The prognostic value of cytotoxic T-lymphocyte antigen 4 in cancers: a systematic review and meta-analysis

The outcomes of studies analyzing the prognostic role of CTLA-4 in cancers are controversial. Therefore, the aim of our meta-analysis was to clarify the correlation between CTLA-4 expression and OS in different cancer cases. Relevant literature was searched using PubMed, EMBASE, Web of Science, and the Cochrane Library. The clinicopathological features, hazard ratio (HR) and 95% confidence intervals (CI) were collected from these studies and were analyzed using Stata version 12.0 software. The pooled HR values showed no significant correlation between CTLA-4 expression levels and OS in relation to tumors (HR: 1.24, 95% CI: 0.98–1.56, I2 = 71.7%, P = 0.000). Further subgroup analyses were conducted and categorized by experimental methods, CTLA-4 sources and cancer types. The survey showed a significant correlation (HR: 1.47, 95% CI: 1.14–1.89) between high expression of CTLA-4 and OS in the SNP subgroup, and subgroups analyzing by PCR (HR: 1.50, 95% CI: 1.20–1.86) and flow cytometry (HR: 2.76, 95% CI: 1.49–5.14). In addition, our analysis observed significant differences between patients and controls in inCTLA-4+CD4+ lymphocytes, surCTLA-4+CD4+ lymphocytes, inCTLA-4+CD8+ lymphocytes, and surCTLA-4+CD8+ lymphocytes. Knowledge of the effects of CTLA-4 could potentially be used to effectively guide appropriate prognosis and therapeutic strategies in cancer patients.

Scientific RepoRts | 7:42913 | DOI: 10.1038/srep42913 production, and cell cycle progression 6 . As a major co-stimulatory receptor highly expressed on the surface of resting T cells, CD28 could be engaged by CTLA-4 to prevent anergy and cell death 8,9 . Stimulation of naive T cells through T cell receptors could cause rapid and transient translocation of intracellular CTLA-4 to cell surfaces and/or extracellular secretions 10 . Constitutive CTLA-4 expressions on T regulatory cell (Treg) reduce the level of B7 ligand on antigen presenting cells, further inhibiting effector T cell immunity 11 . In addition, CTLA-4-expressing cells trans-endocytose ligands on neighboring cells, preventing CD28 co-stimulation 12 .
S-CTLA-4 could also interact with B7, inhibiting the activity of T cells via the interfere with CD28 signaling, and blocking soluble CTLA-4 enhances antigen-driven peripheral blood mononuclear cell responses. However, several studies showed that sCTLA-4 also functions in downregulating the negative signal of CTLA-4 in T-cell responses, which indicated that the immunoregulatory functions of sCTLA-4 might be complicated. It has been established and shown that CTLA-4 protein expression appears to be important for tumors to evade host immune surveillance in cancers. However, the clinical implications of CTLA-4 expression in tumors or immune cells in the tumor microenvironment are still controversial, and the potential for CTLA-4 as a prognostic marker has been complicated by differences in study populations and methods. Thus, our meta-analysis aims to illustrate the correlation between CTLA-4 expressions and overall survival (OS), as well as its prognostic role in cancer patients. We also analyzed the percentage of lymphocyte subsets (CD4+ , CD8+ and CD19+ ) expressing surface and/ of intracellular CTLA-4 in cancer patients and healthy volunteers, aiming to figure out the correlation between lymphocytes CTLA4 locations and their susceptibility to cancers.

Materials and Methods
Literature search. All available published literature concerning the prognostic value of CTLA-4 was systematically searched using electronic database. PubMed, EMBASE, Web of Science, and the Cochrane Library were searched for relevant articles. We also searched for potentially additional studies in references from eligible studies and correlated systematic reviews. The potential eligible literatures were restricted to English publications to facilitate understanding. The search strategy was (CTLA-4 OR cytotoxic T lymphocyte-associated antigen-4) AND (prognosis OR outcome OR survival) AND (cancer OR neoplasm OR carcinoma OR tumor). The final search was performed on July 15, 2016.
