Prognostic value of decreased FOXP1 protein expression in various tumors: a systematic review and meta-analysis

The prognostic value of forkhead box protein P1 (FOXP1) protein expression in tumors remains controversial. Therefore, we conducted a systematic review and meta-analysis, searching the PubMed, Embase and Web of Science databases to identify eligible studies. In total, we analyzed 22 articles that examined 9 tumor types and included 2468 patients. Overall, decreased expression of FOXP1 protein was associated with favorable overall survival (OS) in lymphoma patients (HR = 0.38, 95%CI: 0.30–0.48, p < 0.001). In patients with solid tumors, decreased FOXP1 expression correlated with unfavorable OS (HR = 1.82, 95%CI: 1.18–2.83, p = 0.007). However, when FOXP1 protein expression was nuclear, decreased expression was also associated with favorable OS (HR = 0.53, 95%CI: 0.32–0.86, p = 0.011). Furthermore, decreased FOXP1 expression resulted in the best OS in patients with mucosa-associated lymphoid tissue (MALT) lymphomas (HR = 0.26, 95%CI: 0.11–0.59, p = 0.001), but the worst OS was observed in non-small cell lung cancer (NSCLC) patients (HR = 3.11, 95%CI: 1.87–5.17, p < 0.001). In addition, decreased FOXP1 expression was significantly correlated with an unfavorable relapse-free survival (RFS) in breast cancer patients (HR = 1.93, 95%CI: 1.33–2.80, p = 0.001).

Scientific RepoRts | 6:30437 | DOI: 10.1038/srep30437 Search strategy. We systematically searched in the online PubMed, Embase and Web of Science databases (updated until May 6, 2016) with the restrictions of English language and article format. The following keywords or their combinations were used in the searches: "FOXP1 OR forkhead box protein 1" AND "survival OR prognosis OR prognostic" AND "cancer OR tumor OR tumour OR neoplasm OR neoplasma OR neoplasia OR carcinoma OR cancers OR tumors OR tumours OR neoplasms OR neoplasmas OR neoplasias OR carcinomas OR leukemia OR leukemias OR leukaemia OR leukaemias OR lymphoma OR lymphomas". Additional studies were identified by referring to relevant articles to avoid omissions due to electronic searching.

Study selection criteria.
Eligible studies in our meta-analysis were selected according to the following criteria: (1) full text original studies published in English that measured the FOXP1 protein expression in patients with tumors without restricting the type of cancer; (2) the protein expression was determined by immunohistochemistry (IHC); (3) results included the determination of a correlation between FOXP1 expression and patient survival; (4) the hazard ratios (HRs) and their corresponding 95% confidence intervals (CIs) were either reported or calculated using other information (e.g., survival curves); and (5) when repeated results were reported by the same authors, we included the most complete report. However, patient survival outcomes in this meta-analysis included overall survival (OS), cancer-specific survival (CSS), relapse-free survival (RFS), progression-free survival (PFS), disease-free survival (DFS) and failure-free survival (FFS, which was defined as in Nyman's study 20 that evaluated survival from the date of diagnosis until relapse or death of any cause). Additionally, unpublished studies, meeting abstracts, comments, letters, case reports, literature reviews and meta-analyses were excluded. Quality assessment. In correspondence to a critical review checklist that was proposed by Meta-analysis of Observational Studies in Epidemiology (MOOSE) group issued by Dutch Cochrane Centre 21 and referencing Zhou's study 22 , we used the following quality control criteria: (1) specific definition of study population; (2) specific description of study design; (3) sample size greater than 30; (4) specific definition of survival outcome such as OS, CSS, RFS, PFS, DFS and FFS; (5) specific definition of the cut-off value for decreased FOXP1 protein expression; and (6) sufficient follow-up time.
Data extraction. Two investigators (Jian Xiao and Bixiu He) independently extracted the primary information according to a predefined form, which included the following sub-categories: first author, year of publication, country of study population, tumor type, sample source, test method, location of FOXP1 protein expression, cut-off value, sample size, follow-up time, survival outcome, analysis method and HR estimation. When both multivariate and univariate analyses of the OS results were performed, HRs and their corresponding 95%CIs were extracted preferentially from the multivariate analyses. If HR and its corresponding 95%CI were not directly reported, they were calculated and estimated using the previously reported methods 23 . All disagreements were discussed until a consensus was reached. Statistical analysis. We used STATA 12.0 software (Stata Corporation, College Station, TX, USA) to perform all of the statistical analyses. The extracted HRs and their corresponding 95%CIs were comprehensively calculated to obtain pooled HRs and 95%CIs. If the pooled HR > 1 as well as its 95%CI did not overlap with 1, the decreased expression of the FOXP1 protein would be considered as an indicator for the poor survival prognosis in tumor patients. Analysis of the heterogeneity of the combined HRs was carried out using Cochran's Q test and Higgins' I-squared statistic. Heterogeneity was defined as I 2 > 50% or p < 0.05. If heterogeneity was present, a random-effects model was conducted. If not, the fixed-effects model would be applied. Sensitivity analysis was performed to assess the stability of the results. Furthermore, subgroup analysis and meta-regression were adopted to explore the sources of the heterogeneity. In addition, the publication bias was evaluated by Begg's and Egger's tests. However, all of the p values in our results were two-tailed, and p < 0.05 was considered to be statistically significant.

