Cell Signaling · Unit 5: Signals that build and defend an animal

Antigen Receptors and Checkpoints

How does a receptor with no enzyme and almost no tail switch on a T cell, and what keeps that cell from attacking healthy tissue? This lesson follows the first minute after an antigen receptor finds its target, from two phosphates on a short motif to three pathways you already know, then the second signal and the two brakes, CTLA-4 and PD-1. Afterwards you will be able to explain how checkpoint antibodies and engineered CAR T cells follow from the wiring.

Watch first Two Phosphates, Two Signals · 2:45

The receptor that tells a T cell it has found an infected cell has a tail inside the cell only a few amino acids long, with no enzyme anywhere in it. It is not a kinase, and unlike the cytokine receptors in JAK-STAT signalling it does not even hold one. So how does it signal, and what keeps a cell that is armed to kill from attacking healthy tissue? The answers became two kinds of cancer treatment.

The idea: one set of chains recognises, another set signals

A T cell is a white blood cell that inspects the surfaces of other cells. An antigen is whatever its receptor recognises: for a T cell, a short peptide held in the groove of an MHC protein, a display molecule on the surface of another cell. The T-cell receptor is a pair of chains, alpha and beta, whose outer tips differ from one T cell to the next.

The receptor’s sequence took years to find, because a T cell releases none of it to purify in quantity. In March 1984 the laboratories of Tak Mak in Toronto and Mark Davis at the NIH and Stanford reported the gene for one of its chains in one issue of Nature, and the predicted protein resembled an antibody chain. Two months later Yueh-hsiu Chien, with Davis, showed that the gene is cut and rejoined in T cells, as antibody genes are in B cells, the white blood cells that make antibodies.

The puzzle was on the inside: no kinase, and no catalytic domain of any kind. The receptor travels with partner chains (CD3 gamma, delta and epsilon, and a chain called zeta), which the receptor needs in order to reach the surface.

In 1989 Michael Reth published a short letter in Nature that contained no experiment. He had lined up published tail sequences of partner chains, among them zeta, CD3 gamma and delta, the chain then called MB-1 (now Ig-alpha) of the B-cell receptor, and two chains of the mast-cell IgE receptor. All shared one pattern, written here in the form used today:

Y-x-x-(L or I), then 6 to 8 residues, then Y-x-x-(L or I)

Here x is any amino acid. Reth proposed that, once receptors were cross-linked, the conserved residues from several chains would come together to form a binding site for the proteins that start the signal. The pattern is now called an ITAM (immunoreceptor tyrosine-based activation motif). One T-cell receptor complex carries ten of them: three on each of its two zeta tails and one on each of its four CD3 tails.

A first test came in 1991, of whether the tail alone could signal. Irving and Weiss joined the outer and membrane-crossing parts of an unrelated protein, CD8, to the zeta tail. The hybrid could recognise no antigen, yet clustering it with an antibody set off the T cell’s activation signals. Other laboratories reported similar hybrids that year. The tail alone was enough.

The B-cell receptor follows the same plan: a membrane-bound antibody with a tail three amino acids long (in its IgM form), flanked by Ig-alpha and Ig-beta with one ITAM each. A Src-family kinase, mainly Lyn, phosphorylates the ITAM, and Syk, a relative of ZAP-70 below, reads it.

The first minute: two phosphates, then three branches

So where is the kinase? It is borrowed from a neighbour. Lck is a Src-family tyrosine kinase (see protein interaction modules), and a share of it rides on the inner tail of a co-receptor, CD4 or CD8. The co-receptor binds an unvarying part of the same MHC protein the receptor is reading, which puts Lck beside the CD3 and zeta tails.

In 1988 two groups (Rudd and colleagues, then Veillette and colleagues) reported that Lck comes out of broken-open T cells attached to CD4 and CD8. That did not show it was needed. Weiss’s laboratory had a mutant T-cell line with an intact receptor and no calcium response; the fault lay in its gene for Lck. In 1992 Straus and Weiss put an intact gene back, and the response returned.

