High expression of lncRNA PVT1 independently predicts poor overall survival in patients with primary uveal melanoma

The plasmacytoma variant translocation 1 gene (PVT1) plays an oncogenic role in the initiation and progression of multiple cancers. In this study, by using deep-sequencing data and follow-up data in the Cancer Genome Atlas-Uveal melanomas (TCGA-UVM), we assessed the association between the expression of PVT1 and clinicopathological characteristics of patients with uveal melanoma, the mechanism of its dysregulation and its prognostic value. Results showed that high PVT1 expression group had a higher proportion of epithelioid cell dominant disease (a more malignant histological subtype than spindle cell dominant disease) and more cases of extrascleral extension (a risk factor for metastasis) compared with the low PVT1 expression group. 61 out of 80 cases (76.3%) of primary uveal melanoma had PVT1 amplification in TCGA-UVM. In addition, PVT1 expression was strongly and negatively correlated with its methylation status (Pearson's r = -0.712, Spearman’s r = -0.806). By performing univariate and multivariate analysis, we found that high PVT1 expression was an independent predictor of poor OS in patients with uveal melanoma (HR: 12.015, 95%CI: 1.854–77.876, p = 0.009). Based on these findings, we infer that PVT1 expression is modulated by both DNA amplification and methylation and its expression might serve as a valuable and specific prognostic biomarker in terms of OS in uveal melanoma.


Introduction
Long noncoding RNAs (lncRNAs) are a class of RNA that are longer than 200 nucleotides and do not code for proteins [1]. Previous studies found that this class of RNAs play a pivotal role in regulating gene expression at both transcriptional and post-transcriptional levels [1]. Uveal melanoma, which is also known as ocular melanoma is the most common primary intraocular cancer in adults. Some recent studies found that dysregulated lncRNAs are involved in the pathological development of uveal melanoma [2,3]. For example, hypermethylated in cancer 1 (HIC1) can induce uveal melanoma progression by activating lncRNA-numb [4]. LncRNA CASC15-New-Transcript 1(CANT1) acts as a necessary suppressor of uveal melanoma via triggering the expression of lncRNA JPX and FTX and subsequently inducing the expression of PLOS  lncRNA XIST [5]. HOXA11-AS can increase uveal melanoma cell growth and invasion by interacting with enhancer of zeste homolog 2 (EZH2) to suppress its target p21 protein expression and by sponging miR-124 [3]. The plasmacytoma variant translocation 1 gene (PVT1) has been demonstrated as an oncogenic lncRNA in multiple cancers, including ovarian cancer [6], breast cancer [7], prostate cancer [8], cervical cancer [9], gastric cancer [10] and non-small cell lung cancer [11]. In gastric cancer, high PVT1 expression is an independent prognostic marker for poor overall survival (OS) and disease-free survival (DFS) [10].
PVT1 overexpression promotes melanoma cells proliferation, cell cycle progression, and migration [12]. Mechanistically, PVT1 directly sponges miR-26b, which had been verified as a tumor suppressor in melanoma [13]. These findings suggest that PVT1 may also act as an oncogene in melanoma. However, its association with uveal melanoma, as well as its prognostic value and the mechanism of its dysregulation in uveal melanoma have not been explored. In this study, by using deep-sequencing data and follow-up data in the Cancer Genome Atlas-Uveal melanomas (TCGA-UVM), we found that PVT1 expression is modulated by both DNA amplification and methylation and its high expression independently predicts poor OS in patients with primary uveal melanoma.

Data mining in the Cancer Genome Atlas-Uveal melanomas (UVM)
The level 3 data of patients with primary uveal melanoma in TCGA-UVM or with primary skin melanoma in TCGA-SKCM were downloaded by using the UCSC Xena browser (https:// xenabrowser.net). Heatmap showing PVT1 copy number alterations, RNA expression and DNA methylation (450k) was generated. Regression analysis of the correlation between PVT1 RNA expression and its DNA methylation was examined by using cBioPortal for Cancer Genomics (http://www.cbioportal.org/) [14,15].
Kaplan-Meier curves of overall survival (OS) were generated by GraphPad Prism v6.0 (GraphPad Software Inc.). Patients were grouped according to median PVT1 expression or median PVT1 DNA methylation.

