Conjunctival melanoma copy number alterations and correlation with mutation status, tumor features, and clinical outcome

Abstract Relatively little is known about the genetic aberrations of conjunctival melanomas (CoM) and their correlation with clinical and histomorphological features as well as prognosis. The aim of this large collaborative multicenter study was to determine potential key biomarkers for metastatic risk and any druggable targets for high metastatic risk CoM. Using Affymetrix single nucleotide polymorphism genotyping arrays on 59 CoM, we detected frequent amplifications on chromosome (chr) 6p and deletions on 7q, and characterized mutation‐specific copy number alterations. Deletions on chr 10q11.21‐26.2, a region harboring the tumor suppressor genes, PDCD4, SUFU, NEURL1, PTEN, RASSF4, DMBT1, and C10orf90 and C10orf99, significantly correlated with metastasis (Fisher's exact, p ≤ 0.04), lymphatic invasion (Fisher's exact, p ≤ 0.02), increasing tumor thickness (Mann–Whitney, p ≤ 0.02), and BRAF mutation (Fisher's exact, p ≤ 0.05). This enhanced insight into CoM biology is a step toward identifying patients at risk of metastasis and potential therapeutic targets for systemic disease.

However, the overall prevalence of these CNAs in CoM and their correlation with disease characteristics and prognosis remain unclear. Lake et al did not reveal any association between 6p21.2 gains and histological cell type, age, sex, or survival (Lake et al., 2011). In addition, no correlation between BRAF, NRAS or NF1 mutations and recurrence, metastasis, or mortality was found (Gear et al., 2004;Griewank et al., 2013;Lake et al., 2011;Larsen et al., 2016;Scholz exact, p ≤ 0.05). This enhanced insight into CoM biology is a step toward identifying patients at risk of metastasis and potential therapeutic targets for systemic disease.

K E Y W O R D S
allele-specific copy number, BRAF/NRAS mutation, conjunctival melanoma, copy number alteration, metastasis Significance Conjunctival melanoma (CoM) is a rare but potentially fatal melanoma subtype. We analyzed copy number alterations, and their frequencies, in relation to tumor characteristics and patients' outcomes. We identified recurrent 10q deletions, which correlated with histological features of poor prognosis and metastatic risk. This finding should facilitate future development of disease-specific prognostic and therapeutic models. Sheng et al., 2015). A strong association between BRAF mutation and sun exposure was determined in two reports (p ≤ 0.03; Griewank et al., 2013;Larsen et al., 2016), and with clinical and pathological T1 stage (p = 0.007) in a single study (Larsen et al., 2016). The data are divided with regard to BRAF mutations in relation to age and sex, with some authors reporting a significant correlation with male gender and age younger than 65 years (p ≤ 0.02; Larsen et al., 2016), whereas others did not find any correlation between these parameters (Griewank et al., 2013;Scholz et al., 2018;Sheng et al., 2015).
Hence, there was a clear need to study, in depth, the prevalence of various CNAs and their clinical significance in a large clinically welldefined CoM cohort with a genome-wide approach. This current multicenter collaborative project was established to examine one of the largest CoM cohorts using single nucleotide polymorphism (SNP) genotyping array, with the intention of defining key biomarkers of CoM metastatic risk, and to correlate these with clinico-histological features and clinical outcomes, in order to identify patients at risk of metastasis, and also potential therapeutic targets for systemic disease.  pre-invasive disease (i.e., either termed "primary acquired melanosis" (PAM) or conjunctival melanocytic intraepithelial neoplasia (C-MIN; Damato & Coupland, 2008)

| DNA extraction
For each sample, a 4-μm formalin-fixed paraffin-embedded (FFPE) H&E-stained section was examined to identify areas with greater than 90% tumor cells. DNA was extracted using QIAamp ® DNA Mini Kit (Qiagen, Crawley, UK) following a modified protocol as previously described by Lake et al. (2011). The quality of extracted DNA was determined by a modified multiplex PCR, adapted from van Dongen et al. (2003). PCR products were visualized on 2% agarose gels stained with 1X SYBR Safe (Invitrogen, Paisley, UK) using the BioDoc-It Imaging System (Ultra-Violet Products Ltd., Cambridge, UK).

| SNP array
Affymetrix SNP 6.0 genotyping (Affymetrix, Santa Clara, CA) was performed at Atlas Biolabs (Berlin, Germany). Raw data were analyzed by Partek Genomics Suite ® (PGS) version 6.6 (Partek The SNP data for all tumors can be accessed in the international public repository Gene Expression Omnibus (GEO, http://www.ncbi. nlm.nih.gov/geo; accession number GSE123011).

| Data analysis
Data analysis was undertaken in three separate work packages as outlined below: 1. Global analysis of all CNAs detected. CNAs in the primary and locally recurrent CoMs were initially compared. Gains or losses occurring in at least 40% of the 59 samples were then examined. This was an arbitrary cutoff to reliably identify the most common changes in CoM from the vast spectrum of CNAs obtained by PGS and was chosen after several analyses with higher and lower cutoff levels.

