Ask whether the childhood AML in the KDM3B-haploinsufficiency syndrome cohort carried a KMT2A rearrangement - the one human test this mechanism has, and it is a records request
also called the Diets-Jongmans AML karyotype · does constitutional KDM3B loss predispose to KMT2A rearrangement in people?
Fourteen people born with only one working copy of the KDM3B gene were described in 2019, and two of them developed cancer as children - one of them leukaemia. If the mechanism is right, that leukaemia should carry a broken or duplicated KMT2A gene, and finding out means asking the authors for the karyotype.
The same thing, for a scientist
Germline *KDM3B* haploinsufficiency causes Diets-Jongmans syndrome (DIJOS, MIM 618846): 14 unrelated individuals plus three affected parents with intellectual disability or developmental delay, short stature, infant feeding difficulties, joint hypermobility and a characteristic facial gestalt - and **two developed childhood cancer, one acute myeloid leukaemia and one Hodgkin lymphoma** [E1]. The locus mechanism predicts that constitutional loss of one *KDM3B* allele raises H3K9me1/2 at the *KMT2A* breakpoint cluster region, lowers CTCF occupancy there and predisposes to *KMT2A* amplification and rearrangement [E2]. **So the prediction is specific, the cohort exists, and the test is a records request**: the karyotype, FISH and any NGS panel from that one AML. Neither paper cites the other - `KDM3B[tiab] AND CTCF[tiab]` returns 1 record and `"Diets-Jongmans"[tiab]` returns 8, with no overlap [E3]. **This is not a strong test and it must not be written as one.** Seventeen individuals ascertained by gene matching on a developmental phenotype is not a population; two cancers is not an incidence; and constitutional haploinsufficiency permits lifelong compensation that an siRNA does not. **But a positive result would be this mechanism's first human evidence and the cheapest it will ever be**, and a negative result is informative too, because the mechanism predicts *KMT2A* specifically rather than cancer generally.
The design
Primary question. Did the DIJOS AML carry an 11q23 / KMT2A rearrangement or a KMT2A copy gain?
Route 1, the cheap one. Read the supplementary material of PMID 30929739 - a 2019 AJHG human-genetics paper, where per-individual clinical tables are standard - for the AML case's cytogenetics. If absent, write to the corresponding author. The DIJOS cohort has grown since 2019: "Diets-Jongmans"[tiab] returns 8 records including several 2025-2026 papers 3, so the natural-history literature may already carry the answer or a second case.
Route 2, the systematic one. Ask whether any KDM3B germline variant appears in a paediatric leukaemia germline series, and whether 11q23 status is recorded. This graph already holds PMID 29351919, a Clin Cancer Res series on pathogenic germline mutations in selected children with cancer, which appeared in this beat's KDM3B[tiab] AND syndrome[tiab] set of 13 - it was not read and is the obvious first stop 3.
Route 3, the population one. If a DIJOS registry or a ClinGen/DECIPHER cohort exists, ask for the cancer incidence and the leukaemia subtypes. The mechanism's population prediction is an excess of KMT2A-rearranged leukaemia specifically, not of cancer in general - which is a sharper and more falsifiable prediction than a cancer-predisposition claim.
Controls and interpretation.
| finding | reading |
|---|---|
| AML carries a KMT2A rearrangement or 11q23 amplification | the mechanism's first human evidence. Still n = 1, still needs the cohort. |
| AML has a normal 11q23 | one negative human datum. Does not refute the mechanism - one patient, and the mechanism describes a probabilistic first hit with a latency the authors put at up to 15 years - but it removes the only human support available. |
| cytogenetics unobtainable | record as such. This graph's standing rule is that an identifier and a stated limit are better than an inference. |
| the cohort shows a cancer excess that is not KMT2A-enriched | the sharpest available refutation of locus specificity: constitutional KDM3B loss would be a general cancer predisposition, and the KMT2A-specific mechanism would need to explain why. |
Why this is worth a node rather than a sentence
Because the graph's standing failure mode is the opposite of this. docs/QUEEN_BRIEF.md records entry after entry whose caveat named an uncovered case that turned out to be one query away - a PRIDE accession never swept, an IDH-status supplementary table sitting in an open-access paper, a Cellosaurus query never run. A named-but-unchecked human test is exactly that shape, and naming it in a caveat does not close it.
It also matters that this is the only route to human evidence that exists. The mechanism cannot be tested in a patient prospectively - nobody will give a child a KDM3B inhibitor - and the therapy-induced-leukaemia cohorts that would test the doxorubicin arm are decades-long. Constitutional haploinsufficiency is the natural experiment, it has already happened, and the answer is in somebody's filing cabinet.