The sequence:

  1. Lck phosphorylates both tyrosines of an ITAM.
  2. ZAP-70 docks and is switched on. ZAP-70 is a second tyrosine kinase, with two SH2 domains side by side, spaced to fit the two phosphotyrosines. It binds tightly only when both are phosphorylated, so one stray phosphate does not start a T cell. Docking opens ZAP-70, and Lck then phosphorylates it.
  3. ZAP-70 phosphorylates LAT, a scaffold (adaptors and scaffolds) set in the membrane, whose tyrosines become SH2 docking sites.

Three branches leave LAT.

Calcium to NFAT. PLC-gamma docks on LAT and splits PIP2 into IP3 and DAG (PLC, IP3, DAG and PKC). IP3 empties the calcium store, store-operated channels keep calcium high (calcium signalling), and calmodulin switches on the phosphatase calcineurin. Calcineurin strips phosphates from the transcription factor NFAT, which then enters the nucleus (the phosphorylation switch). The transplant drug cyclosporin works by blocking calcineurin.

DAG to PKC to NF-κB. DAG recruits protein kinase C, in T cells mainly the theta form, which leads through a scaffold complex to the IκB kinase and so to NF-κB (the lesson before this one).

Ras to ERK. Two exchange factors load Ras with GTP: SOS, brought to LAT by Grb2, and RasGRP, switched on by DAG. Raf, MEK and ERK follow (Ras to MAPK), and ERK helps build the transcription factor AP-1.

NFAT, NF-κB and AP-1 meet on the control regions of genes such as the one for interleukin-2, a cytokine that drives T-cell division through a JAK-STAT receptor. In an experiment that triggered the receptor with a light-activated peptide, LAT was phosphorylated within 4 seconds and a calcium rise followed within 6 to 7 seconds (Huse and colleagues, 2007). The steps down to it use most of Units 2 to 4.

One signal is not enough: the second signal and two brakes

Antigen receptors are varied almost at random, and some fit the body’s own peptides, so the receptor is not trusted to decide alone.

Signal two. A second receptor on the T cell, CD28, must bind a B7 protein (B7-1 or B7-2) on the cell presenting the peptide. Presenting cells display far more B7 once they have detected signs of infection. CD28 has no enzyme either: a tyrosine motif in its short tail, once phosphorylated, recruits PI 3-kinase (PI 3-kinase, Akt and insulin) and Grb2. In the classic experiments on cultured T cells, signal one alone did worse than nothing: the cells entered a lasting unresponsive state called anergy.

Brake one: CTLA-4. An activated T cell puts up a further receptor, CTLA-4, a close relative of CD28 that binds the same B7 proteins. Measured on isolated proteins it binds B7 roughly ten times more tightly than CD28 does, and on a cell the gap is wider, because a CTLA-4 dimer grips two B7 molecules at once. It also pulls B7 off the presenting cell and takes it into the T cell to be destroyed (David Sansom’s laboratory, 2011).

Jeffrey Bluestone’s laboratory in 1994, and Matthew Krummel and James Allison in 1995, reported that an antibody blocking CTLA-4 increased T-cell responses in culture, as releasing a brake would. Mice made without CTLA-4, in the laboratories of Arlene Sharpe and of Tak Mak, died within a few weeks of birth, their lymphocytes multiplying unchecked.

Brake two: PD-1. Tasuku Honjo’s laboratory in Kyoto reported the gene for PD-1 (programmed death-1) in 1992. PD-1 also appears on activated T cells; its partner PD-L1 is displayed by many other cells and by many tumours. This brake works from the inside. When PD-L1 binds, two tyrosines on PD-1’s tail are phosphorylated and recruit SHP-2, a tyrosine phosphatase that removes phosphates the activating kinases have just added. In a membrane system rebuilt from purified proteins in 2017, the target SHP-2 reached most readily was CD28; work in T cells suggests the antigen receptor pathway is damped as well.

Mice lacking PD-1 at first looked normal. Bred onto an inbred background and watched as they aged, they developed autoimmunity (an immune attack on the body’s own tissues) resembling lupus, reported in 1999. That is later and milder than the loss of CTLA-4.

Releasing a brake to treat cancer

Many tumours carry altered proteins that T cells can recognise, and many display PD-L1. What happens if you block a brake instead of attacking the tumour? In 1996 Dana Leach, Krummel and Allison reported that mice given an antibody against CTLA-4 rejected tumours that grew in untreated mice. In 2002 the groups of Honjo and Nagahiro Minato reported that mice held tumours back better when PD-1 could not meet PD-L1.

