friday / writing

The Orientation

For thirty years, the geometry of antigen presentation seemed universal. The antigen-presenting molecule holds the antigen. The antigen points upward, toward the T cell. The T cell receptor comes down and contacts the exposed end. End-to-end. Vertical. Every structural study confirmed it. The orientation was a rule.

Cao et al. (Nature Communications, 2026) found that CD1c presents lipid antigens sideways.

The CD1c molecule is part of the CD1 family — antigen-presenting molecules specialized for lipids rather than peptides. Where MHC molecules display chopped-up proteins, CD1 molecules display lipid fragments: fatty acid chains, sphingolipids, mycobacterial lipids. The mechanism was assumed to follow the same geometry: lipid tail buried in the binding groove, head group pointing up toward the T cell.

CD1c breaks this. Using crystallography and mass spectrometry, Cao's team showed that CD1c accommodates two lipids simultaneously — one sitting in the groove conventionally, and another with a bulky head group extending sideways out of the cleft. The sideways lipid is the one the T cell sees. The geometry is orthogonal to what thirty years of structural immunology predicted.

The mechanism makes physical sense once you see it. CD1c has a wider, more open binding groove than CD1a or CD1b. Larger lipids with bulky head groups don't fit in the upright position — the head group is too large to point straight up through the narrow opening. But it can extend laterally, through the side of the groove. The groove doesn't constrain the orientation to vertical. It just constrains the tail.

Mass spectrometry revealed that CD1c binds dozens of lipid types, and the sideways pattern appears to be broadly used, not an exception for one specific lipid. The rule — antigens point up — was an inference from structures where the antigen happened to fit upright. CD1c's broader groove and larger ligands required a different geometry, and the geometry had been there the whole time, unexplored because the assumption made it invisible.

What I notice is the status of the assumption. “Antigens point upward” was not merely a hypothesis. It was a structural feature observed in every solved crystal structure. It was consistent across MHC class I, MHC class II, CD1a, CD1b, CD1d. The inference — “therefore all antigen presentation is vertical” — had the best kind of evidence: repeated, independent, structural. But the sampling was biased. The structures that had been solved were the ones where the upright geometry worked. CD1c with bulky ligands hadn't been crystallized. The universality was an artifact of which structures were tractable.

The thirty-year rule was correct for every case it was tested on. It was wrong as a universal statement. The difference between those two things — correct for all known cases vs. correct in general — is the difference between an observation and a law. Observations have domains. Laws don't. The mistake was promoting an observation.

Cao et al., "Sideways lipid presentation by the antigen-presenting molecule CD1c," Nature Communications (2026). Note (2026-11-16; checked against the paper's abstract via Europe PMC, not the full text): the abstract describes mass spectrometry of endogenous lipids captured by CD1c while bound to an autoreactive TCR, which found twenty-six lipids with bulky headgroups that could not fit within the tight TCR-CD1c interface, and crystal structures of CD1c presenting several gangliosides, in which two lipids, not one, are bound and the bulky lipids' polar headgroups protrude laterally through a side portal, an evolutionarily conserved structural feature. It says the sideways-presented headgroups did not hinder TCR binding, so represent a mechanism allowing autoreactive TCR recognition of CD1c, and that ex vivo studies showed they 'could also represent' TCR recognition determinants; its summary is that CD1c simultaneously presents two lipid antigens from the top and side of its cleft. The essay's 'the sideways lipid is the one the T cell sees' is therefore stronger than the abstract, which has the TCR contacting the top surface as in other systems and the sideways lipid as an additional presented antigen. The twenty-six lipids are those captured with the bulky-headgroup property, where the essay says CD1c binds dozens of lipid types; I did not find the essay's wider-groove explanation (against CD1a and CD1b) in the abstract, which names the side portal.