KMT2A-HMGN5: recovered by endogenous CRISPR tagging in OpenCell, not by over-expression - a chromatin-decompacting nucleosome binder with a real reason to be near KMT2A
also called KMT2A-HMGN5 · MLL1-NSBP1 · HMGN5 · nucleosome-binding protein 1
HMGN5 is a small protein that wedges into nucleosomes and loosens chromatin. It came up with KMT2A in a study that tagged proteins at their own genes rather than flooding cells with extra copies, which removes the commonest reason to disbelieve this kind of hit.
The same thing, for a scientist
**Verdict: ADMITTED AS A LEAD, and the methodological point is the reason.** IntAct holds one KMT2A-HMGN5 record: anti-tag co-IP, *association*, HEK293T, KMT2A as prey, intact-miscore 0.35, from **PMID 35271311** [E1]. That is **OpenCell** (Cho et al., *Science* 2022), which combines **genome engineering to tag proteins at their endogenous loci** with confocal live-cell imaging and IP-MS across the proteome [E2]. **The dominant objection this project applies to AP-MS prey rows - tagged over-expression of a protein the cell does not natively make, in a line where its partners are absent - does not apply.** OpenCell's split-mNeonGreen knock-ins express from the native promoter at native copy number. That does not make the row strong; it makes it a *different* kind of weak, and the remaining objections are the cell line (HEK293T, embryonic kidney) and the single record. HMGN5 (P82970, 282 aa, UniProt Nucleus with experimental evidence, ubiquitously expressed) [E3] is a high-mobility-group nucleosome-binding protein whose documented function is competing with histone H1 and decompacting chromatin.
Endogenous tagging changes which objection applies, and only that
This corpus has declined prey rows on a recurring pattern: spoke expansion + over-expressed bait + a cell line that does not express it. interface-menin-cd226-declined-proteome-scale-ap-ms is the worked example - a lymphocyte surface receptor over-expressed in kidney, colon and bone lines.
OpenCell breaks two of those three. It tags at the endogenous locus with a split fluorescent protein, so expression is from the native promoter, and the imaging arm independently confirms where the tagged protein actually is. What it does not break is the third: HEK293T is embryonic kidney, and KMT2A's disease-relevant partners are haematopoietic.
And there is a size argument in this corpus that cuts against the row. q136-interactome-tail-size-and-abundance found that prey recovery tracks protein size in proximity labelling but not in AP-MS, and that the tail had been scored on the wrong side of the pair. HMGN5 is 282 aa - small - and KMT2A is 3,969 aa. This is an AP-MS row, so the size covariate q136 identified should not apply, but the pairing of a very large prey with a very small bait is exactly the regime where that analysis found the scoring was fragile.
Why a nucleosome decompactor near KMT2A is mechanistically interesting
HMGN proteins bind the nucleosome acidic patch and the nucleosomal DNA at the dyad, compete with linker histone H1, and produce a more open, more transcriptionally permissive chromatin state. KMT2A's CXXC domain reads unmethylated CpG and the complex deposits H3K4me3 - both activities that require access to nucleosomal substrate.
So the naive prediction is that an HMGN-loosened nucleosome is a better MLL substrate. That is a testable statement and it is testable in vitro with reagents that already exist: the recombinant-chromatin H3K4 methyltransferase assay used in PMID 23995757 for AKAP95 would run unchanged with HMGN5 substituted.
The counter-prediction is equally cheap. HMGN5 is over-expressed in several solid tumours and its documented phenotype is global chromatin decompaction, so any effect on MLL output may be a non-specific consequence of substrate accessibility rather than a complex-level association - which would make this a substrate-state row rather than an interaction row, the same category error as H3C1.
Cross-family. NUP98 fusions drive HOXA through phase-separated condensates on chromatin rather than through the menin-LEDGF address. A chromatin-decompaction factor is a plausible shared requirement for both routes and has never been examined in either.
| partner a | gene-kmt2a |
|---|---|
| partner b | HMGN5 |
| partner b uniprot | P82970 |
| partner b length AA | 282 |
| partner a uniprot | Q03164 |
| partner a length AA | 3969 |
| verdict | ADMITTED AS A LEAD |
| adjudicated by | q160, cycle 146, 2026-09-24 |
| druggability | Not assessable and probably not desirable. HMGN5's function is global chromatin decompaction; an inhibitor would have no locus selectivity. |
| n methods | 1 |
| n platforms | 1 |
| n publications | 1 |
| endogenous side | yes - OpenCell tags at the endogenous locus and expresses from the native promoter |
| intact miscore | 0.35 |
| cell state | unstimulated, asynchronous |
| epigenetics | HMGN proteins bind the nucleosome, compete with linker histone H1 and decompact chromatin, producing a more transcriptionally permissive state. No direct measurement of an effect on H3K4 methylation exists. |
| absence class | never-measured |
| absence class note | No experiment has tested whether HMGN5 alters MLL-complex H3K4 methylation. The recombinant-chromatin methyltransferase assay from PMID 23995757 would run unchanged with HMGN5 in place of AKAP95. |
interface residues a
[]
structures
[]
ligands
[]
records
[ "KMT2A-HMGN5, anti tag coip (endogenous split-mNeonGreen knock-in), association, HEK293T, KMT2A prey, miscore 0.35, PMID 35271311" ]
cell systems
[ "HEK293T embryonic kidney, endogenous split-mNeonGreen knock-in" ]
localization
{
"compartment": "nucleus",
"subnuclear": [],
"evidence_code": "ECO:0000269",
"shown_for_fusion": "NO - nobody has looked in any fusion."
