Clinicopathological and Prognostic Significance of PRAME Overexpression in Human Cancer: A Meta-Analysis

Numerous studies have demonstrated that preferentially expressed antigen in melanoma (PRAME) is abnormally expressed in various solid tumours. However, the clinicopathological features and prognostic value of the PRAME expression in patients with cancer remain unclear. Accordingly, we performed a meta-analysis to accurately assess the association of the expression level of PRAME with clinicopathological features and cancer prognosis. Relevant study collection was performed in PubMed, Web of Science, and Embase until 28 February 2020. A total of 14 original studies involving 2,421 patients were included. Our data indicated that the PRAME expression was significantly associated with tumour stage (OR = 1.99, 95% CI: 1.48–2.67, P < 0.001) and positive lymph node metastasis (OR = 3.14, 95% CI: 1.99–4.97, P < 0.001). Pooled results showed that overexpression of PRAME is positively correlated with poor disease-free survival (HR = 1.60, 95% CI: 1.36–1.88, P < 0.001), progression-free survival (HR = 1.88, 95% CI: 1.02–3.46, P = 0.042), metastasis-free survival (HR = 1.86, 95% CI: 1.05–3.31, P = 0.034), and overall survival (HR = 1.75, 95% CI: 1.53–1.99, P < 0.001). In summary, these data are suggesting that PRAME is tumorigenic and may serve as a prognostic biomarker for cancer.


Background
Cancer has become a major public health problem worldwide. According to a previous study, approximately 1,735,350 new cancer cases and 609,640 million cancer deaths occurred in the United States in 2018 [1]. Despite medical and scientific efforts over the past decades, the overall 5-year survival rate remains low because of the malignant progression of tumours [1]. Biomarkers have become important tools for tumour diagnosis, and they also serve as therapeutic targets and can be used to predict clinical outcomes [2]. Therefore, identifying potential novel biomarkers is imperative for refining and optimizing diagnosis, treatment, and prognosis.
Cancer/testis antigens (CTAs) represent a heterogeneous group of tumour antigens and are regarded as ideal potential biomarkers for detection and targets in cancer immunotherapy because of their restricted expression pattern and immunogenicity [3]. In recent years, several CTAs have been studied as target antigens in vaccine clinical trials for various cancer types [4]. Preferentially expressed antigen in melanoma (PRAME) is one of the most immunogenic CTAs discovered to date. The PRAME protein was first described as a tumour antigen in human melanoma by Ikeda et al. [5]. Subsequently, numerous studies reported that overexpression of PRAME was significantly correlated with clinicopathological features in malignant tumours, including medulloblastoma [6], hepatocellular carcinoma [7,8], adenocarcinoma [9], uveal melanoma [10,11], high-grade serous cancer [12], myxoid liposarcoma [13], diffuse large B-cell lymphoma [14], osteosarcoma [15], bladder cancer [16], breast carcinoma [17,18], and neuroblastoma [19]. However, the consistency of the clinicopathological features and prognostic value of PRAME remains unclear.
To our knowledge, this meta-analysis is the first complete overview to clarify the clinicopathological features and prognostic value of PRAME based on the information from all previously published studies.

Search Strategy.
In accordance with the PRISMA statement, we conducted a systematic literature search of PubMed, Web of Science, and Embase using "PRAME" OR "preferentially expressed antigen in melanoma" as keywords. The search concluded in this study was updated in February 2020. We manually reviewed the references of the retrieved articles to identify potentially eligible studies.

Selection Criteria.
Eligible articles met the inclusion criteria based on the following: (1) articles assessed the prognostic or clinical value of PRAME in solid tumours, (2) articles published in English, and (3) articles provided odds ratios (ORs) or information that allowed manual calculation of the 95% confidence intervals (CIs). The articles excluded were (1) duplicated publications; (2) reviews, comments, letters, and conference abstract; (3) and articles without usable data.

Data Extraction.
Two investigators extracted the data from selected studies independently, including the first author's name, year of publication, country, cancer type, number of cases, detection method, tumour stage, number of patients with positive PRAME and negative PRAME expression, HRs, and 95% CIs for DFS, PFS, MFS, and OS. The HRs and 95% CIs were directly extracted from the adequate information in the article; otherwise, we used the Engauge Digitizer version 4.1 software to estimate the survival data from Kaplan-Meier curves.

Statistical
Analysis. The quality of the selected studies was independently evaluated by two reviewers using the Newcastle-Ottawa Scale (NOS). All statistical analyses were performed using the Stata statistical software version 14.0 (Stata Corporation, College Station, TX, USA). Heterogeneity among studies was assessed using Cochrane's Q tests (chi-squared tests) and the I 2 statistic. The fixed-effects model was adopted when there was no obvious heterogeneity (P > 0:1 and I 2 < 50%). Otherwise, a random-effects model was applied. Egger's test was used to assess publication bias. Sensitivity analysis was conducted to investigate the stability of the total pooled results by removing studies one by one. P < 0:05 was considered statistically significant.

