High Cysteinyl Leukotriene Receptor 1 Expression Correlates with Poor Survival of Uveal Melanoma Patients and Cognate Antagonist Drugs Modulate the Growth, Cancer Secretome, and Metabolism of Uveal Melanoma Cells

Simple Summary This research investigates the disease relevance and therapeutic potential of cysteinyl leukotriene receptors in uveal melanoma (UM), a rare eye cancer that often spreads to the liver. Unfortunately, there are no therapies available to stop the spread of UM and patients are often faced with an extremely poor prognosis. We assess whether the cysteinyl leukotriene receptors (CysLT1 and CysLT2) are relevant to the progression of UM. Using UM patient samples, we identified that increased levels of CysLT1 in tumours is associated with reduced patient survival. Using UM cell lines and zebrafish models, we found that drugs targeting CysLT1, but not CysLT2, can alter hallmarks of cancer including cell growth, proliferation, and metabolism. This study is the first to examine the relationship of the CysLT receptors with clinical features of UM. Our data strengthen the importance of CysLT signalling in UM and suggest that antagonism of CysLT1 may be of therapeutic interest in the disease. Abstract Metastatic uveal melanoma (UM) is a rare, but often lethal, form of ocular cancer arising from melanocytes within the uveal tract. UM has a high propensity to spread hematogenously to the liver, with up to 50% of patients developing liver metastases. Unfortunately, once liver metastasis occurs, patient prognosis is extremely poor with as few as 8% of patients surviving beyond two years. There are no standard-of-care therapies available for the treatment of metastatic UM, hence it is a clinical area of urgent unmet need. Here, the clinical relevance and therapeutic potential of cysteinyl leukotriene receptors (CysLT1 and CysLT2) in UM was evaluated. High expression of CYSLTR1 or CYSLTR2 transcripts is significantly associated with poor disease-free survival and poor overall survival in UM patients. Digital pathology analysis identified that high expression of CysLT1 in primary UM is associated with reduced disease-specific survival (p = 0.012; HR 2.76; 95% CI 1.21–6.3) and overall survival (p = 0.011; HR 1.46; 95% CI 0.67–3.17). High CysLT1 expression shows a statistically significant (p = 0.041) correlation with ciliary body involvement, a poor prognostic indicator in UM. Small molecule drugs targeting CysLT1 were vastly superior at exerting anti-cancer phenotypes in UM cell lines and zebrafish xenografts than drugs targeting CysLT2. Quininib, a selective CysLT1 antagonist, significantly inhibits survival (p < 0.0001), long-term proliferation (p < 0.0001), and oxidative phosphorylation (p < 0.001), but not glycolysis, in primary and metastatic UM cell lines. Quininib exerts opposing effects on the secretion of inflammatory markers in primary versus metastatic UM cell lines. Quininib significantly downregulated IL-2 and IL-6 in Mel285 cells (p < 0.05) but significantly upregulated IL-10, IL-1β, IL-2 (p < 0.0001), IL-13, IL-8 (p < 0.001), IL-12p70 and IL-6 (p < 0.05) in OMM2.5 cells. Finally, quininib significantly inhibits tumour growth in orthotopic zebrafish xenograft models of UM. These preclinical data suggest that antagonism of CysLT1, but not CysLT2, may be of therapeutic interest in the treatment of UM.


