Interferon-stimulated gene subtypes as key indicators of immune landscape and survival outcomes in ovarian cancer Authors: Wanjun Luo, Dan Zhang, Zidan Lin, Junran Zhuang, Suiying Liang, Zhihong Huang, Chenfei Zhou Published: 2024 ABSTRACT AbstractPurposeOvarian cancer (OV) remains the most lethal gynecological malignancy, underscoring the critical need for robust prognostic biomarkers to enhance patient outcomes. In this study, we classified OV patients by their interferon-stimulated gene (ISG) expression profiles and investigated the associations between these subtypes, the immune microenvironment, and survival outcomes.MethodsWe employed consensus clustering in the TCGA-OV cohort (n = 376) to classify patients into ISG-related subgroups. Survival analysis, differential gene expression (DESeq), KEGG and GSEA pathway enrichment analyses, genomic variation assessments, immune cell profiling using the CIBERSORT algorithm, and TIDE analysis were conducted in the TCGA-OV cohort. In addition, immune checkpoint marker expressions were assessed using data from the TCGA-OV cohort and multiplex immunofluorescence (mIF) staining on an independent cohort (n = 80).ResultsTwo distinct ISG subtypes were identified: ISG Cluster A and ISG Cluster B. Patients in ISG Cluster B exhibited significantly improved overall survival (OS) (p = 0.0442). A total of 328 dysregulated genes were identified, with Cluster B showing overexpression of immune-related genes and enhanced involvement in immune signaling pathways. ISG Cluster B also presented higher tumor mutation burden (TMB) and an enriched immune profile, including M1 macrophages and CD8 + T cells. TIDE analysis indicated a more favorable response to immune checkpoint inhibitors in this cluster, corroborated by high expressions of PD-L1 and ISG15, which were associated with prolonged OS.ConclusionsOur findings demonstrate that ISG-related subtypes are significantly associated with the immune microenvironment and survival outcomes in OV. The biomarkers identified in this study have the potential to inform precision therapy development, thereby enhancing treatment efficacy and personalized care for OV patients.Supplementary InformationThe online version contains supplementary material available at 10.1007/s12672-024-01617-6.Keywords: Ovarian cancer, Interferon stimulated genes, Immune microenvironment Introduction Ovarian cancer (OV) represents the deadliest form of gynecological cancers [1, 2], with approximately 324,398 women diagnosed and 206,839 fatalities annually worldwide [3]. It ranks as the eighth most common cancer among women globally, exhibiting a 5-year survival rate of 47% post-diagnosis [4]. Despite the presence of innate immunity at both the molecular and T-cell levels in OV [5, 6], immunotherapy for OV has fallen short of expectations. Consequently, the identification of suitable biomarkers is paramount for immune landscape and predicting overall survival (OS) in OV. Type I interferons (IFNs) include IFN-α, IFN-β, IFN-δ, IFN-ε, IFN-κ, IFN-τ, and IFN-ω [7, 8], along with Type II interferon (IFN-γ) [9], which are naturally produced and elevated in response to stress, infections, and cancer. IFNs activate the Janus kinase-signal transducer and activator of transcription (Jak-STAT) signaling pathway to induce the expression of interferon-stimulated genes (ISGs) [10–12], which are essential for antigen presentation, antiviral responses, inflammation, and cancer progression. Additionally, ISGs have shown a strong correlation with the immunoreactive tumor microenvironment, including tumor-infiltrating lymphocytes [13, 14], tumor-associated macrophages [15], and neutrophils [16]. These findings showed that ISGs involved in modulating immune functions [17–19], suggesting that ISGs may act as prognostic markers for OV. Identifying OV subtypes could facilitate the selection of patients who are most likely to benefit from therapy, thereby enhancing patient outcomes. Although numerous predictive models have been developed to characterize OV subtypes [20, 21], ISGs have seldom been considered as potential targets for therapy in OV. Our study identified two distinct ISG subtypes within OV and explored the differences between these subcategories. Moreover, multiplex immunofluorescence (mIF) staining revealed a consistent pattern of ISG15 and PD-L1 expression in OV tissues, with high expression levels of PD-L1 and ISG15 associated with a favorable prognosis. Our findings may provide new insights into ISG biomarkers, thereby advancing the development of precision therapy for OV. Materials and methods This study was approved by Institutional Review Board of Guangdong Provincial