Oncofusion

start / graph / concept-kmt2a-fusion-polarity-is-inverted-in-sarcoma

Explainerstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q92-kmt2a-sarcoma

In leukaemia KMT2A gives the front of the protein; in sarcoma it gives the back. Every consequence follows from that.

also called KMT2A 5-prime versus 3-prime partner · inverted KMT2A fusion polarity · KMT2A as the 3' fusion partner

When KMT2A causes leukaemia it always donates its front end - the part that holds on to menin and finds the right DNA - and gets a new back end from its partner. In the KMT2A sarcomas the arrangement is the other way round, so the fusion protein keeps KMT2A's back end and loses the front end that the leukaemia drugs were designed to attack.

The same thing, for a scientist

The KMT2A fusion literature has one structural invariant: in leukaemia KMT2A is the 5' partner, the breakpoint lies in the exon 8-14 breakpoint cluster region, and the chimera therefore retains the N-terminal AT-hooks, the CXXC domain, the menin-binding motif and both LEDGF integrase-binding motifs while losing the C-terminal SET methyltransferase. In KMT2A-rearranged sarcoma the polarity is reversed: KMT2A is the 3' partner (VIM::KMT2A, LDB1::KMT2A) or an internal insert (YAP1::KMT2A::YAP1, CBX6::KMT2A::PYGO1), and the pathologists who defined the entity state the consequence explicitly - the DNA-binding domain is partially disrupted or totally lost and the SET domain is intact. Translating the published exon breakpoints into residues shows the sharper version: the retained KMT2A begins at exon 2, exon 4 or exon 5 in every defined case, and the menin-binding motif is in exon 1. So the invariant that makes the menin interface universally available in leukaemic KMT2A fusions fails in every sarcoma arrangement with a published breakpoint. This concept exists so that no entry in this graph reasons from 'KMT2A rearrangement' to 'menin-dependent' again.

Go deeper
The two polarities side by side
leukaemic KMT2A fusionsarcoma KMT2A rearrangement
KMT2A's role5' partner, always3' partner or internal insert, always (in the defined arms)
breakpoint in KMT2Aexons 8-14 (breakpoint cluster region)exon 2 (VIM), exons 4-6 (YAP1), exon 4 (LDB1)
KMT2A residues retainedaa 1 to ~1362-1607aa 145-3969 (VIM), aa 1112-1212 only (YAP1), aa 1053-3969 (LDB1)
menin-binding motif (aa 6-25)retainedlost
LEDGF IBM1 (aa 123-134)retainedlost
AT-hooks (aa 169-309)retainedlost in YAP1 arm, retained in VIM and LDB1 arms
CXXC (aa 1147-1195)retainedretained in all
SET (aa 3829-3945)lostretained in VIM and LDB1 arms
partner functiontranscription elongation (SEC, DOT1L complex, P-TEFb)Hippo co-activator, intermediate filament, LIM-domain binding protein, chromobox

Polarity and SET/DNA-binding statement: 1. Breakpoint exons: 234. Exon-to-residue translation: 5.

Why this is more than a bookkeeping difference

Three things this project treats as properties of 'a KMT2A fusion' are in fact properties of the 5'-KMT2A polarity:

  1. Menin dependence. The whole menin-inhibitor franchise targets a motif at aa 6-25. A 3'-KMT2A fusion does not contain it. The drug class has no site, not a weak site.
  2. Loss of catalytic activity. The standard account of KMT2A fusion oncogenesis is that the fusion is a dominant, catalytically dead protein. In the 3'-polarity sarcoma fusions SET is retained, so that account does not transfer, and the therapeutic logic inverts.
  3. Partner-supplied elongation. A 3' KMT2A is not receiving a new C-terminus at all; it is receiving a new N-terminus. Whatever the partner does, it is not replacing the C-terminal effector module.

And one thing that does transfer: the CXXC domain is retained under both polarities, and the sarcoma authors argue from its consistent retention that it is doing work 2. That makes CXXC the only element of KMT2A common to the leukaemic and sarcomatous rearrangements, which is an argument for looking harder at it - and a reason to be careful, because a fixed breakpoint region produces consistent retention whether or not the retained element matters.

