Immunogenic profiling of metastatic uveal melanoma discerns a potential signature related to prognosis

As the most common primary intraocular malignancy among adults (Jager et al. 2020), cancer cells of uveal melanoma (UM) originate from melanocytes within the uveal tract of the eye. In the United States, the mean age-adjusted incidence attains 5.1 per million (Singh et al. 2011). About 85% of the tumor cases derive from the choroid whereas the remaining cases derive from the iris (3–5%) and ciliary body (5–8%) (Damato 2012). The early symptoms of UM are characterized by blurred vision, central dark spots, eye pain, and scleral congestion. Some certain advances have been made in the treatment of UM (Kashyap et al. 2016). As suggested by the National Comprehensive Cancer Network (NCCN) guideline, clinically, UM is usually treated by the combination of enucleation, local resection, stereotactic radiotherapy, brachytherapy, and phototherapy (Rao et al. 2020). However, over the past decades, the mortality rate is still rather high (Spagnolo et al. 2016). Metastasis approximately occurs in 50% of patients diagnosed with primary UM (Kujala et al. 2003), with the liver being the most common area involved. Once metastasis occurs, patients with metastatic UM manifest a median overall survival (OS) of less than 12 months (Patel 2013). Further, the response rates of monotherapy were consistently in the single-digit percentage range in a panel of previous studies while the combined immune checkpoint blockade (ICB) revealed higher response rates and better survival outcomes, albeit at the cost of high immune-related toxicity (Heppt et al. 2017a, 2017b, 2019; Kottschade et al. 2016; Algazi et al. 2016; Zimmer et al. 2015). Tebentafusp is a novel treatment modality and the only approved systemic therapy for uveal melanoma (Balushi et al. 2023). Notably, Tebentafusp achieved the best results compared to combined ICB and other systemic treatments; admittedly, the efficacy of this drug needs to be confirmed in more phase 3 clinical trials (Petzold et al. 2023).

To date, there are few studies on the co-gene expression patterns of UM (Kashyap et al. 2016; Zhang et al. 2014), which may limit the exploration of hub genes associated with the prognosis of the disease.

At present, immunotherapies including ICB, cancer vaccines, and chimeric antigen receptor T cell immunotherapy (CAR-T) have become a novel clinical strategy for treating various malignant tumors (Igarashi and Sasada 2020). Nevertheless, the overall therapeutic effect of immunotherapy for UM patients remains not satisfactory (Fu et al. 2022). The adverse response of UM to immunotherapy highlights the lack of knowledge about how metastatic UM develops immune resistance or evades immune surveillance. Interestingly, previous studies have identified that the special anatomic site of UM may contribute to its distinctive immune resistance (Bronkhorst and Jager 2012; Niederkorn 2012). As an immune-privileged organ, the eye serves to restrict the amount of inflammation caused by immune to occur in this region. To be specific, the eye aqueous humor contains a large number of anti-inflammatory and immunosuppressive cytokines, including transforming growth factor-β (TGF-β) and macrophage migration inhibitory factor (MIF) (Niederkorn 2012). Thus, it is urgent to better figure out the physiologic immunosuppressive microenvironment to improve individualized treatment for patients with UM.

In the present study, we sought to start with characterizing the immunophenotype of UM, and then identify the gene modules associated with metastasis using weighted gene correlation network analysis (WGCNA) analysis. Key genes were further identified by univariate Cox, least absolute shrinkage and selection operator (LASSO)-Cox analysis to construct a risk score. Moreover, CIBERSORT analysis was used to evaluate and quantify the degree of immune cell infiltration among low- and high-risk UM patients, and then a predictive model was constructed by combining clinical variables.

Despite its low incidence, UM is still the most common primary intraocular malignant tumor in adults, and is characterized by its high mortality (> 95%), high metastasis (> 50%), and poor prognosis, making the exploration for effective biomarkers for assessing prognosis crucial (Marseglia et al. 2021). In this study, WGCNA were used to screen out metastasis-related co-expression modules from the GEO database. The Black, Purple, and Blue modules were significantly associated with the progression and metastasis in UM. Then we used univariate Cox, LASSO-Cox regression to screen out four genes associated with prognosis. The GSEA analysis demonstrated that multiple immune-related pathways were upregulated in the high-risk group. Subsequently, we used these four prognosis-related genes to construct a risk score signature, which was found to be an independent factor for predicting OS. Moreover, combined with other age, gender, a clinical predictive model was established. Both the survival curves and ROC analysis results showed the robustness and reliability of the predictive model for prognosis prediction of UM patients. The AUCs of the ROC curve in the training cohort for 1-year, 3-year, and 5-year were 0.930, 0.947, and 0.951, respectively. This predictive model helps simplify clinical individualized monitoring and treatment strategies for UM patients.

