The heterogeneity of PD-L1 protein in gastric cancer: expression and distribution characteristics.

Background: Programmed death receptor ligand 1 (PD-L1) is expressed at different levels in tumour tissues and tumour-infiltrating monocytes (TIMCs). The interpretation of PD-L1 expression in gastric cancer (GC) is more difficult because of its heterogeneity. Methods: The PD-L1 immunohistochemistry (IHC) by E1L3N assay was performed in GC tissues. The level and distributed characteristics of PD-L1 expression were observed to illustrate its heterogeneity both in the cancer tissues and TIMCs. The relationship between PD-L1 level and necrotic features of tumor cells, the number of TIMCs, the distribution of tertiary lymphoid tissue (TLS) in the stroma, and other clinicopathological factors were analysed. A Cox regression model was used to assess the prognostic value of PD-L1 expression. Results: Of the 110 GC samples, not only more cases (51/110 cases) could be detected by combined positive score (CPS) for PD-L1 expression compared the other two, tumour positive score (TPS), and mononuclear immune-cell density score (MIDS), but also there were more cases with the high level of PD-L1 expression by CPS, even if with good consistency among them (P < 0.05). The tumour cells with high expression of PD-L1 was prone to show a diffuse distributing, whereas mottled type in the low level. It was noteworthy that the strongly colored tumor cells tended to exhibit a mossy pattern which were distributed along the border between cancer nests and stroma, and the same pattern happened to occur in the positive mesenchymal cells contacting the tumor border, essentially lymphocytes and macrophages. The substantial necrosis in the tumour and the number of TIMCs was analyzed statistically significant correlated with CPS (P < 0.05), while other clinicopathological factors such as histological type, tumour size, invasion depth, TNM stage were uncorrelated. The number and distribution of TLS in the tumour and para-tumoural stroma indirectly affected PD-L1 in GC by associating with the quantity and pattern of TIMCs. Cox regression analysis revealed that the prognosis was poor when PD-L1 was positive. Conclusion: CPS is the best indicator for PD-L1 expression in GC, which tend to be increased expression following a large number of TIMCs and substantial tumour necrosis appeared. Heterogeneity was reflected in the different distributed pattern of PD-L1 expression, especially the mossy-like pantten of the staining tumor cell in the interface between tumour nests and stroma, regardless of the amount and intensity of PD-L1 expression. TLS is valuable for observing microscopic images to influence the quantity and pattern of TIMCs. CPS can be used as an independent prognostic factor for GC.