Inhibition of programmed cell death-1 (PD-1) and its ligand programmed death ligand 1 (PD-L1) by using an immune checkpoint inhibitor has emerged as a promising immunotherapy for NSCLC. The correlation of PD-L1 expression in tumor cells with treatment outcomes has been reported in many pivotal trials; however, the relationship remains unclear. Here, we demonstrate that those patients with both high density of PD-1–positive CD8 and PD-L1–positive CD4-positive CD25-positive (PD-1^hi PD-L1^hi) regulatory T cells (Tregs) have a better response to PD1/PD-L1 blockade.

In our study between April 1, 2014, and May 30, 2017, a total of 73 NSCLC peripheral blood samples and fresh tumor specimens were collected for study. Of these, 42 large (10-mm³) fresh tumor specimens were obtained from surgical procedures and checked for expression of immunology biomarkers, including PD-L1, PD-1, CD8, CD4, and CD25, in tumor cells and tumor-infiltrating lymphocytes (TILs) by flow cytometry, immunohistochemistry, and immunofluorescence (IF). Moreover, 31 small biopsy specimens from patients who received immunotherapy (pembrolizumab or nivolumab) were analyzed by immunohistochemistry and IF. The correlation between flow cytometry and IF detected for TILs’ density was evaluated by Spearman’s rank correlation test; the primary end point was progression-free survival. For the PD-1/PD-L1 blockade assay, the TILs and peripheral blood mononuclear CD8 T cells were cultured (1×10⁵ per well) with anti–PD-1 (clone MIH4), anti–PD-L1 (clone MIH1). The cytotoxic activity of TILs in killing NSCLC cells after stimulation by anti–PD-1 and anti–PD-L1 was measured by a conventional ⁵¹Cr release assay.

We first identified a population of high–PD-L1–expressing CD25-positive CD4-positive T cells (PD-L1^hi Tregs) in the tumor microenvironment. The frequency of PD-L1^hi Tregs was higher in tumor tissue (mean 48.6 ± 14.3% in CD25-positive CD3-positive CD4-positive T cells) than in blood (mean 35.4 ± 10.2% in CD25-positive CD3-positive CD4-positive T cells) and normal tissue (mean 38.6 ± 9.7% in CD25-positive CD3-positive CD4-positive T cells) (p < 0.05), as determined by flow cytometry. The frequency of PD-L1^hi Tregs was positively correlated with PD-1–positive CD8 in Tregs. In addition, the TILs from these patients (PD-1^hi PD-L1^hi) showed PD-1/PD-L1 pathway dependence and could induce a greater killing effect of TILs by PD-1/PD-L1 blockade treatment. The patients with PD-L1–positive NSCLC with PD-1^hi PD-L1^hi TILs showed a better clinical outcome than those with a low frequency of PD-1^hi CD8 or PD-L1^hi Tregs (median progression-free survival not reached versus 2 months).

Our findings suggested that the density of PD-L1–positive CD4-positive CD25-positive Tregs in the tumor microenvironment can serve as a diagnostic factor to supplement PD-L1 expression in tumor cells and predict the response to PD-1/PD-L1 blockade immunotherapy in NSCLC.