People’s Hospital, Guangdong Academy of Medical Sciences, Guangzhou (KY-Q-2021-097-04). All methods were carried out in accordance with relevant guidelines and regulations. Patients Data for OV patients from the TCGA-OV cohort were included in our analysis, encompassing 376 patients with RNA-seq data and 462 patients with Copy Number Variation (CNV) data. Expression and clinical data were retrieved using the TCGAbiolinks package [22], while CNV data were sourced from the Genomic Data Commons (GDC). Additionally, the study incorporated 80 patients diagnosed with OV at Guangdong Provincial People’s Hospital, for mIF validation purposes. All patients in the study had provided written informed consent. Identification of ISG related Molecular subtypes using K-Means analysis A total of 83 interferon-stimulated genes, obtained from the Molecular Signatures Database (MSigDB) [23], were analyzed (Supplementary Table 1 details this gene list). K-means clustering, based on the ISG expression matrix, facilitated the identification of distinct OV subgroups. Functional analyses Differentially expressed genes between the subtypes were identified using a threshold of |log2 Fold Change| ≥1 and a p-value < 0.05 using DESeq R package. The Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis and Gene Set Enrichment Analysis (GSEA) were employed to elucidate differences in OV subtypes [24, 25]. Genomic analysis CNV data from the TCGA-OV cohort were analyzed using maftools [26], and a waterfall plot illustrated these data via the ComplexHeatmap package. Tumor Mutation Burden (TMB) was calculated based on mutations per patient. Immune analysis The abundance of 22 immune cell types was estimated using the CIBERSORT algorithm [27], and the ESTIMATE algorithm was applied to derive stromal and immune scores [28]. TIDE analysis The Tumor Immune Dysfunction and Exclusion (TIDE) algorithm predicted potential responses to immune checkpoint blockers (ICBs) by evaluating mechanisms of tumor immune evasion and the likelihood of immune system avoidance [29]. mIF staining All experiments using human tissue samples were performed in accordance with relevant guidelines and regulations. Formalin-fixed paraffin-embedded (FFPE) tissues from 80 ovarian cancer patients were subjected to mIF staining according to the Opal 7-Color IHC kit instructions (Akoya Biosciences, Cat. No. NEL811001KT). Following deparaffinization and antigen retrieval in sodium citrate, tissues were stained with antibodies against PD-L1 (Cat: 13684S) and ISG15 (Cat: ab285367), plus DAPI. Vectra Polaris Quantitative Pathology Imaging Systems captured the stained slides, analyzed using Phenochart Image Analysis Software (Akoya Biosciences) version 1.0.12. Statistical analyses Statistical analyses were performed using R (version 3.5.2) and GraphPad Prism (version 9.0). Survival analysis utilized the Kaplan-Meier method. Gene expression and TMB comparisons between groups were conducted via two-tailed Student’s t-test, while the Kruskal-Wallis test assessed CIBERSORT immune infiltration data. Correlation analysis was conducted using the Pearson correlation coefficient to evaluate the relationships between two genes. Differences between the subgroups were analyzed using the HOME for Researchers platform. A p-value < 0.05 was considered statistically significant. Results Determination of two ISG-Related molecular subtypes Utilizing consensus clustering based on ISG expressions, we classified the TCGA-OV cohort into two ISG-related subgroups, finding K = 2 to offer optimal clustering stability (Fig. 1A and B). Among the 376 OV patients analyzed, 246 were categorized into ISG Cluster A, and 130 into ISG Cluster B. Kaplan-Meier survival analysis revealed that patients in ISG Cluster B exhibited significantly better OS (p = 0.0442, Fig. 1C). Table 1 showed that neoplasm histologic grade 3 in ISG Cluster B showed an increasing trend, and there was no difference between the two groups in age, FIGO staging, and residual illness rates. These results suggest that OV patients in the TCGA-OV cohort can be stratified into two distinct subgroups based on ISG expressions, each with markedly different survival rates. Identification of Two ISG-Related Molecular Subtypes. A Consensus clustering indicates K = 2 as the optimal subgroup distinction. B Optimal clustering stability is achieved at K = 2, as demonstrated in consensus clustering. C Kaplan-Meier survival analysis comparing OS between ISG Cluster A and Cluster B Table 1Correlation between clinicopathologic variables and ISG-related subtypes in TCGA-OVVariablesISG-related subtypesp valueCluster ACluster BTotal245130Age (mean ± SD)59.69 ± 11.6659.53 ± 11.100.860Racelist0.155White208118American indian or Alaska native11Asian74Black or African american226FIGO Stage0.740Stage I20Stage II148Stage III189104Stage IV4018Neoplasm Histologic Grade0.074G110G23111G3205118Venous Invasion0.872Negative2813Positive4321Lymphatic Invasion0.378Negative3414Positive7031Tumor Residual Disease0.494No Macroscopic disease4324Macroscopic disease18088 Correlation between clinicopathologic variables and ISG-related subtypes in TCGA-OV Differentially expressed genes and functional analyses The DESeq analysis differentiated ISG Cluster A from Cluster B by identifying 328 dysregulated genes (Fig. 2A, Supplementary Table 2). FGF17 and CDH16 were predominantly expressed in ISG Cluster A, whereas ISG15 and IFIT1 showed marked overexpression in ISG Cluster B (Fig. 2A). Previous studies have reported associations between FGF17 and CDH16 and epithelial-mesenchymal transition, while ISG15 and IFIT1 are linked to interferon-mediated signaling pathways. Analysis of Differentially Expressed Genes and Functional Pathways in ISG Molecular Subtypes. A Volcano plot highlights DEGs between ISG Cluster A and Cluster B, applying a threshold of |log2 Fold Change| > 1 and P < 0.05 in the TCGA-OV cohort. B Bar plot illustrates KEGG pathway enrichment between ISG subtypes, with bar length representing -log10(p-adjusted value). C, D GSEA delineates distinct signaling pathways active in ISG Cluster A versus Cluster B Subsequent KEGG and GSEA analyses explored potential signaling pathways. KEGG enrichment analysis showed DEGs in ISG Cluster A were mainly involved in the calcium signaling pathway and actin cytoskeleton regulation (Fig. 2B). Conversely, DEGs in ISG Cluster B were significantly enriched in immune-related signaling pathways, including the JAK-STAT signaling pathway and RIG-I-like receptor signaling pathway (Fig. 2B). GSEA highlighted enrichment in the Wnt signaling pathway and drug metabolism pathways for ISG Cluster A, whereas ISG Cluster B demonstrated enrichment in pathways critical for immune response, including the Toll-like receptor (TLR) signaling pathway (Fig. 2C and D). Genomic variations of ISG-Related molecular subtypes Oncoplot analysis revealed significant genomic variations between the two ISG-related subtypes (Fig. 3A), with mutations in genes such as COL6A3 and BCLAF1 more prevalent in ISG Cluster B (Fig. 3B). TMB was significantly higher in ISG Cluster B than in ISG Cluster A (p = 0.0405, Fig. 3C). Genomic Variations within ISG-Related Molecular Subtypes. A Analysis of somatic mutations across ISG-related subgroups in the TCGA-OV dataset. B Comparison of differential somatic mutations between ISG-related subgroups. C TMB comparison across ISG-related subgroups, illustrating genomic instability differences. The p values in (C) were calculated by unpaired two-tailed Student’s t test. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, no significance Distinct Immune Cell statuses in the ISG Molecular subtypes Immune analysis, using the CIBERSORT algorithm, revealed a significant enrichment of M1 macrophages, CD8+ T cells, and other immune cells in ISG Cluster B, while M0 macrophages and resting dendritic cells were more common in ISG Cluster A (Fig. 4A). Moreover, ISG Cluster B patients had significantly higher immune scores as determined by the ESTIMATE algorithm (Fig. 4B), and marked differences in the expression of ITGAX, CD8A, and CD80 between the subtypes (Fig. 4C). Immune Landscape of ISG Molecular Subtypes. A CIBERSORT-derived estimates of 22 immune cell types within ISG-related subtypes, showing median values and interquartile ranges. B Comparison of stromal and immune scores between ISG-related subtypes, indicating variations in tumor microenvironment composition. C Expression differences in ITGAX, CD8A, and CD80 between ISG-related subtypes, highlighting immune marker disparities. The p values in A were calculated by Kruskal-Wallis test. The p values in B, C were calculated by unpaired two-tailed Student’s t test. