Experiment: measure KMT2A-AFF1 spatial proximity in the human fetal progenitor, in the cell and at the time the translocation actually happens
also called Capture Hi-C KMT2A AFF1 fetal liver · 3D FISH partner proximity fetal HSC · liCHi-C on fetal liver LMPP
The idea that KMT2A joins AFF1 because the two genes sit near each other in the nucleus has only ever been measured in cancer cell lines. Nobody has looked in the cell type where the fusion actually forms, and the technology to do it on tiny numbers of cells now exists.
The same thing, for a scientist
The only direct measurement of KMT2A-partner proximity is a FISH experiment in transformed human lines showing 2-3% of transcribing KMT2A alleles close to AF9/MLLT3 or AF4/AFF1, interpreted as shared transcription factories [E1]. Everything else in this graph's proximity argument is extrapolation from mouse pro-B cells [E2] or from genome-wide loop-anchor statistics. The experiment that would ground it is a proximity measurement in human fetal liver HSC/LMPP, the compartment this graph places the fusion's origin in. Two methods make it feasible where it was not a decade ago. Low-input capture Hi-C (liCHi-C) maps and robustly compares promoter interactomes at high resolution from scarce clinical cell populations, and the same method recovers genome-wide structural variants and resolves their breakpoints [E3]. And a reference exists for the comparison arm: promoter capture Hi-C interactomes of 31,253 promoters across 17 human primary haematopoietic cell types, showing promoter interactions are highly cell-type specific and reflect the lineage relationships of the haematopoietic tree [E4] - none of which are fetal. A 2026 in-situ Hi-C study now profiles healthy human B-cell progenitors and B-ALL patient samples including KMT2A::AFF1 [E5], which is the closest thing to an answer and is not one, because a leukaemic KMT2A::AFF1 genome contains the derivative chromosome and therefore measures the consequence of the translocation rather than the geometry that preceded it.
The design1 deeper
Cells. Human fetal liver CD34+ HSC and the lymphoid-primed multipotent progenitor fraction, staged across the first and second trimester; cord blood CD34+ as the next-developmental-step comparator; adult bone marrow CD34+ and a mature B cell as the far comparators. The prediction that makes this worth doing is that contact is HIGHER in fetal liver than in the adult compartments, because nothing else currently explains the age distribution of this fusion.
Assay, primary. Capture Hi-C with baits tiling KMT2A and the seven partner loci that account for more than 90% of the recombinome, run as liCHi-C so that it works on the cell numbers fetal tissue actually yields 3.
Assay, orthogonal. 3D DNA FISH with probes on KMT2A and AFF1, scored as inter-probe distance distributions rather than as a binary colocalisation call, and repeated with and without a transcription inhibitor to test the transcription-factory interpretation of 1.
Readout. Contact frequency between KMT2A and each partner locus, normalised the same way across cell types, plus the rank correlation between contact frequency and clinical partner frequency (specified separately in ana-kmt2a-partner-rank-vs-contact-frequency).
The methodological trap that would sink this if ignored
Interchromosomal contacts are the weakest and least reproducible signal in any Hi-C experiment. Trans contact frequency is small, is dominated by the chosen normalisation, and is sensitive to ligation-noise models in a way that cis contacts are not. A naive trans contact matrix will show KMT2A and AFF1 'in contact' at some rate in every cell type, and the between-cell-type difference will be within the noise unless the experiment is designed for it: matched depth, matched crosslinking, matched fragment resolution, and an explicit set of negative-control loci at the same genomic separation and the same expression level. Without those, this experiment will produce a number, and the number will not mean what the title says.
What a positive and a negative each buy
Positive - contact is higher in fetal liver progenitors than in adult, and partner contact rank tracks partner clinical frequency: 3D architecture would then explain both the age distribution and the partner distribution of this fusion, which nothing currently does. It would also strengthen hyp-kmt2a-aff1-shared-loop-anchor-hub.
Negative - contact is no higher in fetal liver, or does not rank with partner frequency: the contact-first model is demoted for this fusion and the field is pushed toward the break-frequency explanation, which is where the strongest published comparison already puts recurrent oncogenic translocations 6. That is the more likely outcome on current evidence and it would be a genuinely useful negative, because the proximity story is repeated in reviews as though it were settled.
Either way the result is interpretable, which is the test of whether an experiment is worth doing.
