PRDM9 - a germline risk mechanism for childhood ALL that does not reach KMT2A

This is a prediction, not a measurement. No one has seen this protein in this shape. The colours say how much the software trusts itself, and that is worth reading before anything else — the orange stretches are places the model is effectively saying this part has no fixed shape.

• The shared enzymology invites a false connection. PRDM9 and KMT2A are both H3K4 methyltransferases, which makes a mechanistic story easy to tell and easy to believe without data. The one PubMed hit connecting them is a fish paper matching on the letters 'Kmt', which is a fair warning about how this particular link gets manufactured.
• Recording a well-motivated hypothesis as unsupported carries its own risk: a germline risk factor for infant leukaemia would matter enormously if real, and three papers is thin enough that absence of evidence is genuinely weak here. This entry should be revisited if anyone types PRDM9 in a KMT2A-rearranged infant cohort.

1. Do KMT2A breakpoints fall preferentially at PRDM9-dependent meiotic hotspots? This is testable against existing breakpoint maps and appears never to have been asked.
2. Were the infants in the Hussin replication cohort typed for KMT2A rearrangement? If so the decisive stratification may already exist in that dataset.
3. PRDM9 is a H3K4 methyltransferase, as is KMT2A. Is there any non-meiotic overlap in their target chromatin, or is the shared enzymology a coincidence of family membership?
4. Is PRDM9 transcript or protein detectable in any human haematopoietic population? This is a measured-not-analysed question - deposited HSPC and leukaemia RNA-seq would answer it with no wet work - and until it is asked, 'meiosis-restricted' is a literature statement rather than a measurement in the tissue of interest.
5. CTCF is recorded as absent in primary spermatocytes and PRDM9 as present there. Is the complementarity functional - does PRDM9-marked hotspot specification require CTCF's absence - and does anything about that bear on why loop-anchor breakage in somatic cells looks CTCF-driven rather than PRDM9-driven?
6. If PRDM9 is never present in a somatic haematopoietic cell, what is the mechanism behind its replicated association with high hyperdiploid childhood ALL? A germline effect acting in the parental germline rather than in the child's blood would be a different model from the one usually assumed, and it would predict a parent-of-origin signature.

1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   Rare allelic forms of PRDM9 are reported in significant excess among parents of children with B-ALL (44 parental exomes) and among patients relative to controls, with an independent 50-child cohort showing excess in aneuploid and infant B-ALL. The index family had two siblings with HYPERDIPLOID childhood B-ALL and unusual maternal recombination localisation. The authors conclude the data support a potential role in risk of acquiring aneuploidies or genomic rearrangements.

   Rare allelic forms of PRDM9 associated with childhood leukemogenesis · Hussin J, Sinnett D, Casals F, et al. (Awadalla P, senior author) · Genome Research 23(3):419-30 2013
   PMID 23222848DOI
   ▸What this evidence cannot show

   Read as abstract only. Small numbers throughout: 44 parental exomes, a 50-child replication cohort, one index family. 'Rare allelic forms' of PRDM9 is a composite category across a hypervariable minisatellite-containing zinc-finger array, not a single variant, so the effect is not attributable to a defined allele. The infant B-ALL signal is one arm of the replication cohort and is not stratified by KMT2A status in the abstract.

   ▸Evidence that argues against this (1)

   • The independently replicated association is with a DIFFERENT subtype. Woodward 2014 reports PRDM9 allelic variants in high hyperdiploid childhood ALL, which is close to mutually exclusive with KMT2A-rearranged disease. Read together, the two papers point at aneuploidy rather than translocation, which is also the mechanism the Hussin conclusion names first. Woodward EL et al. 2014
2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   No published study links PRDM9 to KMT2A rearrangement. Bounded 2026-09-19: 4 PubMed records for PRDM9[Title/Abstract] AND (KMT2A OR "MLL" OR leukemia OR leukaemia)[Title/Abstract], all read, none reporting such a link. The nearest signal is an excess of rare PRDM9 alleles in aneuploid and infant B-ALL (PMID 23222848); infant B-ALL is mostly KMT2A-rearranged, but that cohort was not genotyped for KMT2A status in the report, so it is a population overlap rather than a link.

   PubMed E-utilities esearch: PRDM9 against KMT2A / MLL-rearranged / mixed lineage leukemia · NCBI PubMed 2026
   ▸What this evidence cannot show

   An absence in a title/abstract index is bounded evidence: it cannot see an analysis reported inside a paper about something else, a negative result that went unpublished, or a preprint outside PubMed. It establishes that the link is not findable, not that nobody has looked.

