HOXA9/HOXA10 is unoccupied and unmethylated at H3K4 in two of the three primary infant KMT2A::AFF1 states, while the SEM cell line carries the strongest AFF1 signal over that locus of any sample in the deposit
HOXA9 is the textbook target gene of this fusion, but in two of the three real infant patient samples the fusion is not on it at all and the gene is switched off. The cell line that laboratories use as a stand-in for these patients has the strongest signal there of anything measured.
The same thing, for a scientist
**Measured over a 20 kb hg38 window bracketing both TSSs (chr7:27,160,000-27,180,000; HOXA9 TSS 27,165,537, HOXA10 TSS 27,174,320 - coordinates read back from UCSC refGene, not typed), as fold over each sample's own mean promoter coverage** [E1]: | track | SEM (line) | p279 (B-ALL) | p318 (AML, same patient as p279) | p454 (B-ALL) | |---|---:|---:|---:|---:| | AFF1-1 | **5.88** | 1.75 | 0.55 | 0.77 | | AFF1-2 | 0.90 | 0.61 | 0.09 | 0.20 | | KMT2A-N1 / N2 | - | 3.16 / 2.12 | **0.15 / 0.24** | **0.21 / 0.39** | | KMT2A-C1 / C2 | - | 1.03 / 1.00 | 0.45 / 0.16 | 0.09 / 0.22 | | ENL R1 / R2 | - | 4.53 / 4.07 | **0.13 / 0.10** | **0.06 / 0.18** | | DOT1L R1 / R2 | - | 1.83 / 2.12 | 0.37 / 0.45 | 0.79 / 0.22 | | H3K4me3 R1 / R2 | - | 0.98 / 0.93 | **0.04 / 0.04** | **0.06 / 0.05** | | IgG | - | 1.01 | 0.00 | 0.35 | **In p318 and p454 the locus is not merely unoccupied, it is off.** H3K4me3 at 0.04-0.06x the sample's own mean promoter is a silent promoter, and KMT2A-N at 0.15-0.39x means neither the fusion nor wild-type KMT2A is there. Neither promoter is called a fusion target by the N/C criterion in either sample. Over the whole HOXA cluster (chr7:27,092,994-27,200,091) KMT2A-N1 is 0.70 in p279 and **0.09 / 0.12** in p318 / p454. **And the fusion is elsewhere in the same nuclei, strongly.** In the same samples and the same units, KMT2A-N1 is 13.5 / 13.0 / 18.9 at the MEIS1 promoter, 16.3 / 13.9 / 18.7 at CDK6, 20.9 / 11.5 / 22.0 at PAX5, and ENL reaches 54 / 26 / 60 at CDK6 [E1]. So this is not a failed experiment - it is a locus-specific absence in samples where the instrument is working loudly 20 kb-scale windows away. **This is the known HOXA-negative subgroup, and it is the majority here.** Infant MLL-AF4 ALL splits into mutually exclusive `iALL-HOXA9` and `iALL-IRX1` subgroups; the IRX1 group is diagnosed younger, is more aggressive and has the worse prognosis [E3]. Two of the three states profiled here sit on the HOXA-negative side. **The cell-line gap is the practical point.** SEM's AFF1-1 signal at HOXA9/HOXA10 is 5.88x its own mean promoter - the fourth-highest of the ten loci tested in that sample, behind CDK6 (16.5), PAX5 (14.7) and MEIS1 (11.3) - while in the untreated infant B-ALL p454 the same locus is at 0.77x for the same antibody. **Any mechanism, drug response or biomarker anchored on HOXA in SEM is anchored on a locus that is silent in an infant B-ALL at diagnosis.**
What this does not say
It does not say HOXA9 is irrelevant to KMT2A::AFF1 biology, and it does not say SEM is a bad model. It says the HOXA axis is present in some infant KMT2A::AFF1 disease and absent in other infant KMT2A::AFF1 disease, that the split is already published as a transcriptional subgroup structure 3, and that this deposit shows the split at the level of fusion occupancy and H3K4 methylation rather than expression. n = 2 patients, 3 states.
p279 is the informative middle case
In p279, the B-ALL before the lineage switch, HOXA9/HOXA10 does carry KMT2A-N at 3.2x and ENL at 4.5x with H3K4me3 at 1.0x - real occupancy, but roughly a fifth of what the same sample shows at CDK6 or PAX5. Then in p318, the AML that the same patient's disease became, every one of those numbers collapses to near zero while CDK6, MEIS1 and PAX5 stay occupied. Within one patient, HOXA is the part of the programme that goes.