Introduction
Uveal melanoma (UM) is a rare, intraocular cancer that metastasises predominantly to the liver in approximately 50% of patients. The primary ocular tumour is usually successfully treated with surgery or radiotherapy [1,2]. However, these treatments have limited success in halting metastatic spread of the disease. Once the cancer has disseminated to the liver, there are limited options available to patients. Overall survival for patients with metastatic UM ranges from 4 to 18 months [3][4][5][6]. UM develops in one of the most capillary-rich tissues of the body and is spread solely through the blood stream, suggesting that angiogenesis and vascular invasion play important roles in UM progression.
In comparison to other solid tumours and skin melanomas, UM has a low mutational burden [7,8] making the disease less sensitive to checkpoint inhibitors. The lack of identified mutations in UM narrows the scope for targeted therapies, with no successful targeted therapies available to date [9]. Activating mutations in GNAQ or GNA11 are found in >80% of all UMs [10], with mutations in CYSLTR2 or PLCB4 likely to account for an additional 8-10% of activating UM mutations [11]. These mutations are mutually exclusive and operate in the same pathway [12], highlighting the importance of CysLT 2 /G αq/11 /PLCB4 signalling in UM oncogenesis. In contrast to cutaneous melanoma [13], targeted therapies for UM, including those targeting the CysLT 2 /G αq/11 /PLCB4 downstream pathways, such as MEK and AKT, failed in early clinical studies [14,15]. associated with disease-free survival (p = 0.03; HR 1.05; 95% CI 1.03-1.07 and p = 0.02; HR 1.35; 95% CI 1.25-1.45, respectively) ( Figure 1A,D) and overall survival (p = 0.002; HR 1.06; 95% CI 1.04-1.08) ( Figure 1B) in UM patients. To resolve a violation of the proportional hazard assumption, the association between CYSLTR2 gene expression and overall survival ( Figure 1E) was stratified by time from 0 to 20 months and beyond 20 months and analysed by the likelihood ratio (LHR) test. Expression of CYSLTR2 is significantly associated with overall survival (p = 0.0001; HR 1.01; 95% CI 1-1.02) ( Figure 1E). When stratified based on high versus low expression of the receptors, all Kaplan-Meier curves showed significant results. This suggests that UM patients with high CYSLTR1 or CYSLTR2 transcript expression have a worse prognosis than those with low expression of the receptors. TCGA-UM samples were then divided into high and low expression of CYSLTR1 and CYSLTR2 using the third quartile as cut-off and interrogated for their association with pathways of interest using the Molecular Signatures Database (MSigDB). Using Gene Set Variation Analysis, enrichment scores were calculated for the association between high expression of either receptor and pathways of interest. Colour values correspond to the median values of the enrichment scores. Samples expressing high CYSLTR1 showed a significantly altered expression of profiles relating to Inflammatory Response, IFN-γ, TNF-α, Angiogenesis, and GPCR signalling ( Figure 1C). Interestingly, the same profiles plus glycolysis showed an altered expression in samples with high expression of CYSLTR2 ( Figure 1F).
Cancers 2020, 12, x 4 of 27 stratified by time from 0 to 20 months and beyond 20 months and analysed by the likelihood ratio (LHR) test. Expression of CYSLTR2 is significantly associated with overall survival (p = 0.0001; HR 1.01; 95% CI 1-1.02) ( Figure 1E). When stratified based on high versus low expression of the receptors, all Kaplan-Meier curves showed significant results. This suggests that UM patients with high CYSLTR1 or CYSLTR2 transcript expression have a worse prognosis than those with low expression of the receptors. TCGA-UM samples were then divided into high and low expression of CYSLTR1 and CYSLTR2 using the third quartile as cut-off and interrogated for their association with pathways of interest using the Molecular Signatures Database (MSigDB). Using Gene Set Variation Analysis, enrichment scores were calculated for the association between high expression of either receptor and pathways of interest. Colour values correspond to the median values of the enrichment scores. Samples expressing high CYSLTR1 showed a significantly altered expression of profiles relating to Inflammatory Response, IFN-γ, TNF-α, Angiogenesis, and GPCR signalling ( Figure 1C). Interestingly, the same profiles plus glycolysis showed an altered expression in samples with high expression of CYSLTR2 ( Figure 1F).   MCP-counter was used to impute an estimated score for the contribution of infiltrated stromal and immune cells in TCGA samples. A positive correlation was identified between CYSTLR1/CYSTLR2 expression and the presence of immune cells, and the ESTIMATE tool was used to calculate a global score for stromal infiltration. When the models were adjusted by the stromal infiltration scores from ESTIMATE, there was no statistically significant relationship between high CYSLTR1 expression and disease-free survival ( Figure S1A) or overall survival (Figure S1B) (p = 0.57 and p = 0.075, respectively). However, high expression of CYSLTR2 maintains a statistically significant relationship with disease-free survival ( Figure S1C) and overall survival ( Figure S1D) (p = 0.004 and p = 0.00027, respectively). This suggests that tumour infiltrate may contribute to high CYSLTR1 expression in UM patients with reduced survival.
Following analysis of gene expression profiles for both receptors, we examined the protein expression and localisation in an independent cohort of patients. The clinical relevance of CysLT receptors in UM was evaluated by analysing the expression of CysLT 1 and CysLT 2 proteins in a tissue microarray (TMA) generated from primary UM of 52 consented patients treated at the Liverpool Ocular Oncology Centre.
To conduct manual analysis for both CysLT 1 and CysLT 2 , scores were assigned based on staining intensity (Figure 2A,D), and the percentage of tumour cells stained combined. Using the median as a cut off, a score of 0-7 was designated as low CysLT 1/2 expression, while a score of <7-12 was designated as high CysLT 1/2 expression. In Kaplan-Meier survival curves generated from median scoring of all primary UM cases, immunohistochemical levels of CysLT 1 or CysLT 2 did not demonstrate a significant association with patient survival from metastatic disease (CysLT 1 p = 0.122; HR 2.04; 95% CI 0.81-5.12, CysLT 2 p = 0.341; HR 1.15; 95% CI 0.58-2.28) ( Figure 2B,E, respectively). However, high CysLT 1 expression showed a robust trend towards reduced patient survival ( Figure 2B). In agreement with TCGA data, manual analysis of the UM TMA at the median cut-off revealed a statistically significant relationship between high CysLT 1 expression and overall survival in UM patients (p = 0.034; HR 2.34; 95% CI 1.04-5.25) ( Figure 2C). High CysLT 1 expression also had a statistically significant relationship with ciliary body involvement in the UM patient cohort (p = 0.041) ( Figure S2D). Ciliary body involvement is a feature of the disease associated with metastatic risk Cancers 2020, 12, 2950 6 of 27 and poor patient prognosis [39,40]. High expression of CysLT 2 was not significantly associated with overall survival (p = 0.697; HR 1.15; 95% CI 0.58-2.28) ( Figure 2F). There was no statistically significant relationship between high or low CysLT 1 expression and other clinical features of disease (extraocular extension, monosomy 3, periodic acid-Schiff positive loops, or epithelioid cell morphology) examined ( Figure S2D). Manual analysis did not identify a statistically significant relationship between high or low CysLT 2 expression and clinical features of the disease examined ( Figure S2D). A wider scoring range achieved by digital pathology analysis of this TMA ( Figure 3A) strengthened the relationship between high CysLT1 expression and patient survival. With a 3rd quartile segregation, high expression of CysLT1 is significantly associated with reduced melanomaspecific survival (p = 0.0012; HR 2.76; 95% CI 1.21-6.3) and reduced overall survival (p = 0.0011; HR 2.76; 95% CI 1.21-6.3) in this primary UM cohort ( Figure 3B,C, respectively). In agreement with manual analysis, there was no significant relationship between CysLT2 expression and patient outcomes ( Figure 3E,F).
Gene expression data from TCGA suggest that high expression of CYSLTR2 is significantly  A wider scoring range achieved by digital pathology analysis of this TMA ( Figure 3A) strengthened the relationship between high CysLT 1 expression and patient survival. With a 3rd quartile segregation, high expression of CysLT 1 is significantly associated with reduced melanoma-specific survival (p = 0.0012; HR 2.76; 95% CI 1.21-6.3) and reduced overall survival (p = 0.0011; HR 2.76; 95% CI Cancers 2020, 12, 2950 7 of 27 1.21-6.3) in this primary UM cohort ( Figure 3B,C, respectively). In agreement with manual analysis, there was no significant relationship between CysLT 2 expression and patient outcomes ( Figure 3E,F).
Cancers 2020, 12, x 7 of 27 CysLT1 expression and melanoma-specific survival, as well as overall survival, was confirmed by this data. Interestingly, high expression of CysLT1 was significantly associated with ciliary body involvement, further suggesting a potential link between receptor expression and patient prognosis.

