The value of CD9 as a prognostic predictor for various human malignant neoplasms: a meta-analysis

Recent studies indicate that the correlation in the expression level of CD9 and the prognosis differs from tumor types. This meta-analysis was performed to illustrate the predictive value of CD9. This analysis included 13 studies involving a total of 1231 patients. We used PubMed, Cochrane Library, Web of Science and Embase for comprehensive literature search and manual search for related bibliography. To probe into the relationships between CD9 expression and patient prognosis, pooled hazard ratios (HRs) with 95% condence intervals (CIs) were used to describe patient survival and disease recurrence. This meta-analysis contained 13 eligible studies. High expression of CD9 indicate worse overall survival (OS) (the polled HR = 1.14, 95%CI: 0.66–1.99), although the result was not statistically signicant (P = 0.638).Moreover high expression of CD9 could not result in the worse disease-free survival (DFS) / recurrence-free survival (RFS) of the tumor patients (the pooled HR = 0.67, 95% CI: 0.24–1.87; P = 0.445).


Background
Tetraspranin is a kind of protein which has four transmembrane domains, a small extracellular loop (termed SEL), a large extracellular loop (known as LEL), and short intracellular N-and C-terminal tails (1). Tetraspanin acts as a cell surface organizer by recruiting particular partner proteins to tetraspanin-enriched micro domains to roll into the meditation of various of cellular activities (2), such as signal transduction, cell growth, cell adhesion, cell motility and differentiation, and sperm-egg fusion (3)(4)(5).
CD9, also known as tetraspanin 29, Motility-Related Protein-1 or Multidrug resistance protein-1 (MRP-1), is widely expressed and expressed in a variety of hematopoietic cells and non-hematopoietic cells (6)(7)(8). SFQ, a site in the LEL of CD9, is essential for CD9 functions in sperm-egg fusion. CD9 is widely expressed in various normal and cancer tissues (9). Some studies have found that CD9 expression is a common pathogenic switch that directly leads to chronic glomerulonephritis and focal segmental glomerular sclerosis in humans and mice (10). CD9 can also participate in the MDR mechanism by binding to ATP binding cassette (ABC) transporters such as p-glycoprotein (P-gp/MDR1) (11).
Through the hydrolysis of ATP, CD9 actively removes substrates from cytoplasm and lead to cell drug resistance (12,13). CD9 has been found remarkably up regulated in some kinds of tumor tissues. There are various researches indicate that the high level of CD9 expression is associate with poor prognosis (14)(15)(16)(17). However, other researchers hold the opposite opinion (18)(19)(20)(21). some studies even revealed that there was no signi cant relationship between CD9 expression and patient's prognosis (22)(23)(24). Results of these studies are inconsistent. It is necessary to combine these researches and conduct this analysis to evaluate the ability of CD9 predicting the prognosis of human malignant neoplasms.
The purpose of this study was to assess the overall survival risk of CD9 for survival in cancer patients. The possibility to regard CD9 as a prognosis predicter for clinical treatment and statistics was also discussed.

Criteria of inclusion and exclusion
The study meets following criteria: (1) trials were for human cancers; (2) the relationship between CD9 and overall survival has been evaluated; (3) the measurement of the CD9 expression in tumor tissue was done; (4) OS and the hazard ratios (HR) could be directly extracted or indirectly calculated. Articles that met the following criteria were excluded: (1) researches were not for human cancer; (2) reviews, comments, economic analyses, conference abstracts, case reports, animal studies, and laboratory studies; (3) critical information about survival outcomes were missing; (4) HR and 95% CI could not be obtained according to the available data.

Date extraction
According to the needs, we extracted the following data from quali ed research: (1) the Author's name, the publication year; (2) the characteristics of the patients, including age, gender, country, ethnicity, tumor type, pathologic type, and sample type; (3) follow-up time; (4) Sample detection method and the expression levels of CD9; (5)HRs of elevated CD9 expression on overall survival (OS), recurrencefree survival (RFS) and diseasefree survival (DFS) with 95% con dence intervals (CIs) and P values. If the data were not provided intuitively but only in the form of a Kaplan-Meier curve, the data was extracted from the survival curves and the HRs were estimated using the method described earlier (25). statistical methods Q statistics was used to measure the heterogeneity of these studies. The evaluation criteria were as follows: I 2 value between 50% and 100% suggested high heterogeneity, 25%-50% suggested low heterogeneity, and 0-25% suggested no heterogeneity. A xed-effect model would be used in combination with HRs and 95% CI when I 2 < 50% and P > 0.10; A random-effect model would be chosen if I 2 > 50% and P < 0.10. Then subgroup analysis would be conducting.
The Egger's and Begg's bias indicator test would be performed to evaluate the publication bias. A two-sided P value < 0.05 was considered statistically signi cant. STATA Statistical Software Version 12.0 (Stata Corp, College Station, TX, USA) and Excel software 2019 were used to conduct all calculation.

