Data preparation, processing, general analysis, and verification of basic experiments in this study are displayed in the flow chart ( Figure 1 ). Primarily, transcriptome data, somatic mutation profiles, and clinical characteristics of 380 OC tumor cases were extracted from The Cancer Genome Atlas (TCGA) databases (https://portal.gdc.cancer.gov/). Besides, we further downloaded the gene expression profiles and corresponding clinical features of OC patients from the International Cancer Genome Consortium (ICGC) OV-AU (Ovarian cancer-Australia) database to verify the survival results in our study (https://icgc.org/). The clinical information of OC patients in TCGA and ICGC datasets are displayed in Table S1 . Then, KM Plotter online database (http://kmplot.com) was exploited to analyze the prognostic value of gene expression in OC. The OC cases were divided equally into two different expression groups to analyze the overall survival (OS). In order to demonstrate the prognostic value for immunotherapy, we ultimately included an immunotherapeutic dataset called IMvigor210 cohort from Mariathasan et al. in this study, which contained the intervention information of treatment using anti-PD-L1 antibody in urothelial cancer patients (32).

ESTIMATE package with R software (version 4.0.5) was applied to determine the proportion of immune, stromal, and immune-stromal components in TME of each OC patient. Each component was presented by means of ImmuneScore, StromalScore, and ESTIMATEScore, respectively (33). The higher the score, the higher the proportion. Then, 380 OC samples were separated equally into two cohorts (high-ImmuneScore group and low-ImmuneScore group) grounded on the median value of ImmuneScore. Additionally, immune-related differentially expressed genes (DEGs) were recognized via “limma” package through the comparison of high-ImmuneScore with low-ImmuneScore cohorts, with the filtering quality of false discovery rate (FDR) < 0.05 and absolute value of log2 fold change (FC) > 0.58. Simultaneously, gene expression profiles and somatic mutation data were combined into a tumor genome to recognize genomic instability-related DEGs and the following steps were conducted in turn: (i) the cumulative amount of somatic mutations per OC sample was calculated and then sorted in descending order, which meant the higher the total number of somatic mutations, the more the chromosomal mutation sites; (ii) the genomic unstable (GU) cohort contains the top 25% of samples while the genomic stable (GS) cohort contains the last 25%; (iii) genomic instability-related DEGs were identified based on the comparison of the transcriptome information between the GU and GS cohorts using “limma” package (absolute value of log2 FC > 0.58 and FDR < 0.05).

R package “survival” and “survminer” were loaded to perform survival analysis (34). The Kaplan–Meier method was taken to evaluate the survival rate, and the log-rank test was simultaneously employed to analyze the survival difference by comparing different expression cohorts (high- vs. low-expression groups) with p-value < 0.05. In this study, 361 OC samples were picked out from the total of 380 TCGA patients via the following exclusion criteria: (i) exclude the samples whose survival time is less than 30 days; and (ii) exclude the samples with incomplete clinical characteristics. The 93 OC patients from ICGC were also selected according to the above filtering standards. Additionally, the package “ggpubr” was employed to analyze the relationship between gene expression and clinicopathological features, taking Kruskal–Wallis rank sum test as the method to assess the statistical significance. Moreover, the KM Plotter online database (GSE15622 of 35 samples and GSE14764 of 80 samples) was also employed to verify the correlation of gene expression with survival rate in OC.

Functional enrichment analysis of DEGs was conducted to determine the significantly enriched Gene Ontology (GO) terms through R packages “clusterProfiler”, “enrichplot”, and “ggplot2”. The terms satisfying the standards (both p-value < 0.05 and q-value < 0.05) were considered as significant. Moreover, the HALLMARK gene sets and C2, CP, and KEGG.v7.2 symbols were performed between different gene expression cohorts through GSEA using the gsea software (version: 4.0.3). The gene sets were thought to be significant with the level of 5% (NOM p-value and FDR q-value).

