Rachel E. Doherty1 • Mohammed Alfawaz1,2 • Jessica Francis1 • Beckie Lijka-Jones1 • Karen Sisley1
1Academic Unit of Ophthalmology & Orthoptics, Department of Oncology & Metabolism, The Medical School, The University of Sheffield, Sheffield, UK; 2Department of Clinical Laboratories, College of Applied Medical Sciences, Northern Borders University, Arar, Saudi Arabia.
Abstract: Uveal melanomas (UMs) comprise only 3% of all melanomas, but they are the most common primary intraocular malignancy in adults. The disease is associated with high mortality, with the liver being the most common site for secondary tumors. Genetic studies performed over the last 30 years have provided a wealth of information on the changes found in primary posterior UM (ciliary body and choroid), but less is known about the specific alterations of the rarer and the more benign iris melanomas, or the metastatic lesions. Early cytogenetic studies identified consistent chromosomal abnormalities, including monosomy 3 (M3), gain of the long arm of chromosome 8, and changes affecting both arms of chromosomes 1 and 6. More recently, specific genetic mutations have been related to UM. Prominent among these are mutations of guanine nucleotide-binding protein Q polypeptide/guanine nucleotide-binding protein alpha-11 (GNAQ/GNA11), which are mutually exclusive and occur in approximately 90% of posterior UMs. Other mutations such as BRCA1-associated protein (BAP1), splicing factor 3B subunit 1 (SF3B1), eukaryotic translation initiation factor 1A, X-linked (EIF1AX), and telomerase reverse transcriptase promoter (TERTp) have also been associated with UM. There are clear relationships between cytogenetic alterations and prognosis, and M3, 8q+ 6p+, and 1p- can be considered as biomarkers. An improved understanding has also been gained regarding the impact of genetic mutations, but ultimately the underlying drivers of the most predictive changes are poorly understood. This review discusses the cytogenetic alterations and gene mutations of UM and the relationship they have, if any, with the outcome.
Keywords: Biomarkers; Cytogenetic; Mutations; Uveal melanoma
Author for correspondence: Karen Sisley, Academic Unit of Ophthalmology & Orthoptics, Department of Oncology & Metabolism, The Medical School, The University of Sheffield, Sheffield, UK. E-mail: k.sisley@sheffield.ac.uk
Doi: http://dx.doi.org/10.15586/codon.noncutaneousmelanoma.2018.ch2
In: Noncutaneous Melanoma. Jeffrey F. Scott and Meg R. Gerstenblith (Editors), Codon Publications, Brisbane, Australia. ISBN: 978-0-9944381-5-7; Doi: http://dx.doi.org/10.15586/codon.noncutaneousmelanoma.2018
Copyright: The Authors.
Licence: This open access article is licenced under Creative Commons Attribution-NonCommercial 4.0 International (CC BY 4.0). https://creativecommons.org/licenses/by-nc/4.0/
Uveal melanoma (UM) is the most common intraocular malignancy in adults accounting for 80% of all noncutaneous melanomas (1). UM arises from melanocytes within the iris, choroid, and ciliary body collectively termed the uvea or uveal tract, and can be separated into two groups depending on where they are located. Anterior UM arises within the iris, accounting for up to 10% of all UMs. These tumors tend to be benign and are rarely seen to metastasize (2). Posterior UM within the choroid or ciliary body accounts for the remaining 90% of cases and are often highly aggressive and they frequently metastasize. There have been far fewer studies of the genetics of iris melanoma or the metastatic lesions, and they will be reviewed separately after consideration of the genetic changes of posterior (ciliary body and choroid) melanomas. It is also important to bear in mind that UM and cutaneous melanoma (CM) are distinct genetic entities, and they vary not only in their chromosomal changes but also in their mutational signature.
