HGSOC samples were categorized into low- and high-risk groups according to their median risk score. The Kaplan-Meier curve revealed that the high-risk group samples exhibited poorer overall survival (OS) and progression-free survival (PFS) as opposed to the low-risk group samples, demonstrating that the prognostic signature of the risk score is efficacious for predicting survival. Notably, the data from the GEO (GSE32062) database was used to validate this finding (Fig. 2A,B). In addition, by applying receiver operating characteristic curve analysis on the TCGA data, it was discovered that risk signature was a strong predictor for 1-year (AUC = 0.646), three-year (AUC = 0.634), and five-year survivals (AUC = 0.685). The validity of this discovery was greatly increased by incorporating data gathered from the GEO database (AUC = 0.640, 0.595, 0.628) (Fig. 2C). Univariate Cox analysis revealed that risk signature (HR = 1.783; 95% CI = 1.440–2.207; p 0.001), and tumor residual disease were both greatly associated with a dismal OS. The results of the Multivariate Cox analysis illustrated that high-risk signature in TCGA database (HR = 1.829; 95% CI = 1.420–2.290; p 0.001) and in GEO database (HR = 1.287; 95% CI = 1.099–1.506; p = 0.002) exhibited an independent correlation with a worse OS (Fig. 3A,B). Consequently, our risk signature has the potential to serve as an independent prognostic marker for HGSOC, as shown by this finding.

Fig. 2.

The prognostic value of risk signature. (A) In the GEO and TCGA datasets, low-risk group patients had a favorable OS rate as opposed to those in the high-risk group. (B) In the TCGA and GEO datasets, low-risk group patients had a longer PFS as opposed to the high-risk group patients. (C) The ROC curve for 1-, 3-, and 5-year OS of HGSOC patients in the GEO and TCGA datasets.

Fig. 3.

Evaluation of the independent prognostic significance and GSEA enrichment analysis of risk signature. (A,B) Cox analyses, both univariate and multivariate, were used to determine the independent prognostic significance of the risk signature with regard to OS in glioma patients based on the CGGA and TCGA datasets. (C) Relationships between risk signature and stage in HGSOC. (D) Relationships between risk signature and TCGA subgroup in HGSOC. (E,F) GO and KEGG were evaluated utilizing the GSEA. **p 0.01, ***p 0.001, ${}^{\text{ns}}$p $>$ 0.05.

Then, we explored relationships between risk signature and stage in HGSOC. According to the data from the TCGA and GEO databases, HGSOC patients showed significantly higher risk scores with the high stage as opposed to the low stage group patients (Fig. 3C). Interestingly, we discovered that the immunoreactive group had a lower risk score compared to patients in the other TCGA subgroups (Fig. 3D). These findings identify that risk signature is crucial to anti-tumor immunity.

Gene set enrichment analysis was performed on the GO and KEGG databases utilizing the MSigDB database (FDR 0.05). NES was conducted to determine the GO and signaling pathways that were strongly enriched. In this research, antigen processing, as well as peptide antigen presentation through immune response regulating signaling pathway, MHC class I, positive modulation of T cell receptor signaling pathway, response to type I interferon, positive modulation of lymphocyte chemotaxis, antigen processing and presentation, JAK_STAT signaling pathway, and cell adhesion molecules (CAMs), were enriched in low-risk phenotype (Fig. 3E,F).

Depending on the proportion of cytotoxic immune cells infiltrating TME, the tumor can be described as either immunologically active ‘inflamed’ or immunologically passive ‘non-inflamed’ [46]. Risk signature was observed to exhibit a negative correlation with a vast proportion of immunomodulators (Fig. 4A). Most of the MHC molecules were downmodulated in the high-risk group, indicating that the ability to present and process antigens had been reduced. Three key chemokines (CXCR3, CXCL10, and CXCL9), which are required to recruit CD8+ T cells into the TME in HGSOC, were shown to be upmodulated in the low-risk group. Further investigation revealed that paired receptors and other chemokines such as CXCL13, CXCL11, CCL5, XCL2, and CCL4, exhibited a negative association with risk signature. These receptors and chemokines stimulate the mobilization of effector TIICs, including antigen-presenting cells and CD8+ T cells.