At the same time, R package CIBERSOT (35, 36) and ssGSEA (37) algorithms were occupied to determine TICs proportion of all OC samples and cell immune responses between high- and low-expression cohorts. Additionally, ssGSEA using “GSVA” (R package) was specifically employed to calculate the proportion of TICs among different expression groups and assess immune-related biological functions as well. The immune gene sets were acquired from the MSigDB database (https://www.gsea-msigdb.org/). Moreover, ICPs were extracted from the previously published articles, whose correlation with gene expression was further analyzed with p-value < 0.05. Besides, IPS of OC patients was retrieved from the Cancer Immunome Database (TCIA, https://tcia.at/home) and its relationship with gene expression was then assessed with the significance quality of p-value < 0.05. Moreover, the IMvigor210 cohort was selected in our study to further verify the prognostic value of the biomarker for immunotherapy response, similarly with p-value < 0.05 as the significance quality.

Somatic mutation data of OC samples was stored in the MAF form and differentially mutated genes were simultaneously determined between the different expression cohorts via MAFTOOLS package (38). The TMB score of each OC case was computed according to the following formula: (total mutation/total covered bases) × 10⁶ (39). DDR genes and m6A-related regulators were gathered from many published literatures and the association of these regulators with gene expression was analyzed further (40–44).

Five OC specimens and five normal ovarian tissues were provided by Shanghai First Maternity and Infant Hospital, which were then collected and conserved in liquid nitrogen before use. The diagnoses of acquired samples were all carefully certified and checked by experienced pathologists, which were consistent with the diagnostic principles of the latest World Health Organization Classification. All specimens have obtained the informed consents with permission of Medical Ethics Committee of Shanghai First Maternity and Infant Hospital. Half of the tissues were stored in liquid nitrogen and the other half were fixed in formalin and embedded in paraffin for histological analysis.

Human OC cell lines HEY and the human leukemic cell line, THP-1 cells were purchased from ATCC (Manassas, VA, USA). Then, cells were carefully preserved using RPMI 1640 medium (Servicebio, China) replenished with 10% fetal bovine serum (FBS, Biological Industries, Israel) and 1% penicillin/streptomycin (New Cell & Molecular Biotech Co, China) in a humid incubator containing 5% CO₂ at 37°C. The THP-1 cells (5×10⁴ cells/100 μl) were seeded in six-well plates (Corning, USA) and differentiated into macrophages by adding 100 ng/ml of phorbol-12-myristate-13-acetate (PMA, MCE, China) for 48 h. After THP-1 cells were differentiated into M0 macrophages, macrophages were treated with 100 ng/ml lipopolysaccharide (LPS) (Peprotech, USA) for 48 h to differentiate into M1 macrophages. A transwell device (Corning, USA) with a 0.4-μm porous membrane was used for coculture treatments. HEY cells were seeded onto the upper chamber of the Transwell apparatus while M0 and M1 macrophages were seeded at a density of 2 × 10⁵ per well of the six-well plate. After 24 h of co-culture, the macrophages were collected for the succeeding analysis.

Additionally, total RNA from clinical tissues and cells was extracted by TRIzol (Invitrogen, USA), which was further reversely transcribed into cDNA using 5X ALL-IN-One RT Master Mix kit (Applied Biological Materials Inc, Canada) after assessing the purity and concentration. Moreover, real-time PCR was conducted by TB Green Premix Ex Taq kit (Takara, Japan) and GAPDH served as the internal control for all PCR reactions. As shown in Table 1 , the primers used in this study were exhibited.