Eligibility criteria. Two investigators (LQ and HP) initially assessed the relevance by checking the titles and abstracts of the articles. The further eligibility criteria for our meta-analyze were as follows: (1) trials was dealing with human cancers, (2) an association between CTLA-4 and OS, or (3) dealing with CTLA-4 protein present on the surface and/or intercellular of lymphocyte, (4) Hazard ratios (HR) could be extracted directly or calculated indirectly in those articles dealing with OS, and (5) the studies represented original articles. Duplicate articles were excluded by checking the title, author name and study detail. We contacted authors for further details when data were not provided in articles. All the disagreements were resolved through discussion. Data extraction. Two investigators (LQ and HP) independently conducted the data extraction using a pre-determined form. The following information was collected when analyzing the association between CTLA-4 and OS: the first author, publication year, country, number of included patients, tumor type, stage, source of CTLA-4, detective method, analytic method, cut-off criteria, cut-off value, and survival data (Table 1). When analyzing the sur-and/or in-CTLA4 lymphocytes, we extracted the number of cases and controls, and available genotype frequencies information additionally. Incomplete data was determined by the accrual periods, the median follow-up periods, analysis and submission dates, as described by Tierney et al. 13 All the disagreements were resolved through discussion.

Statistical analysis.
To clarify the correlation between CTLA-4 and clinical data, we calculated pooled estimates of HR which were extracted in eligible articles 13 . Clinical data of patients including tumor histology, stage, smoking history, sex, and age were collected and analyzed in the article. A sensitivity analysis was performed to investigate the heterogeneity for all analyses which can be reflected by P-values and I 2 values. Random-effect model was applied when there appeared to have heterogeneity between studies (I 2 > 50% or P < 0.1), otherwise, the fixed-effect model was used. The sources of heterogeneity can be identified by subgroup stratification analyses according to experimental methods, CTLA-4 sources and cancer types. Notably, the analysis can only be performed when more than two articles were included in each subgroup. Publication bias was detected using Begg's and Egger's tests 14,15 . P < 0.05 was considered statistically significant. All data were analyzed using Stata software (version 12.0).

Results
Characteristics of eligible studies. Our search retrieved 331 articles after initial searching for pooled estimates of HR, which is utilized to explore the correlations between CTLA-4 expression and clinical data of cancer patients. We also searched for studies dealing with sur-CTLA4 and/or intercellular CTLA4 expression of lymphocytes. After glancing the titles and abstracts of the articles, 50 studies were remained whose full text was reviewed. Among the articles, 13 studies were excluded due to editorial, letters, reviews and meta-analysis, 5 articles were excluded due to duplication, 26 articles were excluded due to irrelevance to the prognostic value or expression position of CTLA-4, and/or inadequate data to calculate HRs and CIs. Finally, a total of 14 publications were finally selected for analyzing the prognostic value of CTLA-4, and 2 other studies were relevant to CTLA-4 protein expression position (Fig. 1). Some studies involved several subtypes, so they were analyzed repeatedly. The characteristics of these studies are listed in Table 1. A total of 2,932 patients were statistically calculated for OS in 14 studies 5,7,16-27 . The median number of patients in the articles was 151 (range 24-780). Eligible articles consisted of 3 articles for lung cancer, 4 for malignant hematologic diseases, and 7 other tumor types. Five studies were reported for stages I-III, six for stages I-IV, two for II-III and only one for stage III-IV, six more didn't mention the stage information. In terms of CTLA-4 detection, seven eligible articles used immunohistochemistry (IHC) to assess CTLA-4 expression, six used polymerase chain reaction (PCR), one used a microarray database, and the last one used both ELISA and IHC. Cut-off values were used to divide the expression of CTLA-4 into high and low levels, among which median or mean levels were mostly chosen in these studies. The majority of studies utilized median value as cut-offs, three used a ROC curve and six didn't mentioned. In addition, in the 2 eligible studies for expression position of CTLA-4 protein in lymphocytes, one study is for lung cancer and the other for laryngeal carcinoma. The following variants were evaluated when analyzing sur-/in-CTLA-4 lymphocytes: sur-/in-CTLA-4 + CD4+ lymphocytes, sur-/in-CTLA-4+ CD8+ , and sur-/in-CTLA-4+ CD19+ lymphocyte. We also conducted subgroup analysis by variants and CTLA-4 protein expression region.
