Oncofusion

start / graph / concept-q143-the-kmt2a-n-terminal-antibody-reports-wild-type-plus-fusion-and-the-c-terminal-one-reports-wild-type-alone

Explainerstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q143-menin-under-drug-in-infants

The KMT2A N-terminal antibody reports wild-type plus fusion and the C-terminal antibody reports wild-type alone - which makes an N-versus-C difference a fusion-specific readout with no tag and no engineering

Normal KMT2A is cut in two by an enzyme, and the two halves stay stuck together, so an antibody against either end finds the normal protein. A fusion keeps only the front half. So the front-end antibody sees normal protein AND fusion; the back-end antibody sees only normal protein; and the difference between them is the fusion.

The same thing, for a scientist

Taspase 1 cleaves KMT2A at two conserved sites into an N-terminal ~320 kDa fragment (N320) and a C-terminal ~180 kDa fragment (C180) which re-associate into a stable heterodimer through the FYRN/FYRC interaction. UniProt models this explicitly, carrying separate subcellular-location comments for `MLL cleavage product N320` and `MLL cleavage product C180` alongside the full-length entry. A KMT2A fusion breaks within the breakpoint cluster region, 5' of the FYRN domain, so the fusion protein retains the N-terminal AT-hooks, the CXXC domain and the menin-binding motifs and loses everything the C-terminal antibody recognises. Consequently **an N-terminal antibody reports (wild-type + fusion) and a C-terminal antibody reports (wild-type only)**, and the sites where N is high and C is not are fusion sites. This is a free internal control available in any primary patient sample - no degron, no HA tag, no isogenic line - and nothing in this corpus had used it before.

Go deeper
Why it is worth naming as a concept

Every direct measurement of fusion occupancy this project holds comes from an engineered system: an HA tag on a KMT2A::MLLT3 degron line, a FLAG tag, an isogenic knock-in. Those cannot be made in a patient sample. The N/C difference can, and GSE252378 runs four antibodies across the two ends in every primary sample - N1 Millipore 05-764 (mouse), N2 CST D2M7U (rabbit), C1 Santa Cruz H10 (mouse), C2 Millipore 05-765 (mouse) - which makes the comparison a two-reagent-per-end design rather than a single-antibody claim.

The four ways it can fail, stated before it is used
  1. Epitope accessibility, not epitope absence. A C-terminal epitope can be masked at a crowded promoter. A low C reading is therefore consistent with the fusion interpretation but does not prove it.
  2. The reciprocal product. A balanced translocation also produces a partner::KMT2A-C reciprocal, which carries the C-terminal epitope and is expressed in a minority of cases; where it is expressed, the C channel is not purely wild-type.
  3. Antibody class. Three of the four reagents here are mouse and one is rabbit, and the two ends are not matched for host, so an N-versus-C difference partly reflects a mouse-versus-rabbit difference. The N pair spans both hosts, which is the mitigating design feature.
  4. It is a difference of two noisy channels, so it is only usable at the top of a ranking where both have real signal. At k = 1000 promoters the N and C channels here converge (accessibility-corrected enrichment 1.32 and 1.02); at k = 50 they separate (3.12 and 1.10).
What it is not

It is not a measure of fusion abundance, because wild-type KMT2A contributes to the N channel and the contribution is not subtractable without knowing the relative amounts. It is a measure of where the N-terminal epitope is present without the C-terminal one, which is a site-level statement.

Facts and figures
mechanismtaspase 1 cleavage of KMT2A into N320 + C180, which re-associate via FYRN/FYRC
why it mattersa fusion-specific occupancy readout that needs no tag, no degron and no engineered line, so it works in primary patient material
novelty in this corpusno prior entry uses the N/C difference as a fusion-specific readout
uniprotQ03164
length AA3969
reads
{
  "N_terminal_antibody": "wild-type KMT2A + N-terminal KMT2A fusion",
  "C_terminal_antibody": "wild-type KMT2A only (plus any expressed reciprocal product)"
}
reagents in this deposit
{
  "N1": "Millipore 05-764 (mouse)",
  "N2": "CST D2M7U (rabbit)",
  "C1": "Santa Cruz H10 (mouse)",
  "C2": "Millipore 05-765 (mouse)"
}
failure modes
[
  "epitope masking rather than epitope absence",
  "an expressed reciprocal partner::KMT2A-C product carries the C epitope",
  "host species is not balanced between the two ends",
  "usable only at the top of a ranking, where both channels have signal"
]
How this could bite us back
  • If the N/C difference becomes a pharmacodynamic readout, remember it reports the N-terminal epitope's presence, not the fusion's activity; a fusion held in an inactive state still carries its N-terminus.
  • A therapy that degrades the fusion but leaves wild-type KMT2A would shrink the N/C difference; a therapy that displaces both would not, and the two look identical in a C-only assay.
Still open
  1. How large is the wild-type contribution to the N channel in a primary KMT2A-r sample? Without that the N/C difference is a site-level statement and not an abundance one.
  2. Does the reciprocal MLLT1::KMT2A-C product get expressed in this patient? The deposit carries a fusion exon-junction mapping file for MLL-ENL but the reciprocal was not checked here.
  3. Could an N/C difference be calibrated against an HA-tagged degron line, where the true fusion occupancy is known, to turn it into a quantitative readout?
Every source behind this entry (3)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    Wild-type KMT2A is cleaved by taspase 1 into N320 and C180 fragments that re-associate, so a C-terminal antibody reports wild-type while an N-terminal antibody reports wild-type plus any N-terminal fusion.

    What this evidence cannot show

    UniProt's two fragment location comments are UNCODED (no ECO:0000269), so they are curator statements rather than pointers to a specific experiment; the taspase-1 cleavage itself is established primary literature.

    Evidence that argues against this (1)
    • UniProt's coded experimental citations for KMT2A localisation are a ZMYND8/NuRD DNA-damage screen and a KSHV latency paper rather than KMT2A localisation studies, so the database's pointers are not the founding observations and should not be cited as such. UniProt Consortium 2026
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    GSE252378 runs two N-terminal and two C-terminal KMT2A antibodies on every primary patient sample, making the N/C difference a two-reagent-per-end design.

    What this evidence cannot show

    Reagents read from `!Sample_characteristics_ch1` antibody lines. Three of four are mouse monoclonals and one (N2, CST D2M7U) is rabbit, so host is not balanced between the two ends.

    Evidence that argues against this (1)
    • Two antibodies per terminus is a stronger design than one, but three of the four reagents here are mouse monoclonals and the hosts are not balanced across the two ends, so part of any N-versus-C difference is a host and clone difference rather than an epitope difference. Janssens DH et al. 2024
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    In this deposit the N/C separation is present at the top of the ranking and absent lower down, so the control is usable only where both channels carry signal.

    analyses/menin-under-revumenib-in-an-infant-patient-pair - pre-registered re-analysis of GSE252378 samples 148752 and 152985 · Q143-menin-under-drug-in-infants (Oncofusion Knowledge Graph) · this repository 2026
    What this evidence cannot show

    Measured in one sample (148752). The convergence at k = 1000 is expected for any difference-of-two-noisy-channels measure and is not evidence against the concept.

    Evidence that argues against this (1)
    • A fusion that occupies many weak sites rather than few strong ones would be invisible to a control that only works at the top of the ranking, so this method is biased toward the strongest fusion sites and says nothing about the long tail. Olsen SN et al. 2022
Connected to