Wilms' tumor 1 (WT1) expression and prognosis in solid cancer patients: a systematic review and meta-analysis

Though proposed as a promising target antigen for cancer immunotherapy, the prognostic value of Wilms' tumor 1 (WT1) in solid tumors remains inconclusive. Here, we report a systematic review and meta-analysis of the association between WT1 expression and prognosis in solid tumors. PubMed, Web of Science and Google Scholar were searched to identify studies exploring the impact of WT1 on clinical outcomes, including overall survival (OS), disease-specific survival (DSS), disease-free survival (DFS), relapse/recurrence-free survival (RFS) or progression-free survival (PFS), in solid cancer patients. Hazard ratio (HR) and 95% confidence interval (CI) were applied to assess the strength of these associations. Finally, a total of 29 eligible studies with 4090 patients were identified for qualitative analysis, and 22 studies with 3620 patients were enrolled for quantitative synthesis. Overall, positive expression of WT1 was significantly associated with worse OS (metaHR = 1.48, 95% CI = 1.11–1.97) and DFS/RFS/PFS (metaHR = 2.14, 95% CI = 1.42–3.21). Subgroup analyses showed that WT1 positive expression could independently predict unfavorable DFS/RFS/PFS (metaHR = 1.86, 95%CI = 1.04–3.35). In summary, our study suggests that WT1 may be a potential marker to predict DFS/RFS/PFS in solid tumor patients. Further studies are needed to confirm the role of WT1 expression in clinical practice.

C ancer is a major cause of mortality worldwide 1 . It is well-known that the balance between oncogenes and tumor suppressor genes is a prime determinant of normal development versus tumorigenesis 2,3 . Therefore, identification and characterization of these genes is of primary importance in understanding the onset and progression of cancer, and ultimately leading to recognition of potential markers and specific targets for prevention and individualized treatment of cancer.
The Wilms' tumor 1 (WT1), located at chromosome 11p13, was first cloned in 1990 as a suppressor in Wilms' tumor 4 . Subsequent research indicated that WT1 might play an oncologic role in hematologic malignancies and a variety of solid tumors, including leukemia, breast cancer, ovarian cancer, glioblastoma and soft tissue sarcoma, etc [5][6][7][8][9] . It has been shown that WT1 was overexpressed in a number of cancer cells, and more importantly, knockdown of WT1 by shRNA could induce mitochondrial damage and then inhibit malignant cell growth 10 . In addition to its prognostic role in a variety cancer types [11][12][13][14][15] , WT1 was also recognized as a promising target for immunotherapy based on its unique features 16,17 , and preliminary results from clinical vaccine trials revealed WT1's untapped potential to induce cancer immunity with minimal side effects [18][19][20] .
Despite the clinical development of WT1-targeted therapies, the prognostic value of WT1 expression in solid tumors remains unclear. It would clearly be desirable to explore whether the tumors manifesting WT1 expression are associated with worse outcome in patients with solid tumors. Therefore, we conducted the present systematic review and meta-analysis in order to appraise the prognostic value of WT1 in solid tumors with the goal of gaining insights into the clinical implications.

Results
Description of Studies. A total of 29 studies 9,11-15,21-43 with 32 datasets and 4090 patients assessing the relationship between WT1 expression and prognosis of solid tumors were initially identified (Figure 1), and the details of selected studies are shown in Table 1. Twenty-one datasets evaluated carcinoma (9 for ovarian cancer, 5 for breast cancer, 3 for non-small-cell lung cancer [NSCLC], 2 for endometrial cancer, 1 for colorectal cancer and 1 for hepatocellular carcinoma) and 11 datasets evaluated non-carcinoma (2 for soft tissue sarcoma). Twenty-nine datasets reported an adverse impact of WT1 positive expression on prognosis regardless of the statistic power, while, 3 datasets (2 for NSCLC and 1 for soft tissue sarcoma) reported opposite results. Twenty-four datasets used antigen-based methods (21 for immunohistochemistry[IHC], 2 for ELISA and 1 for western blot), and 8 datasets used mRNA-based method (5 for reverse transcription polymerase chain reaction[RT-PCR] and 3 for gene microarray). Thirteen datasets were carried out in Europe, 14 in Asia and 5 in North of America. Twenty datasets had information on overall survival (OS), 5 had information on diseasespecific survival (DSS) and 18 had information on disease-free survival (DFS)/relapse or recurrence-free survival (RFS)/progression-free survival (PFS). The median follow-up period of the 12 studies with definite follow-up time was 47 months.
For quantitative synthesis, 3 studies 38,40,41 were excluded due to limited sample size ,40, and 2 studies 42,43 were rejected for insuf-ficient data, 1 study 36 was excluded for unclear data. Finally, 1 study 27 and 1 dataset 32 were rejected because the samples reported were not tissues-based. Therefore, the consolidated sample for our analysis consisted of 22 studies with 24 datasets. Among the 24 datasets, 14 addressed OS, 5 referred to DSS, and 12 pertained to DFS/RFS/PFS. A total of 3620 patients were enrolled in these studies, and the median sample size was 99, which was used as the criterion for subgroup analysis by sample size.
Publication bias. The shapes of the funnel plots did not reveal evidence of obvious asymmetry for OS ( Figure 5A), DSS ( Figure 5B) and DFS/RFS/PFS ( Figure 5C) analyses, which was confirmed by the Begg's and Egger's tests ( Table 2). The results of publication bias for studies using antigen-based methods to detect WT1 expression are shown in Table 3.
Sensitivity analysis. Sensitivity analysis was performed by sequential removal of individual studies to assess the influence of each study on the metaHRs. Results showed that metaHRs were not statistically altered (data not shown).