Publication bias. Egger's and Begg's tests were examined to detect publication bias for our article. Of the 14 studies correlated with OS information, the p value was 0.277 for Egger's test and 0.651 for Begg's test, which suggested that this study had no publication bias when evaluating OS for CTLA-4 ( Fig. 2). In addition, we evaluated the influence of each study on the overall meta-analysis estimate by sensitivity analysis, which indicated that all data correlated with prognostic role of CTLA-4 expressions in all cancer patients were stable with OS as the endpoint in our analysis (Fig. 3).

Discussion
CTLA-4 is significant for tumors to evade host immune surveillance, and has been implicated in immune dysregulation of lung cancer 28 , cervical cancer 29 , breast cancer 30 , skin cancer 31 , gastric cancer 32 , colorectal cancer 33 , B cell chronic lymphocytic leukemia and non-Hodgkin's lymphoma 34 . However, the prognostic role and clinical application of CTLA-4 in tumors are still controversial, which might due to the differences in experiential methods and study populations. For example, non-small cell lung cancers (NSCLC) with CTLA4 overexpression were associated with a reduced death rate 16 . Higher clinical stage and promoted axillary lymph node metastasis can be found in breast cancer patients with higher CTLA-4 mRNA levels 35 . CTLA4 downregulation led to a significant increase in the proliferation and survival of chronic lymphocytic leukemia cells 1 . However, little evidence showed correlations between CTLA-4 and prognosis in cancer patient. Thus, our meta-analysis aimed to illustrate the prognostic role of CTLA-4 expression in several of cancers. CTLA-4 have been discovered to be expressed in B lymphocytes 36 , CD4 and CD8 subsets of T cells and monocytes 37,38 , which could be wildly detected in tumors, lymph nodes, and spleen. However, CTLA-4 could not be detected in a series of non-lymphoid tissues. CTLA-4 consists of two isoforms: a full-length membrane-bound receptor isoform and sCTLA-4, which consists of only the extracellular domain 7,39 . Studies indicated that the expression of CTLA-4 isoforms was at the same level in CD4 cells, while CD8 cells appear to express nearly 2.5-fold more full-length product with respect to sCTLA-4 40 .
To our knowledge, this is the first systematic review and meta-analysis to explore and report the prognosis value of CTLA-4 in various cancers. The pooled HR of 14 eligible studies for OS in our meta-analysis concluded that no significant correlation was showed between CTLA-4 expression and OS in tumors (HR: 1.24, 95% CI: 0.86-1.56). This result didn't in accordance with the previous findings that CTLA-4 might be a favorable/negative indicator of cancers on OS, which may due to the controversy and complexity of the function, expression, and location in different isoforms of CTLA-4. Nevertheless, our analysis indicated that sCTLA-4 subgroup, analyzing by ELISA, has a strong tendency to be associated with good prognosis in malignant tumor, though no significance was showed in this article (HR: 0.63, 95% CI: 0.37-1.08). While the cytoplasmic tail is shorter than that of the full-length form of CTLA-4 41-43 , sCTLA-4 contains the extracellular MYPPPY motif which plays a crucial role in binding B7 molecule 44 . Thereby, sCTLA-4 maintains the function of binding CD80/CD86 natural ligands expressed on antigen-presenting cells (APCs). Nevertheless, the immunoregulatory functions of sCTLA-4 seem to be quite complicated. On one hand, sCTLA-4 is capable to bind CD80/CD86 and participate in the inhibitory regulation pathway to suppress T cell activation, which is similar with the full-length CTLA-4 45 . On the   observed to be sharply reduced following B7/CTLA-4/CD28 cross-linkage, while IL-10 and TGF-β increased. All cytokine changes could participate in the regulation of cell-mediated autoimmunity 46,47 . Besides, several studies indicated that the expression of sCTLA-4 is associated with proinflammatory cytokine levels. Grohmann et al. found that sCTLA-4 expression can be enhanced by interferon beta-1a (IFN-β 1a) in mononuclear cells of healthy individuals 48 . The research showed no significant difference in the intensity analysis of full-length CTLA-4 from incubated mononuclear cells, whereas the cDNA intensity of sCTLA-4 was significantly enhanced following IFN-β 1a stimulation. This result verified the immunomodulatory