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, no significance Distinct immune functions in the ISG molecular subtypes TIDE analysis suggested a more favorable response to immune checkpoint inhibitor therapy in ISG Cluster B, as indicated by lower TIDE and Exclusion scores (Fig. 5A). This cluster also showed significantly higher expression of CD274, CTLA4, PDCD1, and LAG3 (Fig. 5B), underscoring the crucial role of the expression of ISGs in modulating immune functions. High expression level of these markers in the ISG15high subgroup further corroborate their consistency (Fig. 5C). Predictive Value of ISG-Related Molecular Subtypes for Immunotherapy Efficacy. A TIDE analysis comparing ISG-related subtypes, suggesting differential immunotherapy outcomes. B Expression levels of CD274, CTLA4, PDCD1, and LAG3 differentiate ISG-related subtypes, implicating immune checkpoint regulation. C Comparison of CD274, CTLA4, PDCD1, and LAG3 expression between ISG15low and ISG15high subgroups within the TCGA-OV cohort. D Immunohistochemistry validates ISG15 and PD-L1 expression patterns in ovarian cancer samples. Scale bars, 50 μm. E Multiplex immunofluorescence staining shows PD-L1 expression differences between ISG15low and ISG15high subtypes in OV tissues. F Multiplex immunofluorescence staining shows the correlation between PD-L1 and ISG15 expression in OV tissues. G Kaplan-Meier curves illustrate overall survival differences between ISG15highPD-L1high and ISG15lowPD-L1low subtypes in OV, assessed by multiplex immunofluorescence staining. The p values in A, B were calculated by unpaired two-tailed Student’s t test. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, no significance The high expression of PD-L1 in ISG15high subgroup was verified by mIF staining, which indicated consistency with the TCGA database (Fig. 5D and E). We also observed a positive correlation between ISG15 and PD-L1 through mIF staining (R = 0.605, p < 0.001) (Fig. 5F). Moreover, patients with high PD-L1/ISG15 expression had significantly prolonged overall survival compared to those with low expression (p = 0.0109, Fig. 5G), but two groups showed no difference in age, FIGO staging, and histology (Table 2). Table 2Correlation between clinicopathologic variables and ISG15highPDL1high and ISG15lowPDL1low subtypes in validation cohortVariablesValidation cohortp valueISG15highPDL1highISG15lowPDL1lowTotal2625Age (mean ± SD)59.69 ± 11.6659.53 ± 11.100.828Histology>0.999Serous carcinoma2221Endometrioid carcinoma34Clear cell carcinoma10FIGO Stage0.555Stage I20Stage II12Stage III1618Stage IV75Neoplasm Histologic Grade>0.999G110G232G32123Tumor Residual Disease>0.999No Macroscopic disease66Macroscopic disease2019 Correlation between clinicopathologic variables and ISG15highPDL1high and ISG15lowPDL1low subtypes in validation cohort In summary, these findings suggest that the identified ISG-related subtypes are significantly associated with the immune landscape and survival outcomes in OV. Discussion The tumor immune microenvironment plays a critical role in clinical outcomes for cancer patients [30, 31]. However, there is a notable lack of robust molecular markers that assess the relationship between immune cell presence and cancer outcomes [5]. Previous studies have highlighted that ISGs are instrumental in remodeling the tumor immune microenvironment, which can significantly influence the efficacy of therapeutic interventions and patient survival outcomes [32, 33]. Our study provides compelling evidence that ISG molecular subtypes may modulate clinical prognosis in OV through their effects on the tumor immune microenvironment. Specifically, we observed that distinct ISG expression profiles correlate with varying levels of immune cell infiltration, including CD8+ T cells and M1 macrophages, which are associated with enhanced anti-tumor responses. These findings suggest that high expression of ISGs can create a more immunogenic tumor microenvironment, potentially leading to improved patient outcomes. Moreover, by demonstrating the relationship between ISG subtypes and immune cell dynamics, our research underscores the necessity of integrating immune profiling into prognostic assessments. Understanding how ISGs contribute to the immune landscape of OV may pave the way for the development of targeted therapies that enhance immune responses and improve clinical outcomes. Utilizing consensus clustering on TCGA-OV data, we identified two distinct ISG subgroups based on ISG expressions. Our analyses reveal that ISG Cluster B, characterized by the upregulation of ISG15 and IFIT1, is associated with a favorable prognosis and enriched in immune-related signaling pathways. This subgroup exhibited significant immune cell infiltration and a heightened response to immunotherapy. Furthermore, our study demonstrates consistent expression trends for PD-L1 and ISG15 through mIF staining, linking high expression of these markers to an improved prognosis in OV. These insights are pivotal for the design of precision therapy strategies in OV. ISG15 and IFIT1 have been reported to be secreted by tumor cells [34], fibroblasts [35], myeloid cells [36], and even T cells [37]. ISG15, a ubiquitin-like protein, is implicated in the innate immune response to viral infections and plays a vital role