PD-1–Targeted Immunotherapy: The Ligand Matters

Immune checkpoint therapies have revolutionized the clinical management of many malignancies. Since 2011, mAbs targeting key inhibitory pathways, such as programmed cell death protein 1 (PD-1), CTLA-4, and lymphocyte activation gene 3, have received Food and Drug Administration approvals. In the United States, as of 2019, over one-third of patients are now eligible to receive these drugs as part of standard-of-care therapy (1). Among immune checkpoint inhibitors, mAbs binding either the inhibitory receptor PD-1 or its primary ligand PD-L1 are the most widely used in the clinic to date.

In this Pillars of Immunology article, we highlight work by Latchman et al. (2) describing the discovery of PD-L2, a second ligand for PD-1. Although agents targeting the PD-L1/PD-1 pathway have received much attention in the context of immuno-oncology, clinically used anti–PD-1 mAbs also target the PD-L2/PD-1 pathway. Whether targeting PD-L2/PD-1 is of significance in human patients remains unknown. However, understanding the historical context of PD-L1 and PD-L2 is helpful for considering this question. The genesis of work on these pathways originated from work by Honjo et al., who identified PD-1 in 1992 (3) and later showed that it is a negative immune-regulatory receptor (4). In 1999, the first ligand for PD-1 was reported when Dong et al. (5) cloned a novel B7 family member (B7-H1), and, shortly thereafter, Freeman et al. (6) identified the same gene and determined the encoded protein to be the ligand for PD-1, hence naming it PD-L1. In 2002, Iwai et al. reported that administration of an Ab against PD-L1 to tumor-bearing mice could induce antitumor immunity (7). The combination of target/ligand identification and the functional role of the PD-1 pathway in autoimmunity and cancer precipitated intense research into harnessing this pathway for therapy, culminating in Food and Drug Administration approval for the first PD-1–targeting mAbs in 2014 (nivolumab and pembrolizumab).

Shortly after identifying PD-L1 as a ligand for PD-1, Freeman, Sharpe, and colleagues set out to discover additional ligands. Using a PD-L1 homology-based approach, Latchman et al. (2) cloned a murine dendritic cell cDNA encoding a protein that shared ∼38% identity with PD-L1 and named it PD-L2. Human PD-L2 was also identified. Of note, Tseng et al. (8) also reported this second PD-1 ligand almost simultaneously and designated it B7-DC because of its dendritic cell–selective expression. Latchman et al. (2) confirmed that the identified gene product, PD-L2, is indeed a second ligand for PD-1 using cell lines stably transfected with PD-L2 and staining with recombinant PD-1–Ig. Analysis of PD-L1 and PD-L2 mRNA expression in various human and mouse tissues revealed similar yet distinct expression patterns between the two ligands. Interestingly, compared with human PD-L1, PD-L2 expression was more prevalent in several tissues, such as heart, pancreas, lung, and liver, with PD-L2 expression being more readily detectable in human tissues than in their murine counterparts, an important point to keep in mind when translating preclinical studies to the clinic and vice versa.

Latchman et al. (2) not only characterized the expression of PD-L2 and confirmed its biochemical interaction with PD-1 but also determined through a set of in vitro experiments that the interaction between PD-1 and PD-L2 can indeed inhibit TCR-mediated responses. Using cocultures of PD-L2–transfected APCs with preactivated CD4⁺ T cells, they demonstrated that engagement of PD-1 by PD-L2 not only inhibits TCR-driven proliferation of T cells but also blunts secretion of various effector cytokines in the absence as well as the presence of CD28 receptor signaling. Demonstrating relevance for cancer immunity, Tanegashima et al. (9) found that when transfected into tumor cells, PD-L2 can impair antitumor immunity in mice. Although much work has focused on the immune-inhibitory properties of PD-L1, in their initial publication describing PD-L2, data from Tseng et al. (8) suggested that PD-L2 has costimulatory properties. These seemingly divergent outcomes of PD-L2 engagement might be related to different experimental setups or cell-specific parameters, or they could also be explained by a second receptor for PD-L2. In 2014, Freeman and colleagues reported that repulsive guidance molecule b is a binding partner for PD-L2, and, furthermore, blockade of the PD-L2/repulsive guidance molecule b interaction impairs the initial expansion of T cells and respiratory tolerance induction (10). A recent study also suggests that glycosylation of PD-L2 may facilitate its direct interaction with epidermal growth factor receptor and hinder epidermal growth factor receptor–targeted therapies (11). PD-L2 can thus have diverse regulatory activities, with the biological consequences of most of these interactions being poorly understood.

Immunotherapeutic targeting of the PD-1 pathway has so far focused on abrogating the interaction of PD-1 and PD-L1, with therapeutic Abs targeting either molecule now being widely used in the clinic. Focus on the PD-1/PD-L1 axis has been driven mainly by preclinical studies demonstrating limited PD-L2 expression in most murine tissues and models, as well as by the impressive preclinical and clinical responses targeting this signaling axis, including with the use of anti–PD-L1 mAbs. Hence, PD-L2 has traditionally been considered a minor ligand for PD-1 due to its limited expression on mostly professional APCs. The data by Latchman et al. (2) regarding PD-L2 mRNA expression seem to suggest important species-specific differences between humans and mice, with a wider range of human tissues expressing PD-L2. Of note, significant levels of PD-L2 expression have been reported for some human tumors, with expression patterns correlating well in many but not all cases with PD-L1. Furthermore, in at least one study, PD-L2 expression was independently associated with an improved response to anti–PD-1 therapy (12). PD-L2 was found to be commonly expressed in the stromal compartment, including immune cells, but also on tumor and endothelial cells, although with higher variability between tumor types.

The homology to PD-L1 and similar bioactivity upon PD-1 engagement as shown by Latchman et al. (2) suggested that PD-1 engages both its ligands in a similar manner, a fact that later was confirmed through structural biology approaches (13). Interestingly, although PD-L1 and PD-L2 both engage PD-1 in a similar manner, the amino acid composition of their interaction sites surrounding a central tryptophan residue differs, resulting in an ∼4-fold stronger affinity for PD-L2 than for PD-L1. The shared binding site of PD-1 for both its ligands also has implications for the therapeutic activity of PD-1–targeting mAbs, which will abrogate not only PD-L1 but also PD-L2 engagement. In contrast to PD-L1–targeting mAbs, anti–PD-1 mAbs represent the stone capable of killing two birds (PD-L1 and PD-L2). Of note, no trial to date has directly compared the clinical efficacy of PD-1– and PD-L1–targeting agents, hindering the direct assessment of combined abrogation of PD-L1– and PD-L2–mediated signals through anti–PD-1 mAbs versus sole abrogation of PD-L1–mediated signals. A recent meta-analysis of over 11,000 patients with cancer aimed to assess the aforementioned differences in therapeutic activity. The authors analyzed clinical responses across a total of 19 randomized clinical trials that were selected on the basis of matching/mirroring clinical characteristics to minimize biases between studies (14). Importantly, this analysis revealed a superior overall and progression-free survival of patients treated with anti–PD-1 compared with anti–PD-L1 therapy, suggesting that PD-L2 can significantly inhibit antitumor immune responses and thus impede clinical responses. Overall, the study by Latchman et al. (2) revealed a second ligand for the inhibitory receptor PD-1, thus substantially advancing our knowledge of the biology of this pathway whose clinical significance remains to be fully elucidated.