Role of TSP-1 as prognostic marker in various cancers: a systematic review and meta-analysis

Background Published studies present conflicting data regarding the impact of Thrombospondin-1 (TSP-1) expression on prognosis of various cancers. We performed this meta-analysis to illustrate the preliminary predictive value of TSP-1. Methods Twenty-four studies with a total of 2379 patients were included. A comprehensive literature search was performed by using PubMed, Cochrane Library, Web of Science, Embase, and hand searches were also conducted of relevant bibliographies. Pooled hazard ratios (HRs) with 95% confidence intervals (CIs) for patient survival and disease recurrence were initially identified to explore relationships between TSP-1 expression and patient prognosis. Results A total of 24 eligible studies were included in this meta-analysis. Our results showed that high level of TSP-1 was correlated significantly with poor overall survival (OS) (HR = 1.40, 95% CI: 1.17 ~ 1.68; P<0.001). However, high TSP-1 expression predicted no significant impact on progression-free survival (PFS)/ metastasis-free survival (MFS) (HR = 1.35, 95%CI: 0.87–2.10; P = 0.176) and disease-free survival (DFS)/ recurrence-free survival (RFS) (HR = 1.40, 95%CI: 0.77–2.53; P = 0.271). In addition, we performed subgroup analyses which showed that high TSP-1 expression predicted poor prognosis in breast cancer and gynecological cancer. Additionally, the relatively small number of studies on PFS/MFS and DFS/RFS is a limitation. The data extracted through Kaplan-Meier curves may not be accurate. Moreover, only English articles were included in this article, which may lead to deviations in the results. Conclusions Our findings indicated high TSP-1 expression may act as a promising biomarker of poor prognosis in cancers, especially in breast cancer and gynecological cancer.


Background
Thrombospondin-1 (TSP-1) is one of the thrombospondin gene family, which be composed of five matricellular proteins includes TSP-1, − 2, − 3, − 4, and − 5 [1,2]. Thrombospondin gene family plays a role in extensive physiological and pathological processes, including development, angiogenesis, inflammation and neoplasia [3]. This TSP family interact with a variety of membrane proteins on the cell surface, such as proteoglycans, integrins, CD36 and CD47 [4,5]. Thus, during tissue development and remodeling, TSP family control cellular phenotype and extracellular matrix structure [3].
TSP-1 is a multifunctional matrix glycoprotein which is synthesized and secreted by platelets, endothelial cells and smooth muscle cells. Under the transmission microscope, TSP-1 is composed of three identical peptide chains, each showing a spherical amino at one end and a spherical carboxyl at the other end, connected by a slender rod-like arm in the middle [6]. Because of multiple functional domains, TSP-1 can mediate the interaction between cell and cell, cell and extracellular matrix. Therefore, TSP-1 is a kind of glycoprotein with a wide range of biological effects, such as activating transforming growth factor-β, inhibiting angiogenesis, antitumor activity, participating in tissue repair and so on [7,8]. TSP-1 was originally found in platelets, but now it has been shown to play an important role in carcinogenesis [9,10]. Besides its direct role in regulating the behavior of tumor cells, TSP-1 also shows function in tumor vessels [11]. Taken together, TSP-1 can regulate the growth, adhesion and migration of tumor cells [12].
The function of TSP 1 remains disputable in angiogenesis and tumor progression. In some cancers, TSP 1 has been deemed to be an inhibitor of both processes [13][14][15], while in others, it has been considered a stimulator [16][17][18]. Some research concluded that the actual function of TSP 1 was organ specific [19].
To address this issue, we performed meta-analysis to comprehensively assess the overall risk of TSP-1 for survival in patients with various types of cancers. Furthermore, we attempt to evaluate the value of TSP-1 as a prognostic marker in the aspect of clinical features and statistics.

Search strategy
Original studies aimed to analyze the predictive value of TSP-1 in multiple human malignant neoplasms. PubMed, Web of Science, Cochrane Library and Embase were searched up from inception to December 14, 2019 using the following key words: "Thrombospondin 1", "Neoplasm", "prognosis", "survival", "recurrence", "death", "incidence", "mortality". The search strategy used a combination of Medical Subject Headings and thesaurus terms.