in cancer, influencing energy metabolism [38], immune cell infiltration [39–43], and chemoresistance [44]. Fan et al. demonstrated that ISG15, produced by tumor cells, could enhance cytotoxic T cell-mediated tumor killing, which is consistent with the CIBERSORT analysis showing increased CD8+ T cell infiltration in the ISG Cluster B group [39]. Moreover, free ISG15 has been reported to bolster the tumor immunity of CD8+ T cells [40, 41]. However, some studies have reported that tumor cells overexpressing ISG15 are associated with M2 macrophage recruitment, which is contrary to the results obtained from CIBERSORT [42, 43]. IFIT1, induced by interferon therapy, exhibits dual roles in tumors, enhancing NK cell function via the IFIT1-STING-STAT4 pathway [45] while also promoting metastasis in certain cancers [46]. PD-L1 immunotherapy represents a cornerstone of tumor immunotherapy. Mechanistically, ISGs are instrumental in reshaping the tumor immune microenvironment, regulating MHC-I and PD-L1 to facilitate immune cell infiltrations [47, 48]. Our work confirms the relationship between ISG15 and PD-L1 expression, associating high expression levels with favorable OV prognosis. This study underscores the potential of ISG-related biomarkers in combination therapies with ICIs. Moreover, ISG Cluster B exhibited increased TMB, suggesting potential markers of genomic instability and predictors of immunotherapy success [49]. Our findings elucidate the ISG landscape’ s link to the tumor immune microenvironment, essential for strategizing therapy in OV. Nonetheless, there are several limitations in our study. This investigation primarily relies on bioinformatics analyses to establish correlations between gene expression and clinical outcomes, which are supported by our mIF data. However, more functional validation experiments are needed to confirm the biological significance of the identified ISG subtypes. Additionally, while our study suggests that ISG expressions can inform patient stratification, it lacks robust evidence on how these findings could be translated into clinical practice. Outlining potential therapeutic strategies or clinical trial designs based on the identified subtypes would enhance the impact of our study. Conclusions In conclusion, our study employed k-means clustering to delineate two molecular subtypes of ovarian cancer, underscoring the pivotal role of ISG expressions in modulating the immune-related signaling pathways and the tumor microenvironment, thereby contributing to improved prognoses. Furthermore, our research validated a consistent correlation between ISG15 and PD-L1 expression, establishing that high levels of PD-L1 and ISG15 expression are associated with favorable outcomes in OV tissues, as evidenced by mIF staining. This work offers novel insights into the utility of ISGs as biomarkers and underscores their potential in enhancing the efficacy of precision therapeutic approaches for OV. Supplementary Information Supplementary Material 1. Supplementary Material 2. Author contributions All authors participated in this research, including conception and design (LWJ, ZCF), data acquisition (ZD, LZD, ZJR, LSY), data analysis and interpretation (LWJ, ZD, HZH), material support (ZJR, LSY), study supervision (HZH, ZCF), and drafting the article or critically revising (LWJ, ZD, HZH, ZCF). Funding This study was supported by a grant from the Science and Technology Foundation of Guangzhou City (202102021057). Data availability The TCGA-OV datasets analysed during the current study are available in the GDC Data Portal repository, https://portal.gdc.cancer.gov/. The datasets generated during the current study for mIF validation are not publicly available due to the privacy of patients but are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate This study was approved by Institutional Review Board of Guangdong Provincial People’s Hospital, Guangdong Academy of Medical Sciences, Guangzhou (KY-Q-2021-097-04). Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Wanjun Luo and Dan Zhang have contributed equally to this work. Contributor Information Zhihong Huang, Email: huangweijun2001@126.com. Chenfei Zhou, Email: zhouchenfei@gdph.org.cn. References Associated Data Supplementary Materials Supplementary Material 1. Supplementary Material 2. Data Availability Statement The TCGA-OV datasets analysed during the current study are available in the GDC Data Portal repository, https://portal.gdc.cancer.gov/. The datasets generated during the current study for mIF validation are not publicly available due to the privacy of patients but are available from the corresponding author on reasonable request.