KMT2A::USP2 is invisible to karyotype and often to FISH, so these patients are misclassified as fusion-negative
also called cryptic KMT2A-USP2 · FISH-negative KMT2A rearrangement · under-ascertainment of minor-BCR fusions
Because this fusion is made by a flip inside one chromosome rather than a swap between two, the standard chromosome test looks normal and the usual fluorescent probe test is often normal too. Children with a high-risk leukaemia are therefore filed as having no fusion at all, and treated accordingly.
The same thing, for a scientist
KMT2A::USP2 arises by intrachromosomal inversion at 11q23.3 over roughly 1 Mbp, so there is no derivative chromosome for karyotype to show, and a KMT2A break-apart FISH probe set gives one of two misleading patterns: loss of the 3' signal (attributable to the frequent 3' deletions, present in 23.8% of minor-BCR cases) or a completely normal wild-type pattern [E1]. An independent three-case series states that approximately half of such cases cannot be resolved by standard cytogenetic approaches and are likely under-recognised, and resolved all three only by mate-pair sequencing [E2]. The consequence is not academic: the founding cohort exists *because* these patients were the residue left over when MP-PCR and LDI-PCR failed, and 23% of that residue was minor-BCR disease [E1]. The risk is therefore systematic misclassification of a high-risk genotype - frequent CNS disease, 76% day-33 MRD positivity, poor prednisone response - as fusion-negative. Modern targeted RNA sequencing does detect it: three of four laboratories in a 2025 Australian inter-laboratory exchange identified KMT2A::USP2 correctly [E3].
The four ways this fusion hides
- Karyotype. An intrachromosomal inversion spanning ~1 Mbp produces no visible derivative and, at the resolution of banding, no visible change at all. Two of the three cases in the independent series had an otherwise unremarkable karyotype (one normal, one with trisomy 10 as the sole abnormality) 2.
- FISH, normal pattern. A break-apart probe whose signals both lie 5' and 3' of the inversion's KMT2A breakpoint can reassemble into a wild-type-looking pattern. One of the three independent cases had a normal KMT2A FISH result 2, and the founding series reports a normal wild-type FISH pattern as one of two recurring appearances 1.
- FISH, 3'-signal loss. The other recurring pattern, driven by the 3' deletions found in 23.8% of minor-BCR cases 1. This is reportable as 'atypical' but is easily written off as a probe artefact.
- Partner-directed PCR. Multiplex PCR is built to catch the eight commonest partners, which is roughly 90% of KMT2A-r patients, and USP2 is not among them; LDI-PCR covers the rest but is anchored on the major BCR 1. So the assay that would name the partner is looking in the wrong place.
What the mis-ascertainment costs
- Risk stratification. KMT2A rearrangement is a stratifying lesion in essentially every paediatric ALL protocol. A child whose KMT2A::USP2 is missed is stratified as though they had none, in a genotype where 76% were MRD-positive at day 33 and 42% of those assessed had a poor prednisone response 1.
- MRD monitoring. Patient-specific genomic breakpoint sequences are exactly what the DCAL group uses for MRD, and the 2019 paper's stated purpose is to make that possible for these patients 1. Without the breakpoint, there is no fusion-based MRD assay.
- Trial access. Menin-inhibitor trials enrol on KMT2A rearrangement. A patient scored fusion-negative by karyotype and FISH is not eligible, for a drug class that targets the retained N-terminal module their fusion has.
- Epidemiology. Every published frequency for this partner is downstream of an assay that misses it, which is why the same centre's count went from 17 (rank 14) in 2019 to 30 (rank 9) in 2023 14.
What actually detects it
| method | detects KMT2A::USP2? | source |
|---|---|---|
| G-banded karyotype | no | 2 |
| KMT2A break-apart FISH | unreliable - normal or 3'-loss patterns | 12 |
| Multiplex PCR for the 8 commonest partners | no, USP2 is not on the panel | 1 |
| LDI-PCR | designed around the major BCR; these are the cases it failed | 1 |
| Targeted capture NGS across the whole KMT2A gene | yes - this is how it was found | 1 |
| Mate-pair sequencing | yes | 2 |
| Whole-transcriptome sequencing | yes - how the first AML case was found | 5 |
| Targeted RNA-seq fusion panels | yes in 3 of 4 laboratories tested | 3 |
The practical rule that falls out: a KMT2A-rearranged case whose partner cannot be named is not a partner-less case, it is an un-sequenced one, and roughly one in four of those is minor-BCR disease 1.
