The density and functional status of CD8+T cells are of crucial importance for determining patient survival and the response to immune checkpoint inhibitors. the prognostic significance. The prognoses were assessed by log-rank test. == Results == PD-L1 clones SP142 and 288 displayed great concordance by qualitative ( = 0.816, 0.810 for total cells and tumor cells at the 5% cut-off) and quantitative analyses (R2= 0.7991, 0.8187 for positive percentage and H-score). PD-L1 clone SP142 showed the highest positivity in immune/stromal cells staining (18.41%) compared to 288 (7.62%), while clone E1L3N showed poor staining in both tumor and immune/stromal cells. Clone SP142, but not 288 and E1L3N, predicted a worse prognosis at the 5% cut-off (p= 0.0243). Both the clone SP142 and 288 had high inter-pathologist correlation for LY2140023 (LY404039) tumor staining (R2= 0.9805 and R2= 0.9853), but a moderate correlation for stromal/immune cell staining (R2= 0.5653 and R2= 0.5745). Furthermore, a higher density of PD-L1+CD8+T cells was correlated with a shorter survival time (R2= 0.0909,p= 0.0352). == Conclusions == PD-L1 antibody clone SP142 was superior in cell staining, particularly in immune/stromal cell and prognosis. These findings are important for selection of PD-L1 antibody clones in the future diagnostic test. == Electronic supplementary material == The online version of this article (10.1186/s13000-018-0766-0) contains supplementary material, which is available to authorized users. Keywords:Programmed cell death ligand 1, Immunohistochemistry, H-score, Multiplexed immunofluorescence == Background == Positive PD-L1 (programmed death-ligand 1) expression by immunohistochemistry (IHC), which has been mainly used to determine PD-L1 status, is a prerequisite for PD-1 blockade therapy. Patients with PD-L1 positive expression show a higher overall response rate than those with PD-L1 negative expression in gastric cancer [1]. Recently, FDA approved the PD-L1 IHC 22C3 PharmDx kit (Dako North America), the PD-L1 288 PharmDx kit (Dako North America) and the PD-L1 SP142 Ventana test (Ventana Medical Systems Inc) as a diagnostic test for pembrolizumab, nivolumab and atezolizumab, respectively. Furthermore, PD-L1 antibody clone E1L3N was also used in some anti-PD-1 clinical trials. There were two problems in the assessment of PD-L1 expression: first, the same antibody clone showed a different staining ability in different tumors and different antibody clones showed the different staining abilities in the same tumor, especially in immune/stromal FLJ20285 cells [26]. Second, the cut-off value LY2140023 (LY404039) was varied including 1, 5 and 10% cut-offs. PD-L1-positive staining ranges LY2140023 (LY404039) from 17 to 72% in gastric cancer (GC), and this dramatic difference might be due to the use of diverse antibody clones and the lack of a consensus regarding assessment criteria [714]. According to the recent clinical trials (NCT01848834andNCT02335411), 40 to 55% [15] of GCs were PD-L1-positive using the 22C3 monoclonal antibody at a 1% cut-off value (including tumor cells and stromal or immune cells). Our study aimed to qualitatively and quantitatively compare the expression of PD-L1 on tumor cells and immune/stromal cells in GC using three PD-L1 antibody clones. Five-year overall survival (OS) rates were compared based on the staining patterns of the three antibodies at the 1, 5 and 10% cut-off values. Then, the antibody clone SP142 was selected to characterize PD-L1 expression on CD8+T cells and analyze the correlation with prognosis in formalin-fixed paraffin-embedded (FFPE) tissue samples using quantitatively multiplexed immunofluorescence. == Methods == == Patients == A consecutive series of 315 surgically resected samples of primary advanced GC was obtained from the Xijing Digestive Hospital from July 1, 2011 to July LY2140023 (LY404039) 1, 2012. All patients were diagnosed with advanced GC (stages I-III) by pathologists based on hematoxylin and eosin (H&E) staining. Patients did not receive any treatment before surgery, while most patients received chemotherapy after surgery. The clinical parameters evaluated for each patient included sex, age, tumor location, depth, differentiation, tumor staging, and vascular and nerve invasion. Tumor staging was based on the 8th American Joint Committee on Cancer (AJCC) criteria. Five-year OS rates were calculated from the date of the first operation to the date of death from any cause or survival 5 years later. == Immunohistochemistry (IHC) == FFPE tissue specimens were collected from 315 patients, and three 5-m consecutive sections were cut from each specimen. PD-L1 staining using three primary antibodies was compared: clone SP142 (1:100; Spring Bioscience Corp, rabbit IgG), clone 288 (1:300; Abcam, rabbit IgG) and clone E1L3N (1:200; Cell Signaling Technology Inc., rabbit IgG). PD-L1 antibody clone SP142 applied using the Ventana Benchmark platform and clones 288 and E1L3N were applied using the Leica Bond platform. Bond Epitope Retrieval ER2 Solution and ER1 Solution were used for E1L3N and 288 antigen retrieval, respectively. The normal tonsil tissue was used as a positive control.