17. The plates that break the rules · Prof. Kwon's TCR-HLA and CD4/CD8, First Steps

Up to lesson 16 we have used three statements as rules.

① Plates differ between people

HLA-A alone has 5,839 alleles (lesson 8).

② The groove holds a peptide

Nine residues on class I, 13-17 on class II (lesson 2).

③ Coreceptors split the classes

CD8 for class I, CD4 for class II (lessons 5 and 9).

This lesson looks at plates that break all three. Not odd exceptions — a substantial share of human T cells look this way.

There are 377 alleles, yet the groove barely moves. The names are many; the molecule is one.

Python Runs in your browser

# Is HLA-E really the same in everybody? Same method as lesson 8.
import thla

def alpha1(gene):
    d = thla.hla(gene, domain="α1")
    return d["entropy"].mean(), int((d["n_aa"] > 1).sum()), len(d)

print(f"{'gene':<8}{'alleles':>10}{'alpha1 entropy':>17}{'variable pos':>14}")
print("-" * 50)
for g in ["A", "B", "C", "E"]:
    ent, var, tot = alpha1(g)
    print(f"{g:<8}{thla.facts()['hla_n'][g]:>10,}{ent:>17.4f}{var:>10}/{tot}")
print()
print(f"The HLA-A groove is {alpha1('A')[0]/alpha1('E')[0]:.0f} times more "
      f"variable than the HLA-E groove.")

E*01:03 is nearly all of it — the human population has effectively two meaningful variants.

Python Runs in your browser

# Is HLA-E in our table at all
import thla

d = thla.vdj()
e = d[d["mhc_a"].str.startswith("HLA-E")]
print(f"rows restricted by HLA-E : {len(e):,}   {100*len(e)/len(d):.2f}% of everything")
print()
print("plate names")
for a, n in e["mhc_a"].value_counts().head(4).items():
    print(f"   {a:<22}{n:>7,}")
print()
print("fragments")
for ep, n in e["epitope"].value_counts().head(5).items():
    print(f"   {ep:<14}{n:>6,}")

Overlay the per-position variability of a classical and a non-classical plate. Two worlds split apart in one picture.

Loading…

Plates that hold something other than a peptide

🦠

MR1 — a vitamin metabolite

It carries the small chemicals bacteria produce while making riboflavin (vitamin B2). The cells that read them are MAIT cells, and they are 1-10% of human blood T cells.

🧈

CD1 — lipids

Its groove is deep and fat-friendly, so it carries lipids and glycolipids. Cells that read lipids on CD1d are iNKT cells.

🧬

HLA-E — somebody else's signal sequence

It carries the signal peptides of other HLA molecules. Which makes HLA-E a dashboard reporting whether the other plates are being built properly.

MR1 and CD1 are also nearly identical between people. So all three are plates that work the same way in everybody.

T cells are not the only readers of HLA-E

HLA-E has a second audience: NK cells.

Plenty of HLA-E on the surface → NK NKG2A binds → the NK cell stops

So a cell can put NK cells to sleep merely by displaying HLA-E — and tumours use the move. The antibody that cuts this axis is monalizumab (anti-NKG2A).

Lesson 18 picks this up again. A tumour that clears away classical class I becomes invisible to T cells but a target for NK cells — unless it leaves HLA-E behind, which puts the NK cells to sleep as well.

🤔 Check yourself

On what grounds do we call HLA-E 'nearly monomorphic' when it has 377 alleles?

🤔 Check yourself

What does MR1 carry in its groove?

🤔 Check yourself

Why are non-classical plates attractive for cell therapy?