Prognostic significance of CD103+ immune cells in solid tumor: a systemic review and meta-analysis

CD103 is a transmembrane heterodimer complex that mediates cell adhesion, migration, and lymphocyte homing of cell through interaction with E-cadherin. Recently, CD103+ immune cells in human carcinoma has been investigated as a prognostic factor, however, the correlation between CD103+ immune cells and survival are still elusive. Therefore, a meta-analysis was performed to determine the prognostic value of CD103+ immune cells in solid tumor. Studies relevant to the subject was searched from PubMed, Embase, and Web of Science. Ten studies including 2,824 patients were eligible for the analysis. Tumors positive for CD103+ immune cells were associated with favorable overall survival, disease-free survival, and disease-specific survival. Subgroup analysis revealed that assessing CD103+ immune cells in epithelial and total (both epithelial and stromal) areas or using whole slide section were associated with good prognosis. Furthermore, stromal CD103+ immune cells or CD103+ immune cells evaluated by tissue microarrays were not always significantly prognostic. In conclusion, these results show that CD103+ immune cells are associated with prognosis in solid tumor. However, the region of assessment and selection of material for the evaluation could affect the value of CD103 as a prognostic biomarker.

CD103, also known as integrin αEβ7 (ITGAE), is a transmembrane heterodimer complex that is involved in cell to cell or cell to matrix interaction 1 . CD103 mediates cell adhesion, migration, and lymphocyte homing of cell through interaction with E-cadherin, which is expressed in epithelial cells 2 . Therefore, CD103 is considered a hall mark of specific subset of immune cells that resides within the epithelium of mucosal organs including genital tract, stomach, lung, and skin [3][4][5][6] . The immune cells that express CD103 includes effector memory CD8+ T cells, regulatory CD8+ T cells (Tregs), and CD11c-high/MHC class II-high dendritic cells [6][7][8][9] .
In recent years, the importance of CD103+ immune cells in cancer has grown due to its intimate relationship with immune checkpoint and BRAF blockade. In ovarian cancer, intraepithelial CD8+ tumor-infiltrating lymphocytes (TILs) that expressed CD103 were suspected to cause exhausted phenotype by chronic stimulation that leads to resistance in immune therapy, and CD103 was widely co-expressed with PD-1 10 . In rodent melanoma, expansion and activation of CD103+ dendritic cells were the only antigen presenting cells that transported intact antigen which led to enhanced response of PD-L1 and BRAF target therapies 11 . CD103+ immune cells were also a predictive marker of immunotherapy in lung cancer and increased expression was observed following anti-PD-1 therapy in ovarian cancer and melanoma [12][13][14] .
A number of studies suggests that CD103+ TIL in human malignancy is associated with prognosis 15-18 . However, this characteristic of CD103+ immune cells in cancer are still elusive with the exception of a meta-analysis of ovarian cancer 19 . Some studies define CD103 as predictive markers of intraepithelial CD8+ T cells or Tregs 20,21 , but other studies show that CD103+ immune cells are present within the stroma 22 and could be considered to be associated with survival 23 . Moreover, recent studies have shown that the usage of tissue microarray (TMA) instead of whole tissue section slides could have issues when evaluating TIL within tumors 24,25 but a comparison between TMA and whole section has not been addressed for CD103+ immune cells.
In this study, we conducted a meta-analysis to determine the prognostic significance of CD103+ immune cells in solid tumors. The aim of this study is to clarify the prognostic role of CD103 expression in immune cell as a biomarker across multiple tumors and to verify which region within tumor (epithelial, stromal, or both) and material (TMA or whole section) should be used in the assessment of the marker.
Publication bias and sensitivity analysis. Begg's funnel plot and Egg's test were used to evaluate potential publication bias (Fig. 3). According to Egg's test, OS and DFS did not show publication bias (P value = 0.090 and P value = 0.600, respectively) but DSS was significant for publication bias (P value = 0.029). Removing a single study from the meta-analysis did not have significant effect on the overall conclusion apart from a single study from Wang et al. that involved urothelial cell carcinoma (Fig. 4).