Four genes included in the predictive model have not been adequately studied in UM. As indicated by previous study, downregulation of OTUD3 was associated with poor prognosis in patients with esophageal cancer. The specific mechanism may be that OTUD3 directly interacts with zinc finger protein ZFP36, and stabilizes ZFP36 by inhibiting FBXW7-mediated polyubiquitination of K48 junction (Ntunzwenimana et al. 2021). Xie et al. (2021b) found that OTUD3 was significantly overexpressed in hepatocellular carcinoma, and high expression of OTUD3 was correlated with tumor size, distant metastasis, and poor TNM stage. Wang et al. (2021) analyzed the relation between OTUD3 and lymphatic metastasis, and found that nicotine-mediated OTUD3 downregulation inhibited VEGF-C mRNA decay to promote lymphatic metastasis of human esophageal cancer. Qiu et al. (2023) found that acetylation degradation of KAT8 could affect colorectal cancer invasion and migration. Dong et al. (2019) found that MYST1/KAT8/MOF promoted the progression of glioblastoma by activating epidermal growth factor receptor (EGFR) signaling. Previous study also identified KAT8 as a key molecule important for cancer cell survival. To be specific, they found that KAT8 regulated G2/M cell cycle arrest through AKT/ERK-cyclin D1 signaling (Zhang et al. 2013). In addition, LAMTOR3 expression was significantly decreased in renal clear cell carcinoma compared with normal renal tissue. LAMTOR3 may be a potential marker for the diagnosis and treatment of renal clear cell carcinoma (Gong et al. 2022). Song et al. (2015) found that LAMTOR3 polymorphisms may be a potential biomarker of genetic susceptibility to gastric cancer. In a mouse model of pancreatic cancer, the siRNA-FA-PEG-COL nanoparticles against LAMTOR, which strongly inhibited retroperitoneal invasion and significantly inhibited peritoneal dissemination compared to the other nanoparticles, improved prognosis of the mice (Taniuchi et al. 2019).

Two clinical variables, age and gender, were also included. Kaliki et al. (2013) found that the older patients with UM had worse prognosis. In addition, they divided UM patients into young adults (< 20 years old), middle-aged (21–60 years old), and senior citizens (> 60), and found that younger people had a lower risk of metastasis compared to middle-aged and older adults. In a survival analysis of 119 patients, Rietschel et al. (2005) found that women had a significantly lower risk of disease-specific death than men. Similarly, a study of 723 patients with UM found that male patients had a worse prognosis compared to female patients (Zloto et al. 2013). Therefore, two variables including age and gender were included in the clinical predictive model for further analysis. Previous studies have tried to reveal the immune infiltration pattern of this special tumor (Luo and Ma 2020; Pan et al. 2020). In this study, we found that γδT infiltration levels were significantly higher in high-risk UM samples than in low-risk ones (P < 0.05). Consistent with previous reports (Wang et al. 2020; Lei and Zhang 2021), γδT were identified to be related to poor prognosis in UM patient. Previous studies also have focused on prognosis prediction based on the gene expression profiles of UM. For instance, Vaquero-Garcia et al. (2017) used variables including chromosome characteristics, age, sex, tumor site, and tumor size to predict patients’ prognosis, and the accuracy of model attained 85%. Using TCGA as the training cohort and GSE22138 as the validation cohort, (Luo and Ma 2020) identified 21 prognostic genes related to microenvironment. Xue et al. (2019) conducted univariate and multivariate Cox analysis on the TCGA cohort to establish an 18-gene prognostic model for patients with OS. NDUFB9, NDUFV2, Cyc1, and CTNNB1 were screened out by Choi et al. (2020) as prognostic predictors of UM patients. Luo et al. (2020) performed univariate Cox regression and LASSO-Cox to construct ten gene signatures, and used the GSE22138 cohort to verify the signatures. Compared to Luo et al., the model in our paper is more accurate in predicting 1-year and 3-year OS (AUCs of 0.892 and 0.737, respectively).

Our study had some strengths. First, we identified metastasis-related genes by WGCNA, and then constructed a gene signature based on these genes for better prognosis prediction. Notably, metastasis is significantly associated with the survival of patients. To the best of our knowledge, there is limited literature on the immune signature of metastatic UM tumors (Qin et al. 2017; Rothermel et al. 2016; Krishna et al. 2017; Javed et al. 2017). Second, the model was derived from the GSE22138 cohort and verified by the external data set TCGA cohort. Third, clinical variables including age and gender were added together with risk score to establish a model for predicting UM prognosis.

Admittedly, there are some limitations in our current study. First, these results were obtained from public databases, and therefore need to be validated by experiment. Second, the sample size is not large enough, and further data training of large multi-center samples is needed to improve the performance of the model.

In conclusion, through WGCNA, our study demonstrated that three co-expression modules were associated with metastasis among UM patients. Through LASSO-Cox analysis, four genes were finally identified for inclusion in the risk score. Furthermore, patients in high-risk group have significantly higher levels of an immune infiltrate (γδT cell) compared with patients in low-risk group. Notably, this model needs to be validated in a larger dataset, and may be conducive to therapeutic decision-making in the clinical setting.