Menin inhibition collapses the serine/one-carbon/ASNS programme in KMT2A::AFF1 cells - PHGDH -3.24, PSAT1 -2.40, ASNS -1.72, flat in a wild-type line - so it should re-sensitise KMT2A-rearranged ALL to asparaginase, a combination PubMed has zero records for

This panel was NOT pre-declared: it was assembled after the 24-gene concordance list pointed at ATF4, so it is a confirmatory look and not a discovery screen, and it should be discounted accordingly relative to the pre-registered HOXA/MEIS1 panel. The wild-type contrast uses a DIFFERENT COMPOUND (MI-2-2) because GSE127506 has no KMT2A-wild-type line, and MI-2-2's whole-transcriptome effect in REH was small anyway - so 'flat in wild-type' is the weakest link in the chain. Transcript abundance is not flux: a fall in PHGDH and PSAT1 mRNA is not a measurement of serine synthesis, and a fall in ASNS mRNA is not a measurement of asparagine synthetase activity or protein. No p-values. Day 7 is late enough that a slowing or dying population could reduce biosynthetic transcripts as a consequence rather than a cause.

• The whole programme falling together is as consistent with cells shutting down biosynthesis because they are arrested or dying as with a specific menin-dependent transcriptional effect. At day 7 under an effective drug, reduced translation and growth lower amino-acid demand, and the ATF4 axis reports exactly that. A growth-matched control, or the day-2 arm, would discriminate - and this evidence item uses day 7 because the effect sizes are larger, which is precisely the choice that makes the confound worse.
• ATF4 itself moves only -0.86, less than most of its targets. That is compatible with ATF4 acting post-transcriptionally (it is classically regulated by translation, not transcript level) but it also means this analysis has no direct evidence that ATF4 activity changed at all. The signature is being read off the targets.

Abstract-level; the full text was not read, so which serine-pathway genes moved and by how much in that system is not established here. Ewing sarcoma is a different disease, lineage and fusion, and EWS-FLI1 has no menin-binding motif analogous to KMT2A's MBM - so the shared node in that system is menin itself rather than a menin-binding fusion, which is a different architecture from KMT2A::AFF1. Verified live by efetch 2026-09-22; already cited twice in this graph, both times in a sarcoma-risk context rather than as a mechanistic precedent.

• If menin drives serine biosynthesis wherever menin is active, then menin inhibition should collapse this programme in cells with no KMT2A fusion too - and in the one wild-type line available it does not. Either the leukaemic version is fusion-dependent while the Ewing version is not, or the wild-type null is an artefact of testing it with the weak compound. Both readings are live and the experiment that separates them is named below.

Abstract-level, and it is the ONLY record returned by ASNS[tiab] AND MLL[tiab]. It is AML rather than infant B-ALL, EVI1 rather than menin, and the mechanism by which its metabolic reprogramming alters asparaginase sensitivity is not established from the abstract - it may not run through ASNS at all. It is cited here as evidence that the axis is live in MLL-rearranged disease, not as support for the specific mechanism proposed. Uncited elsewhere in this graph; verified live by efetch 2026-09-22.

• The direction is the useful part and also the trap: EVI1 overexpression INCREASES asparaginase sensitivity, so in that system the oncogenic driver makes cells MORE vulnerable to asparaginase, not less. If the analogy held tightly it would predict that removing the driver's programme - which is what menin inhibition does here - makes cells LESS sensitive, the opposite of this hypothesis. The two are reconciled only through ASNS specifically, and that reconciliation is assumed rather than shown.

A parsed zero bounds INDEXED TITLES AND ABSTRACTS ONLY. This project has already been caught by exactly this: a genuinely reported NUTM1 fusion partner returned a parsed zero because the fact lived in a review's table. A menin-inhibitor/asparaginase combination tested inside a drug-screen supplement, a conference abstract, a trial protocol or an unpublished negative result would not appear in these counts. So 'nobody has published it in a title or abstract' is the claim, not 'nobody has tried it'.

• An absence of publications on a plausible and cheap combination, in a field with several approved and investigational menin inhibitors and a seventy-year-old chemotherapeutic, is weak evidence of novelty and might be evidence that it has been tried and did not work. Combination screens are routinely run and negative arms routinely go unreported.

Day 2 is early for a 7-day experiment but it is not an immediate-early timepoint, so this bounds the arrest confound rather than eliminating it - cells can be slowing by 48 hours. The d2/d7 fractions are ratios of two noisy estimates and are unstable where the d7 effect is small; they are reported only for genes with a d7 effect above 0.5 log2. ASNS specifically is -0.84 at day 2, BELOW the 1 log2 threshold used everywhere else in this analysis, so the gene the hypothesis turns on is the weaker part of this evidence item. And an early effect is still transcript, not flux.

• The temporal argument can be turned around. That the biosynthetic enzymes move with the canonical targets at day 2 is equally consistent with both being part of one direct transcriptional programme AND with both being early consequences of a proliferation arrest that begins before day 2 - VTP-50469's IC50 is 20-60 nM and the dose here is 330 nM, so growth effects at 48 hours are expected. What would settle it is a growth-matched control or a 6-12 hour timepoint, and neither exists in this series.
• If ATF4 loss at day 2 drove the serine enzymes, the enzymes should lag ATF4 rather than lead it - and PHGDH (-1.08), PSAT1 (-1.05) and CBS (-1.86) are all LARGER at day 2 than ATF4 itself (-0.51). That ordering fits a shared upstream cause better than a linear ATF4-to-enzyme cascade, and ATF4 is in any case regulated translationally rather than transcriptionally, so its mRNA is a poor proxy for its activity.

ASNS - the gene this hypothesis turns on - is the WEAKEST replicated member at -1.16 raw and -0.83 after housekeeping adjustment. The serine enzymes upstream replicate far more strongly than the asparaginase-relevant step does. Housekeeping is not flat in that matrix (ACTB -0.73), the exposure duration is unstated, MV4;11 is myeloid/biphenotypic rather than B-lymphoid, and it is not infant-derived either. See the full analysis for the raw-versus-adjusted table and the PCGF1/BCOR arms.

• The non-replication of TRIB3, ATF3, SESN2 and DDIT3 removes the framing this hypothesis was originally built on. The ATF4-signature identification came from a 24-gene concordance list in which those genes were central; two of them now reverse sign in an independent line. The hypothesis survives on the serine/ASNS arm, which is the part that matters for asparaginase, but the mechanistic story attached to it is weaker than when it was written this morning.
• A global housekeeping drift in the replicating dataset is itself evidence for the arrest explanation this hypothesis has never been able to exclude. Two independent datasets now show a coordinated fall in biosynthetic transcripts under an effective drug, and in one of them the housekeeping genes fall too.