In the Thymus, Co-stimulation Kills

In the periphery, co-stimulation rescues a T cell. In the thymus, the same molecules kill it. A peripheral T cell that engages its receptor without CD80 and CD86 becomes anergic; engage them and it responds. Put the identical ligands on medullary thymic epithelium, and strong engagement is what marks a thymocyte for deletion. That inversion looks like a contradiction, and it is the fastest way into what the thymus is actually doing — because the organ is not reading specificity at all. It is reading a quantity.

Thymic T-cell development and central tolerance — 39 clickable nodes from the double-negative stages through β-selection, positive selection in the cortex, deletion in the medulla, Treg diversion and egress. Open the interactive version to click any protein through to its ELISA kit or In Vivo antibody.

The thymus measures a quantity, not a specificity

The selection rules are usually taught as a list of outcomes, which makes them sound arbitrary. They are not. A thymocyte whose receptor binds self-peptide-MHC too weakly receives no survival signal and dies by neglect. One that binds too strongly is deleted. The narrow band in between is what leaves. Everything else on this map is machinery for measuring where a given cell falls on that single axis — LCK starts the signal, ZAP-70 converts it to a biochemical quantity, LAT branches it, and THEMIS tunes the gain.

Two nodes make that quantity visible, and they are the ones worth putting in an experiment. CD5 rises in proportion to the TCR signal a thymocyte has received and then damps it, so a CD5-high cell is one that saw a lot of self. Nur77 does the same job faster and more cleanly: the protein accumulates with signal strength regardless of which ligand delivered it, which is why Nur77 reporter mice are what made the affinity model visible rather than merely plausible. Neither tells you what a cell recognises. Both tell you how hard it was pushed, which is the variable that actually decides its fate.

The arithmetic is brutal. Something close to ninety per cent of thymocytes die by neglect, before deletion has removed anything at all. Failure is the default outcome of this organ, and the interesting cells are the rare ones that clear a low bar without clearing a high one.

Four stages, two markers, and the point of no return

Thymus-seeding progenitors arrive expressing CD117, and at that stage they are still not committed — a DN1 cell retains myeloid and B-cell potential. What closes the door is Notch. DLL4 on cortical epithelium engages Notch1 on the thymocyte, and the consequence of removing that signal is not a smaller thymus but B cells developing inside it. Site does not determine fate; the ligand does.

The four double-negative stages are staged on just two markers, CD44 and CD25, which between them give DN1 through DN4 without any need for receptor sequencing. Running underneath is IL-7 from the thymic stroma acting on CD127. It is worth being precise about what IL-7 does here: it does not instruct lineage, it keeps the cell alive long enough to finish rearranging. Because the stroma makes a limited amount, thymocyte numbers are set by niche capacity rather than by how fast the cells divide.

β-selection spends nothing until one question is answered

RAG1 cuts the DNA that generates the repertoire, and the same enzyme that creates diversity creates double-strand breaks — which is why DN3 is also the stage most prone to translocation. Once TCRβ has rearranged, it pairs with pre-Tα and signals through CD3ε. The checkpoint asks a single question: did rearrangement work? Only cells that can answer yes are permitted to proliferate, which is an efficient design — the expensive expansion happens after the quality check, not before. CD24 falls as cells mature through this, so together with CD5 it lets you stage a thymocyte without knowing its specificity.

The cortex selects on peptides your body never makes again

This is the part that surprises people who know peripheral immunology well. Cortical thymic epithelial cells run β5t, a proteasome subunit found in no other cell type in the body, and use cathepsin L in place of the usual class II editing enzymes. Both generate peptide repertoires that no peripheral cell can reproduce. Positive selection is therefore deliberately conducted against a shadow repertoire — a cell selected in the cortex is not thereby guaranteed to be activated by the same tissue outside it.

THEMIS makes the low-affinity reading possible at all, damping the signal just enough that a weak interaction registers as survival rather than as nothing. Lineage choice then follows from which MHC class the receptor could engage: MHC class I for CD8, MHC class II for CD4. The co-receptor a mature cell carries is a record of how it was selected. If you want the downstream signalling in more detail, the TCR signalling map picks up where this one leaves off, and the antigen processing and presentation map covers the peripheral loading pathway that β5t and cathepsin L deviate from.

In the medulla, AIRE shows the cell the rest of the body

Positive selection switches on CCR7, and CCR7 is what physically drags the survivor from cortex to medulla. Only cells that passed the first test ever sit the second — the geography enforces the order. There, AIRE does something no other transcription factor does: it forces medullary epithelium to transcribe insulin, thyroglobulin and myelin basic protein, tissue-restricted genes in demonstrably the wrong tissue, so that developing T cells meet the body before they are allowed near it. CD40 from the thymocyte licenses that maturation, and XCL1 recruits thymic dendritic cells to cross-present what those rare mTECs made, because one AIRE-expressing cell cannot physically meet every thymocyte.

The consequence of losing AIRE is not a subtle immune defect. It is APECED — multi-organ autoimmunity against exactly the tissues whose genes AIRE would have expressed. And this is where the co-stimulation inversion resolves: CD80 and CD86 here are not rescuing anything, they are raising the signal a self-reactive cell receives above the deletion threshold. Nur77 records that the signal was strong; BIM carries it out. Delete Bim and self-reactive clones walk into the periphery intact, which is the single cleanest demonstration that negative selection is apoptosis and not anergy.

The third fate, and why it needs IL-2 the cell cannot make

A thymocyte receiving an intermediate signal — too strong to simply survive, not quite strong enough to delete — has a third option. It can be diverted into FOXP3 and leave as a regulatory T cell, already carrying CTLA-4 and TGF-β rather than acquiring them later. This requires CD28, which is why CD28-deficient mice lose most of their thymic Tregs, and it requires IL-2 that the developing Treg does not make itself.

That last point has a practical consequence that is easy to miss: because the IL-2 comes from neighbouring thymocytes, Treg development is competitive and the niche is small. It is not a cell-autonomous decision. What happens to these cells afterwards is the subject of the Treg suppression and IL-2 axis map. And the exit itself is a switch rather than a gradient: KLF2 turns on the whole export programme at once, S1PR1 and CD62L together, while CD69 has to come off first because it antagonises S1PR1. That is the mechanism fingolimod exploits from the other direction.

What you can block, and what you can only measure

Fourteen of the thirty-nine nodes on this map carry a functional-grade In Vivo antibody, and unusually they are spread across every stage rather than clustered in one band. That makes the pathway tractable in a way most developmental maps are not — you can intervene at thymic entry, at β-selection, at either selection step, at the Treg branch point, or at egress.

The intracellular machinery — RAG1, LCK, ZAP-70, LAT, THEMIS, β5t, cathepsin L, AIRE, Nur77, BIM, FOXP3, KLF2, S1PR1 — is measurable rather than blockable, and the map says so rather than implying otherwise. One warning that is genuinely load-bearing on this pathway and not a boilerplate caution: LPS causes acute thymic atrophy on its own. A contaminated antibody will shrink and distort the organ you are trying to study, so endotoxin grade here is part of the experimental design, not a purchasing detail. For depleting strategies across other compartments, the In Vivo depletion atlas maps which antibody removes which population.

For research use only; not for use in diagnostic or therapeutic procedures. Explore the full library of interactive pathway diagrams.