| kind | delivery |
|---|---|
| severity | High for the individual patient - misclassification of a high-risk genotype as fusion-negative changes risk stratification, forfeits fusion-based MRD monitoring and removes eligibility for rearrangement-selected trials. Moderate at population level, because the genotype is rare. |
| mitigation | Treat an unnameable partner as an indication for sequencing, not as a negative result. Whole-KMT2A capture NGS, mate-pair sequencing, whole-transcriptome sequencing or a targeted RNA-seq fusion panel all detect it; karyotype, break-apart FISH and partner-directed multiplex PCR do not reliably. For MRD, a genomic (not transcript) breakpoint is required, so an RNA-based detection still needs a DNA follow-up. |
| conditional yield | Among KMT2A-rearranged patients whose breakpoint MP-PCR and LDI-PCR cannot resolve, 21/93 (23%) were minor-BCR and 17/93 (18%) were KMT2A::USP2. |
affects
[ "fusion-kmt2a-usp2", "fusion-kmt2a-usp8", "breakpoint-kmt2a-minor-bcr", "drug-revumenib" ]
detection matrix
{
"karyotype": "no",
"kmt2a_breakapart_fish": "unreliable: normal pattern or 3'-signal loss",
"multiplex_pcr_8_partners": "no - USP2 is not on the panel",
"ldi_pcr": "no - anchored on the major BCR; these are its failures",
"targeted_capture_ngs_whole_kmt2a": "yes",
"mate_pair_sequencing": "yes",
"whole_transcriptome_sequencing": "yes",
"targeted_rna_seq_fusion_panel": "yes in 3 of 4 laboratories tested"
}- The fix - sequence everything - is not free. Whole-KMT2A capture or mate-pair sequencing is not available in most paediatric centres worldwide, and the genotype is commonest in infants, who are concentrated in exactly the settings least able to run it. A recommendation to sequence is a recommendation that widens an existing inequity unless it comes with access.
- Detecting more of these patients will lower the apparent aggressiveness of the genotype, because the cases found today are enriched for those that declared themselves. Any prognostic figure in this graph for KMT2A::USP2 will drift as ascertainment improves, and should not be used as a fixed number.
- An RNA-based positive without a genomic breakpoint gives the diagnosis but not the MRD assay, so a laboratory can 'solve' the case and still leave the patient without monitoring.
- What fraction of all KMT2A::USP2 patients does break-apart FISH actually detect? Every existing series is ascertained from FISH failures, so the sensitivity of FISH for this fusion has never been measured against a sequencing gold standard in an unselected cohort.
- How many patients currently classified as 'KMT2A-rearranged, partner unknown' in cooperative-group trial databases are minor-BCR cases? This is answerable by re-sequencing banked material and would fix the epidemiology in one study.
- Does a targeted RNA-seq-positive, genomic-breakpoint-unknown patient have any route to fusion-based MRD monitoring, or do they need a separate DNA capture step?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In diagnostic FISH, KMT2A::USP2 cases showed two patterns - loss of the 3' probe signal, or a normal wild-type pattern - and the fusion was found only by whole-gene targeted capture NGS in patients whose breakpoints multiplex PCR and LDI-PCR had failed to resolve; 3' deletions were present in 23.8% of minor-BCR cases.
Human MLL/KMT2A gene exhibits a second breakpoint cluster region for recurrent MLL-USP2 fusions. · Meyer C, Lopes BA, Caye-Eude A, Cavé H, Arfeuille C, et al., Marschalek R · Leukemia 2019What this evidence cannot show
The FISH patterns are reported in a supplementary table and figure that were not opened in this session; the two-pattern statement is taken from the main text. The cohort is by construction the set of cases other assays failed on, so it cannot estimate what fraction of all KMT2A::USP2 cases FISH would miss - only that FISH missed these. Full main text read (PMC6756029).
Evidence that argues against this (1)
- The detectability picture has moved since 2019. A 2025 multi-laboratory exchange found three of four laboratories detected KMT2A::USP2 by routine targeted RNA-seq, so 'cryptic' is now specific to karyotype and FISH rather than a general property of the fusion. Li EW et al. 2025
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Three infant and childhood leukaemia cases with KMT2A-USP2 or KMT2A-USP8 were resolvable only by mate-pair sequencing - one had a normal karyotype with atypical KMT2A FISH, one a normal KMT2A FISH result, one an atypical FISH with a 15q deletion - and approximately half of such cases cannot be resolved by standard cytogenetic approaches and are likely under-recognised.
