CCDC32::CBX3, a previously unreported second fusion found alongside KMT2A::AFDN in a relapsing AML - and the only record in PubMed that pairs CBX3 with KMT2A at all
also called CCDC32/CBX3 · CCDC32-CBX3
In a young adult whose leukaemia kept coming back, sequencing found a second, previously unknown gene fusion sitting alongside the KMT2A fusion that caused the disease. The second fusion involves CBX3, a protein this project already tracks for a different reason, and it changed the activity of both genes it is made from.
The same thing, for a scientist
Reported in a 2025 *Drug Resistance Updates* study of 16 chromosomal-rearrangement AML patients monitored through diagnosis, refractoriness and relapse. In a FISH-diagnosed **KMT2A (11q23) / AFDN (AF6, 6q27)** rearrangement, RNA-sequencing identified an **unknown CCDC32 (15q15.1) / CBX3 (7p15.2) fusion in both the newly diagnosed and the relapsed samples**, described by the authors as previously unreported in KMT2A/AFDN-rearranged AML. Quantitative PCR showed the fusion significantly affected wild-type expression of **both** parent genes at relapse. The authors propose a triple biomarker - the KMT2A/AFDN rearrangement, the CCDC32/CBX3 fusion, and chimeric RNA variants with distinct breakpoints - confirmed in a 21-year-old male with rapidly relapsed/refractory AML [E1]. This entry exists for two reasons. First, **CBX3 is HP1-gamma**, already tracked in this graph's resection and H3K9 layers, and this is the first record anywhere connecting it to a KMT2A rearrangement in a patient. Second, it was found by a spelling control: `CBX3[tiab] AND KMT2A[tiab]` returns exactly one record and this is it, while the bounded absence that surveyed this protein queried only `HP1gamma[tiab] AND MLL[tiab]` and never the approved symbol against the approved symbol [E2].
Why a CBX3 fusion in a KMT2A-rearranged AML is worth an entry here
CBX3/HP1-gamma appears in this graph twice already: in the H3K9 reader layer, and in the resection-factor set whose occupancy at the KMT2A locus has been the subject of three bounded absences and a concurrent signal-level re-read. A somatic event that alters CBX3 expression in a KMT2A-rearranged patient is the first patient-level datum connecting the two, and it arrived as a co-occurring lesion rather than as an occupancy measurement - a different kind of evidence than anything the occupancy beats have been looking for.
The honest weight is small: n=1 patient with the fusion, within a 16-patient series, in a partner (AFDN) that is not the index fusion, reported once and not replicated. It is a lead, not a mechanism.
How it was found, which is the methodological point
absence-no-resection-factor-localised-at-the-kmt2a-locus bounds CBX3/HP1-gamma with, among 21 queries, HP1gamma[tiab] AND MLL[tiab] = 0. This beat ran the spelling controls the third proven failure mode calls for: "HP1-gamma"[tiab] AND MLL[tiab] = 0, "HP1 gamma"[tiab] AND MLL[tiab] = 0, HP1gamma[tiab] AND MLL[tiab] = 0 - and then CBX3[tiab] AND MLL[tiab] = 1 and CBX3[tiab] AND KMT2A[tiab] = 1, the same record both times 2. The approved symbol was the one that retrieved it.
The absence itself survives: this is not an occupancy measurement and does not place CBX3 at the KMT2A locus. But a bounded absence about a protein that never queried the protein's approved symbol against the approved symbol of the partner gene is a bound with a hole in it, and this is what was in the hole.