The initial investigations exploring the genetic background of UM were almost entirely cytogenetic, and they quickly identified and confirmed that the chromosome changes in UM are not random, and unlike many other solid tumors were relatively few and characteristic in their manner of alteration (3–5). M3, the loss of one copy of chromosome 3, was the most frequent alteration, observed in between 40 and 50% of cases. Also involved was chromosome 8, where simultaneous loss of the short arm (p) and gain of the long arm (q) resulted in the formation of a characteristic isochromosome of 8q (i(8q)). Around 45% of UMs were found to have both M3 and i(8q), with the association most often observed in melanomas with a ciliary body component (Figure 1) (3–15). Some UMs however only show 8q+ without M3.
Figure 1 Analysis of chromosome alterations in the same primary uveal melanoma showing the characteristic changes of loss for chromosome 3, or monosomy 3 (M3) and i(8q), as indicated by arrows in the karyotype (a). In the array CGH profile (b), M3 is identified by the red line left of the chromosome ideogram and the i(8q) by the respective loss of 8p (red) and gain of 8q (blue).
Other nonrandom abnormalities are gains of 6p and deletions of 1p, found in approximately 40 and 25% of cases, respectively (3, 4, 7–9, 11–15). Deletions of 6q are also observed, and more recent investigations using fluorescence-based methodologies, such as multiplex fluorescence in situ hybridization (MFISH) and array comparative genomic hybridization (aCGH), have shown the combined frequency of changes affecting both arms of chromosome 6 in the order of 60–70%, making chromosome 6 the most altered chromosome in UM (6, 7). The reason why abnormalities of chromosome 6 were under-reported in the earlier cytogenetic studies is that, unlike the gross whole arm changes affecting 1p, M3, and 8q+, alterations of chromosome 6 are often more covert with subtle rearrangements only disclosed through the use of MFISH or aCGH. Other nonrandom changes present in approximately 10–20% of tumors affect chromosomes 9, 10, and 11, while correlated with M3 and i(8q) are deletions of 16q and gain of 21 (8, 10–15). Although changes of all chromosomes have been reported in UM, the relative simplicity of karyotypic alterations, in comparison with other solid tumors, has allowed for a rapid and clear correlation to be observed between specific aberrations and clinical outcomes.
There have been a number of mutations associated with posterior UM and with the increasing use of rapid high throughput analysis, such as next-generation sequencing, it is likely that more genes will be identified with mutations of relevance to UM. The frequencies at which these mutations occur vary, and their role and importance for UM for the most part remains to be determined. Overall, even deep sequencing studies have shown that UM is a malignancy with a low mutation rate (16). The most common mutations detected however are ubiquitous among posterior UM and involve the guanosine nucleotide-binding protein Q polypeptide (GNAQ) and its paralogue, guanosine nucleotide-binding protein alpha-11 (GNA11).
Mutations of GNAQ and GNA11 are not found in CM, but in posterior UM, mainly targeting codon 209, resulting in an amino acid change from a glutamine to a proline or leucine within the Ras-like domain. Point mutations at codon 209 in either GNAQ or GNA11 are mutually exclusive and have been found to occur in up to 90% of UMs (17–20). There are also rarer mutations that target codon 183 which are present in around 5% of cases, although these usually only occur in the absence of codon 209 mutations (19). GNAQ and GNA11 code for heterotrimeric Gq-proteins containing three subunits, α, β, and ƴ that facilitate the coupling of transmembrane receptors to intracellular pathways; under normal circumstances, the binding of a ligand to the receptor activates Gq-proteins. When guanonine-5’diphosphate (GDP) is bound to the α subunit, then the G-protein is in an inactive state; however, when guanonine-5’triphosphate (GTP) binds, the β-ƴ complex is released and in turn activates phospholipase C (PLC). PLC triggers a kinase cascade, resulting in the transcription of pro-proliferative and anti-apoptotic genes, including the upregulation of a key growth regulatory pathway, the mitogen-activated protein kinase (MAPK) pathway (21–23). Mutations at codon 209 of GNAQ and GNA11 target the Ras-like domain of the alpha-q subunit, and the loss of glutamine (essential for intrinsic GTP hydrolysis activity) results in GTP being irreversibly bound to the α subunit and hence constitutive activation of the Gq-protein (17, 18, 20). How these mutations advance UM is still under consideration, but mutations to codon 183 are thought to be less effective in activating the MAPK pathway in comparison to mutations occurring at codon 209. In one instance, mutations of GNA11 at both codon 209 and codon 183 were found in the same lesion (19). Although in cancer the MAPK pathway is reported as the most consistently upregulated pathway (24–26), the true significance of mutations leading to its upregulation in UM need to be clarified. As activating mutations to GNAQ or GNA11 have been found to be present at all stages of UM progression (including benign lesions), they are thought to be an early, if not initiating, event in UM (17, 27).