Discussions
The WT1 gene, which encodes a protein consisting of four zinc finger domains at the C terminus and an glutamine and proline-rich domain at the N terminus, plays an important role in cell proliferation, differentiation, apoptosis, organ development and the maintenance of several adult tissues 44,45 . Though recognized as a classic tumor suppressor gene in Wilms' tumor, there is a growing body of evidence demonstrating that wild-type WT1 is expressed in a variety of tumors arising from different tissues that normally do not express WT1 [46][47][48] . It is possible that WT1 could inhibit cell apoptosis by transcriptional activation and/or upregulation of proto-oncogenes 10 . Recent work has also revealed that WT1 is a key regulator in overcoming senescence downstream of KRAS signaling 49 , and is involved in the apoptotic response to cytotoxic stress 50 , further demonstrating its oncogenic effects. Moreover, WT1 could also promote invasion, migration and metastasis 43,51,52 , facilitate angiogenesis 53,54 and confer drug resistance to cancer cells 55,56 .
Because of its implications in the etiology of cancer, clinical values of WT1 expression are the subject of increasing scrutiny. To date, the role of WT1 expression as a potential prognostic marker has been extensively explored in a number of solid tumors as listed in Table 1. In addition, considering its unique features 16,18 , much attention has been given to the use of WT1 peptides in eliciting an immune response. In phase I and II clinical trials for solid tumors, two patients obtained a complete remission, three had a partial remission, and five showed a decrease in tumor marker or tumor size 18 . Nevertheless, the clinical implications of WT1 expression are still largely unknown, and the prognostic value of WT1 expression in solid tumors remains controversial. Therefore, we conducted the present systemic review and meta-analysis to clarify this issue and to provide more evidence for facilitating the development of anti-WT1 target research.
The present study included a total of 29 eligible studies with 32 datasets and 4090 patients for qualitative analysis. Ovarian cancer ranked first among the reported cancer types, and IHC was most commonly used for detection. For most of these studies, positive expression of WT1 was linked with an unfavorable impact on the prognosis. However, two of three studies for NSCLC (one using ELISA, and the other one using RT-PCR) revealed that patients with high WT1 expression had a better prognosis than controls, which might be attributed to differences in detection methods as well as cellular context 49 . For meta-analysis, we excluded studies using ELISA as a detection method because the source WT1 IgG antibody in plasma was largely unclear and controversial 27,32 . Finally, 22 studies with 24 datasets and 3620 patients were subjected to quantitative analysis. Overall, WT1 positive expression showed significant association with worse OS and DFS/RFS/PFS, and borderline association with poor DSS, in solid cancer patients. Multivariate analyses indicated that WT1 positive expression was an independent unfavorable predictor for DFS/RFS/PFS in solid cancer patients, and the correlation became more prominent if we only focused on WT1 protein expression. Taking together, we can infer that WT1 may be of limited prognostic value for all-caused and disease-specific mortality in solid tumor patients. However, it could be a potentially useful marker to predict the risk of relapse and progression for these patients.
Since ovarian cancer was the most common cancer type reported in the current meta-analysis, we conducted overall and subgroup (uni/multivariate) analyses to explore the correlations of WT1 expression with prognosis in these patients. Results showed that WT1 positive expression could only predict poor outcomes for ovarian cancer in univariate model, but not in multivariate model. Restricting studies to antigen-based methods did not alter the results, which indicated that WT1 was not an independent prognostic marker for ovarian cancer. In fact, two large scale studies with a sample size $500 11,24 also found that WT1 might be of limited prognostic value in ovarian cancer patients under multivariable model. There are several distinct subtypes of ovarian cancer, which develop differently and respond differently to chemotherapy 11 . At the same time, WT1 was expressed differently among these subtypes, mostly in serous carcinoma 11,22,24,29,33 . Therefore, the negative prognostic significance of WT1 in the entire cohort might reflect subtype differences in expression. It is obviously desirable to investigate the prognostic value of WT1 in each subtype individually rather than treating ovarian cancer as a homogeneous malignancy; however, since it is not feasible at present to secure the specific data regarding WT1 expression status and patients' survival data in each ovarian cancer subtype, we were unable to do this potentially important individualized analysis. Therefore, studies attempting to clarify the specific association between WT1 expression and prognosis in subtypes of ovarian cancer are strongly encouraged, as they offer the distinct possibility of advancing our insights into the clinical implications of WT1 expression in ovarian cancer.
It should be noted that there are some limitations to the analyses presented here. First, publication bias is a concern, because more positive results tended to be published, thus potentially exaggerating the association between WT1 expression and poor outcomes. Second, despite intensive effort, it was not possible to communicate with some investigators as we attempted to indentify all relevant date. In addition, because of limited number of studies, we were unable to conduct analyses for certain types of cancer, such as non-carcinoma cancer, and subtypes of specific cancer, such as ovarian cancer; in addition, we were also not allowed to do subgroup analyses specific to DFS/RFS/PFS. Third, some HRs were estimated indirectly as previously reported 57 , however, these data were less reliable than direct data from the original literature. Finally, though subgroup analysis by detection method was done, we were unable to carry out stratified  analysis according to cutoff values of WT1 expression owing to great methodological variation among studies we selected here. Clearly, it is urgent to establish a uniform and validated method for assessment of WT1 expression to reduce the substantial heterogeneity currently present in clinical reporting protocols.
In summary, our study suggests that WT1 may be a potential marker to predict the risk of relapse and progression in solid cancer patients, but the indicative value of WT1 for all-caused and diseasespecific mortality in these patients might be limited. Further data from large multicenter prospective studies are needed to validate the clinical importance of WT1.

