PD-1 Mediated Inhibition of T Cell Activation: CD28 as a Primary Target
Introduction to PD-1 and T Cell Regulation
T Cell Activation Requirements:
T cells are activated through a dual signaling mechanism:
Antigen-specific signals: Originating from the T cell receptor (TCR).
Antigen-independent signals: Provided by co-signaling receptors.
Role of Co-signaling Receptors:
Costimulatory receptors: Deliver positive signals vital for the full activation of naive T cells.
Coinhibitory receptors: Attenuate the strength of T cell signaling.
Key function: Act as checkpoints to prevent uncontrolled T cell activation.
Crucial for maintaining peripheral tolerance and immune homeostasis, especially during infections.
PD-1: A Key Coinhibitory Receptor:
Programmed cell death–1 (PD-1) is a prominent coinhibitory receptor.
Binds to two ligands: PD-L1 and PD-L2.
Ligands are expressed by a diverse array of immune and nonimmune cells.
Regulation of PD-L1 expression: Often induced by interferon-$\gamma$ (IFN\gamma), therefore indirectly controlled by activated T cells that secrete this cytokine.
Regulation of PD-1 expression: T cell activation itself increases PD-1 expression on the T cells.
Consequence (Negative Feedback Loop): During chronic viral infection, T cells can become progressively "exhausted." This exhaustion is partly due to a homeostatic negative feedback loop, driven by heightened expression of both PD-1 and PD-L1.
Clinical Relevance of PD-1 in Cancer Immunotherapy:
The interaction between PD-1 and its ligands has been demonstrated to inhibit effector T cell activity against human cancers.
Therapeutic Strategy: Antibodies designed to block the PD-L1–PD-1 axis have shown significant and durable clinical benefits across various cancer indications.
Predictive Biomarker: Efficacy often correlates with evidence of pre-existing anticancer immunity, indicated by PD-L1 expression.
Intriguing Observation: Clinical benefit frequently correlates more strongly with PD-L1 expression by tumor-infiltrating immune cells (e.g., lymphocytes, monocytic cells, dendritic cells which express CD28 ligands) rather than by the tumor cells themselves (which generally do not express CD28 ligands).
Previously Understood Mechanisms of PD-1 Inhibition:
Despite its importance, the precise mechanism of PD-1–mediated inhibition of T cell function has been poorly understood.
Early Findings: Binding of PD-1 to PD-L1 leads to phosphorylation of two specific tyrosines within the PD-1 cytoplasmic domain.
Recruitment of Effectors: Phosphorylated PD-1 was thought to recruit, directly or indirectly, several molecules:
Cytosolic tyrosine phosphatases: Shp2 and Shp1.
TCR-phosphorylating kinase: Lck.
Inhibitory tyrosine kinase: Csk.
Uncertainty of Direct Targets: Defining the direct targets of these inhibitory effectors is crucial but remained unclear.
Hypothesized Pathways: Previous studies suggested PD-1 activation suppressed TCR signaling, CD28 costimulatory signaling, ICOS costimulatory signaling, or a combination.
Common Effectors: Reported decreases in phosphorylation of signaling molecules like ERK, Vav, PLC$\gamma$, and PI3K, but these are general effectors common to both TCR and costimulatory pathways, and thus not necessarily direct targets of PD-1.
Aim of the Study
To identify the immediate and direct targets of PD-1–bound phosphatase(s) by employing a combination of
in vitrobiochemical reconstitution and cell-based experiments.
Experimental Approach: Biochemical Reconstitution System
System Design: A cell-free reconstitution system was used, where the cytoplasmic domain of PD-1 was bound to the surface of large unilamellar vesicles (LUVs), designed to mimic the T cell plasma membrane (as depicted in Fig. 1A).
Characterization of PD-1 Phosphorylation:
Kinase Identification: Lck, but not Csk, was found to efficiently phosphorylate PD-1
in vitrousing a fluorescence resonance energy transfer (FRET)–based assay (also shown within the overall system setup in Fig. 1A). Csk could weakly phosphorylate PD-1 but notably slowed down PD-1 phosphorylation in the presence of Lck, likely due to its known ability to inhibit Lck (as detailed in fig. S1).Conclusion: Lck is identified as the major PD-1 kinase.