Harbour and colleagues were the first to describe the inactivation of the Breast Cancer 1 (BRCA1)-associated protein 1 (BAP1) gene in more than 80% of metastatic UMs, suggesting that this may be a UM tumor suppression gene (28). This gene is located on chromosome 3p21.1 and therefore a copy is often deleted in UM due to the frequent occurrence of M3. The BAP1 gene encodes for a nuclear localized deubiquitinase with an N-terminal ubiquitin carboxyl hydrolase domain and C-terminal nuclear localization signal domain. Mutations in the BAP1 gene can prematurely terminate BAP1 protein and also affect the ubiquitin carboxyl-terminal hydrolase domain altering its deubiquitinase activity (29). The BAP1 protein interacts with various proteins, including the tumor suppressor gene BRCA1, and thus BAP1 protein plays essential roles in maintaining genome stability, epigenetic modification, transcription regulation, and in the response to DNA damage. Mutations of BAP1 (unlike those of GNAQ/GNA11) are not restricted to melanocytic lesions, and reports in other cancers, as well as UM, have suggested that the action of BAP1 is wide ranging, directly affecting proliferation by stalling cells in S phase and down-regulating the E2F transcription factor (30, 31). Other studies suggest BAP1 protein dysregulation has implications in pluripotency (32). Furthermore, BAP1 may represent the first clearly identified predisposing gene among hereditary forms of UM. There are many reports of familial UM, often with no clearly identified genetic link (33–36). Germ line mutations among younger patients with UM, and a correlation with BAP1 mutations in families where there has been a predisposition for renal cancers and mesothelioma, suggest that BAP1 may indeed be a predisposing gene among a certain class of hereditary UM patients (37–39).
Mutations of the splicing factor 3B subunit 1 (SF3B1) gene have been found in 10–20% of UM cases but virtually exclusively in the subset of UM without M3 (40–42). In CM, mutations in SF3B1 are rare, occurring in only 1% of patients (43), but like GNAQ/GNA11, SF3B1 mutations have been reported in other forms of melanoma (44, 45). SF3B1 gene is located at chromosome 2q33 and encodes for a subunit of a large complex responsible for processing precursor mRNA (spliceosome), where it ensures correct splicing by retaining pre-mRNA to define the site for splicing (46). Mutations of SF3B1 can therefore lead to alternative splicing events for multiple genes, with different mutations in UM and other cancers, producing alternate splice variants of a number of genes. For example, in leukemia, and breast and pancreatic cancer, a hotspot at codon 700 is reported (47–50), but for UM and mucosal melanomas mutations at codon 625 of exon 14 predominate (44). The full biological significance of this aberrant splicing of multiple genes remains to be determined (40, 51, 52), but in leukemia mutations to SF3B1 alter the DNA damage response, leading to an increase in DNA damage (53).
Mutations to eukaryotic translation initiation factor 1A, X-linked (EIF1AX), a gene located on chromosome Xp22, occur in approximately 13% of posterior UM. In a similar pattern to SF3B1 mutations, EIF1AX mutations are also mainly restricted to UM without M3 (41). Furthermore, mutations of EIF1AX appear to be mutually exclusive to SF3B1 in UM and other melanocytic lesions (41, 54); although reported in thyroid cancer, they only occur in a subset of tumors without V-Raf Murine Sarcoma Viral Oncogene Homolog B1 (BRAF) mutations (commonly mutated in CM) (55, 56). It is of interest that EIF1AX (similar to SF3B1) also encodes for a protein which interacts with mRNA, with a role in initiating translation, through a combination of recognition of target mRNA and stabilization of the ribosome, to prepare mRNA for translation (54, 56). In UM, mutant EIF1AX has been confirmed to result in aberrant translation (57), but whether mutations of both SF3B1 and EIF1AX have biological significance due to a generalized deregulation of translation, or a focalized effect targeting the same or similar genes, has not been shown. It is however considered that both SF3B1 and EIF1AX are later events in UM progression, while their mutual exclusivity suggests that they provide alternative evolutionary pathways for disomy-3 (D3) UM, with no numerical abnormalities of chromosome 3 (58).