Methods
This study was performed according to the guidelines and recommendations for improving the quality of reporting of medical research such as REMARK and PRISMA 58,59 .
Search strategy. A systematic literature search was performed to identify articles regarding WT1 and prognosis of solid tumors. The Pubmed, Web of Science, Google Scholar were used simultaneously, with the combination of terms ''WT1 or Wilms' tumor 1 or Wilms' tumor gene 1 or Wilms' tumor protein 1 or Wilms' tumor suppressor gene 1'' and ''survival or prognosis or mortality or death'' and ''neoplasm or cancer or tumor or carcinoma'' (up to 8 August 2014). All articles were initially reviewed by abstract and title browsing to select the relevant reports, which were subjected to further screening.
Study selection criteria. Studies that were included in the systemic review had to meet all of the following criteria: 1) the publication explored the relation between WT1 expression and solid tumor prognosis, such as OS, DSS, DFS, RFS and PFS, 2) if one study reported multi-datasets based on different detection methods, cutoff values or populations, datasets would been recognized individually, and 3) publication language was confined to English. For qualitative meta-analysis, additional criteria were needed: 1) there were sufficient, clear, and available data to extract or estimate 57 the individual HR and 95% CI, 2) sample size was 40 or more, 3) WT1 expression was detected in cancer tissues.
Data extraction. Three authors (X.W. Qi, F. Zhang and H. Wu) independently extracted information using predefined criteria. The following information of each study was collected: first author, reference year, patient source, cancer type, number of patients, detection method, number and percentage of WT1 positive expression and its cutoff value, median and range of follow-up time, outcome endpoint, univariate or multivariate HR and 95%CI for WT1 positive expression (exposed group) versus WT1 negative expression (unexposed group). If univariate and multivariate HR and 95%CI were both reported, we preferred the multivariate result. For the studies with inadequate or unclear information, the authors were contacted by E-mail if possible. However, information from those authors was not forthcoming.
Statistical analysis. MetaHR and 95% CI were applied to assess the association between WT1 expression and outcomes of solid cancer patients. Outcome endpoints were divided into three groups, OS, DSS and DFS/RFS/PFS, based on the data acquired in the current study and previous report 60 . An observed metaHR .1 implied increased hazard of all-caused mortality for OS, disease-specific mortality for DSS, and relapse/progression for DFS/RFS/PFS, respectively, in patients with WT1 positive expression. Subgroup analyses were done according to cancer type (carcinoma and non-carcinoma; for carcinoma, ovarian cancer, breast cancer, endometrial cancer and others), uni/multivariate model, detection method (antigenbased and mRNA-based), sample size (,median value and $median value) and patient source (Europe, Asia and North of America). Because it was the most common method used to detect WT1 expression, we conducted overall and subgroup analyses restricted to studies with antigen-based methods. In addition, we also did subgroup analysis by uni/multivariate approaches to explore whether WT1 was an independent prognostic factor for patients with ovarian cancer, since it was the most extensively studied type of solid cancer in the current meta-analysis. Heterogeneity assumption was assessed by chi-based Q-test and I 2 test. The heterogeneity was considered statistically significant if P , 0.05. If heterogeneity across studies was not identified, the fixed-effects model (Mantel-Haenszel) was used. Otherwise, the random effects model (DerSimonian and Laird) was used. Publication bias was analyzed by performing funnel plots qualitatively, and estimated by Begg's and Egger's test quantitatively. Sensitivity analysis was carried out by removing each study at a time to evaluate the stability of the results. Two-sided P , 0.05 was considered statistically significant. The metaHRs and funnel plots were generated by Review Manager Version 5.3 (The Nordic Cochrane Centre, Copenhagen, Denmark). The analysis of publication bias and sensitivity was performed using STATA version 11.0 (StataCorp, College Station, TX, USA).