Cryptic and atypical KMT2A-USP2 and KMT2A-USP8 rearrangements identified by mate pair sequencing in infant and childhood leukemia. · Blackburn PR, Smadbeck JB, Znoyko I, Webley MR, Pitel BA, Vasmatzis G, Xu X, Greipp PT, Hoppman NL, Ketterling RP, Baughn LB, Lindsey KG, Schandl CA, Wolff DJ, Peterson JF · Genes, Chromosomes & Cancer 2020What this evidence cannot show
Three patients across two institutions, selected because they were diagnostically difficult, so the 'approximately half' figure is a summary of prior literature rather than a measurement here. Mate-pair sequencing is not a routine method in most laboratories, so the finding establishes detectability in principle rather than a practical recommendation. Abstract read in full.
Evidence that argues against this (1)
- A series assembled from cases that were hard to diagnose cannot estimate how often the diagnosis is hard. The denominator - all KMT2A::USP2 patients, including any that FISH did catch - is unknown and unknowable from this design. Meyer C et al. 2019
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
In a four-laboratory Australian sample-exchange programme using targeted RNA sequencing, three of four laboratories identified all five reference fusions including KMT2A::USP2; the platforms were Archer FusionPlex or Qiagen Fusion XP.
Advancing diagnostic targeted RNA sequencing for haematological malignancies within an Australian sample exchange program. · Li EW, Tiong IS, Gould C, Catalano A, Aung H, Seymour L, White D, Thompson E, Blombery P, Iland H, Wright D · Pathology 2025What this evidence cannot show
Six samples, four laboratories - an inter-laboratory concordance exercise, not a sensitivity study, and a single KMT2A::USP2 sample. The one laboratory that missed a fusion missed a different one (ETV6::SRR). All four laboratories missed an IGH::EPOR rearrangement, which shows the same panels fail on regulatory-element rearrangements. Abstract read in full.
Evidence that argues against this (1)
- Targeted RNA-seq detects the fusion transcript but not the genomic breakpoint, and the genomic breakpoint is what a patient-specific MRD assay needs. Detection and monitoring are different problems, and solving one does not solve the other. Meyer C et al. 2019
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The same diagnostic centre's count for USP2 rose from 17 patients (rank 14) in 2019 to 30 patients (rank 9) in 2023 as sequencing replaced PCR-based screening.
The KMT2A recombinome of acute leukemias in 2023. · Meyer C, Larghero P, Almeida Lopes B, Burmeister T, Gröger D, et al., Marschalek R · Leukemia 2023What this evidence cannot show
The rise conflates four more years of accrual with the change of method, and neither paper separates the two. A rank change from 14 to 9 over four years is consistent with either. Counts read from the published articles in this session.
Evidence that argues against this (1)
- Accrual alone could produce the same rise: four additional years at a growing referral centre would add cases of every partner, and ranks shift when the denominators move. Attributing the change to assay sensitivity is an inference the papers do not make. Meyer C et al. 2019
- E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
The first reported KMT2A-USP2 infant AML was identified only by whole-transcriptome sequencing after conventional screening for known translocation partners failed.
Whole transcriptome sequencing reveals a KMT2A-USP2 fusion in infant acute myeloid leukemia. · Ikeda J, Shiba N, Tsujimoto SI, Yoshida M, Nakabayashi K, Ogata-Kawata H, Okamura K, Takeuchi M, Osumi T, Tomizawa D, Hata K, Kiyokawa N, Ito S, Kato M · Genes, Chromosomes & Cancer 2019What this evidence cannot show
One patient. The paper documents that conventional screening failed in that patient; it does not quantify how often conventional screening fails. Abstract read in full.
Evidence that argues against this (1)
- Single cases reach publication partly because they were diagnostically difficult, so a case report is a biased estimator of how cryptic a lesion is. The unbiased statement available is the cohort one: 23% of PCR-unresolvable KMT2A-r cases were minor-BCR disease. Meyer C et al. 2019
inversion geometry plus 3' deletions make karyotype and break-apart FISH unreliable; roughly half of cases unresolvable by standard cytogenetics
the 23% conditional yield is the operational consequence of the cryptic phenotype
menin-inhibitor trials enrol on KMT2A rearrangement; a FISH-negative patient is ineligible for a drug targeting the module their fusion retains
- ← related toKMT2A::USP8, t(11;15)(q23;q21) - three patients, and the second instance of a deubiquitinase partner
partially: the one detailed case WAS visible to karyotype and microarray, so the cryptic profile does not transfer wholesale from USP2
the second failure mode: patients whose fusion is never found cannot meet a rearrangement-based inclusion criterion
the individual-patient face of under-ascertainment: no partner named means no risk stratification in paediatric AML