   ▸Evidence that argues against this (1)

   • This project has twice asserted 'nobody has looked' and been wrong - the endocrine endpoints had an FDA post-marketing requirement and a label, and 'never measured for KMT2A' was refuted by a 1998 cord-blood screen of 65 samples. A search returning nothing is a reason to ask someone in the field, not to conclude the question is open. The entry is kept at 'not findable' for that reason. NCBI PubMed 2026
3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   UniProtKB Q9NQV7 (PRDM9_HUMAN), histone-lysine N-methyltransferase PRDM9, primary gene name PRDM9, 894 aa.

   UniProtKB Q9NQV7 (PRDM9_HUMAN), histone-lysine N-methyltransferase PRDM9 · UniProt Consortium 2026
   ▸What this evidence cannot show

   Resolved 2026-09-19 via gene_exact:PRDM9 with reviewed:true, primary gene name read back rather than assumed from the query term. Note that the canonical sequence does not capture PRDM9's defining feature: the C-terminal zinc-finger array is a hypervariable minisatellite and allele length varies between individuals, so a single reference length is a convenience rather than a description.

   ▸Evidence that argues against this (1)

   • A single canonical accession is the wrong representation for this protein, and the functional consequence is measurable. Humanising the DNA-binding zinc-finger domain of PRDM9 in C57BL/6 mice repositioned double-strand-break hotspots genome-wide and completely restored fertility in male hybrids; altering one Prdm9 allele changed the behaviour of breaks controlled by the other at chromosome-wide scales. The zinc-finger array is therefore not a detail the canonical sequence abstracts away - it is the part that determines what the protein does, it differs between species and between human alleles, and any analysis keyed to Q9NQV7's canonical 894-aa sequence is keyed to one allele of a hypervariable minisatellite. Davies B et al. 2016
4. E_q89_prdm9_loc1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   HUMAN PRDM9 HAS NO EXPERIMENTAL LOCALISATION IN UNIPROT. Q9NQV7 (PRDM9_HUMAN, primary gene PRDM9, synonym PFM6, 894 aa, read back 2026-09-21) carries two location values, 'Nucleus' and 'Chromosome', and BOTH are ECO:0000250 BY SIMILARITY to Q96EQ9. Q96EQ9 resolves to PRDM9_MOUSE, primary gene Prdm9, 843 aa, synonyms Hst1 and Meisetz - the mouse orthologue. The uncoded note on the human record reads: 'Localizes in nuclei of pre-leptotene, leptotene, and early to mid-zygotene spermatocytes.'

   UniProtKB Q9NQV7 (PRDM9_HUMAN) SUBCELLULAR LOCATION comment and note, fetched 2026-09-21 · UniProt Consortium · UniProtKB 2026
   ▸What this evidence cannot show

   A statement about the record. The similarity source was RESOLVED rather than assumed - a Q96EQ9 that turned out to be a human paralogue would have meant something different from a mouse orthologue, and this beat fetched it (E_q89_prdm9_loc2) rather than inferring from the accession. Absence of a human experiment in UniProt does not prove none exists in the literature; the bounded queries in data.localization_bounded_absence are what bounds that, and they are bounded by titles and abstracts.

   ▸Evidence that argues against this (1)

   • A critic could reasonably say that demanding a human localisation experiment for PRDM9 is precious, since the mouse evidence is solid (three ECO:0000269 attributions for Nucleus) and PRDM9's meiotic function is conserved enough that human hotspot maps are predicted from its zinc-finger array. That is right as far as the NUCLEAR call goes. Where it fails is the question this graph is actually asking. Human PRDM9 is famously NOT like mouse PRDM9 in the respect that matters most - its zinc-finger array is hypervariable between human individuals, which is the whole basis of the childhood-ALL allelic association this entry documents - so 'the mouse does it' is a weaker argument for this gene than for almost any other. And the mouse evidence is entirely spermatocyte evidence, so it cannot speak at all to whether a somatic human cell ever contains this protein, which is the only question that would make PRDM9 relevant to KMT2A breakage. UniProt Consortium 2026
5. E_q89_prdm9_loc2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   THE MOUSE RECORD CONFIRMS THE RESTRICTION RATHER THAN RELIEVING IT. UniProt Q96EQ9 (PRDM9_MOUSE, primary gene Prdm9, synonyms Hst1 and Meisetz, 843 aa, read back 2026-09-21) carries 'Nucleus' on three ECO:0000269 attributions (PubMed:25894966, 27932493, 29478809) and 'Chromosome' on one (PubMed:29478809), with the same uncoded note: 'Localizes in nuclei of pre-leptotene, leptotene, and early to mid-zygotene spermatocytes.' Every localisation statement about PRDM9 in either species is a statement about meiotic prophase I.