The drugs are antibodies that cover a brake: ipilimumab for CTLA-4, approved in 2011 for advanced melanoma, a skin cancer, then nivolumab and pembrolizumab for PD-1 in 2014. None is aimed at the tumour cell. Each binds a receptor on the patient’s own T cells, and a response that was being shut down runs on. Allison and Honjo shared the 2018 Nobel Prize.

The price is what the mice predicted. With a brake released, T cells also attack healthy tissue: inflammation of the gut, skin, liver and hormone glands is common, generally more so with CTLA-4 blockade. And many tumours do not respond at all.

Building a receptor from parts

Watch Antibody Outside, Zeta Tail Inside · 1:30

Blocking a brake helps only if the patient already has T cells that recognise the tumour. The other approach gives T cells a new receptor.

In 1989 Gideon Gross, Tova Waks and Zelig Eshhar joined the variable regions of an antibody to the constant regions of T-cell receptor chains, and T cells carrying the hybrid responded to the antibody’s target. Yoshihisa Kuwana and colleagues had published the same kind of construct in 1987. An antibody grips a whole folded protein directly, so such a receptor needs no peptide and no MHC protein. Later designs, from Eshhar’s laboratory and others, were single chains: an antibody fragment outside and the zeta tail inside. This is a chimeric antigen receptor, or CAR.

Design problem

Build a receptor that makes a T cell kill any cell that carries a surface protein of your choice. You may take parts from an antibody and from the T-cell receptor complex. Say which part does the recognising, which gives the first signal and which gives the second. Then name one way your receptor could harm the patient.

One way to do it

Recognition. Take the two variable regions of an antibody against your target and join them with a short linker into a single-chain fragment. Add a hinge and a membrane-crossing segment.

First signal. Attach the zeta tail, with its three ITAMs. When the receptors cluster on a target cell, Lck phosphorylates the ITAMs, ZAP-70 docks, and everything downstream of LAT runs as usual.

Second signal. The first receptors of this kind stopped there, and in early cancer trials the cells often failed to multiply and persist: signal one without signal two. The fix, worked out by several groups including Michel Sadelain’s, was to add the cytoplasmic domain of a co-stimulatory receptor, CD28 or 4-1BB, to the same chain, so that one binding event delivers both signals.

Carl June’s laboratory put such a receptor, aimed at the B-cell protein CD19, into the T cells of patients with leukaemia, a blood cancer. In 2011 it reported an adult in whom the cells multiplied more than a thousandfold.

The harm. The receptor cannot tell a cancer cell from a healthy cell that carries the same protein. CD19 is on normal B cells too, so the treatment removes them and the patient may need antibodies supplied. A second risk: so many T cells fire at once that the cytokines they release inflame the whole body.

How we measure it

Chimeric receptor (domain swap)

Join the outside of one membrane protein to the cytoplasmic tail of another and express the hybrid in a cell. If clustering the hybrid gives the response, that tail is enough to signal. This is how the zeta tail was shown to carry the signal, and it is the principle behind a CAR.

Mutant cell line and rescue

Mutagenise a T-cell line, pick out cells that no longer respond when the receptor is stimulated, and find the damaged gene. Then put an intact copy back. If the response returns, the gene is required, which is how Lck was shown to be needed for the calcium response.

Blocking antibody

An antibody that binds a surface receptor can keep its ligand off without triggering it. If a response rises when a receptor is blocked, the receptor was a brake. The same reagent, given to an animal, tests whether releasing that brake changes the course of a tumour.

Knockout mouse

Delete one gene in the germ line and see what the animal can no longer do. Mice without CTLA-4 and mice without PD-1 showed what each brake holds back. The result can depend on age and genetic background, as it did for PD-1.

Finish line

Test yourself

8 questions

Practice and MCAT-style, with explained answers.

Next · Lesson 30

Ion Channels as Receptors

A chilli pepper and a hot stove open the same protein. How does its message reach your brain in milliseconds?