Why the 2026 B-ALL Hi-C study does not answer it
Ghebrechristos et al. profiled chromatin architecture by in-situ Hi-C and RNA-seq in healthy human B-cell progenitors and in B-ALL patient samples carrying ETV6::RUNX1, KMT2A::AFF1 and BCR::ABL1, and report subtype-specific architectural changes that partly reflect the differentiation stage of the disease 5. This is the right technology in nearly the right cells, and it cannot answer the question asked here for a structural reason: in a KMT2A::AFF1 blast the two loci are covalently joined on a derivative chromosome, so their contact frequency is one by construction. The healthy B-cell progenitor arm of that study is the informative one, and B-cell progenitors are downstream of the fetal liver compartment where this fusion is thought to arise. The experiment is the healthy arm, run on fetal tissue, with partner loci as baits.
| design | Capture Hi-C (liCHi-C, custom baits tiling KMT2A and the seven commonest partner loci) plus orthogonal 3D DNA FISH, across a developmental panel of human haematopoietic progenitors |
|---|---|
| readout | interchromosomal contact frequency KMT2A-partner, per cell type, with rank correlation against clinical partner frequency |
| controls | matched sequencing depth, crosslinking and fragment resolution across cell types; negative-control locus pairs matched for genomic separation and expression level; transcription-inhibitor arm to test the transcription-factory interpretation of E1 |
| prediction if contact first | contact higher in fetal liver than adult compartments; partner contact rank correlates with clinical partner frequency |
| prediction if break frequency dominates | no fetal-versus-adult difference and no rank correlation; partner distribution is then set by break rate and by selection on the product |
| known trap | interchromosomal contacts are the weakest signal in Hi-C and are dominated by normalisation choice; without matched depth and negative-control locus pairs the experiment returns an uninterpretable number |
| closest existing measurement | FISH, 2-3% of transcribing KMT2A alleles near AF9 or AF4, in transformed human lines (E1) |
| cost tier | high |
| timeline | 12-18 months including fetal tissue access and ethics |
cell panel
[ "human fetal liver CD34+ HSC (first and second trimester)", "human fetal liver LMPP", "cord blood CD34+", "adult bone marrow CD34+", "mature B cell" ]
- A positive result would explain where and when the fusion forms and would still offer nothing to intervene on, because nuclear architecture in a fetus is not a druggable variable.
- A negative result removes the field's tidiest story about partner choice and replaces it with break rate plus selection, which is harder to study and harder to publish, so the negative is likelier to go unreported than the positive.
- Fetal tissue work carries ethical and regulatory constraints that vary by jurisdiction and can stop the experiment after the reagents are built.
- Is KMT2A-AFF1 contact frequency higher in human fetal liver HSC/LMPP than in adult bone marrow HSC? This is the question and it has no measurement of any kind.
- Does the transcription-factory interpretation survive a transcription-inhibitor arm? The 2-3% figure is conditioned on the allele being transcribed, and the causal direction has never been tested.
- Can a custom bait set on intragenic breakpoint cluster regions, rather than promoters, be made to work in liCHi-C at fetal-tissue cell numbers? This is the feasibility crux and it is a wet-lab optimisation question, not a biology one.
- Would the healthy B-cell-progenitor arm of the 2026 in-situ Hi-C dataset already contain a usable KMT2A-AFF1 trans contact estimate? If the data are deposited, this is a reanalysis rather than a new experiment, and it should be checked before any tissue is requested.
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
2-3% of KMT2A alleles undergoing transcription do so in close proximity to AF9/MLLT3 or AF4/AFF1, consistent with the genes sharing transcription factories, supporting a model in which TOP2B-introduced breaks in co-located transcribing genes are joined illegitimately.
Model for MLL translocations in therapy-related leukemia involving topoisomerase IIbeta-mediated DNA strand breaks and gene proximity · Cowell IG, Sondka Z, Smith K, et al. · Proceedings of the National Academy of Sciences of the United States of America 2012What this evidence cannot show
FISH colocalisation in transformed and immortalised human cell lines, scored as a binary proximity call. It is the only direct measurement of this geometry for this fusion and it is in the wrong cells. 2-3% is a fraction of transcribing alleles, not a contact probability comparable to a Hi-C statistic.
Evidence that argues against this (1)
- The same paper reports that topoisomerase II alpha and beta occupancy and measured etoposide-induced cleavage do NOT explain why KMT2A breakpoints cluster where they do, so the proximity model is offered by its own authors as a partial account that requires factors other than local enzyme concentration. Reading the 2-3% figure as an explanation of partner choice over-reads the paper. Cowell IG et al. 2012
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In G1-arrested mouse pro-B cells with DSB numbers normalised genome-wide by ionising radiation, translocation frequency is directly related to pre-existing Hi-C spatial proximity, in cis and within subchromosomal domains.
Spatial organization of the mouse genome and its role in recurrent chromosomal translocations · Zhang Y, McCord RP, Ho YJ, et al. · Cell 2012What this evidence cannot show
Mouse, one cell type, and the geometry term is isolated precisely by removing the break-frequency term with irradiation. That design choice is legitimate for the question it asks and removes the variable that dominates recurrent oncogenic translocation.
Evidence that argues against this (1)
- Measuring contact frequency and break frequency in the same system gives a partition that places KMT2A-rearranged disease on the other side: without targeted damage, translocation frequency tracks contact frequency, but translocations associated with recurrent site-directed damage are proportional to the RATE OF BREAK FORMATION, and the authors extend that conclusion specifically to the translocations driving B-cell malignancies. Proximity sets a background rate; it does not decide which pairs become recurrent disease. Hakim O et al. 2012
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Low input capture Hi-C (liCHi-C) maps and robustly compares promoter interactomes at high resolution from scarce clinical cell populations, was applied to the normal and malignant human haematopoietic hierarchy, and can additionally uncover genome-wide structural variants and resolve their breakpoints.