Specificity of Shp2 Recruitment by PD-1:
Binding Partners: Phosphorylated PD-1 directly bound Shp2, but significantly, not Shp1, Csk, SHIP-1, or other tested SH2 domain–containing proteins (quantitatively demonstrated in Fig. 1B).
Quantitative Selectivity: A full titration experiment revealed a 29-fold selectivity of PD-1 towards full-length Shp2 over Shp1 (as shown in fig. S2A). This aligns with qualitative cellular studies. Although the tandem SH2 domains of Shp1 and Shp2 bound phosphorylated PD-1 with indistinguishable affinities (illustrated in fig. S2B), the observed selectivity for full-length Shp2 suggests a tighter autoinhibited conformation for Shp1 compared to Shp2.
Role of Tyrosine Residues: Mutation of either tyrosine (Y224 or Y248) in the cytosolic tail of PD-1 resulted in a partial defect in Shp2 binding, and mutation of both tyrosines completely eliminated binding (as depicted in Fig. 1C and further supported by fig. S3).
Correction to Previous Reports: This quantitative, direct binding assay contradicted earlier qualitative co-IP studies that suggested Y224 was dispensable for Shp2 binding (28, 29).
Conclusion: Both tyrosines in the PD-1 cytosolic domain are crucial for optimal Shp2 binding.
Overall Conclusion: Shp2 is the major effector of PD-1, and Lck-mediated dual phosphorylation of PD-1 is essential for optimal Shp2 recruitment.
Specificity for PD-1: Shp2 recruitment was not observed for other signaling receptors tested, including CD3$\zeta$, CD3$\varepsilon$, CD28, ICOS, DAP10, CD226, CD96, TIGIT, and CTLA4 (as shown through binding assays in Fig. 1D, E).
Note on CTLA4: This finding is significant as CTLA4 was previously reported to co-IP with Shp2 and was thought to suppress T cell signaling partly via Shp2 (30, 31). The current data suggests Shp2 does not directly bind CTLA4, implying other bridging proteins are likely involved in their association
in vivo.Overall: These results highlight an unexpected and high binding specificity of Shp2 for phosphorylated PD-1.
Stability and Feedback Regulation of the PD-1–Shp2 Complex:
Complex Disassembly: Experiments using the FRET assay showed that ATP-triggered phosphorylation of PD-1 led to rapid recruitment of Shp2 (as seen in Fig. 1G) and activation of its phosphatase activity (quantified in fig. S4). Termination of Lck activity by rapid ATP depletion (using apyrase) resulted in complete dissociation of Shp2 (also demonstrated in Fig. 1G).
Conclusion: Shp2 actively dephosphorylates PD-1, which destabilizes the PD-1–Shp2 complex. Continuous Lck kinase activity is required to activate and maintain PD-1–Shp2 mediated inhibitory signaling.
Positive-Negative Feedback Loop: A slow, spontaneous disassembly of the PD-1–Shp2 complex was observed even before Lck activity was terminated (shown in Fig. 1G). This disassembly was not due to ATP depletion, as it persisted even after further ATP addition (as illustrated in Fig. 1H).
Mechanism: This suggests that the activation of Shp2 upon binding to PD-1 allows Shp2 to overcome Lck's activity, leading to a gradual net dephosphorylation of PD-1.
Functional Significance: This Lck, PD-1, and Shp2 network forms a positive-negative feedback loop, enabling the system to rapidly revert to an inactive state in the absence of continued PD-1 ligation or Lck activation.
Identifying Substrates of PD-1–Shp2 Complex
Reconstitution System for Substrate Identification:
To understand how the T cell network responds to varying levels of PD-1 (as seen in development, activation, and exhaustion), a diverse set of T cell signaling components were reconstituted (the overall system design is illustrated in Fig. 2A).