Study Selection and Characteristics.
The flow diagram is shown in Figure 1; a total of 1,149 articles were identified for initial evaluation, and 1,087 manuscripts were excluded from analysis because of irrelevant titles or duplicates. We excluded 32 citations based on their abstracts, leaving 30 studies for further full-text review. After further reading of the studies, we excluded 16 studies because of insufficient survival data. Ultimately, 14 studies meeting the selection criteria were finally included in this meta-analysis.
The main features of the 14 eligible studies are summarized in Table 1. The publication years of the eligible studies ranged from 2004 to 2019. The total number of patients was 2,421 with a mean of 172.9 (range, 51-576 patients). Among the 14 studies, one focused on medulloblastoma, two on hepa-tocellular carcinoma, one on adenocarcinoma, two on uveal melanoma, one on high-grade serous cancer, one on myxoid liposarcoma, one on diffuse large B-cell lymphoma, one on osteosarcoma, one on bladder cancer, two on breast carcinoma, and one on neuroblastoma. Eleven of the studies evaluated PRAME at the gene level, while four conducted evaluation at the protein level. Patients were dichotomized into two groups with positive and negative PRAME expression. The quality of the selected studies was assessed by using the Newcastle-Ottawa Scale and found to range from 6 to 8, indicating that the studies were of good quality.

PRAME Expression and Clinicopathological Parameters.
Nine articles investigated the associations between PRAME expression and tumour stage. There were 604 cases of III-IV stage and 559 cancer cases of I-II stage ( Table 1). Since the low heterogeneity (I 2 = 43:4%, P = 0:079), we used the fixedeffects model to pool data. The stratified data showed that the PRAME overexpression was significantly associated with tumour stage (OR = 1:99, 95% CI: 1.48-2.67, P < 0:001) ( Figure 2(a)). The results suggest that the increased expression of PRAME was markedly higher in the III-IV stage group than in the I-II stage group.
The results indicate that the increased expression of PRAME was markedly higher in the positive lymphatic metastasis group than in the negative lymphatic metastasis group.

Correlation between the PRAME Expression and
Disease-Free Survival (DFS). There were four studies, comprising a total of 858 patients, provided data for us to analyse the correlation between PRAME and DFS. Since the low heterogeneity (I 2 = 28:9%, P = 0:239), we used the fixed-effects model to pool data. As seen in Figure 3(a), the data indicated that the overexpression of PRAME had an obvious impact on DFS (HR = 1:60, 95% CI: 1.36-1.88, P < 0:001). The overall results suggest that PRAME overexpression is an indicator of poor DFS in patients with cancer.

Correlation between the PRAME Expression and
Progression-Free Survival (PFS). There were three studies, comprising a total of 603 patients, provided data for us to analyse the correlation between PRAME and PFS. Since the low heterogeneity (I 2 = 0:0%, P = 0:375), we used the fixedeffects model to pool data. As seen in Figure 3(b), the data indicated that the overexpression of PRAME had an obvious impact on PFS (HR = 1:88, 95% CI: 1.02-3.46, P = 0:042). The overall results suggest that PRAME overexpression is an indicator of poor PFS in patients with cancer.

Correlation between the PRAME Expression and
Metastasis-Free Survival (MFS). There were four studies, comprising a total of 689 patients, provided data for us to analyse the correlation between PRAME and MFS. Since the obvious heterogeneity (I 2 = 89:7%, P ≤ 0:001), we used the random-effects model to pool data. As seen in Figure 3(c), the data indicated that the overexpression of PRAME had an obvious impact on MFS (HR = 1:86, 95% CI: 1.05-3.31, P = 0:034). The overall results suggest that PRAME overexpression is an indicator of poor MFS in patients with cancer.

Correlation between the PRAME Expression and Overall
Survival (OS). There were twelve studies, comprising a total of 1,978 patients, provided data for us to analyse the correlation between PRAME and OS. Since the low heterogeneity (I 2 = 46:5%, P = 0:038), we used the fixed-effects model to pool data. As seen in Figure 3(d), the data indicated that the overexpression of PRAME had an obvious impact on OS (HR = 1:75, 95% CI: 1.53-1.99, P < 0:001). The overall results suggest that PRAME overexpression is an indicator of poor OS in patients with cancer.
3.3. Publication Bias. Publication bias was evaluated by funnel plot and Egger's test. Because of the small sample size of LNM, DFS, PFS, and MFS, we had not conducted publication bias analysis. The results indicated no significant publication bias was observed in this analysis (Figure 4).

Sensitivity
Analysis. Sensitivity analysis was performed to assess the influence of each study omission on the overall outcome. The results suggested that there was no significant influence of the pooled HR by omitting any study, indicating our analysis was robust ( Figure 5).

Discussion
There is growing evidence that the overexpression of PRAME can contribute to the differences in the prognostic outcome between different kinds of solid tumors. To the best of our knowledge, this study describes the first meta-analysis to analyse the association between the level of PRAME and Full-text articles excluded: without usable data (n = 16) Figure 1: Flow diagram of the study selection in the meta-analysis. In this study, we included 14 retrospective studies with a total of 2,421 patients. Remarkable positive associations were identified between the PRAME expression and clinicopathological characteristics, including advanced clinical stage (P < 0:001) and positive lymph node metastasis (P < 0:001).

