The one experiment that settles this beat: dose-matched menin and DOT1L arms in RS4;11 with H3K4me3 ChIP at PHGDH, plus 13C-serine flux - and it answers the cycle-58 question at the same time
also called dose-matched menin vs DOT1L · H3K4me3 at PHGDH in RS4;11 · serine flux under VTP-50469
One experiment in cells two laboratories already grow would decide whether the serine result in this beat is real biology or an artefact of comparing two drugs at arbitrary doses. Match the doses so both drugs hit the leukaemia's own marker genes equally hard, then look at the serine genes, at the chromatin mark on the PHGDH gene, and at whether the cells actually make less serine.
The same thing, for a scientist
**Four arms, one cell line, one readout set, and it answers three live questions in this graph at once.** (1) The dose-artefact question, asked by `ana-does-atf4s-own-transcript-move-under-a-menin-inhibitor` and still open. (2) The compound-selectivity question this beat's analysis raises. (3) The chromatin attribution in `hyp-the-serine-arm-has-a-second-chromatin-proximal-menin-input-that-is-not-atf4`. It also supplies the one thing the whole serine line of work lacks: **a flux measurement**, which exists in Ewing sarcoma [E2] and has never been made in a KMT2A-rearranged cell. **Design.** RS4;11, plus SEM as a second KMT2A::AFF1 line and **NALM-6 as the KMT2A-wild-type line** (not REH - REH is below the array detection floor for GPT2 at 5.17 log2 RMA and for CBS on all three probes, so it cannot control those genes [E1]). VTP-50469 or revumenib at a fixed dose; EPZ5676 or pinometostat **titrated to match the menin arm on MEIS1 and MEF2C by qPCR** - matching on the canonical target programme is the whole point, and no published experiment has done it. Harvest at 6, 12, 24 h, day 2 and day 7, because both public series sample only day 2 and day 7 and that is far too late to separate a direct effect from arrest. **Readouts.** (i) qPCR or RNA-seq on the pre-registered panel; (ii) **H3K4me3 and H3K79me2 ChIP-qPCR at the PHGDH, CBS, GPT2, PSAT1, CHAC1, TRIB3 and MEIS1 promoters**, with KMT2A and menin ChIP in the same chromatin; (iii) **13C-glucose tracing** for serine and glycine synthesis; (iv) ATF4 protein and phospho-eIF2alpha western, which is `exp-phospho-eif2alpha-and-atf4-protein-under-vtp-50469-in-rs4-11` and should be folded in rather than run separately; (v) medium amino-acid composition **recorded**, because no existing series records it and that omission is a standing alternative explanation for this entire beat. **The decisive contrast is one number:** PHGDH's log2 fold change in the dose-matched DOT1L arm. If it is still 1.5-1.7 log2 shallower than the menin arm while CHAC1 matches within 0.1, the two-input model holds. If it matches the menin arm, the compound-selectivity finding was a dose artefact and this beat's headline reduces to the MOLM13 observation.
Why the control line has to change
Every wild-type contrast in this beat's array data comes from REH under MI-2-2, and REH fails as a control for two of the three menin-selective genes 1:
| gene | RS4;11 baseline | REH baseline | usable? |
|---|---|---|---|
| PHGDH | 11.41, 11.06 | 11.66, 10.58 | yes |
| PSAT1 | 10.35, 11.67 | 8.59, 10.65 | yes |
| SHMT2 | 11.74, 13.41, 11.72 | 10.95, 13.09, 10.89 | yes |
| GPT2 | 10.78 | 5.17 | NO - below the 6.0 floor |
| CBS | 4.83, 4.07, 3.86 | 4.84, 3.85, 3.92 | NO - below floor in BOTH lines |
| ALDH1L2 | 6.47 (1 of 3 probes) | 3.14 | NO |
And separately, the potent compound has never been run in any KMT2A-wild-type line at all - GSE127506/GSE127508 contains none. So the fusion-selectivity of this entire programme rests on a weak compound in one line that cannot measure three of its members.
NALM-6 is the replacement candidate: KMT2A wild type, B-lineage, and already used in the PSAT1 literature alongside REH 3. Whichever line is chosen, its baseline for every panel gene must be printed before the experiment is interpreted - which is how these three failures were found rather than assumed.