And the honest counterweight, stated because this entry is easy to over-sell. DIJOS is described in its own founding paper as an intellectual-disability syndrome, not as a cancer-predisposition syndrome, and the two cancers are reported as a notable incidental observation. If constitutional KDM3B loss were a strong KMT2A-rearrangement predisposition, one would expect the clinical literature to have noticed by now - and the absence of that noticing is itself weak evidence against a large effect 1.
| design | A records and literature request, in three escalating routes, against an existing human cohort. Route 1: read the supplementary material of PMID 30929739 and then the 7 later Diets-Jongmans records for the AML case's karyotype, FISH and NGS panel; if absent, write to the corresponding author. Route 2: check paediatric-cancer germline series for KDM3B variants and recorded 11q23 status, starting with PMID 29351919. Route 3: if a DIJOS registry, ClinGen or DECIPHER cohort exists, request cancer incidence and leukaemia subtype distribution. |
|---|---|
| readout | presence or absence of an 11q23 / KMT2A rearrangement or KMT2A copy gain in the Diets-Jongmans AML case; secondarily, the KMT2A-subtype distribution of any cancers across the enlarged cohort |
| controls | The mechanism predicts KMT2A SPECIFICALLY, not cancer generally, so the informative control is subtype: a cancer excess that is not KMT2A-enriched is a sharper refutation of locus specificity than no excess at all. The founding paper's non-responsive loci (MLLT1, MLLT3) and responsive loci (TCF3, AFF3) give a second axis - TCF3-rearranged ALL in this cohort would support the axis without supporting KMT2A specificity. |
| cost tier | low |
| timeline | days for the literature and supplementary sweep; weeks for correspondence; months if a registry request is needed |
| n available | 1 AML (plus 1 Hodgkin lymphoma) among 17 individuals in the founding report; unknown in the 7 later records |
| statistical power | NONE at n = 1. This is a single-case check whose value is that a positive result is the mechanism's first human evidence and a negative result removes its only human support. It is not, and must not be reported as, a test of predisposition. |
| why a negative does not refute | the mechanism describes a probabilistic first hit requiring an unidentified second hit, with a latency the authors put at up to 15 years; and constitutional haploinsufficiency permits lifelong compensation that acute siRNA depletion does not |
| what this beat did not do | PMID 30929739's full text and supplementary material were NOT retrieved, and neither were the 7 later Diets-Jongmans records or PMID 29351919. Recorded as a bounded rejection rather than an absence: they were enumerated and not read, in this cycle, for time. |
| absence class | measured-not-searched |
tests hypotheses
[ "hyp-kdm3-family-sits-on-both-ends-of-this-disease" ]
- A positive result here would be quoted as evidence that KDM3B is a leukaemia-predisposition gene, and at n = 1 in a cohort ascertained on a developmental phenotype it would not support that. Any positive finding must be reported with its denominator and its ascertainment in the same sentence.
- This is somebody's child's medical record. The request must go through the original investigators and whatever consent covered the 2019 publication, and this graph should hold only what is published or formally shared - never a re-identifiable detail.
- A negative result is the more likely outcome and will be tempting to omit. It is the more valuable of the two to record, because it removes the mechanism's only human support and nothing else will.
- Has the Diets-Jongmans cohort grown enough for a cancer-incidence statement? Eight PubMed records now exist including 2025-2026 papers, and a natural-history series with 40 or 60 individuals would convert this single-case check into a real predisposition question.
- Do any paediatric-leukaemia germline series carry KDM3B variants with 11q23 status recorded? PMID 29351919 is the obvious first stop and was not read in this cycle.
- Is there a Kdm3b-heterozygous mouse with a lifetime cytogenetic endpoint? The mouse work in this mechanism is acute drug exposure over days; a constitutional heterozygote aged out with Kmt2a FISH would be the animal counterpart of this human question and nobody has done it.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
De novo and inherited pathogenic KDM3B variants - missense and truncating, indicating haploinsufficiency - were identified by exome sequencing and gene matching in 14 unrelated individuals and three affected parents with varying intellectual disability or developmental delay and short stature, sharing feeding difficulties in infancy, joint hypermobility and characteristic facial features (wide mouth, pointed chin, long ears, low columella) demonstrable as a facial gestalt by a hybrid facial-recognition model; and NOTABLY two individuals developed cancer in childhood, acute myeloid leukaemia and Hodgkin lymphoma.
