The concept of T cell costimulation has evolved over time. The two-signal model for T cell activation was proposed by Kevin Lafferty as a model for the activation of naive T cells (1). According to this model, T cells require two signals to become fully activated.The first signal, which gives specificity to the immune response, is provided by the interaction of antigenic peptide -MHC complex with the T cell receptor (TCR). The second,antigen-independent costimulatory signal, is delivered to T cells by antigen-presenting cells (APCs) to promote T cell clonal expansion, cytokine secretion, and effector function. In the absence of the second signal,antigen-specific lymphocytes fail to respond effectively and are functionally inactivated, or anergic, and resistant to subsequent activationby the antigen. The critical inhibitory function of cytotoxic T lymphocyte–associated antigen 4 (CTLA-4, also known as CD152) was revealed by the fatal lymphoproliferative phenotype of Ctla4-/- mice (2, 3). This function demonstrated that T cell pathways could provide negative as well as positive second signals and provided the first indication that negative second signals could regulate T cell tolerance. The discovery of more members of the B7:CD28 family has revealed additional costimulatory pathways that can provide positive and negative second signals to antigenexperienced effector T cells. The functions of these newer pathways have broadened the concept of costimulation. This review focuses on recent advances in our understanding of one of the newer pathways in the B7:CD28 family, the pathway consisting of the programmed death 1 (PD-1; also known as CD279) receptor and its ligands, PD-L1 (B7-H1; CD274) and PD-L2(B7-DC; CD273). The PD-1 receptor wasdiscovered in 1992 as a gene upregulated in a T cell hybridoma undergoing cell death (4).
The important negative regulatory function of PD-1 was revealed by the autoimmuneprone phenotype of Pdcd1-/- mice in 1999(5, 6). Since the ligands for PD-1 were identified in 2000 (7, 8) and 2001 (9, 10), there has been steady progress in understanding the functions of PD-1 and its ligands. Here, we first describe the structure and expression of PD-1, PD-L1, and PD-L2. Next, we review recent advances in understanding PD-1 and PD-L signaling. We then summarize recent studies that identify B7-1 (CD80) as a binding partner for PD-L1 and indicate that PDL1 interactions with B7-1 can lead to bidirectional inhibitory responses in T cells (11). Finally, we discuss our current understanding of the roles played by PD-1 and its ligands in regulating T cell activation and tolerance and consider the therapeutic potential of manipulation of PD-1 and its ligands.
STRUCTURE OF GENES
PD-1 is a 288 amino acid (aa) type I transmembrane protein composed of one immunoglobulin (Ig) superfamily domain, a ∼20 aa stalk, a transmembrane domain, and an intracellular domain of approximately 95 residues containing an immunoreceptor tyrosine-based inhibitory motif (ITIM) as well as an immunoreceptor tyrosine-based switch motif (ITSM). PD-1 is encoded by the Pdcd1 gene on chromosome 1 in mice and chromosome 2 in humans. In both species, Pdcd1 consists of 5 exons. Exon 1 encodes a short signal sequence, whereas exon 2 encodes an Ig domain. The stalk and transmembrane domains make up exon 3, and exon 4 codes for a short 12 aa sequence that marks the beginning of the cytoplasmic domain. Exon 5 contains the C-terminal intracellular residues and a long 3 UTR. Splice variants of PD-1 have been cloned from activated human T cells (12). These transcripts lack exon 2, exon 3, exons 2 and 3, or exons 2 through 4. All these variants, except for the splice variant lacking exon 3 only (PD-1 ex3), are expressed at similar levels as full-length PD-1 in resting peripheral blood mononuclear cells (PBMCs). All variants are significantly induced upon activation of human T cells with anti-CD3 and anti-CD28. The PD-1 ex3 variant encodes an mRNA that lacks the transmembrane domain and resembles soluble CTLA-4, which plays an important role in autoimmunity (13). This variant is enriched in the synovial fluid and sera of patients with rheumatoid arthritis (14). PD-L1 is a 290 aa type I transmembrane protein encoded by the Cd274 gene on mouse chromosome 19 and human chromosome 9. Cd274 comprises seven exons, the first of which is noncoding and contains the 5 �UTR. The next three exons contain the signal sequence, IgV-like domain, and IgC-like domains, respectively. The transmembrane domain and the intracellular domains are contained in the next two exons (exons 5 and 6). The last exon contains intracellular domain
