Spliced and balanced KMT2A::MLLT1 do not automatically make the same oncoprotein - unless you restrict the comparison, which nobody has
also called direct-product matching · MLLT1 junction heterogeneity · exon-2 restriction
It is tempting to treat all patients without a mirror-image gene as one group and all patients with one as another, and to assume both groups build the same fusion protein. They do not: one in six patients builds a shorter fusion protein missing a whole working module, and those patients are all in the second group.
The same thing, for a scientist
The reciprocal-negative class in KMT2A::MLLT1 is generated by breaks upstream of MLLT1 exon 1, which force the direct transcript to splice to MLLT1 exon 2 [E1]. The reciprocal-capable class is generated by breaks inside MLLT1, and those breaks are reported at introns/exons 1, 2, 4, 6 and 7 [E1][E2]. Of 302 t(11;19) patients, 33.4% (n = 101) break in MLLT1 intron 1 - which also gives an exon-2 junction - and 15.9% (n = 48) break downstream of exon 2, producing a fusion protein with the YEATS domain partly (6.0%) or wholly (9.9%) missing [E3]. So a naive spliced-versus-balanced comparison puts 101 patients who make the **same** oncoprotein as the index arm into the same bucket as 48 who make a **different** one. Restricting to exon-2 junctions - 153 against 101 - removes the problem entirely, and the restriction is free because the breakpoint classes are already published. **This entry exists because the unrestricted comparison is the one everyone would run by default, and it is the one that would not be interpretable.**
Why a YEATS-less fusion is a different molecule, not a variant
Deleting the YEATS domain and the sequence downstream of it from MLL-ENL leaves a construct that still supports serial colony replating but fails entirely to cause leukaemia in mice, while point mutations that abolish acetyl-lysine binding only extend disease latency and reduce leukemic stem cell frequency about threefold 3. A fusion protein whose in vivo leukaemogenicity differs that much is not a minor architectural variant. Mixing 48 such patients into a 219-patient comparator arm and calling the arm 'reciprocal-positive KMT2A::MLLT1' attributes to reciprocal status whatever those 48 contribute.
The one residual difference inside the restricted comparison
Even restricted to exon-2 junctions, the two arms are not literally identical at the DNA level: the intergenic-break arm delivers an uninterrupted MLLT1 coding unit to der(11) while the intron-1-break arm splits the MLLT1 allele. The protein made from the fusion allele is the same in both. What may differ is (a) how much wild-type ENL the cell makes, and (b) the length and content of the transcribed region between the KMT2A promoter and MLLT1 exon 2, which in the intergenic class includes an entire intergenic stretch and in the intron-1 class is a single intron. Neither has been measured. Both are answerable from breakpoint sequences already held.
The direction the mismatch pushes
The 48 downstream-break patients are enriched for AML: significantly fewer AML than ALL patients retain the YEATS domain (Chi-square p < 0.0001, df = 1) 3. AML is the lineage in which KMT2A::MLLT1 is independently adverse 4. So folding those 48 into the reciprocal-positive arm would tend to make the reciprocal-positive arm look worse - which is the opposite of the t(4;11) finding that reciprocal-positive patients do better, and would therefore show up as a failure to replicate when it was actually an artefact of arm construction.
| lineage association | significantly fewer AML than ALL patients retain the YEATS domain (chi-square p < 0.0001, df = 1) |
|---|---|
| residual dna level difference | intergenic-break arm delivers an uninterrupted MLLT1 coding unit to der(11); intron-1-break arm splits the MLLT1 allele |
naive comparison
{
"reciprocal_negative": 241,
"reciprocal_positive": 219,
"problem": "the 219 mixes exon-2 junctions with exon-4/6/7 junctions"
}restricted comparison
{
"reciprocal_negative": 153,
"reciprocal_positive": 101,
"excluded": 48,
"matched_on": "MLLT1 exon-2 junction and YEATS retention"
}downstream break classes — the numbers behind the chart
{
"partial_yeats_pct": 6,
"no_yeats_pct": 9.9,
"exon6_7_n": 16
}- A trial that stratified on spliced-versus-balanced status without restricting to exon-2 junctions would put patients whose fusion protein has no acetyl-lysine reader into the comparator arm, and any YEATS-directed agent would then appear to work worse in the arm it was designed for.
- Is the 101 intron-1 class really reciprocal-capable in frame? A break in MLLT1 intron 1 puts only exon 1 on der(19), and whether exon 1 joined to KMT2A exon 12 is in frame has never been checked for this partner.
- How long is the transcribed intergenic stretch in the 153, and does it contain anything - a promoter, a lncRNA, a polyadenylation signal - that would change fusion transcript output? The read-through model assumes it is inert.