BioMed Research International
The results of the overall analysis revealed that the overexpression of PRAME is positively related to poor DFS (P < 0:001), PFS (P = 0:042), MFS (P = 0:034), and OS (P < 0:001). The results of the heterogeneity in the subgroup analyses also showed that our results were stable for all vari-ables. Therefore, these results indicated that PRAME might serve as a clinicopathological and prognostic biomarker for malignancy. Numerous studies have reported PRAME expression in various malignancies. In solid malignancies, including hepatocellular carcinoma [7,8], uveal melanoma [10,11], osteosarcoma [15], bladder cancer [16], and breast carcinoma [18], high PRAME expression correlates with advanced-stage disease and poor survival, whereas in pediatric acute leukemia, PRAME overexpression was found to predict good outcome [20][21][22]. Steinbach et al. [21] found that the overexpression of PRAME was found in 62% (n = 31) of acute myeloid   5 BioMed Research International leukemia patients, and the rates of OS and DFS were higher than in patients with no or low expression (P < 0:05). Moreover, Abdelmalak et al. [20] reported that positive PRAME expressers had a statistically longer DFS and OS (P < 0:001, <0:001, respectively) compared with negative PRAME expressers. These studies suggest that PRAME overexpression is a predictor for better prognostic outcome in acute leukemia, which is completely opposite to the prognostic results of solid tumors in our meta-analysis. However, largescale and prospective cohort studies will ultimately be needed to validate the good prognostic outcome of PRAME overexpression in acute leukemia.
As we all know, DNA methylation is an epigenetic mechanism often affecting gene expression and modifying the function of the genes. PRAME has been reported to be epigenetically regulated by DNA methylation in malignancies. Field et al. [11] found that PRAME is aberrantly hypomethylated and transcriptionally activated in uveal melanomas and is associated with increased metastatic risk. Moreover, PRAME is frequently expressed in epithelial ovarian cancer at the mRNA and protein levels, and DNA methylation is a key mechanism regulating its expression [12]. Additionally, Schenk et al. [23] reported that changes in the methylation pattern in defined parts of the regulatory regions of PRAME   are sufficient for its upregulation in cells. Together, these studies show that the PRAME expression is associated with aberrant hypomethylation of the PRAME promoter and may have therapeutic implications. To note, most of the studies included in this meta-analysis have not identified the association between PRAME promoter DNA hypomethylation and its expression, which warrants further investigations. Although the function of PRAME has not yet been fully elucidated, a number of studies have addressed this issue. It is known that PRAME can bind to the retinoic acid receptor in the presence of retinoic acid, which suggests that it serves as a transcription regulator of nuclear receptor signaling [24]. PRAME knockdown was shown to decrease the proliferation of cancer cells [7,15,24,25]. Tan et al. [15] found that PRAME siRNA knockdown significantly suppressed the proliferation, colony formation, and G1 cell cycle arrest in primary osteosarcoma. Oehler et al. [25] reported that PRAME inhibits myeloid tumor cell differentiation in leukemic progenitor cells. Downregulation of PRAME suppresses proliferation and promotes apoptosis in hepatocellular carci-noma through the activation of the P53-mediated pathway [7]. Additionally, Orlando et al. [6] reported that high PRAME mRNA expression correlates significantly with a worse OS and PRAME-specific TCR may represent a promising innovative approach for treating medulloblastoma patients. To date, PRAME is one of the most immunogenic CTAs discovered thus far, regarding as an attractive potential immunotherapy target [4,[26][27][28]. The vaccination approaches using PRAME as the target are currently undergoing clinical trials (trial numbers NCT01149343, NCT01853878, and NCT00423254) [27].
There were several limitations in this study that should be acknowledged. Firstly, the articles included were from only three databases (PubMed, Web of Science, and Embase); thus, the data collection may be incomplete. Secondly, different criteria were applied in these studies for defining PRAME positive or negative, because of the lack of uniform cut-off values in PRAME expression. Thirdly, HRs with 95% CIs were calculated by digitizing and extracting from the survival curves in several papers, which inevitably brought minor  Meta-analysis estimates,given named study is omitted Zhu (2018) Pan (2017) Gezgin (2017) Oyama (2016) Mitsuhashi (2014) Tan (2012) Dyrskjt (2012) Doolan (2008) Oberthuer ( (2018) Pan (2017) Gezgin (2017) Oyama (2016) Mitsuhashi (2014) Tan (2012) Doolan (2008) Domennico (2018) Field (2016) Zhang (2016) lura (2015) Epping (2008) Meta-analysis estimates,given named study is omitted 7 BioMed Research International statistical deviations. Furthermore, the effects of some factors, such as age and gender, were not considered in this analysis because of insufficient data. Therefore, high-quality studies are urgently needed to draw more accurate conclusions.
Our meta-analysis provided credible evidence that the overexpression of PRAME was significantly related to the TNM stage, LNM, and poor prognosis. In addition, PRAME might serve as an attractive therapeutic target in the treatment of malignant tumors. However, considering the limitations of individual study, large-scale and prospective cohort studies will ultimately be needed to validate the results of our study.