residues plus the 3UTR. The intracellular domain of PD-L1 is short, only about 30 aa, and highly conserved in all reported species.There is no known function for the intracellular tail of PD-L1. There is one reported splice variant of PDL1 in humans (15) consisting of a sequence lacking the IgV-like domain encoded in exon 2. This mutant should not be able to bind PD- 1, although the function of this splice variant has not yet been reported. No splice variants have been identified for mouse PD-L1. PD-L2 is a type I transmembrane protein encoded by the Pdcd1lg2 gene adjacent to Cd274 and separated by only 23 kb of intervening genomic DNA in mouse and 42 kb in human. The gene comprises six exons in mouse and seven in human. Exon 1 is noncoding, whereas the second exon contains the signal sequence. The IgV-like domain is composed of exon 3, the IgC-like domain is exon 4, and exon 5 contains a short stalk, transmembrane region, and the beginning of the cytoplasmic domain. In mouse exon 5, there is a stop codon that results in a cytoplasmic domain of only 4 aa. In human, exon 6 and 7 contain an additional coding region resulting in a cytoplasmic domain of 30 aa. The longer form of the cytoplasmic domain is found in human, macaque, chimp, dog, cow, pig, and horse but is lost in mouse and rat. The long form of the cytoplasmic domain has no appreciable signaling motifs but is conserved across diverse species, suggesting that the cytoplasmic tail of PD-L2 may have a functional role. There are three PD-L2 splice variants identified from activated human PBMCs (9, 16). Analogous to the splice variant described in PD-L1, one form drops out the IgV-like exon and presumably loses the capacity to bind PD-1 (9). A second form (called type II) drops out the IgC-like domain. Another form (called type III) loses the IgC-like domain and the transmembrane residues but preserves the intracellular residues. The type II form would be expected to bind PD-1, as most binding activity resides in the IgV-like domain (17), and the type III form might represent a soluble ligand for PD-1.STRUCTURE OF PROTEINS
The unbound three-dimensional structure of PD-1 has been obtained by X-ray crystallography and shows that the β-strands of the Ig superfamily fold are well conserved between CTLA-4 and PD-1 (root mean squared deviation of 1.5 A comparing common ˚ α carbons) (18). The CDR3 loop in PD-1 is loosely ordered and does not have conserved amino acids, unlike the binding interface of CTLA- 4, which is centered on the MYPPPY motif of the CDR3 loop and is highly ordered owing to the consecutive prolines in this motif. None of the PD-1 CDR3 amino acids was found to be important for binding PD-L by scanning mutagenesis. Biophysical studies have addressed the question of self-association of PD-1. Fluorescence resonance energy transfer analyses of full-length PD-1 expressed in CHO cells and analytical ultracentrifugation on a soluble extracellular PD-1 IgVlike domain showed PD-1 to be monomeric. Whether PD-1 requires dimerization to transduce signals is unclear.Molecular models of PD-L1 and PD-L2 have been generated on the basis of the crystal structures of B7-1 and B7-2 to guide sitedirected alanine scanning mutagenesis (17).
The binding interfaces of both PD-L1 andjunction with anti-CD3. These non-PD-1.EXPRESSION OF PD-1
AND ITS LIGANDS
PD-1 can be expressed on T cells, B cells, natural killer T cells, activated monocytes, and dendritic cells (DCs) (
Figure 1). PD-1 is not expressed on resting T cells but is inducibly expressed after activation (19). Although PD-1 cell surface protein expression can be detected within 24 h of stimulation, functional effects of PD-1 ligation are observed within a few hours following T cell activation (20).Ligation of TCR or BCR can upregulate PD-1 on lymphocytes, and the level of mRNA transcription does not strictly correlate with protein production (21). In normal human reactive lymphoid tissue, PD-1 is expressed on germinal center–associated T cells (133). PD-1 compartmentalization in intracellular stores has been described in a regulatory T cell population (22, 22a). PD-1 is inducibly expressed on APCs on myeloid CD11c + DCs and monocytes in humans (23), but its function on these cells is not clear. There are no data to support a function for PD-1 in the absence of antigen receptor signaling. The two PD-1 ligands differ in their expression patterns. PD-L1 is constitutively expressed on mouse T and B cells, DCs, macrophages, mesenchymal stem cells, and bone marrow–derived mast cells (24). PD-L1 expression is also found on a wide range of nonhematopoietic cells (see Figure 1)and is upregulated on a number of cell types after activation. Both type I and type II interferons (IFNs) upregulate PD-L1 (25, 26). Analyses of the human PD-L1 promoter demonstrate that both constitutive and inducible PD-L1 expression are dependent on two IFN regulatory factor-1 (IRF-1) binding sites that are between 200 and 320 bp upstream of the transcriptional start site (27). These IRF-1 binding sites are also found in mouse, although their importance has not been directly tested. Several studies have examined which signaling pathways are required for PD-L1 expression by using pharmacological inhibitors. PD-L1 expression in cell lines is decreased when MyD88, TRAF6, and MEK are inhibited (28). JAK2 has also been implicated in PD-L1 induction (27, 28). Loss or inhibition of phosphatase and tensin homolog (PTEN), a cellular phosphatase that modifies phosphatidylinositol 3-kinase (PI3K) and Akt signaling, increases post-transcriptional PD-L1 expression in cancers (29). PD-L2 expression is much more restricted than PD-L1 expression. PD-L2 is inducibly expressed on DCs, macrophages, and bone marrow–derived mast cells. PD-L2 is also expressed on 50% to 70% of resting peritoneal B1 cells, but not on conventional B2 B cells (30). PD-L2 expression on B1 cells tracks with a restricted VH usage that is skewed toward VH11/VH12. PD-L2+ B1 cells bind phosphatidylcholine and may be important for innate immune responses against bacterial antigens. Less is known about transcriptional regulation of PD-L2. Its induction by IFN-γ is partially dependent on NF-κB (31). PD-L2 can also be induced on monocytes and macrophages by GM-CSF, IL-4, and IFN-γ (24, 32).
Signaling Through PD-1
Signaling through costimulatory receptors primarily functions to modify antigen receptor signaling. PD-1 typically has greater effects on cytokine production than on cellular proliferation, with significant effects on IFN-γ, TNF-α, and IL-2 production. PD-1-mediated inhibitory signals depend on the strength of the TCR signal, with greater inhibition delivered at low levels of TCR stimulation. This reduction can be overcome by costimulation through CD28 (8) or IL-2 (33). PD-1 may exert its effects on cell differentiation and survival directly by inhibiting early activation events that are positively regulated by CD28 or indirectly through IL-2(33). Both CD28 and IL-2 promote cell expansion and survival through effects on antiapoptotic, cell cycle, and cytokine genes. IL-2 withdrawal can lead to cell death, another process in which PD-1 has been implicated.There is strong evidence that PD-1 ligation inhibits the induction of the cell survival factor Bcl-xL (20). PD-1 inhibits the expression of transcription factors associated with effector cell function, including GATA-3, Tbet, and Eomes (34). Further studies are required to determine whether PD-1-mediated inhibition is related to its ability to counteract cell survival signals and effector differentiation mediated through CD28, IL-2, Bcl-xL,or a combination of these factors. PD-1 is phosphorylated on its two intracellular tyrosines upon ligand engagement,and then binds phosphatases that downregulate antigen receptor signaling through direct dephosphorylation of signaling intermediates. Two phosphatases, SH2-domain containing tyrosine phosphatase 1 (SHP-1) and SHP-2, can bind to the ITIM and ITSM motifs of PD-1 (35, 36). PD-1 inhibitory function is lost when the ITSM alone is mutated, demonstrating that this tyrosine plays the primary functional role of PD-1 inhibition (20, 36). The association between SHP-1 and PD-1 appears to be weaker than the interaction of PD-1 with SHP-2. Together, these studies suggest that PD-1 functions by recruiting SHP-2, and possibly SHP-1, to the antigenreceptor signaling complex (35).While the binding of SHP-2 to PD-1 is significantly enhanced by PD-1 ligation (20), proximity of PD-1 to the antigen receptor appears to be important for inhibition by PD-1. PD-1 ligation inhibits antigen receptor signaling only in cis and not in trans, indicating that PD-1 ligation must occur close to the site of antigen receptor engagement (37). CTLA-4 moves from an intracellular store to the immunological synapse between an APC and lymphocyte after antigen recognition, depending on the strength of signal (38).In contrast, PD-1 redistributes from uniform cell surface expression to the synapse during T cell–APC interactions (22a). PD-1 could exert its inhibitory effects by bringing SHP-2 into the synapse during antigen receptor signaling, and cross-linking of PD-1 and CD3 increases the amount of SHP-2, but not SHP-1, associated with PD-1 (39). PD-1 ligation inhibits PI3K activity and downstream activation of Akt. In contrast, CTLA-4 inhibits Akt activation but does not alter PI3K activity, indicating