A number of other mutations have been serially identified in UM, but the incidence is substantially lower than those already discussed, and their relevance may reflect “background noise” rather than driver mutations. Of these, the telomerase reverse transcriptase (TERT) gene (located on chromosome 5p15) has been widely reported in many cancers (59–62), but in UM it only has a low mutation frequency of 1% (63, 64). TERT has a multifunctional role including, among others, in the maintenance of telomere length, cell-cycle control, and DNA damage response (59, 65–67). Since most studies suggest that UMs have a low mutational burden (68, 69), it is possible that TERT mutations are bystanders in UM.
Other mutations occurring in less than 5% of UMs have recently been reported (68, 69). Of these, Cysteinyl Leukotriene Receptor 2 (CYSLTR2) and Phospholipase C Beta 4 (PLCB4) are mutually exclusive to GNAQ/GNA11 mutations with presumably the same functional consequences, and they, in combination with others, serve to further classify the mutational signature of UM into potentially four or more subgroups (68). Additional studies are required to clarify the significance of these mutations and whether they complement or drive UM progression.
Even from early genetic investigations, it was apparent that recurrent alterations in UM could be used to stratify and determine prognosis, with poor prognosis being survival of 7 years or less (70, 71). Most of the consistent genetic changes/ mutations in UM are predictive of outcome to a greater or lesser extent. There is however one notable exception; mutations of GNAQ/GNA11 (in almost 95% of UM) may vary in their frequency and presence of rare versus common mutations but still have no clear link to patient outcome (17–19, 68). In comparison, simple chromosome changes such as M3 and 8q+ correlate with clinical parameters of poor prognosis such as larger tumor size, epithelioid cell type, and ciliary body involvement (4, 70–72). Both M3 and 8q+ have repeatedly been shown to strongly correlate with a poorer outcome, especially when they occur together in the same tumor (6, 70, 73, 74) and as such form the basis of many biomarker tests. Furthermore, it is clear that the amplification of the long arm of chromosome 8, in the form of an i(8q), is not a static process and within individual UM there will be a level of heterogeneity with some cells showing just a gain of chromosome 8, while others may have one or many more copies of i(8q) (70, 75). Importantly, this drive toward acquiring additional copies of chromosome 8q is linked to not only a poor prognosis but also a significantly shorter survival (6, 70, 74, 76). Deletion of chromosome 1p is also considered a poor prognostic indicator but may be limited as a biomarker to UM where it is associated with M3 (77, 78). As M3 is such a powerful predictor of poor prognosis, it is not altogether surprising that the alternative scenario of D3 is related to improved survival, while much rarer partial deletions of chromosome 3 appear to correlate with an intermediate prognosis (79–81). Finally, there is some ambivalence over the ascribing of prognosis based upon 6p gain. In most UMs, 6p+ is found in tumors with D3 and is therefore indicative of a good outcome (82). Gain of 6p however is not entirely restricted to D3 tumors and can be found as a later change in tumors with M3 and i(8q) (9, 10, 75, 77, 83). Under these circumstances, it is decidedly associated with a poor outcome; therefore, careful consideration of the circumstances must be made if reliance is just placed on gain of 6p alone.
Although mutations of GNAQ/GNA11 are not predictive, other mutations do confer a prognostic significance, or as a minimum clearly associate with other known indicators of prognosis. As an example, there is some ambivalence over mutations of BAP1, which although clearly segregating with M3 in UM, at a frequency of 40–80% and initially reported in over 80% of metastatic tumors, do not themselves seem to predict a poor prognosis (28, 37, 68). The reason may be that mutations of BAP1 are not restricted to limited hotspots, and recent evidence suggests that they are no more frequent among metastasizing UM compared with nonmetastasizing UM (84). There does seem to be a clear consensus that BAP1 mutations do show a correlation with reduced expression of BAP1 protein and more specifically that assessment of the protein by immunohistochemistry provides a more robust indication of outcome compared with mutational analysis alone (85).