effects of sCTLA-4 by IFN-β 1a and the influence are selectively 48 . In addition, it has been shown that sCTLA-4, as well as the recombinant sCTLA-4 protein (CTLA4-Ig), is capable to suppress the mixed leukocyte reaction in dose-dependent manners 36,45,49 . Studies 50,51 indicated that CTLA4-Ig plays crucial roles in T-cell inhibition via inducing different types of APCs including dendritic cells 52 . The correlation of sCTLA-4 and proinflammatory cytokine levels also certified the immunoregulatory capability of sCTLA-4 in vivo 53 . Subsequently, more subgroup analyses were conducted to further investigate the prognostic role of CTLA-4. Interestingly, the pooled HR for OS was 1.47 (95% CI: 1.14-1.89) in SNP subgroup, which suggested that CTLA-4 + 49AA genotype was an independent adverse indicator for cancer prognosis. The CTLA-4 gene has more than one hundred SNPs, among which the most studied phenotypes was + 49 A/G located in exon 1, and CT60 AA genotype, located in the 30-untranslated region of the gene 54 . The CTLA-4 + 49AA genotype was identified in this study to be an independent adverse prognostic indicator in cancer patients. In the meanwhile, patients with CTLA-4 + 49AA genotype had significantly shorter survival time than those with the GG or GA genotype. The effect of CTLA-4 + 49AA genotype on the development, progression and prognosis of cancers may due to the interference with the transcriptional activity of CTLA-4 mRNA level 55 . The 49AA genotype is associated with increased expression of CTLA-4 mRNA and protein, which has enhanced interaction with B7 ligands and an enhanced effect on T cell inhibition 55,56 . In addition, compared with other genotypes, patients with the CTLA-4 + 49AA genotype had significantly increased CTLA-4 expression on peripheral blood mononuclear cells, and decreased interleukin-2 expression which is crucial in T cell growth 57 . Thus, tumor patients with CTLA-4 + 49AA genotype would be expected to have an enhanced suppression on T cell activity and immune function, which contribute to the poor prognosis of cancers.
The subgroups categorized by cancer types showed significant correlations between CTLA-4 and OS in nasopharyngeal carcinoma, esophageal carcinoma, malignant hematologic diseases, and glioblastoma, most of which had a strong pooled HR. In addition, we observed a favorable effect of CTLA-4 overexpression on OS in malignant pleural mesothelioma, which may due to the complexity of CTLA-4 functions. Salvi et al. 16 suggested that CTLA-4 could mediate inhibitory function on tumor cells, comparable with its suppression effect for T cells. Recent studies showed that early disseminated NSCLC cells could produce CTLA-4 and subsequently suppress proliferation and/or enhance apoptosis of cancer cells via engagement with B7 ligands [58][59][60] . In this regard, CTLA-4 was associated with good clinical outcome and longer survival due to its direct antiproliferative and proapoptotic effects in cancers. However, Huang et al. 5  Our analysis also observed increased expression of sur-and in-CTLA-4 expressing lymphocytes in cancer patients. Researches indicated that intracellular CTLA-4 consists of the majority of molecules present of lymphocytes, while only less than 1% expressed on the surface (surCTLA-4). This observation is in consistent with the dynamic of CTLA-4 expression. After lymphocytes being antigen-specific activated, CTLA-4 expressed on the surface is rapidly internalized intracellular compartment of lymphocytes by clathrin-dependent endocytosis 61 . After reactivation, intracellular CTLA-4 transferred to the surface and exhibited its inhibitory effect of T cells, while the internalization could be inhibited via the formation and co-internalization of TCR complex 62,63 . Several of studies have notified the correlation between CTLA-4 lymphocyte variants and its susceptibility to autoimmunity and cancer 64 . Erfani et al. indicated that surCTLA-4 in lymphocyte subsets of patients with laryngeal squamous cell carcinoma or non-small cell lung cancer was higher in comparison to healthy controls 65,66 . The differences were significant in inCTLA-4+ CD4+ lymphocytes, surCTLA-4+ CD4+ lymphocytes, inCTLA-4+ CD8+ lymphocytes, and surCTLA-44+ CD8+ lymphocytes. However, the results also have restrictions which may due to the limited studies associated with intercellular CTLA4 and/or that present on the surface or of lymphocyte.
