Uveal melanoma (UVM) is a cancer of the eye, specifically involving three uveal melanocytic cell types: the iris, the ciliary body, and the choroid (collectively referred to as the uvea) [1
]. Etiology of uveal melanoma is distinctively different from that of skin melanoma (cutaneous melanoma, CM) and mucosal melanoma despite all melanomas arising from melanocyte cells that produce melanin. Solar UV radiation (UVR) is an environmental risk factor for CM, while the UVR effect on UVM remains under debate. Nevertheless, the eye color and skin color are phenotypic risk factors for both UVM and CM. Like CM, UVM is also positively associated with a higher social economics status as indicated by area-based socioeconomic measures [2
Epidemiology of UVM was published previously based on data from cancer registries [5
]. However, there was inaccuracy of classifications due to heterogenous tumor types in some previous reports. Particularly, some studies included retinoblastoma, while other studies included other cancer types which arise from the orbital structure [6
]. Although these other types of eye cancers count for ~20% of all eye cancers, there is a concern that the trend of this rare cancer is not accurately reflected.
Sex difference in UVM is documented, with men showing higher incidence rates than women in general [1
], which is similar to that in CM [8
]. We have previously found that sex difference in CM is age-dependent, with older men showing a higher incidence rate than older women while younger women show higher incidence than younger men [8
]. The age dividing line is approximately around women’s menopause (~50 years). The age-dependent sex difference may not only reflect the behavior difference of men and women at different ages (i.e., exposure to indoor and outdoor UV radiation), it may also reflect an intrinsic difference in pathophysiological aspects of the disease etiology such as changes in the sex hormone levels. Direct evidence of involvement of sex hormone and their receptors is limited or controversial in the literature for both CM and UVM [10
]. Therefore, the first part of this study attempts to analyze age-specific UVM incidence rates in men and women, in order to develop a hypothesis for mechanistic explanation in incidence and prognosis.
Genetic and somatic mutations are important causes for UVM [12
]. Monosomy 3 (including loss of BAP1 copy number, or loss-of-function mutations in BAP1 gene) causes multiple cancer phenotypes including UVM [13
]. Loss of BAP1 drives metastasis and is associated with poorer survival of UVM patients [14
]. Somatic DNA mutations in UVM include GNAQ, GNA11, PLCB4, SF3B1, SRSF2, EIF1AX, CNKSR3, CYSLTR2, and YAP1 [16
]. Among these genes, about 83% of UVM tumors have mutations in either GNAQ or GNA11 [18
]. Although nearly mutually exclusive, mutations in GNAQ and GNA11 in metastatic UVM are presented at different rates, with GNA11 mutations more frequently associated with metastatic UVMs [18
]. PLCB4 and CYSLTR2 mutations are usually present in UVMs that lack GNAQ or GNA11 mutations, and occur in small percentage of UVMs [20
]. PLCB4 encodes a phospholipase C, while CYSLTR2 (Cysteinyl-Leukotriene Receptor 2) encodes a G protein-coupled receptor. CYSTLTR2 and PLCB4 can initiate mutations along with GNAQ and GNA11, while BAP1, EIF1AX, and SF3B1 can promote mutations. SF3B1 and SRSF2 both encode splicing factors and play key roles in the alternative splicing of mRNA, which affects cell cycle progression and cell death [21
]. CNKSR3 amplification is associated with better survival of UVM [23
]. The EIF1AX is located on the X chromosome and encodes a eukaryotic translation initiation factor 1A. Frequently mutated in a number of cancer types including carcinomas and UVM, EIF1AX is considered a novel oncogenic driver [24
]. Molecularly, EIF1AX is essential for the assembly of 43S pre-initiation ribonucleoprotein complexes for protein synthesis [26
]; a mutant form of EIF1AX was able to increase general protein synthesis in thyroid carcinoma [26
], which is consistent with higher protein synthesis demand in cancer cells. The YAP1 gene is well studied for its function in promoting tumorigenesis. In uveal melanoma, YAP1 acts downstream of GNAQ/GNA11 signaling to promote cell proliferation [27
]. To the best of our knowledge, there have been no reports on sex difference in the above-mentioned mutations, or systematic analysis of gene expression difference in uveal melanoma.