In general terms, both SF3B1 and EIF1AX mutations can be considered as indicators of a better prognosis, presumably due to their preponderance among D3 UM (42, 63). Mutations of SF3B1 should however be considered as an indicator of a qualified good prognosis, as approximately 80% of UM with SF3B1 mutations have been shown to develop later metastasis, despite being D3 (68, 86). The reason why SF3B1 mutations may identify a later metastatic onset among classically good prognosis D3 UM could be their association with partial gains of 8q, while EIF1AX mutations do not (68). The gain of 8q is consistently predictive of a poorer outcome, but usually when clearly identified in the form of an i(8q), representing whole arm gain of 8q (6, 70). Less is known about the prognostic relationship for partial 8q, partly because few studies have distinguished differences in the manner of acquisition for 8q gain, and where explored, no reduced survival was reported (6), possibly because additional follow-up is required to detect patients who will succumb to late onset metastasis in this subgroup.
In addition to chromosomal and mutational biomarkers, UM can be classified into poor prognostic groups on the basis of expression of a target set of genes (initially 26 but later refined to 15 genes) that delineate UM into low risk (class 1) and high risk (class 2) (87, 88). Using this data set, class 1 UM patients are ascribed a survival of 95% at around 7 years, while class 2 had approximately only a 30% survival. Further stratification of these subgroups can be achieved when expression of preferentially expressed antigen in melanoma (PRAME) is considered, as increased expression in class 1 UM identifies those with a very low rate of metastasis (89).
Most UM patients who develop metastatic disease will do so in the liver, often to the exclusion of other sites. There have however been comparatively few genetic studies of the lesions themselves, and not surprisingly those reported are mainly hepatic metastases (5, 90–92). Changes found in the metastases mirror those reported for primary UM, with the single most consistent and frequent change being the 8q gain. It is however of interest that BAP1 mutations and other predictors of poor outcome, such as M3 and 1p-, are not more frequently represented among metastatic lesions. Conversely, 6p gain, considered as an indicator essentially of good prognosis, is also reported in metastatic lesions (90, 93), whereas reports suggest that although GNAQ/GNA11 are not predictive of prognosis, GNA11 mutations are more frequent among metastases (94). Of course, some of these observations require further validation once more information on the metastases themselves becomes available.
The constitutive activation of the MAPK pathway through mutation has been identified as an early event in many cancers (17, 95, 96). As most posterior UMs have activating mutations of GNAQ/GNA11, it is clearly an early, if not initiating, event. Studies have however reported that melanocytic nevi, melanocytomas, and blue nevi also have mutations of GNAQ/GNA11, and in these benign lesions it is purported to associate with an intermediate state with potential to transform and become malignant (97). In relation to posterior UM, the findings of studies on choroidal nevi (98) and anterior iris UM are perhaps more pertinent (see below). As iris melanomas are, in the majority of cases, manifestly benign, the presence of GNAQ/GNA11 mutations suggest that in UM they are initiating events, increasing proliferation by upregulating the MAPK pathway, but not necessarily inducing transformation to malignancy. As such, GNAQ/GNA11 mutations act as a rung on a ladder toward malignancy but other changes are also necessary to proceed.
At the forefront of these additional alterations is M3, with many cytogenetic/ molecular studies identifying it as an early and sometimes the only other change in addition to GNAQ/GNA11 (99). Equally, 6p+ when associated with D3 tumors appears to fulfill the same role in providing the next step in tumor progression (82). Mutations of BAP1 in M3 UM, and SF3B1 and EIF1AX in D3 UM, are closely linked to the cytogenetic changes of M3 and 6p+, respectively. There are few studies that have made detailed comparisons, so information on the timing of these events is limited. BAP1 mutations are however reported as subsequent to M3 (68), and as mutations of both SF3B1 and EIF1AX are reported as being related to 6p+ in UM, it can be assumed that they are secondary to that of 6p gain (69, 86). In terms of gross abnormalities, there is almost universal agreement that 8q gain is a later event and its high frequency among metastatic lesions acts as confirmation of the relationship with advanced UM (70, 77, 82, 83, 99, 100). The proposed sequence with relationship to outcome and clinical parameters is presented in Figure 2.