Two sources that look contradictory and are not: 'DNA binding domain lost' versus 'CxxC preserved'

Recorded explicitly because it was raised as a contradiction in this entry, and the resolution is a real finding rather than a wording fix.

  • Kao et al., verbatim: "In SEF cases, KMT2A is presumably the 3' fusion partner gene and the most common breakpoints were observed in exons 4-5, resulting in a partially disrupted or totally lost DNA binding domain and intact SET domain." 1
  • Massoth et al., verbatim (abstract): "with preservation of KMT2A CxxC-binding domain in the YAP1-KMT2A-YAP1 fusion and concurrent deletions of corresponding exons in KMT2A." 2

They are not opposite claims. Two sentences earlier, Kao uses the same phrase in the plural for the leukaemic case - "retained N-terminal DNA binding domains" - which in KMT2A means the AT-hooks and the CXXC together. A fusion beginning at KMT2A exon 4 or 5 loses all three AT-hooks (which are entirely within exon 3) and keeps the CXXC, which is precisely "partially disrupted". Kao never says the CXXC is lost.

And the "totally lost" half has a named case. Massoth's Table 1 footnote records a breakpoint inside KMT2A exon 5 at amino acid 1143, within a 'pre-CxxC' region (aa 1115-1154) that the authors state accounts for part of the multivalent bond with the PAF1 complex. The CXXC zinc finger starts at 1147. So one allele truncates it 6.

The correct combined statement, which is sharper than either source alone: across the sarcoma arrangements the retained KMT2A fragment keeps the CpG address (CXXC) in the usual configurations and discards the menin and LEDGF tether (MBM aa 6-25 and IBM1 aa 123-134 in exon 1, IBM2 aa 147-152 in exon 2) in every single one. The two are dissociated, and it is the tether that goes.

The one place the invariant is soft

Among the 16 sarcomas in the pan-sarcoma screen whose KMT2A partners were unique and non-recurrent, KMT2A was the 3' partner in eight and the 5' partner in seven, with one case carrying separate partners on its 5' and 3' fragments and no reciprocal transcripts anywhere in the group 2. So 5'-KMT2A sarcoma rearrangements do exist - in the group the authors call 'of unclear significance', with 44% TP53 mutation and a median age of 57. The invariant holds for the defined entities (YAP1, VIM) and not for KMT2A breaks in sarcoma generally. That is the correct scope, and stating it more broadly would be wrong.