APOBEC3A overexpression plasmids were ordered from Public Protein/Plasmid Library. HEY cells were transfected for 48 h with APOBEC3A overexpression plasmids (APOBEC3A-pcDNA3.1, 700 ng/μl) and control plasmids (pcDNA3.1, 500 ng/μl). Then, we followed the instructions of manufacturer and transfected both plasmids into HEY cells using Lipofectamine 2000 reagent (Invitrogen, USA). The transfection efficiency was verified by RT-qPCR after 48 h of transfection. Besides, the cell proliferation reagent CCK-8 (GeneView, America) was applied to approximately measure the cell viability. After plating the cells in the 96-well microtiter plates (Coring, NY, USA), 10 µl of CCK-8 reagent was added to each well and the cells were incubated for 2 h. Then, the value of optical density (OD) was acquired at 450 nm to determine the cell viability. Following the protocol, we further conducted the EdU assay through a Cell-Light EdU Cell Proliferation Kit (RiboBio, China). After transfection for 48 h, the transfected cells were then incubated with 50 mM EdU for another 3 h. Subsequently, 4% paraformaldehyde and Apollo Dye Solution were respectively utilized to fix the cells and stain the proliferating cells, along with Hoechst 33342 to mount the cells. After labeling, we photographed and enumerated the EdU-positive cells via the microscope under five indiscriminately selected visions. As for cell cycle detection, the transfected cells were fixed with 70% ethanol at 4°C for 24 h. Then, the cells were washed with PBS and stained with propidium oxide (PI) for 20 min in a dark room. Finally, the cell cycle was detected by flow cytometry, and the proportion of cells in the phase of G0/G1, S, or G2/M phases was compared. Transwell assay was mainly performed to measure the migration capacity of cells. Firstly, the concentration of transfected cells was adjusted to 2 × 10⁵ cells/ml. Secondly, 150 μl of cell suspension was added in the upper chamber, and 800 μl of RPMI-1640 medium involving 20% FBS in the lower chamber (the bottom of the 24-well plate). After incubating for 24 h under conditions of 37°C with 5% CO₂, cotton swabs were then used to remove the upper layer cells of the membrane. The upper chamber was further fixed and stained respectively by methanol and crystal violet for 15 and 30 min, and then washed with PBS. Five visual fields were randomly chosen to photograph and enumerate the number of migrated cells under a microscope.

IHC staining was employed to investigate the expression of APOBEC3A protein in formalin-fixed, paraffin-embedded tissue sections as described previously. Primary antibody against APOBEC3A (#AP20219a, abcepta) was used overnight at 4°C. Slides were then incubated for another 1 h using secondary antibody (#PK-8501, Vector Lab, USA). The complex was detected using Rabbit IgG mini-PLUS Kit visualized with DAB complex (#PK-8501, Vector Lab, USA). Besides, hematoxylin was applied to counterstain the nuclei. Sections were visualized under a microscope (10× or 40×). For immunofluorescence, the transfected cells were plated onto coverslips for 24 h and fixed for 20 min with 4% paraformaldehyde. After permeabilization for 30 min with 0.5% Triton X-100, cells were incubated in blocking buffer. Primary antibody against H2Ax (1:500, #ab229914, abcam, US) was utilized overnight at 4°C, then washed three times with PBS for 3 min. Alexa-488 goat anti-rabbit (1:500, #GB25303, Servicebio, China) was used as secondary antibody for 60 min and stained with 4’,6-diamidino-2-phenylindole (DAPI) for 15 min at room temperature. Finally, cells were regarded as positive if at least one focus was visible using a 40× objective in a confocal microscope.

This study was systematically conducted through the following analytical processes ( Figure 1 ). Firstly, we acquired the transcriptome information, somatic mutation profiles, and clinical data of 380 OC patients from the TCGA database, through which data immune-related DEGs and genomic-instability DEGs were recognized using the difference analysis. Secondly, APOBEC3A was further determined via intersection analysis of those above DEGs. Thirdly, three aspects of APOBEC3A were comprehensively analyzed, including immunity analysis, mutation analysis, and experimental verification. Among them, immunity analysis contained the correlation of APOBEC3A expression with TICs, ICPs, and IPS. Somatic analysis was about the correlation of APOBEC3A with TMB, somatic mutation, DDR genes, and m6A regulators. As for experimental verification, we have conducted the IHC, qPCR, FCM, Edu staining, and immunofluorescence to confirm the bioinformatical results of APOBEC3A.