Statistical analysis
Statistical analysis was performed by using SPSS 19.0 (SPSS Inc.) and GraphPad Prism v6.0. Comparison of the clinicopathological features between high and low PVT1 expression groups was performed using χ 2 tests. Log-rank test was performed to assess the difference between the survival curves. Prognostic values were analyzed by univariate and multivariate Cox regression models. Welch's t-test was conducted to compare PVT1 expression between different copy number alterations. P < 0.05 was considered to be statistically significant.

High expression of lncRNA PVT1 is associated with malignant behaviors of uveal melanoma
The association between PVT1 expression and the clinicopathological parameters was summarized in Table 1 The expression of PVT1 is modulated by DNA amplification and methylation in uveal melanoma By using the deep sequencing data in TCGA, we tried to explore the mechanisms of PVT1 dysregulation in uveal melanoma. Gene-level thresholded GISTIC2-processed copy-number data, which defines genetic changes as homozygous deletion (-2), heterozygous loss (-1), copy-neutral (0), low-level copy gain (+1), high-level amplification (+2) were downloaded from the Xena browser. Among the 80 cases of primary uveal melanoma, 14 cases (17.5%) had PVT1 high-amplification (+2) and 47 cases (58.8%) had amplification (+1) (Fig 1A). The amplification was associated with significantly higher expression of PVT1 RNA ( Fig 1B). However, no significant difference was observed between the +2 and +1 group ( Fig 1B). Then, we characterized the correlation between PVT1 expression and its DNA methylation (Fig 1A and 1C). Heatmap and following regression analysis revealed a strong negative correlation between PVT1 expression and DNA methylation (Pearson's r = -0.712, Spearman's r = -0.806) (Fig 1A  and 1C). High expression of PVT1 independently predicts poor OS in patients with primary uveal melanoma Since the expression of PVT1 was associated with malignant behaviors of uveal melanoma, we determine to assess its prognostic value. By generating Kaplan-Meier curves of OS, we found that high PVT1 expression was associated with significantly shorter OS (p<0.0001) (Fig 2A). Interestingly, as a mechanism of PVT1 dysregulation, low PVT1 DNA methylation was also associated with unfavorable OS (p<0.0001) (Fig 2B). In univariate analysis, we found that epithelioid cell dominant uveal melanoma, extrascleral extension, high PVT1 expression and low PVT1 DNA methylation were associated with unfavorable OS (

PVT1 expression is not associated with OS in skin cutaneous melanoma
To further verify the specificity of the prognostic value of PVT1 in uveal melanoma, we also examined its expression profile and the association with OS in patients with primary skin cutaneous melanoma in TCGA. DNA copy number alteration data (N = 366) were only available in patients with metastatic skin melanoma (N = 368). PVT1 DNA amplification was less common in metastatic skin melanoma (190/366, 51.9%) than in uveal melanoma (61/80, 76.3%) (Fig 3A). 21 cases even had PVT1 heterozygous loss (Fig 3A). Elevated PVT1 transcription was also observed in DNA amplification group (Fig 3B). In comparison, heterozygous loss did not necessarily result in PVT1 decrease (Fig 3B). DNA methylation was weakly and negatively correlated with PVT1 expression in skin melanoma (Pearson's r = -0.352, Spearman's r = -0.480) (Fig 3C). 357 metastatic patients and 102 primary patients with intact OS data were included in Table 3. According to the best cut-off of PVT1 expression, these patients were divided into high PVT1 (N = 230, which include 163 metastatic cases and 67 primary cases) and low PVT1 (N = 229, which include 194 metastatic cases and 35 primary cases) expression groups. The association between PVT1 expression and the clinicopathological parameters in this group of patients was summarized in Table 3. The high PVT1 expression group had a higher ratio of primary tumor (67/230) compared with the low PVT1 expression group (35/229) (p = 0.0004) ( Table 3). No significant difference in the other parameters was observed between the high and low PVT1 expression groups (Table 3). Log-rank test of OS curves indicated that PVT1 expression was not related to OS, no matter in metastatic melanoma (p = 0.13, Fig 4A) or in primary melanoma (p = 0.98, Fig 4B).