Components Included:
Receptor Cytosolic Domains: PD-1, TCR, CD28, ICOS (another costimulatory receptor).
Tyrosine Kinases: Lck, ZAP70 (a key cytosolic tyrosine kinase that propagates the TCR signal), and in some experiments, the inhibitory kinase Csk.
Downstream Adaptor Proteins: LAT, Gads, SLP76.
Regulatory Subunit of Type I PI3K: p85$\alpha$ (known to be recruited by phosphorylated costimulatory receptors).
Physiological Levels: All protein components were reconstituted at concentrations close to their physiological levels (detailed in fig. S6 and table S1) onto LUVs or added in solution to mimic T cell conditions.
Assay Method: A reaction cascade involving phosphorylation, dephosphorylation, and protein-protein interactions was initiated by ATP addition. The levels of PD-1 on the LUVs were systematically titrated, and the susceptibility of each component to dephosphorylation was measured using phosphotyrosine Western blots (examples shown in Fig. 2B).
Key Finding: CD28 as the Most Sensitive Target:
CD28 was identified as the most sensitive target of PD-1–Shp2, in stark contrast to the TCR or its associated components.
Quantitative Dephosphorylation: CD28 was very efficiently dephosphorylated, exhibiting a 50\% inhibitory concentration (IC_{50}) of approximately 96 \text{ PD-1 molecules/mm}^2 (summarized in table S2, with raw data and dose-response curves presented in Fig. 2B, C).
Comparison with TCR Signaling Components and Lck:
TCR Pathway Targets: PD-1–Shp2 dephosphorylated TCR signaling components (including the TCR intrinsic signaling subunit CD3$\zeta$, associated kinase ZAP70, and downstream adaptors LAT and SLP76) only to a minor extent. Their 50\% dephosphorylation occurred at significantly higher PD-1 concentrations, exceeding 1000 \text{ molecules/mm}^2 (see table S2).
Lck as a Target: Lck, the kinase responsible for phosphorylating TCR, CD28, and PD-1, was the second most susceptible target. Both its activating (Y394) and inhibitory (Y505) tyrosines were approximately 50\% dephosphorylated at similar PD-1 levels (400 to 600 \text{ molecules/mm}^2).
Net Effect: Interestingly, this dephosphorylation pattern suggested a net positive effect of PD-1 on Lck activity, owing to the stronger regulatory impact of the inhibitory tyrosine.
Influence of Csk: The addition of Csk (an Lck-inhibiting kinase) increased the sensitivity of both CD28 and TCR signaling components to PD-1–Shp2, but CD28 consistently remained the most sensitive target (as demonstrated in fig. S7 and table S2).
Sustained Preference: The strong preferential dephosphorylation of CD28 by PD-1–Shp2 was also observed at later time points in the
in vitroreaction (data presented in fig. S8).
Comparison with CD45 Phosphatase:
In contrast to the high specificity of PD-1–Shp2, the transmembrane phosphatase CD45 efficiently dephosphorylated all tested signaling components (as shown in Fig. 2B, C, right panels). CD45 showed only a modest 3 to 4-fold selectivity for CD28 over CD3$\zeta$ and ZAP70 (table S2), highlighting the distinctive specificity of PD-1–Shp2.
Mechanistic Basis of CD28 Sensitivity:
Deconstruction of System: To understand why CD28 is so sensitive, the reconstitution system was broken down into individual modules (conceptually illustrated in fig. S9).
Shp2 Activity: Shp2 alone dephosphorylated CD3$\zeta$ and CD28 with similar activities (as quantified in fig. S9C).
Lck Phosphorylation Rate: However, Lck exhibited a 6-fold higher catalytic rate (kcap) for phosphorylating CD3$\zeta$ compared to CD28 (as shown in fig. S9, D and E).
Explanation: This means CD28 is intrinsically a weaker substrate for Lck phosphorylation, which effectively renders it more vulnerable and sensitive to inhibition by PD-1–Shp2 within the dynamic balance of a kinase-phosphatase network.