Facts and figures
invariantIn leukaemia KMT2A is the 5' partner with a break in exons 8-14, retaining aa 1 onward (menin motif, LEDGF IBMs, AT-hooks, CXXC) and losing SET. In the defined sarcoma arms KMT2A is 3' or internal, retaining CXXC (and in two arms SET) and losing exon 1 and therefore the menin-binding motif.
scopeHolds for the YAP1, VIM, LDB1 and CBX6/PYGO1 arrangements. Does NOT hold for the 16 non-recurrent-partner KMT2A sarcomas, seven of which carried 5'-KMT2A fusions and which the source authors call of unclear significance.
shared elementthe CXXC-type zinc finger (aa 1147-1195, exons 5-6) is the only KMT2A element retained under both polarities
reconciliationKao's 'partially disrupted or totally lost DNA binding domain' and Massoth's 'preservation of KMT2A CxxC-binding domain' are not contradictory: Kao uses 'DNA binding domain(s)' for the AT-hooks plus CXXC collectively, and a fusion starting at exon 4-5 loses all three AT-hooks (exon 3) while keeping the CXXC. The 'totally lost' half is realised in one case with a breakpoint at aa 1143, inside the pre-CxxC PAF1c element. Checked verbatim against both sources 2026-09-21.
sharpest statementThe sarcoma arrangements DISSOCIATE the CpG address from the menin/LEDGF tether and keep the address. MBM (aa 6-25) and IBM1 (aa 123-134) are in exon 1 and IBM2 (aa 147-152) in exon 2, and no defined-arm arrangement contributes either exon; CXXC (aa 1147-1195) is contributed by all of them. Polarity is the MECHANISM by which the tether is lost, not itself the discriminating fact.
kmt2a residues retained by arm
{
  "leukaemic (KMT2A::AFF1 class)": "1 to ~1362-1607",
  "VIM::KMT2A": "145-3969",
  "YAP1::KMT2A::YAP1": "~1112-1212 only",
  "LDB1::KMT2A": "1053-3969 (exons 4-36)"
}
menin motif present
{
  "leukaemic": true,
  "VIM::KMT2A": false,
  "YAP1::KMT2A::YAP1": false,
  "LDB1::KMT2A": false
}
How this could bite us back
  • The whole menin-inhibitor rationale is polarity-dependent, and nothing in a fusion's NAME tells you its polarity: 'KMT2A::YAP1' and 'YAP1::KMT2A' appear in the same papers for the same tumours. A drug decision made from a fusion name rather than from breakpoint exons can be exactly backwards.
  • In the 3'-polarity arms SET is retained, so a KMT2A catalytic or WDR5/WIN inhibitor becomes the rational modality - the opposite of the leukaemia case where SET is already absent. A programme that reused the leukaemia playbook here would pick the one modality with no target.
Still open
  1. Is there any 5'-KMT2A fusion in a solid tumour that defines an entity, as KMT2A::MAML2 does in thymoma? If so, does it behave more like a leukaemic fusion, and is it menin-dependent?
  2. Does the CXXC domain do the same thing under both polarities? Under 5'-KMT2A it sits N-terminal to the partner; in the YAP1 insert it sits between two YAP1 segments with no KMT2A context at all.
  3. Why does the 3' polarity appear in mesenchyme and the 5' polarity in blood? If the difference is which polarity a given lineage can tolerate rather than which breakpoints occur, that is testable by expressing each polarity in both lineages, and nobody has done it.
Every source behind this entry (6)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The authors defining the YAP1 arm state that in KMT2A-rearranged leukaemia KMT2A is the 5' fusion partner with breakpoints between exon 9 and intron 11, retaining the N-terminal DNA-binding domains and losing the C-terminal SET domain, whereas in these sarcomas KMT2A is presumably the 3' partner with breakpoints in exons 4-5, leaving the DNA-binding domain partially disrupted or totally lost and the SET domain intact.

    PMID 31592798DOI
    What this evidence cannot show

    An interpretive statement in a pathology discussion, from exon indices rather than from any protein measurement, and made before the insertional YAP1::KMT2A::YAP1 model existed - so 'KMT2A is presumably the 3' partner' is the authors' own hedge. Read from the PMC full text (route 3, efetch db=pmc, after fullTextXML returned HTTP 500).

    Evidence that argues against this (1)
    • The later insertional model says KMT2A is neither the 5' nor the 3' partner in the YAP1 arm but an internal insert, so 'inverted polarity' is an approximation for that arm. The structural conclusion - the menin-binding exon is absent and the SET domain is not the thing lost - survives either reading. Teku G et al. 2023
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In 34 human KMT2A-fusion-positive sarcomas, the two recurrent arms both place KMT2A downstream or internal (YAP1-KMT2A-YAP1 retaining only the CxxC exons; VIM-KMT2A with breakpoints near VIM exon 4 and KMT2A exon 2), while among the 16 non-recurrent-partner tumours KMT2A was the 3' partner in eight and the 5' partner in seven with no reciprocal transcripts in any case.

    PMID 32461620DOI
    What this evidence cannot show

    A commercial profiling cohort enriched for advanced disease. The 8-versus-7 split in the non-recurrent group comes from fusion-call orientation on a targeted panel, not from cloned transcripts, and those 16 tumours are described by the authors themselves as of unclear significance. Read from the open-access full text.

    Evidence that argues against this (1)
    • That seven of these sarcomas carried 5'-KMT2A fusions is the direct counter-example to a universal polarity claim. Those tumours had diverse histologies, 44% TP53 mutation and a median age of 57, which argues they are passenger events - but 'probably passengers' is an interpretation, not a measurement, and no methylation or expression profile of that group has been published. Warmke LM et al. 2025
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    A novel LDB1::KMT2A fusion in a 19-year-old man's spindle-cell sarcoma joined LDB1 exons 1-10 to KMT2A exons 4-36, confirmed by FISH and RT-PCR, with a concurrent pathogenic BCOR frameshift.