In order to acquire immune-related genes in TME of OC, gene expression profiles were analyzed by means of comparing high- with low-ImmuneScore cohorts. Then, a total of 2,281 immune-related DEGs were identified from the median value of ImmuneScore ( Figure 2A and Table S2 ). Similarly, we calculated the accumulated number of each OC sample’s somatic mutations and sorted the number according to the descending order. Then, OC samples whose cumulative number ranked the top 25% (n = 65) were divided into the GU cohort, and the last 25% (n = 70) into the GS cohort. Additionally, 147 genomic instability-related DEGs were extracted from comparing gene expression profiles of GU patients with GS patients ( Figure 2B and Table S3 ). The above DEGs were all identified by the significance criteria with absolute value of log2 FC > 0.58 and FDR < 0.05. Since cumulative evidence has demonstrated that genomic instability could activate immunological recognition, launch immune responses, and facilitate the dynamic change of TIME, a gene intersection analysis of immune-related DEGs and genomic instability-related DEGs was further performed to see whether there existed any common genes. Interestingly, a total of 69 mutual DEGs were acquired through the intersection analysis ( Figure 2C ). That is, genome instability and immune microenvironment of OC are indeed closely linked at the level of gene expression. Moreover, functional enrichment analysis revealed that the immune-related DEGs were definitely associated with the immunological processes, like T-cell activation and lymphocyte activation regulation ( Figure 2D ). Simultaneously, a similar analytical method was utilized to assess the functional enrichment of genomic instability-related DEGs, indicating that these genes were basically enriched in the following biological processes such as calcium ion homeostasis and interleukin-12 (IL-12) production ( Figure 2E ). These results suggested that the long-lasting genomic instability of patients with OC may cause changes in tumor cell phenotypes, activate immune surveillance, and dynamically change the TME of OC.

After identifying the above 69 genes that were both linked to immune microenvironment and genomic instability of OC patients, we then equally separated tumor cases into different expression cohorts depending on the median level of gene expression, respectively. In addition, Kaplan–Meier analysis was further conducted by comparing high- and low-expression cohorts, demonstrating that there were 10 genes correlated with OS (APOBEC3A, C16orf54, CCL19, HTRA4, IRF4, LILRA5, MS4A1, NUGGC, PDCD1LG2, and SLAMF7). After that, the correlation analysis of these 10 genes’ expression with clinicopathological characteristics was performed, revealing that only APOBEC3A expression was significantly associated with stage classification ( Supplementary Figure 1 ). As for APOBEC3A, OC patients with high expression were found to possess a longer survival time than low expression ( Figure 3A , p = 0.033, by log-rank test). Moreover, APOBEC3A expression was descending along with the advanced stage classification of OC progression ( Figure 3B ). In order to verify APOBEC3A as a protective signature, we then used the OC samples from ICGC database and Kaplan–Meier plotter online database to further assess its prognostic value. The results from the ICGC database illustrated that APOBEC3A upregulation would have an improved OS ( Figure 3C ). In addition, online database analysis also showed that OC patients in high APOBEC3A cohort had a better OS than the low one, which was in accordance with the above results of this study ( Figures 3D, E ). These results exhibited that APOBEC3A could serve as a biomarker to forecast the survival and stage classification of OC patients. Moreover, GSEA focusing on HALLMARK gene sets and KEGG gene sets was employed for different APOBEC3A expression groups ( Table S4 ). On the one hand, for HALLMARK gene sets, the APOBEC3A high-expression cohort genes were basically enriched not only in immune biological processes but also in genomic instability-related pathways, taking the IL2–STAT5 signaling pathway, inflammatory response, DNA repair, and G2M checkpoint for example ( Figure 3F ). On the other hand, similarly, APOBEC3A high-expression cohort genes were also intimately related to immunological functions and genomic instability-associated processes, such as chemokine signaling pathway, JAK-STAT signaling pathway, cell cycle, and cytosolic DNA sensing pathway ( Figure 3G ). Synchronously, there exhibited no enriched pathways for the genes in the APOBEC3A low-expression cohort. The above results illustrated that APOBEC3A expression was indeed strongly correlated with the immune microenvironment and genome instability of OC and could act as a protective biomarker to predict the prognosis of OC patients.