Cysteinyl Leukotriene Receptors are Expressed in Primary and Metastatic Human Uveal Melanoma Cell Lines
To use UM cell lines as in vitro models to investigate the anti-cancer potential of drugs modulating CysLT signalling, we analysed the endogenous expression levels of CysLT1 and CysLT2 receptors between primary versus metastatic UM cell lines ( Figure 4). Mel285 and Mel270 are derived from primary choroidal melanomas and OMM2.5 is derived from a liver metastasis of the same patient as Mel270 [41]. Both Mel270 and OMM2.5 are GNAQ Q209P positive cell lines, while Mel285 is negative for both GNAQ and GNA11 mutations ( Figure 4E) [41]. Real-time PCR confirmed that transcripts for CysLT1 and CysLT2 are abundantly expressed in all three cell lines with no significant differences in expression levels detected ( Figure 4A,B). By Western blotting, prominent 38.5 and 39.6 kDa bands were detected for CysLT1 and CysLT2, respectively, in each cell line ( Figure 4C,D). These data confirmed that Mel270, Mel285, and OMM2.5 cell lines were appropriate to analyse the effects of CysLT receptor antagonists on UM cancer hallmarks in vitro. Gene expression data from TCGA suggest that high expression of CYSLTR2 is significantly linked to overall survival in UM patients. This was not supported by the data generated at the protein level using the UM patient TMA. Data from both TCGA and the UM patient TMA suggest a link between high CysLT 1 expression and patient survival. A significant relationship between high CysLT 1 expression and melanoma-specific survival, as well as overall survival, was confirmed by this data. Interestingly, high expression of CysLT 1 was significantly associated with ciliary body involvement, further suggesting a potential link between receptor expression and patient prognosis.

Cysteinyl Leukotriene Receptors are Expressed in Primary and Metastatic Human Uveal Melanoma Cell Lines
To use UM cell lines as in vitro models to investigate the anti-cancer potential of drugs modulating CysLT signalling, we analysed the endogenous expression levels of CysLT 1 and CysLT 2 receptors between primary versus metastatic UM cell lines ( Figure 4). Mel285 and Mel270 are derived from primary choroidal melanomas and OMM2.5 is derived from a liver metastasis of the same patient as Mel270 [41]. Both Mel270 and OMM2.5 are GNAQ Q209P positive cell lines, while Mel285 is negative for both GNAQ and GNA11 mutations ( Figure 4E) [41]. Real-time PCR confirmed that transcripts for CysLT 1 and CysLT 2 are abundantly expressed in all three cell lines with no significant differences in expression levels detected ( Figure 4A,B). By Western blotting, prominent 38.5 and 39.6 kDa bands were detected for CysLT 1 and CysLT 2 , respectively, in each cell line ( Figure 4C,D). These data confirmed that Mel270, Mel285, and OMM2.5 cell lines were appropriate to analyse the effects of CysLT receptor antagonists on UM cancer hallmarks in vitro.
Our results suggest that drugs targeting CysLT 1 , but not CysLT 2 , selectively alter primary and metastatic UM cell number in a time-and dose-dependent manner.

Quininib Drugs Inhibit Long-Term Uveal Melanoma Cell Proliferation
Clonogenic colony formation assays were conducted in Mel285 and OMM2.5 cell lines to determine if CysLT receptor drugs attenuate long-term UM cell proliferation. Mel270 cells failed to grow when seeded at the low cell density required for this assay and as a result, were excluded. UM cells were treated for 24-96 h and grown for 10 additional days prior to analysis of clone survival ( Figure 6).  Figure 5H), compared to a 95.1% reduced viability observed in Mel285 cells ( Figure 5E) and an 88.8% reduced viability in OMM2.5 cells ( Figure 5F).50 µM 1,4dihydroxy quininib had no significant effect on ARPE-19 cell number at 96 h ( Figure 5H). Our results suggest that drugs targeting CysLT1, but not CysLT2, selectively alter primary and metastatic UM cell number in a time-and dose-dependent manner.

Quininib Drugs Inhibit Long-Term Uveal Melanoma Cell Proliferation
Clonogenic colony formation assays were conducted in Mel285 and OMM2.5 cell lines to determine if CysLT receptor drugs attenuate long-term UM cell proliferation. Mel270 cells failed to grow when seeded at the low cell density required for this assay and as a result, were excluded. UM cells were treated for 24-96 h and grown for 10 additional days prior to analysis of clone survival ( Figure 6).
In summary, the CysLT 1 -selective antagonists quininib and 1,4-dihydroxy quininib are effective at inhibiting long-term proliferation of primary and metastatic UM cell lines. In both cell lines, quininib is more effective following 96 h of treatment. In Mel285 cells, 20 µM quininib is more effective following 24 h treatment while in OMM2.5 cells, 20 µM 1,4-dihydroxy quininib is more effective following 24 h treatment.  Figure S4).
The quininib drugs also produce cell line-dependent effects on the cancer secretome of UM cell lines. For example, in Mel285 cells, quininib decreases the secretion of inflammatory factors (IL-2 and IL-6) and increases the secretion of angiogenic factors (bFGF and VEGF-C). In OMM2.5 cells, a similar upregulation of angiogenic factors was observed following treatment with quininib (Flt-1 and VEGF-A), but in profound contrast quininib increased the secretion of 8 inflammatory factors (IL-10, IL-12p70, IL-13, IL-1β, IL-2, IL-6, IL-8, TNF-α) in OMM2.5 cells.
cells, 24-h treatment with 20 µM quininib (Q1) and 1,4-dihydroxy quininib (Q7) significantly reduced the levels of IL-2 (F) and IL-6 (G). Treatment with 20 µM montelukast or HAMI 3379 had no effect on the secretion of inflammatory markers in either cell line. Conditioned media were collected from three separate experiments and analysed by ELISA (n = 3). All secretions were normalised to total protein content. Statistical analysis was performed by ANOVA with Dunnett's post hoc multiple comparison test. Error bars are mean + S.E. * p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001.  and VEGF-A (D) Treatment with 20 µM montelukast or HAMI 3379 had no effect on the secretion of inflammatory markers in either cell line. Conditioned media were collected from three separate experiments and analysed by ELISA (n = 3). All secretions were normalised to total protein content. Statistical analysis was performed by ANOVA with Dunnett's post hoc multiple comparison test. Error bars are mean + S.E. * p < 0.05, ** p < 0.01, *** p < 0.001.