Figure 2 Correlation of clinicopathological, chromosomal, and genetic biomarkers with prognosis in uveal melanoma and putative sequence of events. There are well-defined clinical, chromosomal, and genetic biomarkers identified in UM that contribute to prognosis in UM. Mutations to GNAQ and its paralogue GNA11 have been identified as an early change in UM and are not associated with prognosis but may represent an initiating event. Poor prognosis is associated with a posterior location, epithelioid cell type, monosomy 3, isochromosome 8q, and loss of 1p. Recent studies have also identified that mutations to BAP1 are indicative of a poor prognosis, whereas mutations to EIF1AX, although considered a later event in UM progression, are in fact associated with low metastatic risk. Mutations to SF3B1 are also thought to occur late in UM progression and are associated with an intermediate prognosis in patients without monosomy 3, who otherwise would have had a good prognosis.
There have been far fewer genetic studies performed on iris melanomas compared to their posterior counterparts, but from the information so far gleaned they appear to represent somewhat a halfway house between posterior UM and CM (101–106). In this respect, iris melanomas are reported as sharing changes with posterior UM such as deletions of 1p, M3, and chromosome 6 alterations (both p and q). Evidence however suggests that iris melanomas also have abnormalities of other chromosomes rarely affected in posterior UM, such as changes of chromosome 9p, an alteration more often related to CM (105), and rearrangements affecting chromosome 18 (105, 106). Equally, iris melanomas are reported as sharing mutations with posterior UM and also CM, with studies identifying mutations of both GNAQ/GNA11 and BRAF in iris melanomas (105, 107). Furthermore, a surprisingly high frequency of EIF1AX mutations has been reported for iris melanomas (108). Mutations have not yet been documented for SF3B1, and it is likely that the small sample size for most of these reports means that the rarer mutations may still be detectable in iris melanomas, once more are sequenced. It is premature to predict how these mutations and chromosome changes impact on iris melanoma development; however, it was observed that BRAF mutations in 9 of the 19 tumors sequenced were more likely to associate with tumor recurrence (107), while in contrast the relatively high frequency of the good prognostic mutations of EIF1AX is perhaps symptomatic of the benign nature of most iris melanomas, or vice versa.
Since genetic studies have become engaged in improving the understanding of the basis of UM, related chromosome changes and genetic mutations have been identified that can be usefully employed as biomarkers to predict prognosis. Whatever methodology is applied, the use of genetic biomarkers is now the most reliable method for detecting outcome, but no single method is infallible. The challenge now is to improve our understanding and to extrapolate how these quintessential alterations act as drivers, and most importantly how their behavior can be counteracted. In this respect, it is highly likely that future studies will be able to fine-tune the relationship between individual genetic mutations and their role in UM. Understanding how gross chromosomal changes, such as those of 1p, M3, 6p+ and 8q+, serve to drive, or differentiate, UM behavior is entirely another ball game. Furthermore, a virtually undescribed landscape that could impact on such changes is emerging, with recent studies suggesting that the well-described genetic biomarkers of UM are only a signpost for epigenetic modulators capable of dividing the broad brush genetic classes into further subgroups, with an impact not yet fully comprehended (68, 69, 109). The genetic landscape of UM, although not as obviously unstable as many cancers, is not however featureless and seemingly is far more complex than first thought.
Conflict of interest: The authors declare no potential conflict of interest with respect to research, authorship, and/or publication of this article.
Copyright and permission statement: To the best of my/our knowledge, the materials included in this chapter do not violate copyright laws. All original sources have been appropriately acknowledged and/or referenced. Where relevant, appropriate permissions have been obtained from the original copyright holder(s).