PD-L1 on Antigen-Presenting Cells Facilitate the Induction of Antigen-Specific Cytotoxic T Lymphocytes

Programmed death-ligand 1 (PD-L1) binds to programmed death-1 (PD-1) on T cells and contributes to T cell exhaustion during chronic infections. Moreover, it has been shown that engagement of PD-1 on T cells by PD-L1 on tumor cells is associated with the immune escape of tumors, and recent clinical trials have highlighted the anti-tumor efficacy of blockade of the PD-1/PD-L1 pathway. Along with costimulatory molecules such as CD80 and CD86, PD-L1 is also highly expressed on dendritic cells (DCs), which are professional antigen-presenting cells (APCs) that are widely used to generate antigen-specific cytotoxic T lymphocytes (CTLs) for adoptive immunotherapy. However, the roles of PD-L1 on APCs have not been well examined. Thus, we evaluated the roles of PD-L1 on APCs in the induction of cytomegalovirus (CMV)-specific CTLs (CMV-CTLs).

First, peripheral blood mononuclear cells (PBMCs) were obtained from five CMV-seropositive healthy donors (HLA-A*02:01, n = 3; A*24:02, n = 1; B*07:02, n = 1). Mature DCs were generated from adherent PBMCs by stimulation with GM-CSF/IL-4 and maturational cytokines (TNF-a, IL-1b, IL-6, and PGE2), and freshly isolated CD3⁺ T cells were stimulated with mature DCs pulsed with CMV pp65-derived HLA-restricted peptides (NLVPMVATV for HLA-A*02:01, QYDPVAALF for A*24:02, and TPRVTGGGAM for B*07:02) in the presence or absence of anti-PD-L1 blocking antibody. After 14 days of culture, the presence of anti-PD-L1 antibody resulted in a less efficient induction of CMV-CTLs (Figure 1), suggesting that PD-L1 might play a positive role in the induction of antigen-specific CTLs. For further evaluation, we generated K562-based artificial APCs (aAPCs), which were retrovirally transduced with HLA class I molecules and various combinations of CD80/86 and PD-L1 (named K562, K562+CD80/86, K562+PD-L1, and K562+CD80/86+PD-L1). CD3⁺ T cells isolated from nine CMV-seropositive healthy donors (HLA-A*02:01, n = 4; A*24:02, n = 4; B*07:02, n = 1) were stimulated weekly for 28 days using HLA-restricted CMV peptide-pulsed K562-based aAPCs. K562+CD80/86+PD-L1 led to significantly higher induction of CMV-CTLs than K562 or K562+CD80/86 (Figure 2). Since the original K562 cells slightly express CD80, we completely knocked out the CD80 expression in the cells using the CRISPR/Cas9 system, and then we transduced the cells with or without PD-L1 (named CD80KO K562+PD-L1 and CD80KO K562, respectively) to determine whether PD-L1 itself has a stimulatory effect on the induction of CMV-CTLs. CD80KO K562 and CD80KO K562+PD-L1 induced the same amount of CMV-CTLs, suggesting that PD-L1 itself does not have a stimulatory effect for the induction of CMV-CTLs. Phenotypic analysis showed that CMV-CTLs induced by K562+CD80/86+PD-L1 contained more CD45RA^-CD62L⁺ central memory T cells, which persist longer and are more effective in vivo after adoptive transfer in CTL therapy, than CMV-CTLs induced by other K562-based aAPCs. In addition, when compared with CMV-CTLs induced by other K562-based aAPCs, CMV-CTLs induced by K562+CD80/86+PD-L1 did not express higher levels of exhaustion markers such as PD-1 and TIM-3, and exhibited similar levels of IFN-gamma and IL-2 production and CD107a expression in response to the specific peptide stimulation. Furthermore, to evaluate the effect of K562-based aAPCs on the induction of tumor antigen-specific CTLs, CD3⁺ T cells were isolated from three HLA-A*24:02-positive healthy donors and stimulated with HLA-A*24:02-restricted WT1 peptide-pulsed K562-based aAPCs weekly. After four stimulations, only K562+CD80/86+PD-L1 could clearly expand the WT1-specific CTLs from all of the three donors (Figure 3).

In conclusion, our findings demonstrate that PD-L1 expressed on APCs along with CD80/86 enhanced the induction of antigen-specific CTLs without causing excessive differentiation or functional exhaustion of the induced CTLs. These results suggest that stimulation with K562+CD80/86 might result in the activation-induced cell death of antigen-specific CTLs and that PD-L1 might be involved in fine-tuning excessive stimulation of CD80/86. Further analyses to determine the mechanisms of our findings are warranted. Our results also highlight the possibility that K562+CD80/86+PD-L1 has therapeutic potential as novel aAPCs for the generation of CTLs for adoptive immunotherapy.