   UniProtKB Q96EQ9 (PRDM9_MOUSE) SUBCELLULAR LOCATION comment and note, fetched 2026-09-21 · UniProt Consortium · UniProtKB 2026
   ▸What this evidence cannot show

   Mouse. The note is uncoded in both records, so the four named substages - the most specific content available - carry no attribution. This beat did not open the three cited mouse papers, so the claim is bounded to what the UniProt record states.

   ▸Evidence that argues against this (1)

   • The inference this entry draws - meiosis-restricted, therefore not available to break a somatic pro-B cell - is the right default and it is not airtight, and the graph should hold the exception open rather than closing the question. PRDM9 is a well-documented CANCER-TESTIS-like antigen class member in the sense that meiotic genes are re-expressed aberrantly in tumours, and this entry's own founding association is with childhood B-ALL, where somebody evidently thought a germline PRDM9 mechanism plausible. An absence of reported somatic expression is not a demonstration of somatic silence, and the archive question was not asked here: nobody in this beat checked whether PRDM9 transcript appears in any deposited haematopoietic RNA-seq, which is a measured-not-analysed question answerable without wet work. The bounded absence recorded is therefore literature-bounded only, and it is weaker than it could cheaply be made. UniProt Consortium 2026

• related to →RAG off-target activity: the dominant mutator in ETV6::RUNX1 ALL, and probably not the cause of KMT2A fusions

  Both are germline/recombination-machinery explanations for KMT2A breakage, and both turn out to be unsupported on search rather than refuted by experiment.

• related to →The KMT2A breakpoint cluster region: an 8.3 kb stretch built to break

  Supplies the decisive test nobody has run: if PRDM9 variation mattered, breakpoints should fall preferentially at PRDM9-dependent hotspots, which is checkable against these maps.

• related to →Topoisomerase II cleavage and TOP2 poisons as a source of KMT2A breaks

  An alternative germline-modifier explanation for the same gap - why this child - that the TOP2 model does not address.

• related to →Why KMT2A picks AFF1: spatial proximity, shared transcription hubs, and the unsettled proximity-versus-damage question

  Another genome-architecture account of breakage; PRDM9 would act on where recombination is licensed rather than on spatial contact.

• related to →A partner's 'normal' compartment is usually read off a different tissue, and for at least two KMT2A partners that reading is wrong in blood cells

  Human PRDM9 has NO experimental localisation: both UniProt values are ECO:0000250 by similarity to the MOUSE orthologue Q96EQ9, and every localisation statement in either species is about pre-leptotene to mid-zygotene spermatocyte nuclei (Q89 2026-09-21)

• related to →CTCF - named in the id of an existing entry in this graph and having no entry of its own, and the protein whose binding sites break preferentially in KMT2A-rearranged patients

  Complementary across meiotic prophase: PRDM9 localises in spermatocyte nuclei at four named substages, while UniProt records CTCF as 'Ubiquitous. Absent in primary spermatocytes' - noted as an observation, not a mechanism (Q89 2026-09-21)

• ← related toMeiotic recombination: hundreds of deliberate DNA breaks per cell, made by a topoisomerase relative, at sites chosen by a single zinc-finger protein

  Counter-evidence (Brick 2012, PMID 22660327; Auton 2013, PMID 24348265): Prdm9-null mice still initiate recombination, relocating hotspots to promoters, and canids lack functional PRDM9 entirely -- PRDM9 positions hotspots and sequesters the machinery away from functional elements rather than being required for them

• related to →Meiotic recombination: hundreds of deliberate DNA breaks per cell, made by a topoisomerase relative, at sites chosen by a single zinc-finger protein

  Counter-evidence (Davies 2016, PMID 26840484): humanising the PRDM9 zinc-finger domain repositions hotspots genome-wide and restores hybrid fertility, and one allele alters breaks controlled by the other -- a single canonical accession represents one allele of a hypervariable minisatellite, not the protein's functional determinant