Discussion
Expression of CD103 has been suggested as a marker of prolonged survival in solid tumors as well as a predictive marker of immunotherapy in recent years. However, the expression of CD103 has been reported in immune cells [6][7][8][9] which limits the comprehensiveness of the subject of individual articles since most studies focus on a single material (TMA or whole slide) and a single region of interest (intraepithelial, stromal, or total areas) when evaluating CD103+ immune cells. In this meta-analysis, for the first time, the correlation between CD103+ immune cells and prognosis has been explored via pooled analysis of solid tumors.
A total of 2,824 patient from 10 different studies was included in the meta-analysis. The results showed that OS (HR = 0.746, 95% CI = 0.583-0.954, P value < 0.020), DFS (HR = 0.800, 95% CI = 0.730-0.876, P value < 0.001), and DSS (HR = 0.571, 95% CI = 0.432-0.755, P value < 0.001) are all indicative measurements of the prognostic value of CD103+ immune cells in solid tumors. Subgroup analysis revealed that CD103+ immune cells were not significantly correlated with OS when TMA studies were pooled and stromal CD103+ immune cells were not prognostic markers of DFS. On the contrary, pooled studies for whole slide or studies that investigated on intraepithelial or total CD103+ immune cells were determined prognostic whenever subgrouping was available. www.nature.com/scientificreports www.nature.com/scientificreports/ The meta-analysis unveils novel characteristics of CD103+ immune cells within solid tumors. Although expression of CD103 was also observed in the stromal area, prognostic impact was only significant when CD103+ immune cells of epithelial or those of both epithelial and stromal (total) areas were accounted for. A pooled survival analysis of studies using TMA where not significantly correlated with OS but still maintained prognostic for DFS and DSS. This might be considered as a short side for TMA, but when reviewing the counting strategies of each article, studies using TMA counted multiple cores with 0.6 mm diameter or a single core with 2.0 mm diameter 16,23,26 , whereas studies that used whole section counted not the entire slide but 3 to 5 microscopic fields within a given slide 22,27 . Moreover, studies that used TMA had a larger average number of patients compared with that of studies with whole section (323.4 vs 186.7). Another caveat is the fact that not enough studies were available to compare the effect of TMA and whole section according to each tumor. Therefore, for better prognostic resolution of CD103, assessing intraepithelial or total CD103+ immune cells and using multiple TMA cores per sample with large diameter could be suggested in a large scale study.
Human malignancy and its microenvironment shows diversity according to genetic background and etiology, and a pooled analysis could undermine unique characteristics of each tumor type. In our current study, subgroup analysis according to tumor type was available in two out of seven different types included in the meta-analysis,  www.nature.com/scientificreports www.nature.com/scientificreports/ and not all tumor types were significantly associated with prognosis. Therefore, although the current study provides meaningful findings in survival-related features of CD103+ immune cells across solid tumors and various materials used for assessment, it should be noted that the correlation between CD103+ immune cells and prognosis could vary among different tumor types. Further investigation is required to fully elucidate clinical value of CD103+ immune cells in prognostic predictions.
Other limitations of this study are as follows. In our study, the subtype of CD103+ immune cells that correlates with prognosis is still poorly determined because none of the included studies used double staining immunohistochemistry or multiplex immunohistochemistry to identify if the immune cells are TIL although most studies defined CD103+ immune cells as CD103+ TIL. Furthermore, relatively small number of studies were available for the pooled analysis, which lead to exclusion of solid tumors of major systems, such as gastrointestinal tract. Since colorectal cancer has shown different prognostic meaning for FOXP3+ Treg compared with other organs in previous studies [28][29][30] , and CD103+ immune cells are widely expressed in Tregs, the inclusion of some organs has the potential to shift the result. Finally, Egg's test determined that the result between CD103+ immune cells and DSS has a publication bias.
In conclusion, CD103+ immune cells are markers of favorable prognosis when effect sizes for survival analysis of solid tumors from various organs where pooled and meta-analyzed, although prognostic feature between tumor types may vary. For further research, to find out the full spectrum of prognostic value and therapeutic potential of CD103, we suggest that immune cell subtyping, optimal tumor region selection, and proper material selection for evaluating CD103+ immune cells that could suffice in producing data for large sample size are necessary.

Materials and Methods
Publication search strategy. Articles relevant with the subject were retrieved from electronic databases, PubMed, Embase, and Web of Science until June 1 st of 2018. The search term were 'CD103' , 'cancer, carcinoma or tumor' , and 'prognosis, prognostic or survival' . Additional articles were retrieved from reference cited in the eligible studies and a meta-analysis article of the ovarian cancer 19 . The results were limited to English articles based on human malignancies. Each source and collecting period of tumor samples were carefully recorded to avoid studies with identical patient population. When duplicate data was found, the most recent iteration of the cohort was included.  www.nature.com/scientificreports www.nature.com/scientificreports/ Inclusion and exclusion criteria. Studies eligible for the inclusion were: (1) studies that measured expression of CD103 by immunohistochemical (IHC) staining in immune cells within any human tumor samples, (2) studies that report the correlation between CD103 expression within the primary tumor and survival, (3) studies that offer hazard ratio (HR) and 95% confidence intervals (CI) directly from the main article or supplementary materials or provide alternative values that could estimate HR and 95% CIs. Exclusion criteria of this study were: (1) studies that were from abstracts of conferences, reviews, and comments, (2) studies that were based on xenograft models or human hematopoietic or lymphoid malignancies, (3) studies that had insufficient pathological or survival data which led to unavailability to extract HR and 95% CIs.
Data extraction and quality assessment. Two reviewers (YK and GK) independently collected data from eligible articles according to the criteria. Newcastle-Ottawa Scale (NOS) was used to evaluate quality of all studies. Scores equal to or higher than 6 out of 9 was defined as qualified. Parameters that were extracted from each study were as follows: surname of first author, year of publication, primary organ and type of cancer, material and regions of assessment for CD103, cut-off value for CD103+ immune cells, type of outcome, and number of analyzed patients. Meta-analysis of CD103+ immune cells were measured by four different outcome endpoints including OS, DFS, and DSS. Survival data were extracted as HR and 95% CIs from Cox univariate analysis. For articles with only Kaplan-Meiere curves, HR and 95% CIs were estimated by using Engauge Digitizer V9.8 (http://digitizer.sourceforge.net) and spreadsheets provided by Tierney et al. 31 and Altman et al. 32 . However, this estimate was only available when number of patients from the research arm and control arm were reported. For this reason, we contacted the authors of articles that missed on this information, which one of the authors responded and provided patient numbers for each arm 23 .
Statistical analysis. The correlation between CD103+ immune cells and prognosis of patients with solid tumor were measured via meta-analysis. Pooled HR and 95% CIs for OS, DFS, and DSS were calculated by using Comprehensive meta-analysis V3.3 (Biostat, Englewood, NJ, USA). Cochran's Q and I 2 were used to determine statistical heterogeneity. Subgroup analysis was performed when multiple studies were in each subgroup. P value <0.05 or I 2 > 50 were considered to have presence of heterogeneity and random effect model was used instead of fixed effect model. Sensitivity analysis was conducted to find out the effect of a single study to the pooled analysis. A graphical funnel plot and Egg's test were used to evaluate potential publication bias. All tests were two-sided and P value < 0.05 was considered as statistically significant, except for tests for heterogeneity.