CysLT1 Antagonists Inhibit the In Vivo Growth of Uveal Melanoma Zebrafish Xenografts
To determine if the in vitro attenuation of UM cell proliferation with CysLT1 antagonists could be reproduced in vivo, we sought to generate mouse and zebrafish UM cell xenografts [49]. In our experience, Mel285 cell lines were not amenable to generate murine xenograft models following subcutaneous, intraocular, intrahepatic, or intravenous implantation. In contrast, the metastatic OMM2.5 cell line produced tumours following subcutaneous, intraocular, or intrahepatic implantation. However, this required lengthy growth times of 7-8 months for the initial subcutaneous cell suspension injection, and 3-4 months for propagation from corresponding subcutaneous fragment implants ( Figure S5). OMM2.5 cells implanted intraocularly give rise to ocular tumours 3-4 months post cell suspension injection ( Figure S5A,B). Similarly, OMM2.5 cells implanted intrahepatically allow tumour growth 3-4 months after cell suspension injection (Figure Data are expressed as mean + SEM. Statistical analysis was carried out using a paired t-test to compare within the same cell line. Data was normalised to cell number, as assessed by crystal violet assay. * p < 0.05; ** p < 0.01; *** p < 0.001.

CysLT 1 Antagonists Inhibit the In Vivo Growth of Uveal Melanoma Zebrafish Xenografts
To determine if the in vitro attenuation of UM cell proliferation with CysLT 1 antagonists could be reproduced in vivo, we sought to generate mouse and zebrafish UM cell xenografts [49]. In our experience, Mel285 cell lines were not amenable to generate murine xenograft models following subcutaneous, intraocular, intrahepatic, or intravenous implantation. In contrast, the metastatic OMM2.5 cell line produced tumours following subcutaneous, intraocular, or intrahepatic implantation. However, this required lengthy growth times of 7-8 months for the initial subcutaneous cell suspension injection, and 3-4 months for propagation from corresponding subcutaneous fragment implants ( Figure S5). OMM2.5 cells implanted intraocularly give rise to ocular tumours 3-4 months post cell suspension injection ( Figure S5A,B). Similarly, OMM2.5 cells implanted intrahepatically allow tumour growth 3-4 months after cell suspension injection ( Figure S5C,D), and 1-2 months after the re-implantation of tumour fragments arising ( Figure S5E). Histological examination of the tumours confirmed the presence of a UM cell phenotype ( Figure S5F-H). Further characterisation of these murine UM xenografts models will ascertain their value in researching the therapeutic effect of UM treatments.
The zebrafish xenograft models proved to be more time-and cost-effective and thus allowed investigation of the effects of the CysLT 1 antagonists on UM cell lines in vivo. Mel285 and OMM2.5 UM cell lines were injected into the perivitelline space or eye of 2 dpf (days post-fertilisation) zebrafish larvae. Following injection, the larvae were treated with the maximum tolerated dose of quininib (3 µM), 1,4-dihyroxy quininib (10 µM) or montelukast (20 µM). Montelukast significantly reduced the tumour size of Mel285 xenografts propagated in the perivitelline space (p = 0.0498, 14.6% reduction) or eye (p < 0.0001, 32.4% reduction) ( Figure 10A). In contrast, montelukast had negligible effect on OMM2.5 xenografts in the perivitelline space but modestly reduced (p = 0.0439 and 25.6% reduction) tumour size in the eye ( Figure 10B). In relation to quininib and 1,4-dihydroxy quininib, the most significant reductions in tumour size were also observed with the Mel285 cells line xenografts in the eye (p < 0.0001 and 26.2% reduction versus p < 0.0001 and 21.7% reduction, respectively) ( Figure 10A). Quininib and 1,4-dihydroxy quininib had negligible effects on OMM2.5 xenografts into the perivitelline space, whereas 1,4-dihydroxy quininib modestly reduced (p = 0.0258 and 18.4% reduction) OMM2.5 xenograft growth in the eye ( Figure 10B). In summary, CysLT 1 antagonists can inhibit the in vivo growth of UM cancer cells; however, the effects are more pronounced with the primary Mel285 compared to the metastatic OMM2.5 cell line. Similarly, the drugs seem to have a more pronounced effect on cells implanted into the eye, the physiologically relevant organ, rather than those implanted into the perivitelline space.
Cancers 2020, 12, x 15 of 27 S5C,D), and 1-2 months after the re-implantation of tumour fragments arising ( Figure S5E). Histological examination of the tumours confirmed the presence of a UM cell phenotype ( Figure S5F-H). Further characterisation of these murine UM xenografts models will ascertain their value in researching the therapeutic effect of UM treatments. The zebrafish xenograft models proved to be more time-and cost-effective and thus allowed investigation of the effects of the CysLT1 antagonists on UM cell lines in vivo. Mel285 and OMM2.5 UM cell lines were injected into the perivitelline space or eye of 2 dpf (days post-fertilisation) zebrafish larvae. Following injection, the larvae were treated with the maximum tolerated dose of quininib (3 µM), 1,4-dihyroxy quininib (10 µM) or montelukast (20 µM). Montelukast significantly reduced the tumour size of Mel285 xenografts propagated in the perivitelline space (p = 0.0498, 14.6% reduction) or eye (p < 0.0001, 32.4% reduction) ( Figure 10A). In contrast, montelukast had negligible effect on OMM2.5 xenografts in the perivitelline space but modestly reduced (p = 0.0439 and 25.6% reduction) tumour size in the eye ( Figure 10B). In relation to quininib and 1,4-dihydroxy quininib, the most significant reductions in tumour size were also observed with the Mel285 cells line xenografts in the eye (p < 0.0001 and 26.2% reduction versus p < 0.0001 and 21.7% reduction, respectively) ( Figure 10A). Quininib and 1,4-dihydroxy quininib had negligible effects on OMM2.5 xenografts into the perivitelline space, whereas 1,4-dihydroxy quininib modestly reduced (p = 0.0258 and 18.4% reduction) OMM2.5 xenograft growth in the eye ( Figure 10B). In summary, CysLT1 antagonists can inhibit the in vivo growth of UM cancer cells; however, the effects are more pronounced with the primary Mel285 compared to the metastatic OMM2.5 cell line. Similarly, the drugs seem to have a more pronounced effect on cells implanted into the eye, the physiologically relevant organ, rather than those implanted into the perivitelline space.