| gene5 | CCDC32 |
|---|---|
| gene3 | CBX3 |
| cytoband | CCDC32 15q15.1 :: CBX3 7p15.2 |
| detection method | RNA sequencing, in a patient whose KMT2A/AFDN rearrangement was diagnosed by FISH |
| co occurring lesion | fusion-kmt2a-afdn |
| n patients | 1 |
| series size | 16 |
| functional consequence reported | significantly altered wild-type expression of both CCDC32 and CBX3 at relapse, by quantitative PCR |
| index case | 21-year-old male, rapidly relapsed/refractory AML |
| prognosis | reported by the authors as part of a proposed triple prognostic biomarker; no survival statistics are quoted here because the abstract does not give one |
| why it is in this graph | CBX3 is HP1-gamma, already tracked in this graph's H3K9 reader and resection layers; this is the first patient-level event connecting CBX3 to a KMT2A rearrangement, and it was found by a spelling control that a bounded absence had not run |
| shown for fusion | KMT2A::AFDN only - nothing in KMT2A::AFF1 or any other partner |
| cell type | AML blasts, patient bone marrow / peripheral blood; the abstract does not specify the compartment |
| cell state | compared across newly diagnosed, refractory and relapsed states in the same patients - the state axis is treatment response |
| localization | not measured |
| subnuclear localization | not measured; CBX3/HP1-gamma's own subnuclear behaviour in this context is unknown |
| epigenetic readout | none in this study; the CBX3 connection is by identity of the gene, not by a chromatin measurement |
diseases
[ "KMT2A::AFDN-rearranged acute myeloid leukaemia" ]
breakpoints
[]
timepoints detected
[ "newly diagnosed", "relapsed" ]
- If a second fusion altering a heterochromatin reader travels with a KMT2A rearrangement in even a minority of patients, a therapy directed only at the menin-KMT2A interface leaves that lesion untouched - and this patient relapsed rapidly. One case cannot support that inference, but it is the shape of the risk and it is why the recurrence question above is the one worth answering first.
- Does the CCDC32::CBX3 fusion recur? One patient in one 16-patient series is a lead; a search of any large paediatric or adult KMT2A-rearranged RNA-seq cohort for the same junction would settle whether it is real and recurrent or a one-off call.
- If CBX3/HP1-gamma dosage is altered in a KMT2A-rearranged blast, does that change H3K9me3 reading at the KMT2A locus itself? That is the question the graph's existing H3K9 reader absences are circling, approached from the genetics rather than from occupancy.
- The fusion was detected at diagnosis AND at relapse but the paper frames it within treatment-driven subclonal evolution. Was its allele fraction stable or did it expand under therapy? The abstract does not say and it decides whether this is a passenger.
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
A previously unreported CCDC32/CBX3 fusion was identified by RNA-seq in a KMT2A/AFDN-rearranged AML at both diagnosis and relapse, and it altered wild-type expression of both parent genes at relapse.
Exploring treatment-driven subclonal evolution of prognostic triple biomarkers: Dual gene fusions and chimeric RNA variants in novel subtypes of acute myeloid leukemia patients with KMT2A rearrangement · Drug Resistance Updates 2025What this evidence cannot show
Read at abstract level only; the full text was not opened by this beat. n=1 patient carrying the fusion within a 16-patient series, no independent replication, no functional experiment, and the partner is AFDN rather than the index AFF1 fusion. The 5-prime/3-prime assignment here follows the authors' CCDC32/CBX3 naming and was not confirmed against a breakpoint coordinate.
Evidence that argues against this (1)
- Novel fusions called from RNA-seq in single relapsed samples are a well-known source of false positives, particularly for read-through and low-expression partners, and the abstract reports confirmation by quantitative PCR of parent-gene expression rather than by orthogonal detection of the fusion junction itself. Until a breakpoint is sequenced this should be treated as a reported call, not an established rearrangement. Haas BJ et al. 2017
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
This is the only PubMed record pairing CBX3 with KMT2A or with MLL at title/abstract level, and the HP1-gamma spellings return zero.
What this evidence cannot show
Run through scripts/pubmed_safe.py on 2026-09-22; every zero carried a non-empty querytranslation and warninglist 'No items found.'. Titles and abstracts only - rule 80 - and the HGNC alias HP1Hs-gamma was not tried.
Evidence that argues against this (1)
- A single record retrieved by the approved symbol and missed by the alias is the weakest possible demonstration of the spelling failure mode: with n=1 it is equally consistent with the alias simply being rarer. The general claim - that a bounded absence should query the approved symbol as well as the historical one - rests on the registry's existing rows, not on this instance. HUGO Gene Nomenclature Committee 2026
records WHY this entry still has no analogous_to edge: it is the only fusion in the corpus with a co_occurring_lesion and the only CBX/HP1-family fusion, so its honest analogy class has one member
- co-occurs with →KMT2A::AFDN (MLL-AF6), t(6;11)(q27;q23)
detected by RNA-seq in the same patient at diagnosis and at relapse; reported as previously unknown in KMT2A/AFDN-rearranged AML
the first patient-level event in this graph connecting CBX3/HP1-gamma to a KMT2A rearrangement; wild-type CBX3 expression was altered at relapse
found by the spelling control that entry did not run: it queried HP1gamma against MLL and never CBX3 against KMT2A, which returns exactly this one record