Discussion
PVT1 has been demonstrated as an oncogenic lncRNA that is usually upregulated in cancer tissues compared with normal tissues. It exerts a carcinogenetic effect via various epigenetic mechanisms such as regulating transcription activity and via acting as miRNAs sponges. For example, it can promote non-small cell lung cancer cell proliferation by recruiting EZH2 to the large tumor suppressor kinase 2 (LATS2) promoter and represses LATS2 transcription [11]. It suppresses miR-146a expression by inducing the methylation of CpG island in its promoter in prostate cancer cells [16]. It can also sponge miR-186 in gastric cancer cells [17], miR-448 in pancreatic cancer cells [18], miR-26b in melanoma cancer cells [13], miR-203 in esophageal squamous cell carcinoma cells [19], thereby contributing to malignant behaviors of these cancers, such as enhanced proliferation, migration, and metastasis. One recent meta-analysis based on 1,443 patients from 15 previous studies found that increased PVT1 expression was significantly associated with positive lymph node metastasis, positive distant metastasis, advanced tumor-node-metastasis stage and poor differentiation grade, but was not related to tumor size in some cancers [20]. In this study, we found that high PVT1 expression was associated with a higher proportion of epithelioid cell dominant disease (a more malignant histological subtype than spindle cell dominant disease) and more cases of lncRNA PVT1 independently predicts poor overall survival of uveal melanoma extrascleral extension (a risk factor for metastasis), suggesting that high PVT1 expression may confer some malignant phenotypes to uveal melanoma. However, we did not find a significant difference in tumor size and thickness between the high and low PVT1 expression groups, which were in consistent with the findings in the meta-analysis.
The mechanisms of PVT1 dysregulation in these cancers are quite complex and far from being fully understood. PVT1 locates at 8q24 in the human genome, a region that is usually amplified in some cancers [8,21]. In gastric cancer, FOXM1 can bind to the promoter region of PVT1 and enhance its transcription [22]. Upregulation of SOX2 can activate PVT1 expression in breast cancer cells via binding to its promoter and promote breast cancer cell growth and invasion [23]. These findings indicate that dysregulated PVT1 may be caused by both genetic and epigenetic alterations. By examining copy number alterations in TCGA-UVM, we found that 61 out of 80 cases (76.3%) of primary uveal melanoma had PVT1 amplification. In addition, the amplification was associated with significantly higher PVT1 RNA expression. These findings supported our hypothesis that genetic amplification is a mechanism of aberrant PVT1 expression in uveal melanoma. Interestingly, by analyzing the DNA methylation status of PVT1, we observed a strong negative correlation between PVT1 expression and its methylation (Pearson's r = -0.712, Spearman's r = -0.806), suggesting that DNA methylation status can also influence PVT1 expression in uveal melanoma. In addition, we also observed different levels of copy number alterations and methylation status between uveal melanoma and skin cutaneous melanoma, which indicate that PVT1 dysregulation might be cancer-specific. Several previous studies also observed the promising prognostic value of PVT1 in multiple types of cancer. In gastrointestinal cancers, elevated PVT1 expression was significantly related to poor OS, DFS, disease-specific survival (DSS) and relapse-free survival (RFS) [10,24]. Its overexpression also serves as an independent prognostic indicator for the OS of patients with small cell lung cancer [25]. By performing univariate and multivariate analysis, we found that high PVT1 expression was an independent predictor of poor OS in patients with uveal melanoma (HR: 12.015, 95%CI: 1.854-77.876, p = 0.009). In comparison, although it acts as an oncogene in skin melanoma, it had no prognostic value in terms of OS. Therefore, the expression of PVT1 might serve as a valuable and specific prognostic biomarker in uveal melanoma.
This study also has some limitations. The most important one is that some patients' characteristics (such as pigmentation) and treatment information was not recorded in the database. In fact, pigmentation has a critical role in melanoma biology [26]. It affects melanoma patients' survival, radiotherapy, chemotherapy and immune therapy [27][28][29][30]. Melanogenesis could simulate HIF-1α expression, thereby conferring malignant behaviors of melanoma cells [30,31]. Inhibition of melanogenesis might enhance the efficacy of radiotherapy and chemotherapy in advanced melanomas [32,33]. Another limitation is the relatively small sample size in the database (N = 80). In addition, due to insufficient data, we were unable to assess the association between PVT1 and DFS among the patients. Therefore, it is meaningful to assess its independent prognostic value in uveal melanoma in a larger cohort in the future.

Conclusion
Aberrant PVT1 expression is associated with malignant behaviors of uveal melanoma and might independently predict poor OS.