Cohesin as an architectural co-lesion class — mutated in AML, associated with NPM1 rather than KMT2A, and leukemogenic on partial loss
also called STAG1/STAG2/SMC1A/SMC3/RAD21 mutations · cohesin complex mutations in leukemia · architectural gene co-mutation
Cohesin is the protein ring that holds DNA loops together. It is broken in a small slice of adult myeloid leukemia, and partly breaking it in mice makes blood stem cells misbehave — but it is not the thing that goes wrong alongside a KMT2A fusion.
The same thing, for a scientist
Cohesin subunit mutations (STAG1, STAG2, SMC1A, SMC3, RAD21) occur in 5.9% of adult AML (23/389 in a uniformly treated NGS cohort), are mutually exclusive of one another, are most frequent in STAG1 (1.8%), STAG2 (1.3%) and SMC3 (1.3%), show a strong association with **NPM1** mutation at similar mutant allele frequencies, and had no effect on overall survival, relapse-free survival or complete remission rate [E1]. Functionally, inducible partial cohesin knockdown in mice does not impair viability but increases progenitor replating capacity, shifts chromatin accessibility and myelomonocytic maturation gene expression, and generates a myeloproliferative-neoplasm-like disease in aged animals — the first demonstration of tumour-suppressor function for the complex [E2]. The relevance to this project is deliberately framed as a **negative plus a warning**: cohesin is not a documented KMT2A-r co-mutation, and its biology tells us that global architectural perturbation is itself oncogenic, which constrains any therapy aimed at loop anchors.
The negative result, stated plainly1 deeper
We looked for cohesin mutations as a co-lesion in KMT2A-rearranged leukemia and did not find a body of evidence for one. What the primary genotype-phenotype literature shows is a strong association with NPM1 mutation, not with KMT2A rearrangement 1. Given that menin inhibitors work in both NPM1-mutant and KMT2A-r leukemia, this is a genuine biological difference between the two menin-dependent diseases that is easy to miss when they are lumped together in a trial.
The cohort is adult (n = 389) and the frequency is 5.9%. In infant KMT2A-r ALL, where the whole genome carries only a handful of somatic mutations, a 6%-frequency co-mutation class from adult AML should not be assumed to transfer at all.
Why a negative belongs in the graph
Because the alternative is that someone later assumes it. The PI's question was what else goes wrong alongside the fusion. The answer for cohesin is: as far as published data show, nothing — and recording that explicitly is worth more than leaving a blank that invites a guess. See ana-noncoding-gap-ledger-kmt2a-r for the full set of cells we checked and found empty.
The warning that does transfer
The mouse genetics are the part that constrains therapy. Cohesin knockdown mice are viable; the phenotype is not death but altered stem-cell homeostasis, gain of replating capacity, changed chromatin accessibility and, with age, a myeloproliferative neoplasm 2. So a drug that partially disables loop architecture systemically would be reproducing a known pre-leukemic genotype.
That is the single most important constraint on the boundary-targeting ideas in genome-ctcf-cohesin-hoxa-tad-boundary: those ideas only survive if the intervention is locus-selective (an ASO against a specific lncRNA, RNase H recruited to one CTCF-binding site) rather than complex-directed. It also means cohesin and CTCF themselves should be treated as anti-targets in this programme, which is the opposite of how a naive 'the fusion needs 3D architecture, so hit architecture' reading would go.
| pathway | cohesin complex / 3D genome architecture |
|---|---|
| mutation frequency adult aml PCT | 5.9 |
| n mutated | 23 |
| n cohort | 389 |
| mutual exclusivity | true |
| co mutation association | NPM1 (strong) |
| association with kmt2a r | none reported in the primary cohort; treated here as an unfilled cell, not a demonstrated absence |
| target status | anti-target: partial loss is pre-leukemic in mouse haematopoiesis |
genes
[ "STAG1", "STAG2", "SMC1A", "SMC3", "RAD21" ]
per gene frequency PCT — the numbers behind the chart
{
"STAG1": 1.8,
"STAG2": 1.3,
"SMC3": 1.3
}prognostic effect
{
"os_hr": 0.98,
"os_ci": "0.56-1.72",
"os_p": 0.94,
"rfs_hr": 0.7,
"rfs_ci": "0.36-1.38",
"rfs_p": 0.3,
"cr_mutated_pct": 83,
"cr_wildtype_pct": 76,
"cr_p": 0.45
}- Cohesin and CTCF should be treated as anti-targets: partial systemic loss reproduces a pre-leukemic mouse genotype, so architecture-directed therapy must be locus-selective or not attempted.
- A cohesin-deficient subclone could be selected by any therapy that depends on intact loop architecture for its effect, including the boundary-directed strategies this beat proposes.