Inclusion criteria and exclusion criteria
We considered articles were considered qualified when they met the following criteria: (1) patients diagnosed with cancers by using pathological methods (2) studies focusing on the relationship between TSP-1 expression and prognosis, (3) survival outcomes with 95% confidence intervals (CIs) and hazard ratios (HRs) that could be extracted directly or indirectly. Excluding criteria were as follows: (1) Not a human study; (2) Not original articles; (3) Unrelated to TSP-1; (4) No clinical parameters; (5) Unrelated to malignant tumor; (6) Not related to prognosis or survival; (7) Insufficient survival data; (8) Overlapping data.

Quality assessment
Quality assessment of the included studies was conducted independently by three researchists (SS, HD and TY) to rule out any discrepancy. Studies for inclusion include the following criteria: (1) the study country and population; (2) definition of study design (3) the samples and pathology information; (4) defiition of measurement of TSP-1; (5) the clinical outcomes and follow-up duration.

Data extraction
All relevant studies were identified by two independent reviewers (TY and HD), and any disagreements were reassessed by a third reviewer (SS). The data elements include the following information: (1) publication year and first author's name; (2) nationality, dominant ethnicity, number of patients, sample type, and main type of pathology; (3) Assay method, follow-up time; (4) TSP-1 expression levels and cut-off values; (5) HRs related to elevated TSP-1 expression for overall survival (OS), recurrence-free survival (RFS), and disease-free survival (DFS). Those indirectly reported HRs and 95% CIs were calculated using graphical survival plots.

Statistical methods
A random or fixed model was selected according the heterogeneity which was measured by the Q statistics and Higgins I-squared statistic (I 2 ). If P < 0.10 or I 2 > 50%, a random-effects model was applied and subgroup ana lyses was carried out to fine the source of heterogeneity; otherwise, a fixed-effects model was adopted [20]. Publication bias was assessed by Egger's test and Begg's funnel plot [21]. To examine the stability and dependability of the overall outcomes, sensitivity analyses were performed by excluding one single study one by one and recalculating their HRs. All P-values were calculated using a two-sided test. We consider P < 0.05 as statistically significant. All statistical analyses were conducted by using Excel software 2016 and Stata version 12.0.

Study selection
The flow chart of study selection process was depicted in Fig. 1. A total of 251 studies were identified from online databases, including PubMed, Cochrane Library, Web of Science and Embase. Based on screening of titles and abstracts, 54 studies were selected for further investigation according to following criteria: repetitive articles, not a human study, not original articles, no clinical parameters, unrelated to TSP-1, unrelated to malignant neoplasms and unrelated to prognosis or survival. Of these 54 studies, 30 were excluded due to insufficient survival data and overlapping data. Ultimately, 24 articles were included for further analysis.

Study characteristics
The major characteristics of the eligible data were summarized in Table 1. We collected basic data on 24 articles published between 2000 and 2019. The metaanalysis included 2379 participants from different regions of the United States, France, Japan, China, India, Greece, the United Kingdom and Norway, including tumors such as breast cancer, liver cancer, ovarian cancer, esophagus cancer, lung cancer, gastric cancer, colon cancer, skin cancer, cervical cancer, oral cancer and bladder cancer.