Continue →

← Lesson 28 · Innate Immunity: Toll-like Receptors and NF-κB

T-cell receptor on a peptide held by an MHC protein — what recognition looks like
T-cell receptor on a peptide held by an MHC protein — what recognition looks like ● LIVE · RCSB PDB 1AO7
open on RCSB PDB ↗
Loading live structure 1AO7 from RCSB PDB…

Find the short peptide lying in the groove of the MHC protein. The receptor's variable loops sit across both the peptide and the rim of the groove, so it reads the two together.

Drag to rotate · scroll or pinch to zoom · right-drag to pan
The whole T-cell receptor with its CD3 partner chains — where the signalling tails begin
The whole T-cell receptor with its CD3 partner chains — where the signalling tails begin ● LIVE · RCSB PDB 6JXR
open on RCSB PDB ↗
Loading live structure 6JXR from RCSB PDB…

Count eight chains: alpha and beta, CD3 epsilon with gamma, CD3 epsilon with delta, and a pair of zeta chains. The cytoplasmic tails that carry the ITAMs are flexible and are not seen in the map.

Drag to rotate · scroll or pinch to zoom · right-drag to pan
The two SH2 domains of ZAP-70 on a doubly phosphorylated ITAM — the reader of two marks
The two SH2 domains of ZAP-70 on a doubly phosphorylated ITAM — the reader of two marks ● LIVE · RCSB PDB 2OQ1
open on RCSB PDB ↗
Loading live structure 2OQ1 from RCSB PDB…

Each SH2 domain holds one phosphotyrosine of the same peptide. The pocket for one of them is built where the two domains meet, which helps explain why the pair binds well only when both tyrosines carry a phosphate.

Drag to rotate · scroll or pinch to zoom · right-drag to pan
CTLA-4 bound to its ligand B7-1 — the brake that outbids CD28
CTLA-4 bound to its ligand B7-1 — the brake that outbids CD28 ● LIVE · RCSB PDB 1I8L
open on RCSB PDB ↗
Loading live structure 1I8L from RCSB PDB…

CTLA-4 and B7-1 are both dimers, and each CTLA-4 dimer reaches two separate B7-1 dimers. Repeated across a membrane, that gives a zipper-like row and a much stronger grip than one-to-one binding.

Drag to rotate · scroll or pinch to zoom · right-drag to pan
PD-1 bound to PD-L1 — the contact the cancer antibodies cover
PD-1 bound to PD-L1 — the contact the cancer antibodies cover ● LIVE · RCSB PDB 4ZQK
open on RCSB PDB ↗
Loading live structure 4ZQK from RCSB PDB…

Two immunoglobulin-like domains meet face to face. Antibodies such as pembrolizumab and nivolumab bind PD-1 on a surface that overlaps this face, so PD-L1 cannot dock.

Drag to rotate · scroll or pinch to zoom · right-drag to pan
Michael Reth and Arthur Weiss

Michael Reth and Arthur Weiss

Reth noticed, by lining up published sequences, the paired-tyrosine motif now called the ITAM in the tails of antigen receptor partner chains. Weiss's laboratory showed that the zeta tail alone can signal, that the kinase Lck is required, and cloned ZAP-70.

By the late 1980s the T-cell receptor had been cloned, and the part of it inside the cell contained no enzyme. In 1989 Michael Reth published a short letter in Nature with no experiment in it. He had set published sequences of partner-chain tails side by side and seen one arrangement in all of them: two tyrosines a fixed short distance apart, each followed two residues later by a leucine or an isoleucine. He proposed that, once receptors were cross-linked, these residues would come together to form a binding site for the proteins that start the signal. In 1991 Irving and Weiss tested whether the zeta tail alone could signal, by joining the outer part of CD8 to the zeta tail; clustering this hybrid activated a T cell that could see no antigen through it. The next year Weiss's laboratory restored the kinase Lck to a mutant T-cell line that could not signal, and the calcium response came back. It also cloned ZAP-70, the kinase that reads the doubly phosphorylated motif. Knowing that one short tail carries the whole signal is what later let engineers attach that tail to an antibody fragment.

✓ verbatim Reth, Nature 338:383 (1989); Irving & Weiss, Cell 64:891 (1991).
Matthew Krummel and James Allison

Matthew Krummel and James Allison

They reported in 1995 that blocking CTLA-4 increases T-cell responses, so it opposes CD28. Allison's laboratory then showed that an antibody against CTLA-4 makes mice reject tumours.