that these coinhibitory receptors function through distinct mechanisms. PD-1 ligation inhibits phosphorylation of CD3ζ, ZAP70, and PKCθ(40). Other costimulatory receptors, such as CD150, bind SHP-2 through interactions of their ITIM/ITSM domains with the adaptor protein SH2 domain–containing molecule1A (SH2D1A, also known as SAP) (41). Unlike the CD150 family of cell surface receptors, the ITIM/ITSM motifs of PD-1 do not bind SH2D1A (20). Therefore, PD-1 inhibition likely functions solely through direct interaction with SHP-2, or possibly SHP-1, to directly inhibit early events in the antigen receptor signaling cascades (Figure 2). PD-1 interaction with SHP-1 and SHP-2 is further supported by a peptide immunoprecipitation study that investigated binding partners for the PD-1 ITIM and ITSM motifs using mass spectroscopy. The PD-1 ITIM/ITSM motifs also associate with Lck and Csk (35). These findings suggest that Csk and/or Lck may mediate the phosphorylation of PD-1 in T cells, similar to Lyn phosphorylation of PD-1 in a B cell line (36). PD-1 ligation also reduces Erk activation, but this effect can be overcome through cytokine receptor signaling, particularly cytokines that activate STAT5, such as IL-2, IL-7, and IL-15 (37). SHP-2 positively regulates Erk phosphorylation by interacting with Gab2 after IL-2R ligation (42). Both activation of Erk, which is specifically inhibited by PD-1 ligation, and activation of STAT5, which can overcome PD-1 inhibition, are demonstrated to be involved in the antiapoptotic and proliferative function of IL-2(43). This suggests a model whereby the association of PD-1 with SHP-2 serves not only to dephosphorylate signaling intermediates, but also perhaps to sequester SHP-2 from its positive signaling role in Erk activation.
Reverse Signaling Through
PD-L1 and PD-L2 PD-L1 and PD-L2 not only may influence responses by engaging PD-1 and modifying
TCR or BCR signaling, but also may deliver signals into PD-L1- or PD-L2-expressing cells. The first indication that PD-L2 may transmit signals into DCs came from studies using a novel, naturally occurring human IgM antibody (sHIgM12), isolated from a patient with Waldenstrom’s macroglobulinemia, that binds both mouse and human PD-L2 (44). Although DCs treated with sHIgM12 do not upregulate MHC II or B7 costimulatory molecules, they produce greater amounts of proinflammatory cytokines, particularly TNF-α and IL-6, and stimulate naive T cell proliferation. Treatment of mice with the sHIgM12 enhances resistance to transplanted B16 melanoma and rapidly induces potent tumor-specific CTL (45–47). sHIgM12 also completely blocked the development of airway inflammatory disease in a mouse model of allergic asthma (48, 49). A recombinantly expressed IgM antibody was generated from sHIgM12 that recapitulates observations with the patient-derived antibody, arguing for a specific effect through PD-L2 ligation (50). Further studies are needed to determine how PD-L2 signals into the DC and how these signals influence immunity or tolerance. Additional evidence for reverse signaling through PD-L1 or PD-L2 on DCs comes from studies of bone marrow–derived DCs cultured with soluble PD-1-Ig fusion protein containing the extracellular domain of mouse PD-1 fused to the constant region of human IgG (51). This soluble PD-1 (sPD-1) inhibited DC activation and increased IL-10 production independent of 2,3-dioxygenase (IDO). These effects could be prevented by preneutralization of sPD-1 with anti-PD-1, suggesting that the acquisition of this suppressive DC phenotype is PD-1-specific and occurs via PD-L1 or PD-L2. These findings are consistent with previous studies in which sPD-1 induced IL-10 production by CD4 T cells (52). Bidirectional signaling of PD-1 and PDL may help clarify some of the contradictory results of studies analyzing the PD-1:PD-L pathway. As is discussed in the next section, the discovery of B7-1 as an additional ligand for PD-L1 also may help explain differences observed in functional studies of PD-1 and PD-L1. The identification of bidirectional interactions between B7-1 and PD-L1 that inhibit T cell responses further underscores the importance of reverse signaling through PDL1 and demonstrates that PD-L1 can signal into T cells. B7-1:PD-L1 Interactions Inhibit
T Cell Responses
A number of lines of evidence have suggested a receptor for PD-L1 or PD-L2, aside from PD-1. B7-1 has recently been identified as a binding partner for PD-L1 (11). Surface plasmon resonance studies demonstrate specific and unique interaction between PD-L1 and B7-1, with an affinity (∼1.7 μM) intermediate between the affinities of B7-1 for CD28(4 μM) and CTLA-4 (0.2 μM), and PD-L1 for