Results
Study selection 580 studies were obtained from PubMed, Cochrane Library, Embase, Web of science and other databases. There were 469 remaining after removing duplicate researches. By reading the title and abstract, 40 researches were chosen. Due to lack of HR, con dence intervals, survival curves, or not the original data, we nally included 13 eligible studies. Figure 1 showed the ow chart of the study selection process.
Characteristics of the included studies Table 1 and Table 2 summarized the main characteristics of the 13 eligible studies. The meta-analysis included 13 studies with a total of 1231 patients. The malignant tumors assessed in this meta-analysis included lung adenocarcinoma, non-small cell lung cancer (NSCLC), breast cancer, pancreatic cancer, diffuse non-Hodgkin's lymphoma, acute lymphoblastic leukemia, adult T-cell leukemia, head and neck squamous cell carcinoma, oral squamous cell carcinoma, gastric cancer and malignant pleural mesothelioma. 8 studies used Immunohistochemistry (IHC) staining to measure CD9 expression. Real time polymerase chain reaction (RTPCR) was used in 4 studies, slot blot analysis and ow cytometry were used to in the left 2 studies respectively. The pathologic types involved in this meta-analysis included squamous cell carcinoma (SCC), adenocarcinoma (Ad), invasive ductal carcinoma, Non-Hodgkin's lymphoma (NHL), stromal tumor and Mesothelioma. These studies were all retrospective in design.

The progression and recurrence of disease associated with the expression of CD9
To better evaluate tumor progression and recurrence, the concepts of Disease-free survival (DFS) and Recurrence-free survival (RFS) were introduced. A total of 6 studies concentrated on DFS/RFS. Due to the high heterogeneity (P = 0.000, I 2 = 89.3%), random-effect model was used to do the analysis (Fig. 2B)

Publication bias
Begg's funnel plot and Egger's test were used to detect the publication bias (Fig. 4). The Begg's funnel plot did not show conspicuous asymmetry, and the Egger's test shown the similar result (P = 0.535). Because the amount of studies of DFS/RFS was less than 10, only Egger's test was performed. Similarly, no publication bias was found (P = 0.651).

Sensitivity analyses
As shown in the Fig. 5A and 5B, no alterations in the results was found in the OS and DFS/RFS studies due to the inclusion of any individual study, which mean that no single study signi cantly impact the pooled HR or the 95% CI.

Discussion
As mentioned above, CD9 is a member of tetraspanin superfamily proteins. It seems that research on CD9 has never stopped, and results have not been consistent. Some studies reported that CD9 could inhibit cancer cells metastasis and invasion via modulating the function of integrins, it could be regarded as a suppressors of metastasis in solid tumors (25)(26)(27). Some researchers hold the different view, they think that CD9 inhibits cancer progression by maintaining normal cell adhesion and motility in malignant mesothelioma (28). However, some people hold completely opposite views on the function of CD9. They think CD9 is not a favorable factor to the prognosis of the cancer patients. CD9 could bind to ATP-binding cassette and lead to chemo resistance in gastric cancer (17). CD9 could also cooperate with heparin-binding epidermal growth factor-like growth factor (HB-EGF) and promote the progress of human gastric cancer. Up to now, more and more functions of CD9 have been discovered, but its physiologic functions and related regulatory mechanisms are still not particularly clear.
The relationship between the expression of CD9 and the OS of the various malignant cancer patients is the key point of this meta-analysis. The pooled HR of OS was 1.14, which means the high level of CD9 indicates a worse prognosis. But it was not statistically signi cant (P = 0.664). The pooled HR of DFS/RFS was 0.67, the high expression of CD9 turned out to be a protective factor, although it was not statistically signi cant too. This phenomenon maybe caused by the different function of CD9. Recent studies have reported that CD9 expression was negatively correlated with the high prevalence of lymph node metastasis. CD9 was also involved in cell growth, adhesion and movement, and the change of CD9 expression may be related to tumor invasion and metastasis (29). CD9 could also act as a metastasis suppressor by neutralizing Aggrus-mediated platelet aggregation as Nakazawa Y reported (30). What's more, the result of DFS/RFS analysis could not be persuasive enough because of the limited number of included researches.
Because of the high heterogeneity, we had to conduct subgroup analysis. The results suggested that high level of CD9 indicate poor prognosis in hematological tumor and digestive system tumor. On the other hand, CD9 could be favorable factor of the squamous cell cancer. As mentioned above, this phenomenon could be caused by different function of CD9. When it came to nationality, China was a risk factor for tumor prognosis. However, no signi cant effect was observed in ethnicities. This difference can be attributed to environmental differences.
We must admit there are still some restrictions that cannot be ignored in this meta-analysis. At rst, there is no standard for the high level of CD9 expression. In some studies, more than 10% are considered positive (16), while in other studies the standard is 30% (31) or 50% (18), and because the method of determination is different, there must be some error between the research results and the real results. Secondly, the number of studies included in the analysis is still too small, especially for the DFS/RFS analysis, and may reduce the statistical power of the result. Thirdly, we had to calculate or extracted data from the survival curves because of the absence of some critical survival information, which would bring some minor differences to the conclusion. At last, these studies did not mention the survival data of MFS/PFS, so we could not conduct research about this aspect. These factors should be taken into account in drawing a conclusion.

Conclusion
To sum up, our meta-analysis result suggested that CD9 could not be regarded as prognostic predictor for multiple human malignant neoplasms. However, in hematopoietic tumors (especially NHL) and digestive system tumors, high CD9 expression can be considered as a risk factor for poor prognosis for OS and DFS/RFS, while for squamous cell carcinoma, high CD9 expression is a protective factor that heralds higher OS and DFS/RFS. As for cancer patients in China, a high level of CD9 is also a risk factor that requires su cient attention.

Declarations
Ethics approval and consent to participate Not applicable.

Consent for publication
Not applicable.

Availability of data and materials
All data generated or analyzed during this study are included in this manuscript.

Competing interests
The authors declare that they have no competing interests.  Selection process of studies for meta-analysis.

Figure 4
Begg's funnel plots of the publication bias. Notes: Begg's funnel plots of the publication bias for overall merged analysis of OS. Each point represents a separate study.