CTLA-4 looks like CD28 and binds the same B7 proteins, so in the early 1990s it was reasonable to think it was a second accelerator. Matthew Krummel, in James Allison's laboratory at Berkeley, tested this with antibodies on cultured T cells. Blocking CTLA-4 raised proliferation, which is what releasing a brake does. Jeffrey Bluestone's laboratory had reported the same conclusion a year earlier, in 1994. Allison's laboratory then asked a different question. Dana Leach injected mice with tumour cells and gave some of them the blocking antibody, and the treated mice rejected their tumours. Allison had the experiment repeated blind, and by his own account he measured the tumours himself without knowing which mice were treated. The paper appeared in 1996. For years, he has written, drug companies turned the idea down. The antibody that finally reached patients, ipilimumab, was approved in 2011, and Allison shared the 2018 Nobel Prize. The work showed that a cancer drug can act on the patient's T cells and never touch the tumour.

✓ verbatim James P. Allison, Nobel Lecture, 7 December 2018; Krummel & Allison, J. Exp. Med. 182:459 (1995).
Tasuku Honjo

Tasuku Honjo born 1942

His laboratory cloned PD-1, showed that mice lacking it develop autoimmune disease, and with collaborators identified its partner PD-L1 and showed in mice that blocking the pair holds tumours back.

In 1991 Yasumasa Ishida, in Tasuku Honjo's laboratory at Kyoto University, was looking for genes switched on in dying cells. He compared the messenger RNAs of two cell lines pushed toward death with those of the same lines left alone, and one new gene came out. The laboratory named it programmed death-1, on a guess about cell death that turned out to be wrong. Mice made without the gene at first looked normal. Bred onto an inbred background and watched for longer, they developed an autoimmune disease with arthritis and inflamed kidneys, reported in 1999. That marked PD-1 as a brake on the immune system. In 2000 Gordon Freeman's group and Honjo's identified its partner, PD-L1, and in 2002 Honjo's and Nagahiro Minato's groups reported that mice held tumours back better when the pair could not meet. Honjo has described how hard it was to find a company willing to develop an antibody. Antibodies against PD-1, nivolumab and pembrolizumab, are now used against many cancers, and Honjo shared the 2018 Nobel Prize with Allison.

✓ verbatim Tasuku Honjo, Nobel Lecture, 7 December 2018; Ishida et al., EMBO J. 11:3887 (1992).
Zelig Eshhar and Carl June

Zelig Eshhar and Carl June

Eshhar's laboratory joined antibody variable regions to T-cell receptor chains and got a working receptor with an antibody's specificity. June's laboratory took a later design, with a co-stimulatory domain added, into patients with leukaemia.

A T cell normally recognises a peptide held by an MHC protein, while an antibody grips whole folded surfaces. In 1989 Gideon Gross, Tova Waks and Zelig Eshhar joined an antibody's variable regions to the constant regions of T-cell receptor chains. A T cell carrying the hybrid was activated by the antibody's target. Yoshihisa Kuwana and colleagues in Japan had published the same kind of construct in 1987. Later designs put the antibody fragment directly on the zeta tail. In early cancer trials, T cells carrying these often failed to persist in patients; they gave signal one without signal two. Adding a co-stimulatory tail from CD28 or 4-1BB fixed that, and several groups worked it out, Michel Sadelain's among them. Carl June's laboratory at the University of Pennsylvania put such a receptor, aimed at the B-cell protein CD19, into patients' own T cells. In 2011 it reported an adult with leukaemia in whom the cells multiplied more than a thousandfold. In 2012 Stephan Grupp's team treated six-year-old Emily Whitehead the same way. Engineered T cells are now an approved treatment for several blood cancers.

✓ verbatim Gross, Waks & Eshhar, Proc. Natl. Acad. Sci. USA 86:10024 (1989); Porter et al., N. Engl. J. Med. 365:725 (2011).
Test yourself
Question 1 of 8 Score 0 / 0

Which part of the T-cell receptor complex carries the ITAMs that Lck phosphorylates?
Sources & Integrity

Every structure, slide, simulation, and clip on this page is listed below with the source it came from. None of the 3D structures is drawn by hand or by AI: each one loads by accession ID from RCSB PDB or PubChem. This panel is generated from the lesson's own data file, so what it lists is what the page loads.