Overall Conclusion: Based on reconstitution at physiological concentrations, CD28, and to a lesser extent Lck, are the primary substrates for dephosphorylation mediated by PD-1–Shp2.
Validation in Intact Cell Systems
Colocalization Studies in Living Cells (TIRF Microscopy):
Experimental Setup: Total internal reflection fluorescence (TIRF) microscopy was used with a supported lipid bilayer functionalized with ovalbumin peptide–MHC class I complex (pMHC, TCR ligand) and B7.1 (CD28 ligand).
Observation 1: Strong Colocalization: PD-1 strongly colocalized with the costimulatory receptor CD28 in plasma membrane microclusters (as visually confirmed in Fig. 3 and dynamically shown in movie S1).
Observation 2: Differential Overlap: There was significantly less overlap between PD-1 and TCR (Pearson correlation coefficient (PCC): 0.69 \pm 0.09) compared to PD-1 and CD28 (PCC: 0.89 \pm 0.05). This difference was statistically highly significant (P < 0.0001, Student's t-test, n = 17 cells).
Observation 3: ICOS Colocalization: Interestingly, the ICOS co-receptor, despite not being a PD-1 substrate, also colocalized more strongly with PD-1 than the TCR did (results supported by fig. S10).
Time Course of Colocalization: Strong colocalization of PD-1 and CD28 commenced immediately upon initial cell-bilayer contact (0 s) and was sustained through T cell spreading (30 s).
Immunological Synapse Formation: Molecules exhibited centripetal movement, eventually segregating into a canonical "bull's eye" pattern: a central TCR island surrounded by CD28 and PD-1, with PD-1 partially excluded from the most TCR-rich zone (145 s) (as illustrated in Fig. 3B).
Mechanism of Cluster Formation: These rapid colocalization and actin-driven flow patterns suggest that PD-1 and CD28 clusters form actively on the plasma membrane itself, rather than from extracellular microvesicles.
PD-L1 Requirement: PD-1 remained diffusely localized without PD-L1 on the bilayer (even with pMHC and B7.1) (as shown in fig. S13 and dynamically visualized in movie S2), indicating that PD-L1 is essential to bring PD-1 and costimulatory receptors into close proximity.
Shp2 Recruitment: PD-1 clusters were confirmed as sites of Shp2 recruitment to the membrane (results shown in fig. S12).
Overall Conclusion: PD-1 and CD28 strongly co-cluster within the same plasma membrane microdomains in stimulated CD8+ T cells.
Verification of CD28 as a Preferential Target in Intact T Cells:
Cell System: Jurkat T cells (lacking PD-1 and PD-L1) and Raji B cell line (APC, lacking PD-1 and PD-L1) were used.
Genetic Modification: Jurkat T cells were lentivirally transduced to express PD-1 (approx. 40 \text{ PD-1 molecules/mm}^2) and Raji B cells to express PD-L1 (PD-L1High: approx. 86 \text{ PD-L1 molecules/mm}^2) (expression levels and setup in table S1, Fig. 4A).
Functional Assay (IL-2 Secretion): PD-1+ Jurkat cells, when stimulated by antigen-loaded PD-L1High Raji B cells, secreted significantly less interleukin-2 (IL-2) than when stimulated with antigen-loaded PD-L1- parental Raji B cells. A 63\% decrease in IL-2 was measured at 24 hours (quantified in Fig. 4B, P < 0.0001), unequivocally demonstrating the inhibitory activity of PD-1 signaling in this cellular system.
Phosphorylation Dynamics:
PD-L1 Titration: The strength of PD-L1–PD-1 signaling was titrated by incubating PD-1-expressing Jurkat T cells with varying ratios of PD-L1High to PD-L1- Raji B cells.
CD28 Dephosphorylation: At 2 minutes post-APC and T cell contact, CD28 phosphorylation decreased in a dose-dependent manner correlating with the percentage of PD-L1High cells (as illustrated in Fig. 4C, D).