Low input capture Hi-C (liCHi-C) identifies promoter-enhancer interactions at high-resolution · Tomas-Daza L, Rovirosa L, Lopez-Marti P, et al. · Nature communications 2023What this evidence cannot show
A promoter-capture method: its baits are promoters, so measuring contact between two intragenic breakpoint cluster regions requires a custom bait design rather than the published one. Its validation is on postnatal clinical samples, not fetal tissue. 'Scarce' still means thousands of cells, not hundreds.
Evidence that argues against this (1)
- A promoter-capture design answers a cis-regulatory question and is the wrong instrument for a trans-contact question without modification: the reference promoter-interactome atlas of 17 primary haematopoietic cell types reports promoter interactions as highly cell-type specific and lineage-structured, which is a statement about enhancer-promoter contacts within chromosomes. Interchromosomal partner proximity is not what either method was built to quantify. Javierre BM et al. 2016
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Promoter capture Hi-C has been used to identify interacting regions of 31,253 promoters across 17 human primary haematopoietic cell types, and promoter interactomes are highly cell-type specific and reflect the lineage relationships of the haematopoietic tree.
Lineage-Specific Genome Architecture Links Enhancers and Non-coding Disease Variants to Target Gene Promoters · Javierre BM, Burren OS, Wilder SP, et al. · Cell 2016What this evidence cannot show
All 17 cell types are postnatal. No fetal liver HSC or fetal LMPP is included, which is precisely the gap this experiment exists to fill. Promoter-anchored, so non-promoter contacts are not systematically assayed.
Evidence that argues against this (1)
- Cell-type specificity of 3D architecture cuts against the simplest version of the loop-anchor fragility model, which reports TOP2B-mediated breakage at CTCF/cohesin loop anchors as largely cell-type-INDEPENDENT. If architecture is lineage-specific but the breakage it causes is not, then measuring architecture in the right cell type may not predict where the breaks are. Canela A et al. 2017
- E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
In-situ Hi-C and RNA sequencing have been used to profile chromatin architecture in healthy human B-cell progenitors and in B-ALL patient samples carrying ETV6::RUNX1, KMT2A::AFF1 and BCR::ABL, showing subtype-specific architectural changes that partly reflect the differentiation stage of the disease.
3D Chromosome Remodeling in B-cell Development and Acute Lymphoblastic Leukemia · Ghebrechristos YE, Evensen NA, Cathelin RS, et al. · Blood cancer discovery 2026What this evidence cannot show
Read from the abstract only. The KMT2A::AFF1 arm is post-translocation leukaemia, where KMT2A and AFF1 sequence is covalently joined on the derivative chromosome, so it cannot report pre-translocation proximity. Healthy B-cell progenitors are downstream of the fetal liver compartment. The paper's own focus is ERG in ETV6::RUNX1, not KMT2A geometry.
Evidence that argues against this (1)
- Architecture measured after transformation is confounded by the fusion protein's own activity on chromatin: in KMT2A-rearranged cells a HOTTIP-dependent R-loop mechanism actively forms CTCF boundaries and TAD structure, and reversing it mitigates leukaemogenesis. Differences seen between a KMT2A::AFF1 blast and a normal progenitor are therefore at least as likely to be effects of the fusion as causes of it. Luo H et al. 2022
- E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Measuring nuclear architecture and DNA break frequency simultaneously in mouse B lymphocytes shows that non-targeted rearrangements reflect contact frequency whereas translocations associated with recurrent site-directed damage are proportional to the rate of break formation, including those driving B-cell malignancies.
DNA damage defines sites of recurrent chromosomal translocations in B lymphocytes · Hakim O, Resch W, Yamane A, et al. · Nature 2012What this evidence cannot show
Mouse B lymphocytes with engineered recurrent damage. The generalisation to a human fetal progenitor with an endogenous fragile site is an extrapolation, and the KMT2A BCR's break rate in that cell has never been measured.
Evidence that argues against this (1)
- The opposite partition is supported by an equally careful design in the same year and the same system class: with DSB numbers normalised genome-wide by irradiation, translocation frequency in mouse pro-B cells tracks pre-existing Hi-C proximity directly. The two studies differ in whether the break-frequency term is removed or measured, which is why this experiment must measure contact WITHOUT assuming which term dominates. Zhang Y et al. 2012
the fetal-progenitor proximity measurement that entry names as its central evidence gap
contact frequency measured in the compartment the fusion is thought to arise in
LMPP fraction included in the developmental panel
measures the shared-hub link of the hypothesis
- related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
pre-translocation geometry of the two loci that make this fusion
a published scHi-C test of the same question, in cell types that are probably not the cell of origin - read before designing the fetal-liver experiment