Since APOBEC3A was found to be intimately associated with immune-related biological functions in this study, we conducted a series of comprehensive bioinformatical analyses to further acknowledge how APOBEC3A expression influences immune microenvironment of OC. Firstly, we analyzed the relationship between APOBEC3A expression with ImmuneScore, and the results demonstrated that OC patients in a higher level of the APOBEC3A group exhibited a higher ImmuneScore ( Figure 4A , p = 3.4e-14). Secondly, the CIBERSORT method was then performed to investigate the relationship between APOBEC3A and specific immune cells, which calculated the proportion of 22 different tumor-infiltrating immune subtypes. A total of five types of TICs were considered as significantly related to APOBEC3A expression, including Macrophage M0, Macrophage M1, Monocytes, T-cell follicular helper, and Neutrophils ( Figure 4B ). Among them, it is worth noting that APOBEC3A expression was positively correlated with the proportion of Macrophage M1 (p < 0.001) and negatively correlated with M0 (p < 0.001). We further analyzed the correlation of immune cells proportion with survival time, illustrating that OC patients with a lower proportion of Macrophage M0 or a higher proportion of Macrophage M1 would have a longer OS ( Figures 4C, D , p = 0.028, p < 0.001, respectively). These results may explain the better survival of the high APOBEC3A expression group. Thirdly, in order to assess the prognostic value of APOBEC3A expression to predict the effect of immunotherapy response, the association between gene expression and ICPs was conducted. The analytical results illustrated that APOBEC3A expression exhibited a positive correlation with ICPs such as PDL1, CTLA4, LAG3, and TIGIT ( Figure 4E ). Besides, the analytical results of IPS with APOBEC3A expression have demonstrated that patients in the high APOBEC3A expression group could have a higher IPS of anti-PD1 and anti-CTLA4 therapy ( Figures 4F–I ), which means a better immunotherapy response. These results have fully illustrated that APOBEC3A could be a favorable signature for effectively forecasting the effect of immunotherapy in OC patients. In addition, we applied ssGSEA to further conduct the correlation analysis on APOBEC3A with immune-related functions and immune cell types, showing that many immunological processes were significantly different in different APOBEC3A expression cohorts ( Figures 4J, K ). In order to further demonstrate the role of APOBEC3A in influencing immune microenvironment of OC, we separated cancer patients into two different APOBEC3A expression cohorts according to the median level and finally identified 187 DEGs by comparing these two groups ( Figure 5A ). In these DEGs, C-X-C motif chemokine 11 (CXCL11) was then selected for the further analysis since its expression changed in a remarkably significant level and its correlation with APOBEC3A was strong ( Figure 5B , R = 0.66, p < 2.2e-16). In addition, the survival analysis showed that high CXCL11 expression had a longer OS ( Figure 5C , p < 0.001, by log-rank test). Similar to the results of APOBEC3A, CXCL11 expression was also found to be positively related to the proportion of Macrophage M1, and M0, the opposite ( Figure 5D , p < 0.001). Moreover, we performed RT-qPCR to further confirm the correlation of APOBEC3A with CXCL11, and the results indicated that upregulation of APOBEC3A expression in HEY cells increased the CXCL11 expression in M0 and M1 macrophages ( Figures 5E, F , p < 0.0001 and p < 0.01, respectively). The above results illustrated that APOBEC3A might increase the secretion level of CXCL11 in M0 and M1 macrophages, and thus a better survival in OC patients.