Discussion
We identified that high CysLT 1 expression is associated with survival of primary UM patients and that antagonism of CysLT 1

alters cancer hallmarks in UM cells in vitro and in vivo.
To our knowledge, this is the first study to investigate the link between the expression of CysLT receptors in UM patient samples and associated clinical data. Gene expression data from TCGA suggest that high expression of CYSLTR1 or CYSLTR2 is significantly linked to disease-free and overall survival in UM patients. This is supported by the known importance of the CysLT 2 /G αq/11 /PLCB4 pathway in UM oncogenesis. CYSLTR2 acts as an oncogene [12], albeit in a small subset of UM. Activation of this receptor, and the associated downstream signalling pathways, are identified drivers of disease progression in early UM [50].
TCGA data suggest that high expression of CYSLTR1 is associated with a poor patient prognosis in UM. This finding was supported by the data generated at the protein level using the patient UM TMA, whereby CysLT 1 was significantly associated with reduced melanoma-specific survival and reduced overall survival in a primary UM patient cohort. CysLT 1 was highly expressed in primary UM, with 52/52 tissue samples staining positive for the receptor.
CysLT 2 was highly expressed in all primary UM examined, with 51/51 tissue samples staining positive for the receptor. All patient samples stained were positive for both CysLT 1 and CysLT 2 . Both receptors are not highly expressed in normal uveal tract [51], which suggests a role in early carcinogenesis. Of note, high CysLT 1 expression is associated with poor prognosis and reduced survival in both colorectal [28] and breast cancer [29], while CysLT 2 has been reported to have an anti-tumorigenic effect in colorectal cancer [52].
Interestingly, high expression of CysLT 1 was also significantly associated with ciliary body involvement, a feature associated with an increased risk of metastasis [53]. This further suggests a potential link between CysLT 1 expression and patient prognosis. Given their link to the MAPK pathway, and the known importance of this pathway in UM, it is not surprising that CysLT receptors are highly expressed in primary UM. Similarly, based on the biology of CysLT receptors it is unsurprising that high expression of both CysLT 1 and CysLT 2 in UM are associated with significant alterations in inflammation and angiogenesis. Indeed, their high level of expression and link to prognostic and clinical features of the disease may suggest that they are associated with malignant transformation of the tumour. Thus, further investigation using additional UM patient samples is warranted to determine if high CysLT 1 or CysLT 2 expression can be statistically associated with patient prognosis and metastatic disease development.
As CysLT receptors are druggable G-protein coupled receptors, we hypothesised that pharmacological antagonists of these receptors may attenuate cancer phenotypes of UM cells in culture. qPCR and western blot analysis confirmed equivalent expression of both receptors in Mel285 and OMM2.5 cell lines suggesting that phenotypic differences observed following drug treatment are not due to differences in CysLT receptor expression. The CysLT 2 antagonist HAMI 3379 was previously tested in Leu129Gln CysLT 2 UM oncogene models but showed limited activity as an inverse agonist in these models of constitutive CysLT 2 activation [54]. To our knowledge, no previous studies have investigated the anti-cancer potential of CysLT 1 antagonists in UM. However, montelukast exerts anti-cancer activity in chronic myeloid leukaemia [55], colorectal cancer [56], lung cancer [57], and is chemopreventative against 14 different cancer types [23]. Likewise, quininib and 1,4 dihydroxy quininib have significant anti-cancer properties in colorectal cancer models [34,35]. Here, we uncover that CysLT 1 antagonists, but not CysLT 2 antagonists, produce significant anti-cancer effect in primary and metastatic UM cell lines through the inhibition of cell survival and cell proliferation.96 hours of drug treatment in short-term cell viability and long-term cell proliferation assays revealed interesting results. The CysLT 2 antagonist HAMI 3379 had no effect on the viability of the primary or metastatic UM cell lines. All three CysLT 1 antagonists significantly reduced viability of the primary and metastatic UM cells, with no intra-drug differences across the cell lines, but with quininib being the most potent in both cell lines. Dacarbazine, a chemotherapeutic clinically used in treatment of metastatic UM, did not reduce the viability of either UM cell line. This is consistent with clinical findings wherein dacarbazine did not offer any survival advantage in treating metastatic disease [6] or in the adjuvant setting after primary tumour resection [58]. Of note, here, all CysLT 1 drugs tested performed significantly better than dacarbazine. In long-term proliferation assays, the CysLT 2 antagonist HAMI 3379 again exerted negligible effect on the primary or metastatic UM cell lines. The CysLT 1 antagonists quininib and 1,4-dihydroxy quininib significantly attenuated proliferation of both cell lines with a seemingly higher efficacy in the primary UM cell line. The same effects of CysLT 1 antagonists are not observed in ARPE-19 cells, suggesting that this is a specific effect in UM cells.
Here, we identify quininib and 1,4-dihydroxy quininib to significantly alter the secretion of cancer-associated inflammatory and angiogenic factors from UM cell lines. In agreement, the anti-angiogenic and anti-inflammatory actions of CysLT 1 antagonists are known in other contexts [32,33,59]. CysLTs modulate vascular permeability through upregulation of VEGF expression [60] and CysLT 1 antagonists modulate vascular permeability through reduction of VEGF expression in mice [61] and in human asthmatic patients [62]. Likewise, the eyes of UM patients often contain increased levels of inflammation-related cytokines in the aqueous humour [63] and several in vitro studies show the expression of soluble inflammatory factors in UM. Our TCGA data analysis also suggests that high expression of CysLT receptors is significantly linked to pro-inflammatory and pro-angiogenic pathways. In relation to secreted inflammatory and angiogenic factors, again montelukast and HAMI 3379 had negligible effects. It has been reported that blocking pro-angiogenic isoforms of VEGF, through inhibition of SRPK1, inhibits melanoma tumour growth in vivo [64]. Quininib and 1,4-dihydroxy resulted in significant, but dramatically differential effects on the cell lines and factors. For example, quininib treatment was associated with a significant increase in VEGF-C secretion from Mel285 cells, but not from OMM2.5 cells. 1,4-dihydroxy quininib had negligible effects on VEGF-C secretion, but significantly increased Flt-1 in Mel285 in contrast to quininib which had no effect. Interestingly, Flt-1 negatively modulates angiogenesis via its actions as a decoy receptor by trapping VEGF and preventing its binding to VEGFR-2 [65]. Further investigation is required to determine if the drug-induced changes in angiogenic factor expression are cause-or-effect for the corresponding reduction in UM cell viability and proliferation.
Profound differences in the secretion of inflammatory markers was also observed with the quininib drugs. For example, quininib significantly reduced the secretion of IL-2 in Mel285 cells, but significantly increased IL-2 secretion from OMM2.5 cells, whereas 1,4 -dihydroxy quininib exerted negligible effects on IL-2 in OMM2.5 cells. Particularly noticeable is the significantly increased secretion of seven inflammatory factors in OMM2.5, but not Mel285, after quininib, but not 1,4-dihydroxy quininib, treatment. Notably, 24 h treatment with quininib reduced the survival and proliferation of Mel285 but not OMM2.5 cells. This is potentially due to a resistance associated up-regulation of inflammatory factors in OMM2.5 cells. Indeed, the basal expression of IL-13 and IL-8 is much higher in OMM2.5 than Mel285 cells. The pattern of IL-2 and IL-6 secretion may explain the similar actions of quininib drugs on proliferation in the Mel285 cells. UM cell lines express IL-2R, and production of the IL-2 ligand by tumour infiltrating lymphocytes and macrophages stimulates tumour cell proliferation [66]. IL-6 stimulates tumour cell proliferation and survival through the inhibition of apoptosis and interference with IL-6R signalling leads to decreased UM cell viability [67]. Quininib and 1,4 dihydroxy quininib result in 3-5-fold reductions in secreted levels of IL-2 and IL-6. Changes in IL-2 and IL-6 are not sufficient to explain effects on the OMM2.5 cells, but anti-proliferative effects in these cells may be confounded by the up-regulation of multiple other inflammatory factors.
The importance of dysregulated metabolism in the initiation and progression of cancer is well understood. Although much is known about metabolic rewiring in cutaneous melanoma, few studies focus solely on the metabolic underpinnings of UM. Oxidative phosphorylation is upregulated in invasive melanoma [68] and the metabolic switch of some melanomas to oxidative phosphorylation has been linked to resistance to inhibitors of the MAPK pathway [69]. While CysLT are predominantly known for their role in inflammation and angiogenesis, they are linked to alterations in respiratory activity. LTD 4 increases mitochondrial metabolic activity and mitochondrial gene transcription in human intestinal epithelial cells and colorectal cancer cells [70]. Significantly decreased oxidative phosphorylation was observed in Mel285 and OMM2.5 cells with quininib or 1,4-dihydroxy quininib and OMM2.5 cells with montelukast. Modulation of oxidative phosphorylation controls proliferation of tumour cells [71] which may explain the effect of CysLT 1 antagonists on UM cells.
In vivo models are preferable to human cell lines to study the complexity of the tumour microenvironment. Zebrafish xenograft models have been established as robust preclinical models in which experimental drugs can be tested [49]. To this end, we created xenograft models using UM cell lines, to determine if the effects of CysLT 1 antagonists on UM cell survival and proliferation can be recapitulated in a more complex system. Our in vitro data are supported by the results generated in zebrafish models whereby CysLT 1 antagonist drugs significantly inhibit the growth of both Mel285 and OMM2.5 zebrafish xenograft models. Interestingly, CysLT 1 antagonists have a greater effect in zebrafish orthoxenograft models, in which the cells are implanted into the corresponding anatomical location. Given that the Mel285 cell line originated in the eye, and the OMM2.5 cell line originated in the liver, this may explain the more substantial effect observed in the Mel285 orthoxenograft model. To further validate our results, we have generated an OMM2.5 cell line-derived orthotopic xenograft mouse model of UM. This and patient-derived UM xenografts are considered the most appropriate pre-clinical models for future studies to evaluate the potential of CysLT 1 antagonist as therapeutics for UM.