TCR Pathway Response: In contrast, ZAP70 showed no dephosphorylation, and CD3$\zeta$ exhibited substantially less dephosphorylation compared to CD28 (also shown in Fig. 4C, D).
Transient Effect: Notably, the PD-L1–PD-1 inhibitory effect on phosphorylation was transient, with significantly less dephosphorylation detected at 10 minutes (evident in Fig. 4C, D). This transient nature likely reflects the negative feedback loop observed
in vitro, where recruited Shp2 dephosphorylates PD-1, thus repressing the inhibitory signal.Lower PD-L1 Levels: Even with a Raji B cell line expressing lower levels of PD-L1 (PD-L1Low: approx. 16 \text{ molecules/mm}^2, characteristic of tumor-infiltrating macrophages and tumor cells), a transient dephosphorylation of CD28 was still observed, with little to no effect on TCR signaling components (as shown in fig. S14B, C, at t = 2 \text{ min}).
Overall Conclusion from Both Systems: Results from both membrane reconstitution and intact cell assays consistently demonstrate that PD-1–Shp2 strongly favors dephosphorylation of the costimulatory receptor CD28 over dephosphorylation of TCR (summarized and compared in fig. S15). While some dephosphorylation of TCR components (SLP76, ZAP70) was observed at high PD-L1 levels, consistent with previous reports, direct and quantitative comparisons showed that TCR dephosphorylation was consistently much weaker than for CD28.
Broader Implications for Cancer Immunology and Immunotherapy
Reframing PD-1 Mechanism of Action:
The study's finding of a preferential inhibition of CD28 costimulatory signaling by PD-1 challenges the conventional view that PD-1 primarily suppresses TCR signaling. This implies a re-evaluation of how PD-1 exerts its primary effects on T cell function.
Significance of CD28 in Antiviral and Antitumor Responses:
Role Beyond Priming: Although CD28 costimulation is most commonly associated with the priming of naive T cells, there is increasing evidence that it also plays a critical role in later stages of T cell immunity in contexts such as cancer and chronic viral infections.
Therapeutic Dependence on CD28: Concurrent research (Kamphorst et al., 44) independently demonstrated that the efficacy of anti–PD-L1/PD-1 therapy in restoring antiviral (e.g., lymphocytic choriomeningitis virus, LCMV) and antitumor T cell responses is dependent on CD28 expression by T cells.
Exhaustion Prevention: Blocking B7.1 and/or B7.2 binding to CD28 completely abolished the ability of anti–PD-L1/PD-1 therapy to prevent T cell exhaustion. These
in vivoobservations strongly support the conclusion that PD-1 primarily exerts its inhibitory effect by regulating CD28 signaling.
Clinical Predictors and Therapeutic Strategies:
PD-L1 Expression in Clinical Response: In certain cancer patients, the inhibition of T cell immunity correlates with PD-L1 upregulation in the tumor bed, often in response to IFN$\gamma$ release.
Improved Biomarker: Significantly, PD-L1 expression by tumor-infiltrating immune cells can be a more robust predictor of clinical response than PD-L1 expression by the tumor cells themselves in some cancer types.
Reactivation of Costimulation: Given that tumor-infiltrating immune cells (e.g., lymphocytes, monocytic cells, dendritic cells) express CD28 ligands, whereas tumor cells generally do not, if PD-1's primary target is CD28, then the therapeutic effect of anti–PD-L1/PD-1 agents is likely mediated by the reactivation of costimulatory molecule signaling on effector T cells, either in addition to or, predominantly, instead of TCR signaling.
Expansion of Early Memory T cells: It is plausible that costimulation is required for the expansion of tumor antigen–specific early memory T cells, a process that is controlled intratumorally by B7.1-expressing antigen-presenting cells (APCs). Recent LCMV experiments further support the idea that an early memory population serves as a key target for the expansion observed in anti–PD-L1/PD-1 therapy.
Future Directions:
These findings underscore the critical need to broadly consider and investigate the roles of other costimulatory molecules, in addition to CD28, in the broader context of antitumor immunity.