- Because cohesin mutations do not affect survival in AML, there is no clinical signal to warn us if a therapy created them; the toxicity would be silent for years.
- Has anyone specifically tested cohesin mutation frequency in KMT2A-rearranged leukemia, in either infants or adults? The St. Jude and TARGET datasets could answer it directly and we found no paper that reports the cross-tabulation.
- Why does cohesin mutation associate with NPM1 but not KMT2A rearrangement, when both diseases are menin-dependent? A real difference here would be an argument against treating NPM1-mutant and KMT2A-r disease as one indication.
- Do KMT2A-fusion leukemias show cohesin *functional* deficiency without mutation, for example through reduced STAG2 expression? That would be invisible to a mutation panel and is untested.
- Is the loop architecture of a KMT2A-r blast unusually robust or unusually fragile to cohesin depletion? A cohesin degron in a KMT2A::AFF1 line would show whether architectural targeting has any selective window at all.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Cohesin complex mutations occur in 5.9% of AML (23/389 uniformly treated patients by next-generation sequencing), are mutually exclusive across STAG1/STAG2/SMC1A/SMC3/RAD21, are most frequent in STAG1 (1.8%), STAG2 (1.3%) and SMC3 (1.3%), are strongly associated with NPM1 mutation at comparable mutant allele frequencies, and do not influence overall survival, relapse-free survival or complete remission rate.
Mutations in the cohesin complex in acute myeloid leukemia: clinical and prognostic implications · Thol F, Bollin R, Gehlhaar M, Walter C, Dugas M, Suchanek KJ, Kirchner A, Huang L, Chaturvedi A, Wichmann M, Wiehlmann L, Shahswar R, Damm F, Göhring G, Schlegelberger B, Schlenk R, Döhner K, Döhner H, Krauter J, Ganser A, Heuser M · Blood 2014What this evidence cannot show
Adult AML only; 23 mutated patients is too few to exclude an association with any individual cytogenetic subgroup, including KMT2A rearrangement, so 'not associated with KMT2A-r' is an absence of reported association rather than a demonstrated absence. Targeted NGS of five genes will miss structural and non-coding lesions of the same pathway.
Evidence that argues against this (1)
- Later reports have described cohesin mutations at higher frequency in specific AML subsets and in pediatric AML, so the 5.9% figure should be read as cohort-specific rather than canonical; this project has not verified those frequencies and does not assert them. Mullenders J et al. 2015
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Inducible shRNA silencing of cohesin subunits in mice does not deplete cell viability but increases haematopoietic progenitor replating capacity, rapidly alters stem-cell homeostasis and myelopoiesis with widespread changes in chromatin accessibility and myelomonocytic maturation gene expression, and produces a myeloproliferative-neoplasm-like disease with extramedullary haematopoiesis and splenomegaly in aged mice.
Cohesin loss alters adult hematopoietic stem cell homeostasis, leading to myeloproliferative neoplasms · Mullenders J, Aranda-Orgilles B, Lhoumaud P, Keller M, Pae J, Wang K, Kayembe C, Rocha PP, Raviram R, Gong Y, Premsrirut PK, Tsirigos A, Bonneau R, Skok JA, Cimmino L, Hoehn D, Aifantis I · Journal of Experimental Medicine 2015What this evidence cannot show
Partial knockdown in mouse, chosen to mimic heterozygous human lesions; the disease produced is myeloproliferative rather than acute leukemia, and no KMT2A fusion is present. Read as a statement about the tolerability of architectural perturbation, not about KMT2A-r pathogenesis.
Evidence that argues against this (1)
- The mouse result is a general haematopoietic phenotype and the human frequency argues the other way for this disease: cohesin mutations occur in roughly 12% of adult AML and cluster with normal-karyotype and NPM1-mutant disease, not with KMT2A rearrangement, while infant KMT2A-r ALL carries on average one rearrangement and 2.5 coding variants in total. An architectural co-lesion has to be present to co-operate. Blood 2014
subunit of the complex that entry treats as an architectural co-lesion class
one of the five cohesin genes that entry enumerates
named in that entry's gene list; no SMC1A-specific KMT2A evidence exists
- ← related toCTCF/cohesin boundaries and the HOXA topological domain — the 3D scaffold the fusion program runs on
Counter-evidence (Rao 2017, PMID 28985562): acute cohesin degradation eliminates ALL loop domains yet does not cause widespread ectopic gene activation, and many loops re-form within an hour of restoration - so TAD collapse is generally not sufficient to collapse a transcriptional programme, and architecture destroyed pharmacologically rebuilds
- ← related toRNA m6A (METTL3/YTHDF/YTHDC1) in KMT2A-rearranged leukemia — the empty cell that turned out not to be empty
both propose that genome architecture rather than sequence is the affected layer
the architectural role this subunit serves