MTT Assay
3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) dye determined cytotoxic effects in cell lines. Cells were trypsinised using trypsin-EDTA (0.05%) (ThermoFisher Scientific, Rockford, IL, USA) and centrifuged at 1200 rpm for 5 min at RT. Cell pellets were re-suspended in complete medium and cells seeded into 96-well plates at 5000 cells/well. After 24 h adherence, cell medium was removed and replaced with the desired drug concentration. 0.5% DMSO in RPMI 1640 (UM cells) or DMEM (ARPE-19 cells) was used as a vehicle control. Cells were incubated for 24 and 96 h with drugs. Drug solution was removed, and the wells washed with PBS before adding 90 µL of serum-free medium and 10 µL MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) dye to each well. The plate was covered and incubated for 2.5 h at 37 • C. Then, 100 µL of 100% DMSO was added to each well to dissolve the formazan crystals. Absorbance values at 570 nm were determined using a SpectraMax ® M2 microplate reader (Molecular Devices Corporation, Sunnyvale, CA, USA).

Colony Formation Assay
1.5 × 10 3 (Mel285) or 9 × 10 3 (OMM2.5) cells were seeded per well of a 6-well plate and allowed adhere for 24 h. Cells were treated with the desired concentration of drug for 24 or 96 h. 0.5% DMSO was used as a vehicle control. Following treatment, the drug solution was removed, and cells grown in complete medium for 10 days. Clones were fixed with 4% paraformaldehyde for 10 min and stained using 0.5% crystal violet (Pro-Lab diagnostics, Wirral, UK #PL700) for 2 h at RT. Clone counting was performed using the GelCount™ system (Oxford Optronix, Abingdon, UK).

Statistical Analysis for Drug Treatment Experiments
Statistical analysis applied GraphPad Prism 7 software (GraphPad, San Diego, CA, USA). Specific statistical tests used are indicated in figure legends. All data are presented as mean ± standard error of the mean (SEM). For all statistical analysis, differences were considered statistically significant at p < 0.05. For MTT assay drug treatments, statistical analysis was performed by ANOVA with Dunnett's post hoc multiple comparison test. Error bars are mean + S.E. * p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001. For clonogenic assay drug treatments, statistical analysis was performed by ANOVA with Dunnett's post hoc multiple comparison test. Error bars are mean + S.E. * p < 0.05; *** p < 0.001; **** p < 0.0001

CysLT 1 and CysLT 2 qPCR
Total RNA was extracted from UM cells using the mirVana™ miRNA Isolation Kit (ThermoFisher Scientific, Rockford, IL, USA) as per the manufacturer's instructions. Briefly, cells were trypsinised and washed by gently resuspending in 1 mL of PBS and pelleting at 1200 rpm prior to lysis and total RNA isolation. Following isolation, total RNA concentration was quantified at 260 nm (Spectrophotometer ND-2000, Thermo Scientific, Wilmington, DE, USA) and samples stored at −80 • C. cDNA was synthesized with the SuperScript II Reverse Transcriptase system (Invitrogen, Carlsbad, CA, USA) or the TaKaRa PrimeScript™ RT Reagent Kit (Takara Bio Europe, Saint-Germain-en-Laye, France), using random hexamers as per the supplier's instructions.