PFS/MFS and DFS/RFS associated with TSP-1 expression
Six studies were included in the PFS/MFS analysis, in which a random-effect model was used because of the significant heterogeneity (p = 0.006, I 2 = 69.2) (Fig. 2B). Our outcomes showed that there was no significant correlation between TSP-1 and PFS/MFS (HR = 1.35; 95%CI: 0.87-2.10; P = 0.176). Likewise, subgroup analyses were stratified for the PFS/MFS group to identify the potential source of heterogeneity and other significant information. In ethnicity subgroup, high expression of TSP-1 was related to unfavorable PFS/MFS in Caucasians (HR = 1.80, 95%CI: 1.34-2.40; P<0.001) (Fig. 4A). Stratifying by the source of HR, high TSP-1 expression revealed a significant relationship with poor PFS/MFS, mainly in the report group (HR = 1.63, 95%CI: 1.24-2.15; P = 0.001) but not in the SC group (Fig. 4B). The subgroup analysis of cancer type indicated that TSP-1   (Fig. 4C). when stratified by main pathological type, analysis in the adenocarcinoma group exhibited a significant correlation between up-regulated expression of TSP-1 and PFS/MFS (HR = 1.80, 95%CI: 1.20-2.71; P = 0.004) (Fig. 4D). Elevated TSP-1 predict  (Fig. 4E). The subgroup analysis in different assay methods had no obvious significance (Fig. 4F). We analyzed tumor recurrence associated with overexpression of TSP-1 by DFS/RFS. Seven studies focused on DFS/RFS analysis, with a high degree of heterogeneity (P = 0.001, I 2 = 73.7) (Fig. 2C). There was no correlation between high level of TSP-1 and poor DFS/RFS, (HR = 1.40, 95%CI: 0.77-2.53; P = 0.271) by random effect model. Furthermore, through the subgroup analyses, we did not observe statistically significant outcomes (Fig.  S1). In summary, no relationship was found between DFS/RFS and TSP-1.

Cumulative meta-analysis
The main function of cumulative meta-analysis is reflecting the dynamic trend of the research results and evaluating the impact of each research on the comprehensive results. All the selected studies were sorted according to the year of publication. (Fig. 5). The relationship between OS and TSP-1 was first statistically significant in 2001. In addition, the corresponding 95% CIs of OS became narrower with the continuous inclusion of studies, suggesting increasing estimated accuracy. On the contrary, as time goes on, the relationship of TSP-1 and DFS/ RFS or PFS/MFS are no longer statistically significant, indicating growing controversy in recent research.

Publication bias
Egger's test and Begg's funnel plot were applied to indicate publication bias in the included studies (Fig. S2). No obvious asymmetry was noticed in funnel plots and the P value of Egger's test also revealed no obvious publication bias. (OS: P = 0.066; DFS/RFS: P = 0.934; PFS/ MFS: P = 0.713).

Sensitivity analysis
In order to ensure the robustness of the above results and evaluate the stability of results, a sensitivity analysis was carried out by Stata 12.0 software. The analyzed result from a fixed model of OS and two random model of DFS/RFS and PFS/MFS demonstrated that no single study considerably influenced the pooled HRs or 95% CIs, suggesting that the results of the present metaanalysis are credible (Fig. 6).