De Novo and Inherited Pathogenic Variants in KDM3B Cause Intellectual Disability, Short Stature, and Facial Dysmorphism. · Diets IJ; van der Donk R; Baltrunaite K; Waanders E; et al. · Am J Hum Genet 2019What this evidence cannot show
**Abstract only; full text and supplementary material not retrieved** - and the supplementary material is precisely where this experiment's answer most likely already sits, which makes that the first step rather than a limitation of the design. Ascertainment is by gene matching on a developmental phenotype, so the cohort is biased towards investigated and severely affected individuals; 17 individuals is not a population denominator and two cancers is not an incidence estimate. **No karyotype, cytogenetics or KMT2A status is given for the AML in the abstract.** The syndrome is classified as an intellectual-disability syndrome, not a cancer-predisposition syndrome. Constitutional haploinsufficiency permits lifelong compensation that acute siRNA depletion does not, so a negative result would not straightforwardly refute the acute-depletion mechanism.
Evidence that argues against this (1)
- The absence of a cancer-predisposition designation for this syndrome is itself weak evidence against a large KMT2A-rearrangement effect: the clinical genetics community has had the syndrome since 2019 and the cohort has grown, and no cancer surveillance recommendation has followed from it in anything this beat could reach. Diets IJ; van der Donk R; Baltrunaite K; Waanders E; et al. 2019
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The mechanism under test predicts that reduced KDM3B raises H3K9me1/2 at the KMT2A breakpoint cluster region, lowers CTCF occupancy at the conserved exon-11 CTCF site, and promotes site-specific KMT2A copy gains and rearrangements; it was shown by siRNA depletion and chemical inhibition in RPE cells, KG1a, a primary AML line, a primary AML organoid and cord-blood CD34+ HSPCs, with the authors describing KDM3B suppression or loss as 'likely just the first step' and noting that therapy-related AML has a latency of up to 15 years and that KDM3B loss alone is unlikely to provide a major cellular fitness advantage.
Epigenetic balance ensures mechanistic control of MLL amplification and rearrangement. · Gray ZH; Chakraborty D; Duttweiler RR; Alekbaeva GD; Murphy SE; Chetal K; Ji F; Ferman BI; Honer MA; Wang Z; et al. · Cell 2023What this evidence cannot show
**Full text read 2026-09-22 via efetch db=pmc on PMC10591855 (Europe PMC fullTextXML HTTP 500; <body> confirmed, 80,638 stripped characters).** The paper's own framing limits what this experiment can test: it describes a first hit that requires an unidentified second hit, with a long latency, and does not claim to explain how an oncogenic fusion is generated. So the human prediction is probabilistic and a single individual's karyotype is a weak test in either direction. The paper does not mention Diets-Jongmans syndrome, constitutional KDM3B variants, or germline predisposition anywhere. All chromatin measurements are in epithelium.
Evidence that argues against this (1)
- The paper's own reasoning cuts against a strong germline prediction: it argues that KDM3B loss alone confers no major fitness advantage to the rearranged cell and that additional hits - ageing, stress exposures, acquired mutations, altered DNA-damage repair - are required. A constitutional heterozygote has one allele's worth of reduction for a lifetime, not the acute near-complete depletion the experiments used. Gray ZH; Chakraborty D; Duttweiler RR; Alekbaeva GD; Murphy SE; et al. 2023
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
The two literatures do not meet: as of 2026-09-22, KDM3B[tiab] AND CTCF[tiab] returns exactly 1 record (37788669), "Diets-Jongmans"[tiab] returns 8 records (several from 2025-2026, so the cohort has grown since the founding paper), KDM3B[tiab] AND syndrome[tiab] returns 13 including PMID 29351919 (a Clin Cancer Res series on pathogenic germline mutations in selected children with cancer), and KDM3B[tiab] AND leukemia[tiab] returns 9 with no overlap with the Diets-Jongmans set.
NCBI E-utilities esearch over PubMed, db=pubmed, run 2026-09-22 by the Q102 beat with every querytranslation read back and ID sets retrieved in full and compared as sets · US National Library of Medicine 2026What this evidence cannot show
Titles and abstracts only, which matters unusually much here: **a karyotype is exactly the kind of fact that lives in a supplementary clinical table and never in an abstract**, so this bound says nothing about whether the answer is already published. PMID 29351919 was found by query and NOT read - it is recorded as a lead and as a bounded rejection, per this project's rule that a paper enumerated and dropped should leave a trace. The 8 Diets-Jongmans records were not screened individually for cytogenetic content.
Evidence that argues against this (1)
- The answer may already be published in a place this query cannot see. The DIJOS literature has grown to 8 records with several very recent ones, and natural-history papers routinely carry per-individual tables; a full-text or supplementary sweep of those 8 should precede any correspondence with an author. US National Library of Medicine 2026
the human natural experiment: constitutional KDM3B haploinsufficiency (Diets-Jongmans syndrome), 2 childhood cancers in 17 individuals, one AML with no reported karyotype
the only human test either end of this hypothesis has
asks whether the mechanism's germline prediction holds in the one cohort where it can be asked