2. Materials and Methods
UVM incidence data source and analysis: US SEER18 research data (1975–2016) was downloaded using the SEERStat software (Version 8.3.8). The selection criteria for UVM followed the International Classification of Diseases for Oncology, third edition (ICD-O-3): “Primary site = C69.2, retina; C69.3, choroid; C69.4, ciliary body”, “Morphology = 8720–8790, nevi and melanomas” and “Behavior = 3, malignant”. US 2000 standard population was used for age-standardization. The annual percentage change of incidence rates was analyzed using the Joinpoint Regression Program, Version 18.104.22.168, downloaded from the SEER website. The age-standardized incidence rates were used for trend analysis. Statistical analysis was carried out by Stata IC13 software (College Station, TX, USA).
UVM genomics data and analysis: The TCGA-UVM data (mutation, copy number variation, mRNA levels normalized by RSEM algorithm [28
], clinical data, and patient information) was downloaded from the GDC Data Portal (https://portal.gdc.cancer.gov/projects/TCGA-UVM
, accessd on 22 July 2021) [17
]. Analysis on individual gene level was carried out by Stata IC13 software. For analysis of sex differentiated gene expression at the genomic level, genes with a RSEM value of less than 1 were removed. The DESeq function (DESeq2 program) was used to determine differential expression between sexes [29
]. All genes that failed to yield a p
value less than 0.05 and a fold change greater than 1.7 were removed. The Benjamini–Hochberg false discovery rate procedure was performed on the trimmed gene list [30
]. All genes that failed to yield a false discovery rate of less than 0.05 were removed. Significant protein-coding genes were then uploaded to the STRING v11 website for functional protein association network analysis [31
]. Significantly enriched pathways and annotated keywords were defined by Benjamini–Hochberg procedure adjusted p
values (i.e., false-discovery rate) of less than 0.05.
The age-dependent sex difference in UVM is summarized by analyzing the SEER data. As reported before, men have a higher incidence of UVM than women, but this difference was caused by the disparity in older age only. At younger age (<40 years) there was no sex difference, unlike the cutaneous melanoma which showed substantial difference at both young and older age [9
]. Over the years, UVM showed a sharp increase from year 1990 to 2000, and then maintained a slow and non-significant increasing trend. A more comprehensive reporting system for UVM cases may be the reason for the increase in reported cases, as many cases were diagnosed outside of cancer centers and may not have been registered (clinical observation by Dr. Mathew Wilson, also [33
]). The younger age group, though, showed significant increase from 1986 to 2016. This trend is especially obvious in young women but not in young men (Figure 1
E,F). The higher cancer incidence in older men is a common phenomenon if all cancer sites are taken into consideration [34
]. The attributing factors are not quite clear, but may be related to both pathophysiological changes (intrinsic changes following aging) and behavior difference in the two sexes; for example, smoking and drinking is more prevalent in men. The intrinsic sex hormone changes may also play an important role, as sex hormones regulate essentially all aspects of cellular activities, which include immune responses, oxidative regulation, and even DNA repair [35
]. Overall, the underlying driving force warrants further investigation.
EIF1AX was identified to express at significant higher levels in female tumors than in male tumors, and it also exhibited an unusual correlation with copy number. A comprehensive study showed that while most genes showed a positive correlation between mRNA level and copy numbers, about 1% of genes showed inversed correlation, i.e., higher copy number was associated with lower mRNA levels [36
]. It is unclear how this gene is regulated; however, it is clear that female tumors showed higher levels of mRNA. A common variant of EIF1AX (A113_splice mutation) found in thyroid cancer is often associated with the RAS oncogene and drives thyroid cancer development [26
]. Thyroid cancer incidence is about three times higher in women than in men [37
], and perhaps EIF1AX plays a role. In UVM, however, further investigation is needed to validate and explain why female tumors express higher levels of EIF1AX. If higher EIF1AX mRNA is a driving force in women, then it may explain why women survive better than men as EIF1AX is an indicator for Class 1 GEP (gene expression profiling) tumors which usually show better overall survival [38
Our STRING network analysis using 82 out of the 93 protein-encoding genes revealed significant GO cellular component functions in interlinked immunoglobin and cell surface/plasma membrane GO components. This is cross-validated by the annotated keyword analysis which revealed that 51.2% (42/82) of the genes in the gene set encode proteins that can be glycosylated. A major function of glycoproteins is their involvement in immune response. These results, therefore, strongly suggest that the sex difference in UVM is perhaps due to differential immune responses in men and women. Over-expression of the IGK, IGLL5, CD79a, and JCHAIN in males also supports that men may show a more inflammatory microenvironment than women, and thus provoke more immune responses to deal with it. This is perhaps due to a more rapid resolution of inflammation in women than in men in general [39
]. Furthermore, the significant protein domain “Early Growth Response” includes EGR1 and EGR2 genes, which are transcriptional factors controlling the TCR-mediated differentiation of natural killer T cells [40
]. Both EGR1 and EGR2 are down-regulated in male tumors, indicating possible fewer NKT cells infiltrated in the tumors from men. These various lines of evidence all point to a more inflammatory microenvironment and a less efficient immune system in men, which may provide a possible molecular mechanistic explanation for the sex disparity in UVM.