    LDB1::KMT2A Fusion in a Spindle-Cell Sarcoma: A Case Report. · Wang Z, Zhang Y, Zhu M, et al. · Genes Chromosomes Cancer 2025
    PMID 40944346DOI
    What this evidence cannot show

    A single case read at abstract level. 'KMT2A exons 4-36' is the most explicit statement in this literature that a sarcoma fusion retains nearly the whole KMT2A coding sequence while starting at exon 4 - i.e. discarding exons 1-3 - but it is n=1 and the concurrent BCOR frameshift means the fusion is not the only lesion.

    Evidence that argues against this (1)
    • The tumour was positive for CD99, SATB2, cyclin D1, BCL2, TLE1, pan-TRK and NKX2.2 with focal BCOR expression, an immunoprofile that overlaps several other sarcoma classes, and it was disease-free at only 5 months. Whether LDB1::KMT2A is a driver, a passenger in a BCOR-driven tumour, or a private event is unresolvable from one case. Nakamura H et al. 2023
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    A KMT2A-rearranged sarcoma in a 47-year-old man carried a novel CBX6::KMT2A::PYGO1 fusion gene - a second three-segment arrangement, with KMT2A internal rather than terminal - and the authors note that YAP1::KMT2A::YAP1 is the commonest fusion gene in this entity followed by VIM::KMT2A.

    KMT2A-rearranged sarcoma with unusual fusion gene CBX6::KMT2A::PYGO1. · Nakamura H, Kukita Y, Wakamatsu T, et al. · Virchows Arch 2023
    PMID 37713130DOI
    What this evidence cannot show

    A single case read at abstract level; the KMT2A exon breakpoints are not given in the abstract, so this case cannot be placed on the exon map. The authors themselves call the arrangement 'highly likely irreproducible'. Follow-up 2 years 8 months with no recurrence after surgery alone.

    Evidence that argues against this (1)
    • Because the abstract gives no KMT2A exon indices, this case cannot be used to test the polarity claim at residue resolution - it supports 'KMT2A internal' by the fusion's name alone. The full text was not obtained. Wang Z et al. 2025
  5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    Computed here (2026-09-21): the KMT2A menin-binding motif lies at aa 6-25 inside exon 1, so any fusion whose KMT2A contribution begins at exon 2 (VIM arm), exon 4 (LDB1) or exons 4-6 (YAP1 arm) cannot contain it, while the CXXC zinc finger at aa 1147-1195 lies in exons 5-6 and is contained in all of them.

    What this evidence cannot show

    A coordinate derivation from reported exon indices, not a protein measurement; it establishes what the reported transcripts can encode. No sarcoma KMT2A fusion protein has been observed by any method in the literature read for this beat, so nothing here is evidence about an expressed protein's actual domain content.

    Evidence that argues against this (1)
    • Reciprocal KMT2A::YAP1 transcripts with 5' breakpoints at KMT2A exons 3, 4 and 6 have been detected in a substantial minority of YAP1-arm cases, and those transcripts DO begin at exon 1. The polarity claim is about the dominant species, not about every transcript in the tumour. Warmke LM et al. 2025
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Massoth's Table 1 footnote records a KMT2A breakpoint inside exon 5 at amino acid 1143, within a 'pre-CxxC' region (aa 1115-1154) that the authors state accounts for a portion of the multivalent bond with the PAF1 complex - four residues before the CXXC zinc finger begins at aa 1147.

    PMID 32461620DOI
    What this evidence cannot show

    A table footnote in one paper, describing one case, read from the open-access full text. The 'pre-CxxC' designation and the aa 1115-1154 boundaries are the authors' own and were not independently checked against a structural source; UniProt annotates no such feature. It is cited here because it is the only residue-level KMT2A breakpoint published in this literature.

    Evidence that argues against this (1)
    • A single case whose breakpoint splits the CXXC region means 'CXXC retained' is a property of the common configurations rather than of the entity. If the aa-1143 tumour behaved like the others, CXXC retention is not doing the work this literature credits it with - and nobody has compared them. Warmke LM et al. 2025
Connected to