Fig. 4.

The low-risk signature shapes an inflamed TME in HGSOC. (A) In HGSOC, there are differences in the expression of 122 immunomodulators (MHC, receptors, and chemokines). (B) Difference between low- and high-risk groups at distinct stages of the cancer immunity cycle. (C) Relationship between the risk score and infiltration levels of severe TIICs, as determined by seven separate algorithms. (D) Differences in the effector genes of CD8+ cells between low- and high-risk groups. (E) Relationship between risk score and CD8+ cell effector genes. *p 0.05, **p 0.01, ***p 0.001, ****p 0.0001, ${}^{\text{ns}}$p $>$ 0.05.

The cancer immunity cycle is a representation of our body’s immunological response to cancer. The functions of the cancer-immunity cycle meticulously represent the ultimate impact of the intricate immunomodulatory interplay in the TME [24, 25]. Most of the cycle steps were shown to be increased in the low-risk group, such as the production of cancer cell antigens (Step 1), priming and activation (Step 3), as well as immune cells transportation to tumors (Step 4) (recruitment of DC, NK cells, Macrophages, Th1 cells, and CD8 T cell) (Fig. 4B). Consequently, the decreased activity of these steps might result in a decrease in the infiltration levels of effector TIICs in the TME. Notably, the low-risk group had a decreased T cell activity in recognizing cancer cells (Step 6). The reason for this phenomenon might lie in the fact that the low-risk group had an elevated level of PD-L1 expression. Step 7 activity (cancerous cell death) was discovered to be upmodulated in the low-risk group.

Following that, we estimated the infiltration levels of TIICs utilizing seven separate algorithms provided by the TIMER website. In our result, risk signature was inversely associated with DC cells, B cells, and CD8+ T cells in the TCGA datasets (Fig. 4C). These results were confirmed in the Geo cohort (Supplementary Fig. 1). Similarly, risk signature was inversely associated with the CD8+ T cells effector genes, which was elevated in the low-risk group (Fig. 4D,E). Moreover, the marker genes of type I IFN response were a high expression in the low-risk group. The risk model was inversely associated with type I IFN response (Fig. 5A,B). Consistently, risk signature was observed to be inversely associated with most of the immune checkpoint inhibitors such as TIGIT, IDO1, TIM-3, LAG-3, PD-1, CTLA-4, and PD-L1 (Fig. 5C,D). Collectively, the low-risk signature shapes an inflamed TME.

Fig. 5.

The immune profile of risk signature in HGSOC. (A) Relationship between risk score and type I IFN response. (B) Differences in the effector genes of type I IFN signature between low- and high-risk groups. (C) Relationship between risk score and 10 inhibitory immune checkpoints. (D) Differences in PD1, PDL-1, CTLA4, LAG3 between low- and high-risk groups. (E) Differences in GEP and CYT between low- and high-risk groups. (F) Differences in TCR and BCR diversity values between low- and high-risk groups. **p 0.01, ***p 0.001.

Meanwhile, we discovered the risk signature strongly correlated with CYT and GEP in HGSOC, which both decreased in the low-risk group (Fig. 5E). Furtherly, we evaluated TCR and BCR repertoires from the TCGA HGSOC cohort. Mean TCR and BCR diversity values, which were measured by Shannon entropy, differed by the risk signature, with the highest diversity in the low-risk score groups (Fig. 5F). A powerful anti-tumor response may be associated with antigen-specific BCR and TCR repertoires that are essential for the identification of malignant cells and pathogens. Our finding indicated risk signature was strongly associated with the inflamed TME and anti-tumor response.