Results
Study selection. The initial database searching identified one hundred and fifty-three potentially relevant records. After the duplicates were removed, fifty-seven records remained. By assessing the full text for eligibility, thirty-five of these studies were excluded because they did not conform to the selection criteria. However, one additional study that also met our selection criteria was obtained from the references of relevant articles. Thus, a total of twenty-two studies were included in this systematic review. Finally, thirty-one datasets were used to perform the meta-analysis (Fig. 1).  (Table 1 and Fig. 1). However, the cut-off value for the decreased expression of FOXP1 protein was inconsistent among these eligible studies (Table 2).

Meta-analysis of OS.
The pooled result from twenty datasets yielded no significant association between decreased FOXP1 protein expression and OS in patients with various tumors (HR = 0.75, 95%CI: 0.48-1.17, p = 0.203) ( Table 3 and Fig. 2). A sensitivity analysis was performed by successively omitting each study, and the results revealed the pooled HRs did not vary substantially after excluding any individual study (Fig. 3  It is interesting that decreased expression of FOXP1 had different prognostic values for lymphomas and solid tumors. To reveal this contradictory phenomenon, we further conducted subgroup analyses for both of these cancer types. As shown in Table 4 for the subgroup analyses results for lymphomas, decreased FOXP1 expression had the best OS in patients with MALT lymphoma (HR = 0.26, 95%CI: 0.11-0.59, p = 0.001). However, decreased FOXP1 protein expression in patients with solid tumors was associated with a significantly worse OS in most of the subgroup categories, and the worst OS was observed in NSCLC patients (HR = 3.11, 95%CI: 1.87-5.17, p < 0.001) ( Table 5).

Meta-regression analysis of OS.
To investigate the source of heterogeneity among OS datasets (I 2 = 84.1%, p < 0.001), we performed meta-regression analyses by choosing variables such as publication year, country, cancer type, sample source, expression location, sample size and analysis method. The results suggested that cancer type (residual I 2 = 6.26%, adjusted R 2 = 100.00%) and expression location (residual I 2 = 80.68%, adjusted R 2 = 24.29%) were the major sources of significant heterogeneity among datasets regarding OS (Supplementary  Table S1). Consequently, as cancer type can almost completely explain the heterogeneity among OS datasets, the subgroup analyses for it showed that the heterogeneities were much lower (Tables 3-5).
Publication bias. As the amount of datasets for meta-analysis of CSS/DFS/FFS/PFS/RFS were fewer (each of them were less than five), we only evaluated the publication bias for the OS meta-analysis. However, both Begg's funnel plot and Egger's linear regression test were used to evaluate the publication bias. The results indicated that no publication bias in all of the OS datasets for all tumor types (p = 0.347 for Begg's test and p = 0.275 for Egger's test). Publication bias also did not exist in the datasets regarding the OS for lymphomas (p = 0.213 for Begg's test and p = 0.291 for Egger's test) or solid tumors (p = 0.602 for Begg's test and p = 0.864 for Egger's test) (Fig. 7).