- Do the 48 YEATS-less patients differ in outcome from the 254? The cohort exists and the comparison has never been made, and it is arguably more tractable than the reciprocal one because the biological difference is larger.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Read in full text (PMC10169636): in spliced fusions the most 3' exon of the 5'-KMT2A usually splices to exon 2 of the partner gene, that being the first exon with a bona fide splice acceptor site; for MLLT1 about 50% of breakpoints (n = 241 of 460) localise upstream of MLLT1 exon 1.
The KMT2A recombinome of acute leukemias in 2023. · Meyer C, Larghero P, Almeida Lopes B, et al. · Leukemia 2023What this evidence cannot show
Read in full text via Europe PMC fullTextXML. The word 'usually' is the authors' own; exceptions to exon-2 splicing in this partner are neither quantified nor excluded.
Evidence that argues against this (1)
- A therapy-related AML line carries an MLL-ENL transcript its authors describe as a novel isoform, so junction usage is not fully enumerated even for the classes that have been catalogued. Ninomiya M et al. 2006
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
MLL/ENL fusion transcripts use ENL breakpoints at exons 2, 4, 6 or 7, all in frame.
Analysis of acute leukemias with MLL/ENL fusion transcripts: identification of two novel breakpoints in ENL. · Fu JF, Liang DC, Shih LY · Am J Clin Pathol 2007What this evidence cannot show
Abstract-level reading; n = 15, so per-junction frequencies are not established here.
Evidence that argues against this (1)
- The larger 302-patient breakpoint analysis gives the distribution at genomic rather than transcript resolution and reports only about 5.3% (n = 16) breaking between exons 6 and 7, so the four junction classes are very unequally populated. Hu H et al. 2023
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Read in full text (PMC11246743): of 302 t(11;19) patients, 50.7% (n = 153) break 5' of MLLT1 and 3' of ACER1, 33.4% (n = 101) break within MLLT1 intron 1, and 15.9% (n = 48) break downstream of exon 2 giving partial (6.0%) or complete (9.9%) exclusion of the YEATS domain, with about 5.3% (n = 16) breaking between exons 6 and 7; significantly fewer AML than ALL patients retain the YEATS domain (Chi-square p < 0.0001, df = 1); mice receiving MLL-ENL transduced cells died of leukaemia with a median survival of 66 days while mice receiving MLL-ENL-delta-YEATS transduced cells failed to develop leukaemia in vivo, although both constructs supported third-round colony replating; and YEATS point mutants extended latency and reduced leukemic stem cell frequency from 1 in 108 to 1 in 355 cells.
The ENL YEATS epigenetic reader domain critically links MLL-ENL to leukemic stem cell frequency in t(11;19) Leukemia. · Hu H, Saha N, Yang Y, Ahmad E, et al. · Leukemia 2023What this evidence cannot show
Read via the PMC HTML page (route 4) after fullTextXML returned HTTP 500 and efetch db=pmc returned HTTP 429. The authors note their delta-YEATS construct (aa 372-559) removes the YEATS domain AND part of the intrinsically disordered region, so the in vivo failure cannot be attributed to loss of the reader alone - which is why the point-mutant result, where leukaemia still develops, is the more specific test of reader function.
Evidence that argues against this (1)
- The point mutants do cause leukaemia, so acetyl-lysine reading is not required for transformation in vivo, only for its efficiency - meaning a YEATS-less fusion is not simply a non-oncogenic molecule and the 48 patients are not a biologically inert subgroup. Hu H et al. 2023
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In 1,256 children with KMT2A-rearranged AML, 19p13.3/KMT2A::MLLT1 was independently associated with adverse outcomes in multivariable analysis.
Optimized cytogenetic risk-group stratification of KMT2A-rearranged pediatric acute myeloid leukemia. · van Weelderen RE, Harrison CJ, Klein K, Jiang Y, et al. · Blood Adv 2024What this evidence cannot show
Abstract-level reading by this entry's author for the independence claim; effect sizes relayed from a delegated worker's full-text reading. Cytogenetic cohort, so breakpoint class within MLLT1 was not measured and the enrichment argument made above joins two studies that share no patients.
Evidence that argues against this (1)
- In childhood non-infant T-ALL, t(11;19)(q23;p13.3) had the highest 5-year EFS of any group in a 629-patient cohort, so the lineage-dependence of this partner's prognosis cuts both ways and the AML enrichment argument does not settle the direction of the artefact. Attarbaschi A et al. 2023
why the comparison must be restricted to exon-2 junctions
adds the assay-dependence: deltaYEATS replates but does not leukaemogenise; point mutants still do
- part of →The KMT2A::MLLT1 chimeric protein: exactly what survives the join on each side, in residue numbers
junction heterogeneity of the chimera