T-Cell Co-Stimulation and Agonist Checkpoint Signalling Explained

Signal 1 tells a T cell what it is looking at; signal 2 tells it whether to care. The same peptide–MHC engagement that launches a productive effector response can just as easily install an anergy programme, and the difference is decided by a second set of receptors acting over the following hours and days. Two structurally unrelated systems do that work — the CD28 family and the TNF-receptor superfamily — and they are not redundant. Reagents that engage them behave nothing like the blocking antibodies most of us are used to.

Signal 1: recognition sets the template

Recognition comes first. Peptide on MHC class I engages the TCR, and since the TCR has no catalytic activity of its own, everything depends on the Src-family kinase LCK phosphorylating the ITAMs of the associated CD3 chains. Those ITAMs recruit ZAP-70, which phosphorylates LAT. LAT has no enzymatic function at all; its job is to become a phosphotyrosine scaffold nucleating PLCγ1, Grb2 and GADS. Mutate its distal tyrosines and the response collapses despite a structurally perfect receptor.

Two arms leave that scaffold. PLCγ1 cleaves PIP₂; IP₃ empties the ER store and sustained Ca²⁺ activates calcineurin, which dephosphorylates NFAT. Diacylglycerol runs through RasGRP and MAPK to build the Fos–Jun dimer AP-1. Their junction matters more than either arm: the two are productive only together. NFAT arriving without an AP-1 partner does not sit idle — it transcribes an anergy programme of Egr2, Egr3 and Cbl-b. Signal 1 alone is worse than no signal.

Signal 2, part one: CD28 and the B7 ligands

CD28 is the only member of its family present constitutively on resting naive T cells, and its two ligands are not interchangeable. CD86 (B7-2) rises within hours of activation and dominates the first encounter; CD80 (B7-1) appears later, is largely dimeric, and binds CTLA-4 with much higher avidity, making it the preferred partner for the inhibitory arm. The CD28 tail is only 41 residues long, but it carries two working motifs.

The phosphorylated YMNM motif binds the p85 subunit of PI3K, generating PIP₃ and recruiting AKT; AKT lifts the TSC1/2 brake on mTOR, and mTORC1 drives the glycolytic shift without which clonal expansion cannot be sustained. The proline-rich motif pulls Grb2 and Itk back into the LAT signalosome, while PKCθ and CARMA1 route to IKK and release RelA-containing NF-κB, which induces Bcl-xL. CD28 amplifies signal 1 rather than replacing it: a cell that has seen it survives contraction, one that has not is deleted.

Signal 2, part two: the TNF-receptor superfamily arm

4-1BB (CD137), CD27 and CD40 are structurally unrelated to CD28: trimeric type-I receptors with no ITAMs and no kinase domain, signalling purely by clustering and recruiting TRAF adaptors. Their ligands 4-1BBL, CD70 and CD40L (CD154) are trimeric, membrane-bound and tightly restricted. CD70 makes the point: it appears on activated dendritic cells for about a day and is then withdrawn, and enforced constitutive expression exhausts the naive T-cell pool rather than improving immunity.

Downstream the pathway splits on adaptor identity. TRAF2 delivers the fast canonical arm, recruiting cIAP1/2 and driving IKK → RelA within minutes. TRAF3 does the opposite at rest: with cIAP it forms the ligase that continuously degrades NIK (MAP3K14), holding it near zero. Ligation redirects that ligase onto TRAF3 itself; NIK accumulates, phosphorylates IKKα, p100 is processed to p52, and RelB enters the nucleus. This arm runs on protein stability, not phosphorylation — slow to start, slow to stop.

So TNFRSF co-stimulation is qualitatively different from CD28 rather than additive with it: 4-1BB holds Bcl-xL high for days and rescues cells that have already begun to contract. CD40 runs the other way entirely — CD40L on the T cell licenses the dendritic cell, which then upregulates CD80, CD86 and 4-1BBL. An agonistic anti-CD40 antibody is therefore not really a T-cell drug at all; it acts on the antigen-presenting cell, and retains activity in settings where anti-4-1BB has nothing to bind.

ICOS, and why timing decides the experiment

ICOS (CD278) belongs to the CD28 family but behaves like neither relative. It binds only ICOSL (B7-H2, CD275), never CD80 or CD86, and its YMFM motif preferentially recruits the p50α regulatory subunit of PI3K, giving a stronger and more sustained PIP₃ signal than CD28 manages. Lacking the Grb2 site, ICOS drives AKT and mTOR hard while producing little IL-2 — hence its dominance in follicular helper differentiation and its modest contribution to primary CD8 expansion.

Expression kinetics are the constraint most in vivo designs underestimate. CD28 is on the surface before the cell has seen anything; ICOS, 4-1BB and CD70 are induced only after activation and peak between roughly 24 and 72 hours. An agonist given with the antigen and the same agonist given two days later are not the same experiment: the first may find almost no target, the second acts on a population that has already committed.

Why agonists are not blockers in reverse

A blocking antibody only has to occupy a site. An agonist has to reproduce the geometry of a trimeric membrane ligand, and for TNFRSF receptors bivalent IgG binding is not enough — higher-order clustering is required, supplied in vivo in trans by FcγRIIB on neighbouring cells. That turns isotype into a functional variable: mouse IgG1 engages FcγRIIB preferentially and gives far stronger anti-CD40 and anti-4-1BB agonism than IgG2a, the subclass you would pick for depletion.

Both directions of that dial have caused trouble. Systemic anti-CD40 produces transaminitis and cytokine release, and clinical anti-4-1BB development was dose-limited by hepatitis. CD28 supplies the other example: superagonist antibodies binding the laterally exposed C''D loop cluster the receptor with no TCR engagement at all, which is the mechanism behind the TGN1412 cytokine storm. Conventional anti-CD28 clones still require signal 1, which is why they are paired with anti-CD3 (145-2C11).

Key targets and matching reagents

Studying co-stimulation and agonist checkpoints in vivo

Almost every node on the outer ring of this map is a surface receptor or its ligand, which makes the axis unusually tractable with functional-grade reagents from the In Vivo antibody range. Experiments divide into three groups.

1. Engaging the co-stimulatory receptors

Anti-4-1BB, anti-CD40, anti-ICOS (CD278), anti-CD70 and anti-CD28 are not equivalent tools. Anti-CD40 acts on the antigen-presenting cell and works where T cells are absent; anti-4-1BB and anti-ICOS need an already-activated T cell; anti-CD28 needs signal 1 alongside it. Keep subclass constant across arms so FcγRIIB crosslinking does not vary silently between groups.

2. Controlling signal 1 and the B7 context

Anti-CD3 (145-2C11) and anti-H-2Kb (Y-3) set or remove signal 1, while anti-CD80 and anti-CD86 separate the two B7 ligands — worth doing individually, since CD86 dominates early priming and CD80 carries most of the CTLA-4 interaction. Low endotoxin matters here specifically: LPS upregulates both B7 ligands and 4-1BBL on its own.

3. Reading out the signalling and transcriptional response

Make the experiment quantitative with the intracellular kits: LCK, ZAP-70 and LAT for proximal integrity; PI3K, AKT and mTOR for the CD28/ICOS metabolic arm; IKKβ, RelA, TRAF2, TRAF3, NIK and RelB to separate canonical from non-canonical NF-κB; and NFAT with AP-1 to test whether both partners are present. Soluble ICOSL, CD40L and CD27 cover the ligand side, with Bcl-xL as the survival endpoint.

Featured products for this pathway

Frequently asked questions

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