Discussion
FOXP1 plays an important role during pathologic tumor development by potentiating Wnt/β -catenin signaling in DLBCL 38 . By repressing S1PR2 signaling, FOXP1 also promotes the survival of DLBCL cells 10 . In addition, FOXP1 negatively regulates androgen receptor signaling in prostate cancer to function as an androgen-responsive transcription factor 39 . Furthermore, FOXP1 still serves as an oncogene through promoting the cancer stem cell-like characteristics of ovarian cancer cells 40 . All of these observations indicate that the FOXP1 protein may have a specific prognostic value for tumor patients. However, thus far, no consistent conclusion has been made [14][15][16]18 .To the best of our knowledge, this is the first meta-analysis examining the prognostic value of decreased FOXP1 protein in various tumors.    Our meta-analysis incorporated 22 eligible studies with 31 datasets. The survival data included OS, PFS, RFS, DFS, CSS and FFS. First, we found no significant association between decreased FOXP1 protein expression and OS in patients with various tumors. When the subgroup analyses were conducted, the pooled results demonstrated that decreased FOXP1 expression was a favorable prognostic factor for lymphomas but an unfavorable factor for solid tumors. However, if the FOXP1 protein expression was located in the nucleus, decreased FOXP1 expression indicated a good OS prognosis. Furthermore, the results showed that decreased FOXP1 expression was correlated with the best OS in patients with MALT lymphoma but associated with  the worst OS in NSCLC patients. Additionally, in patients with solid tumors such as breast cancer, decreased FOXP1 expression was also significantly correlated with an unfavorable RFS. It should be noted that no publication bias was found in this meta-analysis.  Several important implications were confirmed by our study. First, decreased FOXP1 protein expression may be a universal favorable prognostic factor for lymphomas. In this meta-analysis, we included the lymphoma type, such as DLBCL [13][14][15]20,[24][25][26][27][28][29] and MALT lymphoma 14,[32][33][34] , and the results were also confirmed by studies with chronic lymphocytic leukemia 41 . Thus, we speculate that decreased FOXP1 protein expression may have similar prognostic value for all types of lymphoma that originate from lymphocytes. Second, decreased expression of FOXP1 is an unfavorable factor for solid tumors. As the meta-analysis results were pooled from breast cancer 16,30,31 , endometrial cancer 17 , hepatocellular carcinoma 35 , NSCLC 18 , prostate cancer 12 , colorectal cancer 36 and epithelial ovarian cancer 37 , and combined with further evidence from neuroblastoma 11 , we considered that this finding may be applicable to all solid tumors. Third, FOXP1 protein may function as a tumor promoter in lymphomas and act as a tumor suppressor in solid tumors. However, further research into these mechanisms is needed to verify this inference. Additionally, solid tumor patients with decreased FOXP1 protein expression in tumor tissues may indicate sensitivity to chemotherapy. Studies in vitro found that down-regulated FOXP1 expression can improve the sensitivity to chemotherapy in tumor cells 37,40,42 . Thus, we speculate that these situations may also occur in patients with solid tumors. However, more in vivo experiments are needed to confirm our speculation.
In this meta-analysis, we wanted to study the prognostic value of decreased FOXP1 protein expression in various tumors. However, we did not comprehensively evaluate the prognostic impact of overexpressed FOXP1 protein in the tumor patients. The major reason for this is that all of the eligible studies included in our study had defined decreased FOXP1 expression (Table 2), whereas relatively few studies 15,18,35,36 had reported an association between the overexpression of FOXP1 and survival outcome in tumor patients. Therefore, to highlight the key point of decreased FOXP1 expression, we only focused on the prognostic value of decreased FOXP1 protein expression in our current meta-analysis. However, as more original studies regarding the association between the overexpression of FOXP1 and survival outcomes in tumor patients will be conducted, a systematic study on the prognostic value of overexpressed FOXP1 protein in tumor patients can also be performed in the future.
There are some limitations that should be noted in our meta-analysis. The tumor types for both lymphomas and solid tumors included in this meta-analysis are limited, and our results should be cautiously extended to other specific tumor types. We only recruited articles published in English, thus a language bias might exist. Some HRs and their corresponding 95%CIs were extracted from the survival curves. However, these data are less reliable than those directly obtained from survival data. Because of the lack of data, the meta-analysis results regarding the CSS/DFS/FFS/PFS/RFS should be updated when more related studies are completed. Finally, studies regarding various tumors without a consistent cut-off value may be restricted to expand the clinical applicability [43][44][45][46] . Therefore, a unified cut-off value for the decreased FOXP1 protein is warranted. In summary, our meta-analysis suggests that decreased expression of the FOXP1 protein is associated with better survival in patients with lymphomas but poorer survival in patients with solid tumors. However, further prospective studies with larger sample sizes are required to validate the prognostic value of decreased FOXP1 expression in various tumors.