The Cancer Genome Atlas Gene Expression Analysis
Gene expression and clinical data from 80 primary UM included in The Cancer Genome Atlas (TCGA-UM dataset) were collected from the GDC data portal through the R package "TCGAbiolinks". RNA-seq data were downloaded in Fragments Per Kilobase of exon per million fragments Mapped (FPKM) and then converted to log2 scale. The associations between CYSLTR1 and CYSLTR2 gene expression and prognosis were assessed by Cox proportional hazard regression models, adjusted by sex and age (Table S1). To resolve a violation of the proportional hazard assumption, the association between CYSLTR2 gene expression and overall survival ( Figure 1E) was stratified by time from 0-20 months and beyond 20 months and analysed by the likelihood ratio (LHR) test (Table S1). Disease-Free Survival (DFS) and Overall Survival (OS) were used as end points. The third quartile was used as an optimal cut-off to divide samples into High and Low expression categories. Survival probabilities were plotted on a Kaplan-Meier curve and a Log-rank test was used to compare the two groups. Survival analysis was performed with R package "survminer". Disease-free survival is defined as time to metastatic recurrence. Overall survival is defined as death by any cause.
Gene Set Variation Analysis was performed to calculate enrichment scores in functions and pathways "Inflammatory response", "INF-γ", "Glycolysis", "Oxidative Phosphorylation", "TNF-α", "Angiogenesis", and "GPCR signalling" (R package "GSVA"). They were manually selected from the Molecular Signatures Database (MSigDB) which includes gene sets from Hallmarks and Biocarta-curated pathways. Samples were divided by the third quartile gene expression values of CYSLTR1 and CYSLTR2. For each score obtained, differences were assessed using a non-parametric Wilcoxon test. Differences were considered statistically significant when p-value < 0.05.
To infer contributions from stromal infiltration, MCP-counter and ESTIMATE tools were used [72]. MCP-counter (Microenvironment Cell Populations-counter) uses gene expression profiles to quantify the relative abundance of a series of immune and non-immune cell types (T cells, cytotoxic T cells, NK cells, B lineage, monocytic lineage, dendritic cells, neutrophils, endothelial cells, and cancer associated fibroblasts). Estimation of STromal and Immune cells in MAlignant Tumor tissues using Expression data (ESTIMATE) is a tool that infers tumour purity and stromal/immune infiltration using gene expression data. Spearman correlation tests were performed to assess correlation between CYSLTR1 and CYSLTR2 gene expression and the corresponding MCP-counter scores. Stromal scores from ESTIMATE were used to adjust CYSLTR1 and CYSLTR2 expression values and recalculate the survival models ( Figure S1).

Ethics
This study conformed to the principles of the Declaration of Helsinki and Good Clinical Practice guidelines. Approval for the study was obtained from the Health Research Authority (NRES REC ref 16/NW/0380) on the 16th of May 2016, and all patients provided informed consent.
All experiments involving the use of rodents were approved by the Ethical Committee of Animal Experimentation of the Parc Científic de Barcelona (PCB) under the procedure number 9928-P1 approved by the Generalitat de Catalunya.

Tissue Samples
A tissue microarray (TMA) was generated from primary UM samples of 52 consented patients treated at the Liverpool Ocular Oncology Centre, with the primary UM samples being stored within the Liverpool Ocular Oncology Biobank (HTA Licence 12020 and HRA REC 16/NW/0380). Milton Keynes, UK) and a detection kit (Bond Polymer Refine Red Detection Kit; Leica Biosystems, Inc., Buffalo Grove, IL, USA) as previously described [73]. Slides were counterstained with hematoxylin and mounted using DPX mountant (Sigma-Aldrich, St. Louis, MO, USA). Colorectal cancer tissue served as the positive control ( Figure S2A); negative control was omission of the primary antibody ( Figure S2A). Slides were scanned using a slide scanner (Aperio CS2; Leica Biosystems, Inc., Buffalo Grove, IL, USA) and analysed with imaging software (Aperio Image Scope version 11.2; Leica Biosystems, Inc., Buffalo Grove, IL, USA). Each core was scored based on intensity (0, 1, 2, or 3) and percentage of tumour cells stained. The final score was calculated using the following equation: (scoring intensity × % of cells stained)/n number of samples [73]. The IHC-stained slides were scored by three independent investigators (S.E.C., H.K., K.S.). Melanoma-specific survival is defined as death from metastatic melanoma. Overall survival is defined as death by any cause.

Digital Slide Scanning and Automated Image Analysis
Slides were scanned with an Aperio AT2 digital slide scanner (Leica Biosystem, Milton Keynes, UK) with a 20× lens, and the quality of the images was checked manually before the application of the digital algorithm. Automated digital image analysis was performed using the Visiopharm Integrator System (Visiopharm, Hoersholm, Denmark). A cytoplasmic algorithm from the ONCOTOPIX module  Figure 3A).

Statistical Analysis for Immunohistochemistry
Bivariate analysis of high expression of CysLT 1 /CysLT 2 associated with melanoma-specific survival or overall survival was undertaken using the Cox proportional hazards model. Survival time (years) was calculated from the date of first diagnosis until death, or study closure on 29 May 2019. All analyses were carried out using SPSS Statistics v.24 (IBM, Armonk, NY, USA).