To find the relevant genetic alterations, we dissected the somatic mutations between the low and high-risk cohorts. The top 30 most commonly mutated genes in the corresponding cohorts are depicted in Fig. 6A. Then we explored the association between risk signature and measures of DNA damage, including aneuploidy score, homologous recombination deficiency (HRD). The low-risk group correlated negatively with aneuploidy score, positively with HRD (Fig. 6B,C). Due to lower leukocyte infiltration in high aneuploidy samples, aneuploidy was negatively associated with immunological signaling gene expression [36]. These results suggest that the risk signature of the affected immune infiltrations may be associated with aneuploidy. It has been hypothesized that the genetic instability exhibited in HRD tumors alters immunogenicity, making these malignancies highly sensitive to immunotherapy. We next calculated the TMB and MATH by the maftools R package. Despite the fact that the TMB was greater in the low-risk group as opposed to that of the high-risk group, this difference was not significant. No differences in MATH were found between the low- and high-risk groups (Fig. 6D). This result demonstrated that risk signature may serve as an immune marker independent of TMB and MATH.

Fig. 6.

Mapping of mutations found in HGSOC. (A) The top 30 commonly mutated genes between low- and high-METTL7B groups. (B) Differences in AS, including AS_amp and AS_del between low- and high-risk groups. (C) Differences in HDR between low- and high-risk groups. (D) Differences in TMB and MATH between low- and high-risk groups. *p 0.05, **p 0.01.

The commonly used treatment options for ovarian cancer include chemotherapy, targeted therapy, and immunotherapy. Platinum-based chemotherapy is the cornerstone of treating HGSOC. To determine the effect of Risk signature on Platinum-Based Chemoresistance in HGSOC, according to platinum-based response (whether to platinum resistance) and risk score, we categorized the samples in TCGA and GEO databases into four groups, including high-risk score groups with or without platinum resistance and low-risk score groups with or without platinum resistance. We found that no matter whether the risk score in HGSOC patients was high or low, patients who were sensitive about platinum had a longer survival time as opposed to those with platinum resistance. But the low-risk score group was found to have a survival advantage as opposed to the high group in HGSOC patients who were sensitive about platinum, but not those with platinum resistance (Fig. 7A). Moreover, patients with platinum resistance had a higher risk score (Fig. 7B). Furthermore, we estimated the IC50 for each subtype using the prediction model for the three agents developed by Xiaofan Lu [47]. The findings indicated that the low-risk group responded more favorably to chemotherapy as opposed to the high-risk group (cisplatin, p 0.01; paclitaxel, p 0.001; etoposide, p 0.001) (Fig. 7C). Olaparib and Niraparib, the poly(ADP-ribose) polymerase (PARP) inhibitors, were authorized for maintenance treatment of advanced HGSOC after first-line platinum chemotherapy [48]. We then analyzed the effect of risk signature on PARP inhibitors using the GEO database (GSE149960) (Fig. 7D). Results demonstrated that PARP inhibitor-resistance group tended to have higher risk scores than PARP inhibitor-sensitive group (p = 0.03). The findings above revealed that the risk signature can be used to identify potential groups of patients who gain benefit from PARP inhibitors and platinum-based chemotherapy.

Fig. 7.

Risk signature predicts the response of conventional therapy. (A) Kaplan-Meier curves of OS for the high-risk score with or without platinum resistance and low-risk score with or without platinum resistance in the TCGA and GEO cohort. (B) Differences in risk score between platinum resistance and platinum-sensitive groups. (C) Differences in risk score between PARP inhibitors resistance and PARP inhibitors sensitive groups. (D) The boxplots depict the estimated IC50 values for doxorubicin, paclitaxel, and cisplatin for three subtypes, as determined by the GDSC database. (E) The anticipated immunotherapy (TRUE/FALSE) response rate to anti-PD-L1 in low- and high-risk groups in the TCGA cohort. (F) Differences in IPS between low- and high-risk groups. *p 0.05, **p 0.01, ***p 0.001.