Discussion
TSP-1 is a homotrimeric protein which is a member of the thrombospondin gene family [3]. More and more evidence proved that the abnormal expression of TSP-1 is related to the clinical prognosis of cancer patients. Previous research has shown that TSP-1 plays an important role in inhibiting angiogenesis, anti-tumor activity and participating in tissue repair [8]. However, some studies deemed TSP-1 is connected with carcinogenesis [9]. The role of TSP-1 in various cancers has been widely researched, but the conclusions are not consistent.
This is believed to be the first meta-analysis systematically analyzing the association between TSP-1 expression and clinical features of multiple cancers, which included 24 studies with a total of 2379 patients.
Our analysis showed that the pooled HR was 1.40, indicating that the elevated TSP-1 was significantly associated with lower OS (P<0.001). The pooled HR of PFS/ MFS analysis and DFS/RFS analysis was 1.35 (P = 0.176) and 1.40 (P = 0.271) respectively, which demonstrated the consistency of the results, but they were not statistically significant.
Some studies have shown that TSP-1 promotes the development of prostate cancer, and this property is stronger than its anti-angiogenic properties which are mediated by its binding to the CD36 receptor [45,46]. In addition, they found that TSP-1 promotes cell migration by mediating TRPV3 and, in patients, TSP-1 mRNA level in tumor tissue was significantly associated with PSA relapse. Furthermore, there could be a vicious circle in which TSP-1 inhibits angiogenesis and thus increases hypoxia which induces TSP-1 expression in return to speed up cell migration [10]. All the studies consistent with the results from our analysis. Detection of TSP-1 can identify subgroups of high-risk patients with poor outcomes.
In addition, Kang et al. found Sphingosylphosphorylcholine induces TSP-1 secretion which might play an important role in epithelial-mesenchymal transition since migration and invasion are the key indicators of epithelial-mesenchymal transition [47]. Then, the migration and invasion of breast cancer were significantly interrupted when they knocked down TSP-1. Moreover, in breast cancer patients, the high expression of TSP-1 was significantly associated with poor RFS and MFS [10,47,48]. We used subgroup analyses to find out the influence of nationality, dominant ethnicity, main pathological type, disease type, assay method and source of HR on the relationship between TSP-1 and patient prognosis. With regard to the ethnic subgroup analysis, we made a distinction between Asians and Caucasians to clarify the impacts of different genetic backgrounds on the results. Interestingly, the analysis showed a significant association between the high expression of TSP-1 and OS/ PFS/MFS (HR = 1.40, 95%CI: 0.77-2.53) (HR = 1.80, 95%CI: 1.34-2.40; P<0.001) in Caucasians, but not in Asians. The reason may be attributed to differences in genetic background and environmental exposure. Secondly, OS and PFS/MFS of breast cancer and gynecological cancer in disease type subgroup were remarkable poor. Campone et al. deem TSP1 as bad prognostic markers by Kaplan-Meier method and immunohistochemistry (IHC) in triple-negative breast cancers [24]. Some studies have shown that TSP-1 promotes the invasiveness of melanoma which declared TSP-1 is a poor prognostic marker [49]. Nevertheless, other research argued that TSP-1 could significantly inhibit the cell viability of Retinoblastoma cells both in vitro and in vivo [50]. To sum up, the prognostic value of TSP-1 may be different in various cancers. However, more studies are required to confirm the clinical significance of TSP-1 in many samples.
Furthermore, the OS (HR = 1.48; 95%CI: 1.18~1.87; P = 0.001) and PFS/MFS (HR = 1.63, 95%CI: 1.24-2.15; P = 0.001) was significantly worse in "reported" group but not in the SC group. Throughout this literature, 13 of the 24 studies with survival information did not provide the HRs, so we could only gather the data from Kaplan-Meier curves. There may be some slight errors between the accurate data and the estimated HRs and their 95% CIs from the Kaplan-Meier curves, leading to unreliable results. In other subgroups, we did not find any statistical significance.
It should be noted that heterogeneity is a potential and critical issue that should not be ignored when discussing the results of Meta-analysis. We believe that the mild heterogeneity observed in OS is acceptable. Moreover, the heterogeneity of PFS/DFS was greatly reduced by subgroup analysis. Sensitivity analysis was performed to assess the reliability of results which revealed that the pooled HR did not significantly change by omitting any individual Although this study provided a comprehensive metaanalysis for the prognostic role of TSP-1 in multiple cancers, several limitations do exist. First, Heterogeneity was noted among the selected studies. The existence of heterogeneity may be contributing to the unique characteristics, such as the nationality, dominant ethnicity, the main pathological type, disease type, assay method, source of HR and critical values of TSP-1 expression. Second, it was difficult to determine a standard expression cutoff value because of different cancers, varied assay method and diverse detected sample in the included studies. This could result in bias in the effectiveness of TSP-1 as a prognostic factor. Most studies have established a median expression, IHC intensities or a scoring system as the cutoff value. Therefore, pooled outcomes may be greater or lower than the actual value and cause bias in the results. What's more, the relatively small number of studies on PFS/MFS and DFS/RFS is a limitation, further studies with more selected data and enrolled patients are needed. As mentioned above, the data in SC group extracted through Kaplan-Meier curves may not be accurate. Finally, only English articles are accepted in this article, which may cause deviations in the results. These limitations should be addressed in further research and be considered when drawing conclusions.
In conclusion, TSP-1 might serve as an effective index in evaluating the progress and prognosis of multiple cancers, especially in breast and gynecologic cancer, and may be utilized to improve targeted therapies. In order to accurately evaluate the role of TSP-1 as a prognostic factor, more clinical studies are required before TSP-1 is applied, especially for single type of cancer.

Conclusion
In this paper, we firstly evaluated whether TSP-1 was an accurate prognostic prediction for multiple cancers. Our data provided convincing evidence that high expression of TSP-1 was associated with adverse cancer prognosis. However, further in-depth and larger-scale studies are needed to support our results.