Mel285 and OMM2.5 ELISA
Cells were seeded at 1.5 × 10 5 cells per well of a 6-well plate and allowed adhere overnight. Cells were treated with 20 µM of quininib, 1-4-dihydroxy quininib, montelukast, HAMI 3379, or 0.5% DMSO as control. All treatments were conducted in duplicate. Following 24 h treatment, 1 mL of media was removed from each well and stored at −20 • C. Media were processed according to the Meso Scale Discovery (MSD) multiplex protocol. To assess angiogenic and inflammatory secretions from cell conditioned media, a 17-plex ELISA kit separated across three plates was used (Meso Scale Diagnostics, Rockville, MD, USA). The multiplex ELISA determined the secreted levels of; IFN-γ, IL-10, IL-12p70, IL-13, IL-1β, IL-2, IL-4, IL-6, IL-8, TNF-α, bFGF, Flt-1, PIGF, Tie-2, VEGF-C, VEGF-D, and VEGF-A in cell conditioned media. Assays were run as per the manufacturer's recommendation; an overnight supernatant incubation protocol was used for the Pro-inflammatory Panel 1 with the Angiogenesis Panel 1 assay being run on the same day protocol. Cell-conditioned media were run undiluted on all assays as per previous optimisation experiments. Secretion data for all factors were normalised to cell lysate protein content (extracted as described above) by using a BCA protein assay kit (ThermoFisher Scientific, Rockford, IL, USA).

Zebrafish Breeding and Maintenance
All experiments carried out on animals were granted ethical approval by Linköping Experimental Animal Research Ethics Committee under the ethical approval number N89/15. Zebrafish were maintained in a 14-h light, 10-h dark cycle in a recirculating water system at 28 • C. Larvae were produced through natural spawning and maintained as previously described [74].

Zebrafish Cell Line Xenografts
Implantation of Mel285 or OMM2.5 cells into zebrafish embryos followed published protocols [75]. Briefly, cells were labelled for 30 min at 37 • C in 6 mg/mL DiI (Sigma-Aldrich, St. Louis, MO, USA) in PBS followed by washing 3× in PBS. Labelled cells were implanted in the perivitelline space or vitreous of 48 hpf (hours post-fertilisation) Tg(fli1a:EGFP) y1 zebrafish embryos, maintained from the 8-cell stage in 0.003% PTU-containing E3-water (5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl 2 , 0.33 mM MgSO 4 ). Approximately, 200-500 cells in 2-5 nL were implanted in each embryo (20 embryos per group). Embryos were transferred to individual wells of 24-well plates containing 0.5 mL quininib, 1-4-dihydroxy quininib or montelukast at 10 mM final concentration in E3-PTU water. Tumour-bearing embryos were imaged using a fluorescent microscope (SMZ1500, Nikon, Tokyo, Japan) soon after implantation. Embryos with tumour cells erroneously implanted in the yolk, brain, or circulation were removed. Embryos bearing tumours in the perivitelline space or the vitreous were incubated at 36 • C for three days and re-imaged. Relative change in tumour volume was evaluated as the size of the tumours at three days post-implantation (3 dpi) relative to the size immediately after implantation (at 0 dpi).

Seahorse Metabolism Measurements
Mel285 and OMM2.5 were seeded in four wells per treatment group at a density of 12 × 10 3 and 14 × 10 3 cells per well, respectively, in a 24-well cell culture XFe24 microplate (Agilent Technologies, Santa Clara, CA, USA) at a volume of 100 µL RPMI and allowed to adhere at 37 • C and 5% CO 2 for 5 h; then an additional 150 µL of RPMI was added. Twenty-four hours following seeding, cells were treated with 20 µM of quininib, 1-4-dihydroxy quininib or montelukast along with 0.1% and 0.2% DMSO controls. Twenty-four hours following drug treatment, cells were washed with unbuffered DMEM supplemented with 10 mM glucose, 10 mM sodium pyruvate (pH 7.4), and incubated for 1 h at 37 • C in a CO 2 -free incubator. The oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured using a Seahorse Biosciences XFe24 Extracellular Flux Analyser (Agilent Technologies, Santa Clara, CA, USA). Three basal measurements of OCR and ECAR were taken over 24 min consisting of three repeats of mix (three min)/wait (2 min)/measurement (3 min) to establish OCR measurement. Three additional measurements were obtained following the injection of three mitochondrial inhibitors including oligomycin (2 µg/mL) (Sigma-Aldrich, St. Louis, MO, USA), an uncoupling agent carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone (FCCP) (5 µM) (Sigma-Aldrich, St. Louis, MO, USA), and antimycin-A (2 µM) (Sigma-Aldrich, St. Louis, MO, USA), and ATP turnover was calculated by subtracting the OCR post oligomycin injection from baseline OCR prior to oligomycin addition. Proton leak was calculated by subtracting OCR post antimycin-A addition from OCR post oligomycin addition. Maximal respiration was calculated by subtracting OCR post antimycin addition from OCR post FCCP addition. Non-mitochondrial respiration was determined as the OCR value post antimycin-A addition. All measurements were normalised to cell number using the crystal violet assay, transferring the eluted stain to a 96-well plate before reading.

Conclusions
There is an overwhelming, unmet clinical need for targeted therapies for the treatment of UM. The cysteinyl leukotrienes are established as regulators of inflammation and have recently emerged as novel regulators of angiogenesis, two key processes in UM. For the first time, we examine the clinical relevance of CysLT 1 and CysLT 2 expression and in primary UM samples and show a link between high CysLT 1 expression and primary UM patient survival. Our data highlight the involvement of both receptors with clinical features of their disease and reinforces their link to inflammatory and angiogenic pathways. We determined that antagonist drugs of CysLT 1 , but not CysLT 2 , inhibit the survival and proliferation of primary and metastatic UM cells in a specific, time-and dose-dependent manner. This effect is recapitulated in in vivo zebrafish cell line xenograft models. Antagonism of CysLT 1 in UM cells leads to alterations in the secretion of pro-inflammatory and pro-angiogenic secretions and a decrease in oxidative phosphorylation. The importance of the cysteinyl leukotriene receptor signalling pathway, and antagonism of CysLT 1 , in UM should be further explored.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6694/12/10/2950/s1, Figure S1: Analysis of TCGA data using MCP-Counter and ESTIMATE, Figure S2: Patient TMA control tissue. Core-core correlations of manual versus digital pathology analysis. Clinical characteristics of patients included in the TMA, Figure S3: Cell line comparison data for the clonogenic assay, Figure S4: Factors unchanged following treatment in Mel285 and OMM2.5 cells, Figure S5: Generation of OMM2.5 cell line-derived orthotopic xenograft models of UM, Table S1: Statistical analysis of TCGA UM dataset.