More critically, we ultimately selected an immunotherapeutic dataset of urothelial cancer called IMvigor210 cohort from Mariathasan et al. (32) to further demonstrate the prognostic value of using APOBEC3A as a biomarker for ICB therapy such as anti-PD-L1 immunotherapy. From the results, we could clearly see that APOBEC3A upregulation exhibited a significantly improved survival that also provided good proofs for our previous results ( Figure 6A , p = 0.044, by log-rank test). Besides, the correlation of different APOBEC3A expression with different immune phenotypes (desert, excluded, inflamed) was also explored since patients with inflamed immune phenotype were more likely to benefit from immunotherapy, while the other two types exhibited the opposite. The results demonstrated that higher APOBEC3A expression was found to be strongly associated with inflamed immune phenotype while the lower one was found to be associated with the excluded and desert types, which meant that it is more difficult for patients with APOBEC3A downregulation to obtain benefits from immunotherapy ( Figure 6B ). In addition, we further investigated the therapeutic responsiveness and clinical benefits to anti-PD-L1 immunotherapy by comparing different APOBEC3A expression cohorts, and cancer patients with high expression of APOBEC3A were found to have a more positive response to immunotherapy, thus a better clinical outcome ( Figures 6C–E ). The above results strongly indicated that APOBEC3A could represent the immune status of TME, which might play an essential part in forecasting the response to immunotherapy as well.

Apart from the immune-related functions, APOBEC3A was also identified as a genomic instability-related gene. Therefore, a variety of integrative analyses on genomic instability-related processes were performed further. Firstly, we analyzed the association between APOBEC3A expression and somatic mutation count, showing that patients with high APOBEC3A expression had more somatic mutation ( Figure 7A , p = 0.00019). Secondly, correlation analysis of APOBEC3A expression with TMB was also conducted, whose results have demonstrated that the expression of APOBEC3A correlated positively with TMB in OC patients ( Figure 7B , p = 0.00038). As many published literatures reported, TMB could serve as a biomarker to predict immunotherapy response (45). The above results further supported the idea that APOBEC3A could be a possible signature to acknowledge the effect of immunotherapy. Besides, survival analysis has indicated that patients with high TMB would possess a longer OS ( Figure 7C , p = 0.002). When TMB combined with APOBEC3A expression, we could clearly see that high TMB with high APOBEC3A expression had the best prognosis, and contrarily, low TMB with low APOBEC3A expression had the worst ( Figure 7D , p = 0.008). Thirdly, on account that APOBEC3A seemed to play an essential role in genomic instability of OC, we further wondered whether there would exist some differentially mutated genes by comparing APOBEC3A high- and low-expression cohorts. Naturally, such difference analysis focusing on somatic mutation data was applied then and there exactly existed some difference in these two cohorts. Interestingly, the mutation of TP53 and TTN ranked the first and the second in both cohorts ( Figures 7E, F ). Thirty-seven differentially mutated genes were acquired ( Table S5 ), among which Phospholipase C Eta 1 (PLCH1) with high APOBEC3A expression mutated significantly more than the low one ( Figure 7G ). Besides, we analyzed the co-occurring and exclusive mutations of these two expression cohorts, indicating that PLCH1 and TP53 exhibited an exclusive mutation in the high APOBEC3A expression group ( Figures 7H, I ). The results meant that PLCH1 and TP53 might play a similar part in the same pathway with high APOBEC3A expression. TP53 mutation was reported to be connected with poor prognosis and metastasis of OC patients (46, 47). That is, patients with high APOBEC3A expression seemed to have more PLCH1 mutation and relatively less TP53 mutation, which was consistent with previous results that APOBEC3A was a protective factor in OC. Fourthly, correlation analysis of DDR genes and APOBEC3A was then performed since DDR was reported to be closely related to genomic instability. As expected, DDR genes like RAD51, FEN1, and BRCA2 were significantly positive-associated with APOBEC3A ( Figure 7J ). Last but not the least, APOBEC3A was also found to be significantly correlated with m6A regulators such as WTAP, METTL14, ZC3H13, RBM15B, and FMR1 ( Figure 7K ). Emerging evidence has suggested that the expression of m6A regulators was critical for carcinogenesis and cancer development (48), which strengthened the important position of APOBEC3A in OC. The comprehensive analyses confirmed the intimate relationship of APOBEC3A with genomic instability, demonstrating that APOBEC3A had the potential to be a helpful biomarker in OC.