A coordinate that was wrong and how it was caught
The first version of the region scan used two hand-typed intervals that were wrong - PAX5 placed on chr5 instead of chr9, and AFF1 at chr4:87.86-88.06 Mb instead of chr4:86.94-87.14 Mb. Both returned plausible low numbers rather than an error. The PAX5 row read 0.2-0.65x mean promoter, which would have been reported as 'PAX5 is not a fusion target in primary infants' - the opposite of the truth (11.5-22.0x). Every coordinate in the final scan is read back out of the downloaded refGene table and the extract is kept beside the script.
| script | analyses/q142-aff1-in-primary-infants/regions.py |
|---|---|
| status | run |
| result summary | HOXA9/HOXA10 unoccupied and H3K4me3-negative in p318 and p454; occupied at ~1/5 of CDK6 levels in p279; strongest AFF1 locus tested in SEM after CDK6 and PAX5. |
| region hg38 | chr7:27,160,000-27,180,000 |
- A HOXA-expression or HOXA-methylation biomarker for menin-inhibitor response would be uninformative in the subgroup that has no HOXA occupancy to lose, and that subgroup is two of the three states here.
- Menin inhibitors are usually rationalised through HOXA/MEIS1 collapse; MEIS1 is occupied in all three states while HOXA is occupied in one, so the two halves of that rationale do not travel together in primary infant material.
- Is IRX1 occupied by the fusion in p318 and p454? The deposit's bigWigs can answer this and this beat did not.
- Does the HOXA-negative state precede the lineage switch, or is it a consequence of it? p279 is HOXA-positive and p318 is not, which is one patient's worth of evidence for the second.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
Over a 20 kb hg38 window containing the HOXA9 and HOXA10 TSSs, KMT2A N-terminal signal is 0.15-0.39x the sample's own mean promoter coverage in p318 and p454 with H3K4me3 at 0.04-0.06x, while MEIS1, CDK6 and PAX5 in the same samples carry KMT2A-N at 11.5-22.0x; SEM's AFF1 signal at the same HOXA window is 5.88x.
GSE252378 - AutoCUT&RUN and AutoCUT&Tag profiles of 36 MLL-rearranged leukemia samples (171 GSMs, hg38) · Janssens DH, Henikoff S (Fred Hutchinson Cancer Center / Van Andel Institute) · NCBI Gene Expression Omnibus 2024What this evidence cannot show
Computed in this beat by HTTP range reads of the deposited hg38 bigWigs; nothing downloaded. Values are each sample's own region mean divided by its own mean promoter coverage - a within-sample share, because there is no spike-in in this deposit. SEM cannot be compared to the patients in absolute terms and is not. Two patients only. A promoter-anchored window can miss a distal element; the whole-cluster window was also computed and agrees.
Evidence that argues against this (1)
- The classical MLL-AF4 literature places the fusion firmly on the HOXA cluster, and the SEM value here reproduces that; a reader could argue the primary samples are the anomaly - for instance that cryopreserved, low-blast-fraction patient material loses signal at a facultatively heterochromatic cluster. The H3K4me3 floor of 0.04x argues against a technical loss, since H3K4me3 is the most robust track in the deposit (replicate rho 0.97-0.98), but a shared loss mechanism cannot be excluded. Janssens DH et al. 2024
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
In this deposit, p279 and p318 are the pre- and post-lineage-switch states of one patient.
MLL oncoprotein levels influence leukemia lineage identities · Janssens DH, Duran M, Otto DJ, Wu W, Xu Y, Kirkey D, Mullighan CG, Yi JS, Meshinchi S, Sarthy JF, Ahmad K, Henikoff S · Nature Communications 15:9341 2024What this evidence cannot show
From the paper's text and GEO's per-sample description fields; not confirmed against clinical records.
Evidence that argues against this (1)
- If the pairing annotation were wrong, the within-patient framing collapses to a between-patient comparison; the numbers would be unchanged but the interpretation would weaken considerably. Janssens DH et al. 2024
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Infant MLL-AF4 ALL comprises two subgroups with mutually exclusive HOXA9/HOXA10 and IRX1 expression, the IRX1 group being diagnosed younger with a worse prognosis.
HOXA9/IRX1 expression pattern defines two subgroups of infant MLL-AF4-driven acute lymphoblastic leukemia · Symeonidou V, Ottersbach K · Experimental Hematology 93:38-43 2020What this evidence cannot show
A reanalysis of two published RNA-seq datasets, i.e. transcriptome-level and not occupancy-level, and abstract plus conclusions read rather than the full text. It supplies the expected population structure; it does not itself measure chromatin.
Evidence that argues against this (1)
- Subgroup assignment for p279, p318 and p454 was not made - IRX1 occupancy and expression were not measured here - so the mapping of these samples onto the published subgroups is an inference from the HOXA side only. Janssens DH et al. 2024
- related to →HOXA9 — required for survival of established MLL-rearranged leukemia, and dispensable for starting it
unoccupied and H3K4me3-negative in 2 of 3 primary states
SEM carries the strongest HOXA AFF1 signal in the deposit
MEIS1 occupied in all three states while HOXA is not
the HOXA-negative infant subgroup, seen as occupancy
- related to →Plain-English: chromatin and epigenetics — the sticky notes on the DNA, and which ones matter here
H3K4me3 at 0.04x mean promoter is a silent promoter
source deposit
the corrected coordinates that entry's numbers rest on
extends q142 from 3 to 8 sample records and names what is there instead
second instance of the same failure signature in two cycles