PD-1/PD-L1 Checkpoint Signalling: Pathway, Function and Assays

PD-1 does not switch a T cell off; it moves the threshold. Engagement of PD-1 by PD-L1 recruits a single phosphatase into the immunological synapse, and that one recruitment event is enough to collapse calcium flux, ERK activation and the metabolic programme that sustains an effector response. Checkpoint blockade has made the axis famous, but the mechanism is more selective than most summaries admit — and knowing which substrate is dephosphorylated changes how you design a blocking experiment.

The input: PD-L1, PD-L2 and a receptor with two tyrosines

PD-1 (CD279) has two ligands with very different distributions. PD-L1 (B7-H1, CD274) is broadly inducible — on tumour cells, myeloid cells, endothelium and activated lymphocytes — and is driven hard by interferon-γ, the cytokine the attacking T cell itself secretes. PD-L2 (B7-DC, CD273) is far more restricted — largely dendritic cells and macrophages — but binds PD-1 several-fold more tightly. A PD-L1-only blockade therefore leaves an intact PD-L2 arm in any model with a substantial dendritic-cell compartment.

The receptor tail carries two tyrosine motifs, an ITIM and an ITSM, and only the ITSM matters. Mutating the ITSM tyrosine abolishes PD-1 inhibition almost completely; the ITIM mutation is largely tolerated. That asymmetry exists because the phosphorylated ITSM engages both SH2 domains of SHP-2 (PTPN11) simultaneously, prising the N-terminal SH2 domain away from the catalytic pocket and switching the phosphatase on. A single ITIM cannot do that. PD-1 is therefore less an inhibitory receptor than a phosphatase-delivery device.

What gets switched off: the TCR proximal module

The machinery PD-1 acts on starts when the TCR–CD3 complex reads peptide on MHC class I or MHC class II. LCK, held under the co-receptor, phosphorylates the CD3 ITAMs; doubly phosphorylated ITAMs recruit ZAP-70, which phosphorylates LAT. LAT is where the pathway fans out: its phosphotyrosines dock PLCγ1 and GRB2/SOS on one scaffold, so a single clustering event feeds both the calcium and RAS arms.

CD28 ligation by CD80 or CD86 runs in parallel. Its YMNM motif recruits PI3K and GRB2, and its PYAP motif stabilises LCK at the synapse. That is why CD28 is not simply an amplifier: it changes the kinetics of the proximal cascade rather than adding a separate signal. Without it, LAT clusters are short-lived, PKCθ fails to concentrate in the central supramolecular activation cluster, and the cell reads what looks like a partial agonist.

SHP-2 acts on CD28, not the TCR

The least appreciated fact about this pathway is substrate preference. Reconstituted-membrane work showed that SHP-2 recruited to PD-1 dephosphorylates the CD28 tail far more efficiently than CD3ζ or ZAP-70. PD-1 does not primarily blind the T cell to antigen; it removes co-stimulation. That reframes the clinical data — most obviously why blockade fails in CD28-negative T cells, and why the cells that expand after anti-PD-1 therapy are CD28-positive.

PP2A adds a second, mechanistically distinct layer. Where PD-1 chokes off PI3K at source, CTLA-4 recruits PP2A to its own tail and acts further downstream on AKT itself — and it also strips CD80 and CD86 off the antigen-presenting cell by transendocytosis rather than merely competing for them. The two brakes hit the same axis at different points and in different compartments, CTLA-4 during priming in lymphoid tissue and PD-1 in the effector phase at the target site, which is why their blockade combines rather than duplicates.

Downstream collapse: three transcription factors and a metabolic switch

Reduced PLCγ1 activity cuts both second messengers simultaneously. Lower IP₃ means smaller and less sustained Ca²⁺ flux, so calcineurin dephosphorylates less NFAT and nuclear residency is short. Lower DAG means less PKCθ recruitment and weaker IKK activation, so NF-κB stays cytoplasmic; and less RasGRP output means RAS, RAF, MEK and ERK run below the threshold needed to build AP-1. The IL-2 promoter needs all three elements, so partial inhibition at one node costs disproportionate IL-2.