Recently, soluble intact anti-CTLA-4 antibody which is capable to enhance antitumor immunity via greatly boosting T-cell responses to both antigen and superantigen, has become a hotspot 58,62 . The antitumor effects had been wildly verified in murine model of fibosarcoma 67 , colon carcinoma, and metastatic melanoma models 68 . All the animal studies confirmed the safety of blocking antibodies, and provided proof that CTLA-4 inhibition could lead to potent antitumor effects in cancer patients 67,69 . Notably, the sensitivity of poorly immunogenic tumors to anti-CTLA-4 antibodies was low, whereas this condition for animals could be promoted by treatment with cellular vaccines transduced by granulocyte-macrophage colony-stimulating factor (GM-CSF) 69 . These observations promoted several of clinical trials for fully human anti-CTLA-4 monoclonal antibodys (mAbs) including tremelimumab and ipilimumab. However, a Phase III trial concerning tremelimumab was halted due to the limited therapeutic benefit compared with traditional chemotherapy, though it might still be valuable when used as adjuvant combination therapy 70,71 . Ipilimumab was approved in 2011 as antitumor therapy for metastatic melanoma patients, while the treatment is still evaluated for NSCLC, prostate, and pancreatic cancers, among others. Recent Scientific RepoRts | 7:42913 | DOI: 10.1038/srep42913 trials showed that 22-25% of patients responsive to ipilimumab achieved long-term survives beyond 3 years, among whom some suffering from active disease reached for extended periods 5 years 72 . Recently, it is a hot spot in evaluating the therapy of anti-CTLA-4 mAb combined with immunostimulatory agent, including anti-PD-1 mAb 73 , GM-CSF vaccination 74 , and immune stimulatory cytokines including IFN-α 2b 75 and IL-15 76 . Thus, there is no doubt that anti-CTLA-4 mAb therapy, as well as in combination with other immune-activating agents, could offer a significant improvement in antitumor immune effect in cancer patients.
Cut-off values adopted are varies among different patient cohorts that would result in bias. In these studies, median and/or mean values are mostly used in eligible articles. The data extrapolation methods were also correlated with publication bias. In addition, not all articles reported results of multivariate survival analysis which were considered the most reliable data. We collected all the relevant researches through quantities of databases and different search strategies to minimize this bias. However, more of less literature could have been missed during the collection and the publication bias is inevitably. We assessed publication bias in articles correlated with OS for all eligible studies and reported negative, thus the result was acceptable despite publication bias.
This meta-analysis has several limitations. Firstly, rather than individual record which is considered most reliably by some authors, the data in our article was obtained from published literatures 77 . Secondly, our meta-analysis only consists of articles published in English to facilitate understanding. Since the negative results always published in native language, this strategy could benefit the articles with positive conclusion 78 . Furthermore, the number of patients included in subgroups was relatively small, though the whole population is considerably, which could led to the restricted value of this meta-analysis.
In conclusion, the meta-analysis indicated that no significance was found when analyzing the overall effect of CTLA-4 expression on OS in several of cancer cases. Nevertheless, CTLA-4 + 49AA genotype was an independent adverse indicator for cancer prognosis, while the high-expression of sCTLA-4 has a tendency to be correlated with the prolonged OS.