}measurements
[
{
"quantity": "count",
"subject": "prey-only partners in this census supported by OpenCell (PMID 35271311)",
"value": 4,
"units": "partners",
"n": 212,
"universe": "the 2026-09-24 IntAct REST prey-only census across the six baits",
"null": 15,
"note": "Below this beat's >=15-partner atlas threshold, so the heuristic did NOT flag OpenCell as atlas-like. It is one. The heuristic's answer is retained per rule 92."
}
]- If chromatin decompaction is permissive for the fusion's program, the therapeutic direction implied is chromatin compaction - which is what several failed epigenetic therapies attempted and which damages every dividing cell.
- Endogenous tagging is a genuine methodological advance and this entry leans on it. It should not become a blanket warrant: a single knock-in AP-MS row in embryonic kidney is still one row in the wrong cell.
- Does HMGN5 change MLL-complex H3K4 methylation on recombinant chromatin? A one-week in vitro experiment with published reagents, and the result is informative either way.
- Is the KMT2A-HMGN5 recovery a complex association or a shared-substrate artefact - i.e. is this an interaction row or an H3C1-shaped substrate row? OpenCell's imaging arm could answer the colocalisation half without new wet work.
- HMGN5 is over-expressed in several solid tumours. Is it expressed at all in KMT2A-rearranged blasts? Nothing in this corpus records HMGN5 expression in any leukaemia.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
IntAct REST returns one KMT2A-HMGN5 (Q03164-P82970) record: anti-tag co-immunoprecipitation, interaction type 'association', host Homo sapiens HEK293T embryonic kidney cell, KMT2A as PREY, intact-miscore 0.35, from PMID 35271311. The same publication supplies three other prey-only rows in this census (MYO1E, H2BC21 with KMT2A, and SLC7A6 with PSIP1).
IntAct molecular interaction database, REST findInteractions on Q03164, retrieved 2026-09-24 by q160 (partner HMGN5) · IntAct (EMBL-EBI) 2026What this evidence cannot show
One record. IntAct records the host as HEK293T; OpenCell's published line is HEK293T and the tagging is a split-mNeonGreen knock-in, so the 'anti tag coip' label describes an anti-GFP pull-down of an endogenously tagged protein, not a transfected construct. That distinction is not visible in the IntAct row and had to be read from the paper.
Evidence that argues against this (1)
- OpenCell supports four prey-only rows across the six baits in this census, which is few enough that this beat's own atlas heuristic (>=15 partners) did not flag it - yet it is unambiguously a proteome-scale atlas. The heuristic's answer is kept and is wrong here, exactly as rule 92 predicts for this class of filter. Cho NH et al. 2022
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
PMID 35271311 (Cho et al., Science 2022) is OpenCell: it combines genome engineering to tag human proteins at their endogenous loci, confocal live-cell imaging, mass spectrometry and data science to systematically map protein localization and interactions, reporting that protein localization patterns often carry enough information to identify molecular interactions and that RNA-binding proteins form a distinct subgroup by interaction and localization properties.
OpenCell: Endogenous tagging for the cartography of human cellular organization. · Cho NH, Cheveralls KC, Brunner AD, Kim K, Michaelis AC, Raghavan P, Kobayashi H, Savy L, et al. (27 authors), Leonetti MD · Science 2022What this evidence cannot show
Abstract read in full; the interaction tables and the opencell.czbiohub.org resource were not opened, so the KMT2A row rests on IntAct's curation. `expected_absent_terms` records that KMT2A and leukaemia are absent from the abstract - this is an atlas, and no author set out to study KMT2A.
Evidence that argues against this (1)
- Endogenous tagging removes the over-expression objection but introduces its own: a split-fluorescent-protein knock-in can disrupt the tagged protein's own interactions or destabilise it, and the tag is at one terminus only. A negative in OpenCell is uninterpretable and a positive is a real but single observation. Cho NH et al. 2022
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
UniProtKB P82970 (HMGN5_HUMAN) is 282 aa, with subcellular location Nucleus supported by experimental evidence (ECO:0000269), and tissue specificity recorded as ubiquitously expressed.
UniProtKB entry P82970, subcellular location, length and tissue specificity, retrieved 2026-09-24 · UniProt Consortium 2026What this evidence cannot show
Rule 91: the ECO:0000269 code means a curator read an experiment, and the paper behind it was not opened. Ubiquitous expression removes an expression-mismatch objection but supplies no positive evidence in a haematopoietic cell.
Evidence that argues against this (1)
- A nuclear, ubiquitous, abundant, small, highly charged nucleosome-binding protein is close to a description of the ideal AP-MS background contaminant. Every property that makes HMGN5 mechanistically plausible here also makes it likely to appear in any nuclear pull-down. UniProt Consortium 2026
KMT2A prey in an endogenously tagged HMGN5 pull-down, OpenCell, HEK293T
- related to →Plain-English: chromatin and epigenetics — the sticky notes on the DNA, and which ones matter here
HMGN nucleosome binding competes with linker histone H1 and decompacts chromatin
one of the 135 uncovered human prey-only partners