Stoichiometry matters as much as absolute level. NFAT translocating without a partner AP-1 does not produce less of the effector programme — it produces a different one, engaging the anergy and exhaustion modules instead. PD-1 generates exactly that imbalance, because calcium flux is reduced but not abolished while the ERK arm falls away faster. Chronic antigen plus chronic PD-1 engagement is therefore not a weak effector state; it is a distinct differentiation state.

The metabolic arm closes the loop. PD-1 suppresses PI3K and AKT, so mTOR falls and the cell cannot sustain the glycolytic shift effector function demands, defaulting to fatty-acid oxidation. Because AKT is inactive, FOXO1 stays unphosphorylated and nuclear — and nuclear FOXO1 transactivates the PD-1 gene itself. The brake reinforces its own expression, which is why exhausted cells get progressively harder to rescue with blockade alone.

The wider checkpoint network: LAG-3, TIM-3 and Gal-9

LAG-3 (CD223) binds MHC class II with higher affinity than CD4 does, and its inhibitory output depends on a cytoplasmic FxxL motif rather than any ITIM. Functionally it is the checkpoint most tightly coupled to the TCR itself, dissociating LCK from the co-receptor. That wiring is the argument for pairing anti-LAG-3 with anti-PD-1: the two hit non-overlapping steps of the same synapse.

TIM-3 (CD366) is stranger still. It has no ITIM or ITSM, and at rest its cytoplasmic tyrosines are occupied by Bat3, which actively protects the cell from inhibition. Ligation by galectin-9 displaces Bat3 and converts the receptor from a passive to an inhibitory state. This is why TIM-3 expression alone is a poor marker of dysfunction — the receptor only inhibits when its ligand is present, so galectin-9 should be measured alongside it. The four checkpoints are simply not interchangeable, and reagent choice should follow that.

Key targets and matching reagents

Studying PD-1/PD-L1 checkpoint signalling in vivo

Almost every membrane node in this map is addressable with a functional-grade antibody. Experiments fall into three groups.

1. Blocking the axis itself

Anti-PD-1 (RMP1-14) and anti-PD-L1 (10F.9G2) are the standard pair, and running both is informative rather than redundant: receptor blockade frees PD-1 from every ligand, ligand blockade names the compartment supplying the signal. Add anti-PD-L2 whenever dendritic cells or macrophages are prominent, since PD-L2 binds PD-1 with higher affinity and will otherwise sustain inhibition through an apparently successful PD-L1 blockade. All are supplied low-endotoxin and azide-free for repeat dosing.

2. Combination checkpoint blockade

Anti-CTLA-4 (9H10), anti-LAG-3 (CD223) and anti-TIM-3 (CD366) are the obvious partners, and the mechanistic differences above predict what each adds. Fc format matters more here than in most blockade work: anti-CTLA-4 efficacy in mice depends substantially on FcγR-mediated depletion of intratumoural regulatory T cells, so Fc-competent and Fc-silent arms can give opposite answers. Pair with a galectin-9 ELISA before interpreting any TIM-3 result.

3. Dissecting the signal the checkpoint suppresses

To show that a phenotype runs through co-stimulation rather than antigen recognition, titrate anti-CD3 (145-2C11) against anti-CD28, and block CD80 and CD86 separately — the CD28 arm should abolish the benefit of PD-1 blockade. Anti-H-2Kb (Y-3) and MHC class II readouts establish that presentation is intact. Quantify the downstream state with ELISA for ZAP-70, LAT, PLCγ1, ERK, AKT, mTOR, NFAT, NF-κB and FOXO1.

Featured products for this pathway

Frequently asked questions

For research use only. Not for use in diagnostic or therapeutic procedures.