Despite recent advances in the maintenance treatment of HGSOC, patients with locally progressive or metastatic HGSOC have been less successful in achieving satisfactory clinical results. We utilized the TIDE algorithm to estimate immune checkpoint blockade (ICB) treatment response in order to determine possible groups of patients who may gain benefit from immunotherapy. Results illustrated that the low-risk group displayed a significantly improved response to immunotherapy as opposed to the high-risk group (Fig. 7E). The Cancer Immunome Atlas (https://tcia.at/) showed similar results. The low-risk group exhibited a greater response rate to CTLA4 and PD-1 inhibitors, implying that patients who have low-risk scores may gain benefit from immunotherapy (Fig. 7F).

We further performed a Friends analysis of 9 risk genes by using GOSemSim R package to screen a hub gene, CD2 (Fig. 8A) [49]. Notably, the expression of CD2 was greatly decreased in normal tissues compared to HGSOC tissues using Gene Expression Profiling Interactive Analysis (GEPIA) (Fig. 8B,C) [50]. Based on the cellular level, the expression of CD2 was shown to be greater in tumor cell as opposed to normal ovarian cell by PCR (Fig. 8D). In the TCGA database, patients with HGSOC who exhibited low expression of CD2 had obviously poorer OS compared with those who exhibited high CD2 expression (Fig. 8E). Following that, the survival analysis was tested utilizing the GEO database (Fig. 8F).

Fig. 8.

CD2 Expression and Survival Analyses. (A) The boxplots of the estimated factor value for 9 LMDGs based on Friends analysis. (B) Pan-cancer analysis of CD2 expression across the cancerous tissue and the corresponding adjacent normal tissue from TCGA. (C) Differential expression of CD2 in HGSOC. (D) Real-time quantitative PCR analysis of CD2 expression in normal ovarian cells and skov3 ovarian cancer cells. (E,F) Kaplan-Meier curves of OS for CD2 in the GEO and TCGA cohorts. *p 0.05, ***p 0.001.

TIICs of HGSOC samples were utilized to further explore the interaction between CD2 expression and the TME. Seven separate algorithms from the TIMER online platform were employed to analyze these TIICs. Results showed CD8+ T cells, dendritic cells, M1 macrophages strongly correlated with CD2, which offers convincing evidence for the critical function played by CD2 in the TME of HGSOC samples (Supplementary Fig. 2). Interestingly, the radar plot demonstrated that CD2 strongly correlated with immune checkpoint inhibitors (PD1, PDL1, CALT4, IDO1), TCR scores, BCR scores, GEP, and CYT, negatively correlated with TIDE score (Fig. 9A). Moreover, the group displaying elevated CD2 expression was found to have a considerably improved response to immunotherapy as opposed to the low group using the TIDE algorithm (Fig. 9B). Furthermore, multicolor immunofluorescence analysis demonstrated that CD2, PD-1, and CD8 were co-localized in HGSOC tissues, suggesting that CD2 and CD8+ cells spatially interact in the TME and optimizing the CD2 intensity in HGSOC tissue could enhance the effectiveness of immunotherapy by enhancing the anti-tumor immune responses (Fig. 9C,D). In addition, we found that the expression level of CD2 was higher in T cells compared with malignant cells and stromal cells in the OV patients (GSE118828) by using TISCH database (Fig. 9E).

Fig. 9.

Correlations between CD2, the immune infiltration, and the response of immunotherapy in HGSOC. (A) Radar graph of the correlation between CD2 and immune checkpoint inhibitors (PD1, PDL1, CALT4, IDO1), TCR scores, BCR scores, GEP, CYT, and TIDE score. (B) The anticipated immunotherapy (TRUE/FALSE) response rate to anti-PD-L1 in low- and high-risk CD2 in the TCGA cohort. (C) Expression of CD2, CD8, and PD-L1 in our cohort was detected using immunofluorescence in two immune phenotypes based on the expression of CD8+ T cells. (D) Correlation between the CD2 positive rate and CD8 positive rate detected using immunofluorescence. (E) Single cell level analysis evaluating the expression of CD2. ***p 0.001.