Molecular structures live data

  • T-cell receptor on a peptide held by an MHC protein — what recognition looks like — RCSB PDB 1AO7 · coordinates CC0 · Garboczi et al., Nature 384:134 (1996); a human receptor on HLA-A2 holding a viral peptide
  • The whole T-cell receptor with its CD3 partner chains — where the signalling tails begin — RCSB PDB 6JXR · coordinates CC0 · Dong et al., Nature 573:546 (2019); cryo-electron microscopy of the human complex
  • The two SH2 domains of ZAP-70 on a doubly phosphorylated ITAM — the reader of two marks — RCSB PDB 2OQ1 · coordinates CC0 · Hatada et al., Nature 377:32 (1995); tandem SH2 domains with a zeta-chain ITAM peptide
  • CTLA-4 bound to its ligand B7-1 — the brake that outbids CD28 — RCSB PDB 1I8L · coordinates CC0 · Stamper et al., Nature 410:608 (2001)
  • PD-1 bound to PD-L1 — the contact the cancer antibodies cover — RCSB PDB 4ZQK · coordinates CC0 · Zak et al., Structure 23:2341 (2015); human proteins

Scientist profiles & quotes

  • Michael Reth and Arthur Weiss — Reth, Nature 338:383 (1989); Irving & Weiss, Cell 64:891 (1991).
  • Matthew Krummel and James Allison — James P. Allison, Nobel Lecture, 7 December 2018; Krummel & Allison, J. Exp. Med. 182:459 (1995).
  • Tasuku Honjo — Tasuku Honjo, Nobel Lecture, 7 December 2018; Ishida et al., EMBO J. 11:3887 (1992).
  • Zelig Eshhar and Carl June — Gross, Waks & Eshhar, Proc. Natl. Acad. Sci. USA 86:10024 (1989); Porter et al., N. Engl. J. Med. 365:725 (2011).

Quotes are verbatim from the cited source. Profile text is original to this site and was checked against the sources cited.

Animated explainer

Two Phosphates, Two Signals · narration: an AI narrator voice (Gemini, Charon) reading narration written from this lesson's text · animation: AI-generated in code (Manim) by Claude, a schematic with no drawn molecular structures, plots computed from their equations · captions from the narration script.

Section films

  • Antibody Outside, Zeta Tail Inside · narration: an AI narrator voice (Gemini, Charon) reading narration written from this lesson's text · animation AI-generated in code (Manim) by Claude, a schematic with no drawn molecular structures · captions from the narration script.

Images & figures

This lesson carries no raster figures. Every structure is rendered live from RCSB PDB or PubChem, and every plot is drawn from its equation in the browser.

integrity Machines check the mechanics. The build confirms that every local file exists, that every question's answer key points at a real choice, and that any AI-generated scientific figure names the person who checked it; a weekly job re-resolves every accession ID against its database. Questions are original; MCAT-style items are written for this site and are not real AAMC questions. Banned sources (the Tymoczko textbook, publisher decks, copyrighted figures) are not used.

Like I'm five

Your body has guard cells called T cells. A guard needs two things before it attacks: it has to spot a bad sign on another cell, and a helper has to tell it to go ahead. Guards also have brakes so they do not hurt the healthy parts of you. Some cancer medicines take the foot off a brake, and others give the guard a new pair of eyes.
High school

T cells are white blood cells that check other cells for signs of infection or cancer. Each one has a receptor on its surface that recognises one particular fragment of protein displayed by another cell. The receptor cannot send a message into the cell by itself, so partner proteins beside it do that once an enzyme tags them with phosphate groups. To avoid mistakes, the T cell needs a second go-ahead signal from the other cell before it acts, and it carries two brakes, called CTLA-4 and PD-1, that slow it down afterwards. Tumours can press the PD-1 brake to protect themselves. Two cancer treatments come from this: antibody drugs that block a brake, and T cells given a designed receptor that points them at cancer cells.
College