| Identification of CNAs associated with CoM metastatic risk
No oncogenes or TSGs were identified in the CNAs exclusive to CoMMET+ group. In comparison, the TSG PLPP5 (8p11.23) was exclusive to CoMMET− samples but was amplified in 3 and deleted in a further 3 samples (Fisher's exact, p = 1), so was not analyzed further. By comparing the CNAs present in at least 50% of tumors in the two groups, four TSGs on chr 10q24.32-26.2 were significantly deleted in CoMMET+ tumors in contrast to CoMMET− group. These were NEURL1, SUFU, PDCD4, and C10orf90 (Fisher's exact, p ≤ 0.05).
Kaplan-Meier survival curves on the 59 patients estimated this regional deletion to be significantly associated with lower metastasisfree survival (log-rank, p = 0.008, Figure 3)

| Correlation between CNAs and mutation status
Eighteen (

| Examining the effect of deletions on protein expression in selected samples
IHC protein expression localization and intensity scoring of the four significantly deleted TSGs in CoMs that metastasized showed: 1. NEURL1 expression was assessable in the 15 examined CoMs.
Cytoplasmic expression scores ranged between 6 and 12 (mean; 8.4 and median; 8) in five CoMs with NEURL1 deletion, and 4 to 12 (mean and median: 8) in ten tumors with diploid CN.
F I G U R E 3 KM curve of chr 10q24.32-26.2 deletion in relation to metastasis performed on 59 CoM patients. Tumors having the deleted 10q region were significantly associated with reduced metastasis-free survival 3. SUFU expression was assessable in the 14 examined CoMs.
Cytoplasmic staining was identified in 100% of the samples.
Expression staining scores ranged between 4 and 12 (mean and median: 8) for the CoMs with C10orf90 deletion and 3-12 (mean, 7.4; and median, 8) for the diploid CN tumors. Weak expression was also detected in 5% of the nuclei of a single sample with the gene deletion. For all proteins, the difference between the means of scores in normal and deleted CN tumors was not statistically significant (independent t test, p > 0.05); hence, there was no effect of the deleted CNAs on the respective protein expression, Supporting Information prospective genome-sequencing platforms should elucidate such events.

| D ISCUSS I ON
This is the first study to date to characterize, in depth and in such a large clinically well-defined cohort, genome-wide CNAs, their differential frequencies, and relationship to clinico-histomorphological tumor features, BRAF/NRAS mutation, and clinical outcomes using high-resolution SNP array genotyping technology.
Amplifications of 6p21-25 were found in up to 61% of CoM in our cohort. Regional chr 6p amplification was previously identified by our group and others (Griewank et al., 2013;Lake et al., 2011;Swaminathan et al., 2017), and has also been documented in CM (Höglund et al., 2004). The Histone Cluster 1 (6p22.2) was the most common amplification in our study, which is implicated in various cancers and is thought to impact epigenetic and post-transcriptional modification (King, Waxman, & Stauss, 2008), but was not associated with the clinical outcome in the current study.
In this study, we also identified novel regional and arm chromosomal losses on 9q, 16p, 17p, and 19, and more specifically of ASNS (7q21.3) focal deletion detected in 76% of CoM. ASNS expression is regarded as an important biomarker for therapeutic outcome in cancers. Hematological malignancies with ASNS deletions respond favorably to L-asparaginase chemotherapeutics and rarely relapse (Bertuccio et al., 2017), whereas its low expression in colon carcinoma of patients undergoing chemoradiotherapy is associated with poor survival and inferior chemotherapeutic response (Lin et al., 2014). In our study, ASNS did not correlate with the clinical outcome of our patients; however, there is scope to examine its role in the management of the ocular tumor or its secondaries.
Most intriguing in this study are the detected 10q deletions in CoM and their correlation with metastatic risk, of which 10q26.3 was also previously described by our group (Lake et al., 2011). Our work has shown that regional deletions of chr 10q24.  (Griewank et al., 2013). These are well-recognized alterations in CM with implications on the oncogenic PI3K pathway inactivation and targeted therapy application (Isshiki, Elder, Guerry, & Linnenbach, 1993), and, therefore, they are potentially pertinent to CoM druggable targets.
Broad regional chromosomal alterations were characteristic of CoM rather than single gene aberrations. It remains unclear whether such regional anomalies are "bystander" alterations subsequent to single gene loci or non-random events where the oncogenes and TSGs in the affected regions play roles at different stages of oncogenesis (Kwong & Chin, 2014 (Primdahl et al., 2002;Van Loo et al., 2010).
The current study also provides further evidence supporting the similarities between CoM and skin melanoma. BRAF and NRAS mutation frequencies, and their mutual exclusivity, were comparable to CM (Curtin et al., 2005;Davies et al., 2002), and in agreement with previous CoM reports ( Griewank et al., 2013;Larsen et al., 2016;Scholz et al., 2018). We did not detect a correlation between the mutation status and clinical or histopathological features when compared to previous studies (Griewank et al., 2013;Larsen et al., 2016). This could be explained by the difference in analytical methods.
We acknowledge the limitations of our study. First, with variable treatment regimens and follow-up times of the participating patients, it was not possible to assess the relationship between genetic alterations and CoM local recurrence. Nevertheless, the increased risk of metastasis following local CoM recurrence is well-known (Damato & Coupland, 2009). Long-term collaborative studies are essential to determine how radiation and topical chemotherapeutic methods could affect CoM biology. This should be undertaken as a future project when more samples become available. Second, the relation between chr 10q deletions and metastatic death was difficult to assess because of low number of events and the survival of some patients with metastases in our cohort. A prospective reanalysis of chr 10 and any related fatalities is warranted. Third, BRAF and NRAS mutations were tested by different techniques, and, as a result, their rarer mutations might have been missed in a small number of cases.
In addition, it is possible that tumors that were both BRAF wt and NRAS wt harbor NF1 or other RAS mutations, which were reported only after the completion of the current work (Scholz et al., 2018). A follow-up study that incorporates all known CoM mutations is now needed to understand their significance in CoM pathogenesis.
In summary, we here present the most comprehensive profile of CNAs in a clinically well-defined CoM cohort to date and identified potential markers for metastatic risk and prognostication. The ultimate challenge remains to apply our current knowledge in the future development of prognostic models and effective therapeutics in metastatic CoM, as has been achieved in part with skin melanoma.

ACK N OWLED G EM ENTS
We would like to thank the Liverpool Ocular Oncology team,