Another network analysis that is cross-validated by differential gene expression and annotated keyword analysis is the redox-linked disulfide bond. The PDIA2 (PDIp, PDA2) gene belongs to the PDI gene family, which belongs to a larger redox thioredoxin gene family [41
]. PDI enzymes catalyze thiol-disulfide exchange reactions to maintain the correct protein folding and activities; additionally, the disulfide bonds can be formed abnormally under oxidative stress. Men usually exhibit a higher level of oxidative stress than women [42
], suggesting higher levels of oxidation of thiol groups, and requiring more PDI enzymes. However, male tumors showed 4.5-fold lower PDIA2 levels than female tumors (Supplemental Table S1
), which suggests a poorer capacity to cope with oxidative stress. In pancreatic tissue, the PDIA2 targets pancreatic digestive enzymes and prevents formation of inactive aggregates [43
]. The PDIA2 protein is an endoplasmic reticulum-located glycoprotein [44
], exhibiting high affinity with estrogen and serving as a possible intracellular estrogen regulator in vitro and in vivo [46
]. Thus, it is not a surprise that PDIA2 is down-regulated in male tumors as compared to female tumors, as it is expected that female cells may use this enzyme as a local estrogen regulator. These results are consistent with reports that estrogen helps to deal with oxidative stress in women [47
]. Additionally, PDIA2 can directly bind to the human major histocompatibility complex class 1 antigens (HLA-A, B, and C) and play a role in antigen presentation [48
]. Taken together, with the multifunction of PDIA2 in cells, it is likely that the differential expression of this gene provides an important layer of mechanistic explanation for how sex hormones are linked to immunity regulation as well as redox regulation, both of which exhibit substantial difference in men and women.
Another characteristic of uveal melanoma is the production of melanin pigment [49
]. The TCGA-UVM data confirmed that MC1R, MC4R, and MC5R were expressed in uveal melanoma, with MC1R exhibiting the highest expression level (data not shown). The significant sex difference shown in the expression of DCT and POMC was unexpected, with POMC up (3.0 fold) and DCT down (−3.8 fold) in male tumors. POMC gene products include α-MSH, β-MSH, and β-endorphin, playing roles in pain-sensing, pigment synthesis, and immune modulation. The α-MSH peptide binds to MC1R and other receptors to stimulate pigment synthesis and regulate immune responses [50
]. Men showed an average higher level of plasma α-MSH than women [51
], and it is known that human melanocytes can produce local α-MSH [52
]. The role of MRAP2 in melanin signaling is unclear, but loss-of-function MRAP2 variants are associated with obesity [53
], which also involves α-MSH, Mc1R, and MC4R signaling. Overall, it is unclear how the shared melanin and obesity signaling plays roles in UVM. These pathways warrant further investigation.
A major limitation of the genomic data analysis is the small sample size—a total of 80 tumors in the TCGA-UVM dataset. Thus, whether the above-mentioned pathways are indeed reflecting true sex difference needs validation from a larger cohort. Another major limitation of this study remains the population-wide and associative nature of the studies. We have identified two novel mechanisms in which UVM is potentially stimulated, which may be ultimately caused by variations in sex hormone levels. The complex nature of sex hormone biology is challenging to investigate in this study. Detailed molecular studies at cellular level are required to validate the genomics findings. In addition, given the rarity of UVM and the retrospective nature, epidemiological analysis is limited by available data and variables therein.