An antigen receptor separates recognition from signalling. The alpha and beta chains of the T-cell receptor bind a peptide held by an MHC protein but have almost no cytoplasmic tail; the associated CD3 and zeta chains carry ITAMs, pairs of tyrosines at a fixed spacing. The Src-family kinase Lck, carried on the co-receptor CD4 or CD8, phosphorylates both tyrosines. ZAP-70 binds the pair through its tandem SH2 domains, is activated, and phosphorylates the scaffold LAT. From LAT, PLC-gamma produces IP3 (calcium, calcineurin, NFAT) and DAG (protein kinase C, NF-κB), and Ras drives ERK. Full activation also needs CD28 bound to B7 on the presenting cell. Two inhibitory receptors limit the response: CTLA-4 competes for B7, and PD-1 recruits the phosphatase SHP-2. Antibodies that block either one release T cells against tumours, and a chimeric antigen receptor fuses an antibody fragment to co-stimulatory and zeta tails.
Graduate

The receptor complex has ten ITAMs (one each on CD3 gamma, CD3 delta and the two CD3 epsilon chains, three on each zeta), with the consensus YxxL/I-x(6-8)-YxxL/I. Selectivity comes from coincidence detection: the tandem SH2 domains of ZAP-70 bind a doubly phosphorylated ITAM far more tightly than a singly phosphorylated one, and in the crystal structure the pocket for one phosphotyrosine lies at the interface of the two domains. Binding releases ZAP-70 from autoinhibition and Lck phosphorylates it. Children without functional ZAP-70 have a severe immunodeficiency in which CD8 T cells are nearly absent and CD4 T cells are present but do not signal. CTLA-4 acts largely from outside the cell, by higher-affinity, bivalent binding to B7 and by removing B7 from the presenting cell. PD-1 acts from inside, through two phosphorylated tyrosine motifs that recruit SHP-2. An open question is what PD-1 dephosphorylates in a living T cell. In a membrane system rebuilt from purified proteins, CD28 was dephosphorylated at much lower PD-1 densities than the zeta chain. Proteomic work in primary mouse T cells found the antigen receptor and CD28 pathways damped to a similar degree, and T cells lacking SHP-2 still respond to PD-1 blockade, perhaps because the related phosphatase SHP-1 substitutes.
Fourteen pages of one issue

For years nobody could get enough of the T-cell receptor to read its sequence, because a T cell keeps the protein in its own membrane and releases none of it, while an antibody is secreted into blood in quantity. On 8 March 1984 one issue of Nature carried three papers in a row: the human gene from Tak Mak's laboratory in Toronto, with Yusuke Yanagi as first author, then the mouse gene and its predicted protein from Mark Davis's laboratory, with Stephen Hedrick as first author. Both groups had worked backwards from messenger RNAs that T cells make and B cells do not, and both found a protein built like an antibody chain.

— Yanagi et al., Nature 308:145 (1984); Hedrick et al., Nature 308:149 and 308:153 (1984).
Measuring tumours over the holiday

Dana Leach's first experiment with the CTLA-4 antibody gave a striking result: every treated mouse rejected its tumour. James Allison asked for a repeat done blind. By Allison's account, Leach set it up and left for his Christmas vacation, so Allison measured the tumours himself without knowing which mice had been treated. For about two weeks every tumour grew, and he kept recording. Then some stopped growing and shrank away. When it was revealed which mice had been treated, the mice that had cleared their tumours were all in the treated group.

— James P. Allison, Nobel biographical essay and Nobel Lecture (2018); Leach, Krummel & Allison, Science 271:1734 (1996).
Named for a death it does not cause

In 1991 Yasumasa Ishida, in Tasuku Honjo's laboratory in Kyoto, pushed two cell lines toward death and looked for messenger RNAs that only the dying cells made. One new gene came out, and the laboratory named it programmed death-1, hoping it had a part in the death of T cells. It does not. Mice lacking the gene looked normal at first and developed autoimmune disease only later, which showed that the protein restrains T cells that are already active. The name stayed, and antibodies against the protein now treat cancer.

— Ishida et al., EMBO J. 11:3887 (1992); Tasuku Honjo, Nobel Lecture, 7 December 2018.

 ↑ Top