What each outcome means, written before the experiment
| result | conclusion |
|---|---|
| Dose-matched DOT1L still leaves PHGDH/CBS/GPT2 1+ log2 shallower; CHAC1 matched | two-input model confirmed; the serine arm carries a menin-selective component |
| Dose-matched DOT1L matches PHGDH too | dose artefact; this beat's compound-selectivity finding falls and the MOLM13 null is all that survives |
| H3K4me3 falls at PHGDH under menin only | the second input is chromatin-proximal, as in Ewing sarcoma |
| H3K4me3 falls at PHGDH under both, equally | the mark is not the discriminator; look for a factor (MYC is the named candidate) |
| No KMT2A/menin occupancy at PHGDH | an intermediate factor is required and the hypothesis is incomplete rather than wrong |
| 13C serine flux unchanged despite PHGDH -3.24 | the transcript programme has no metabolic consequence in this disease, and the whole serine line of work is a pharmacodynamic observation rather than a biology |
| Serine arm collapses in NALM-6 too | fusion-dependence fails; the programme is an off-fusion drug effect |
Writing the outcome table first is deliberate: this beat's own pre-registered panel exists because the previous serine panel in this graph was assembled after a result pointed at it, and the same discipline applies to interpretation.
| design | Four-arm dose-matched comparison in KMT2A::AFF1 cells. Lines: RS4;11 (primary), SEM (second KMT2A::AFF1), NALM-6 (KMT2A-wild-type B-lineage comparator, baselines to be printed first; NOT REH, which is below the detection floor for GPT2, CBS and ALDH1L2). Arms: (1) VTP-50469 or revumenib at a fixed dose; (2) EPZ5676 or pinometostat TITRATED TO MATCH arm 1 on MEIS1 and MEF2C by qPCR - the load-bearing feature, never done in any published experiment; (3) DMSO; (4) optional MYC degron or BET inhibitor arm, which tests the leading alternative to the chromatin attribution. Timepoints 6 h, 12 h, 24 h, day 2, day 7 - the early points are the ones both public series lack. |
|---|---|
| readout | (i) the Q109 pre-registered 41-symbol panel by qPCR or RNA-seq, every value with its own baseline; (ii) H3K4me3 and H3K79me2 ChIP-qPCR at the PHGDH, CBS, GPT2, PSAT1, CHAC1, TRIB3 and MEIS1 promoters plus KMT2A and menin ChIP in the same chromatin; (iii) 13C-glucose tracing for de novo serine and glycine synthesis; (iv) ATF4 protein and phospho-eIF2alpha western (folds in exp-phospho-eif2alpha-and-atf4-protein-under-vtp-50469-in-rs4-11); (v) the medium's amino-acid composition RECORDED, which no existing series does. |
| controls | MEIS1, MEF2C, PBX3 and HOXA10 as the canonical menin-target positive control - and NOT HOXA9, which produced 8 floor collisions in this beat with DMSO baselines of exactly 0.00 in three of four line/day combinations. ACTB, GAPDH, TBP and RPL13A as housekeeping, reported rather than assumed flat: their VTP-minus-EPZ margins in RS4;11 were -0.03 to +0.06 but in MV4;11 were -0.42 to -0.73. Detection floor printed beside every null, on every probe, in every line - the practice that found the REH/GPT2, REH/STC2 and SEM/SOCS2 control failures. |
| the decisive number | PHGDH's log2 fold change in the DOSE-MATCHED DOT1L arm. Two-input model holds if PHGDH stays 1.5-1.7 log2 shallower than the menin arm while CHAC1 matches within 0.1; dose artefact if PHGDH matches. |
| cost tier | med |
| timeline | 6-10 weeks for the transcript and ChIP arms in cells two laboratories already grow; the 13C tracing adds a mass-spectrometry dependency and perhaps 4 weeks more. |
| why it should be prioritised | It is the only experiment in this beat whose negative result would retire a whole line of reasoning rather than qualify it, and the dose-matching that makes it decisive costs a qPCR titration. |
| shown for fusion | PROPOSED FOR the index fusion KMT2A::AFF1 in RS4;11 and SEM. The chromatin and flux readouts have been performed once, in Ewing sarcoma, which carries no KMT2A fusion - so every mechanistic readout proposed here would be its first measurement in a KMT2A-rearranged cell. |
| absence class | never-measured |
| absence class note | A dose-matched menin-versus-DOT1L comparison does not exist in any public dataset, and no H3K4me3 ChIP at any serine-pathway locus and no serine flux measurement exists in any KMT2A-rearranged cell. GEO was queried for the menin-inhibitor serine question and returned the four series this beat used; none contains a dose-response or a wild-type line under a potent compound. |
what it answers that is already asked here
[ "the dose-artefact question raised as an open question by ana-does-atf4s-own-transcript-move-under-a-menin-inhibitor and unanswered for two cycles", "whether the serine arm's compound-selectivity is real (this beat's analysis)", "whether the second input is chromatin-proximal (this beat's hypothesis)", "whether ATF4 protein falls while its transcript does not (the existing phospho-eIF2alpha experiment entry)", "whether serine FLUX changes in this disease at all - never measured" ]
- **A negative result here retires most of this beat.** If the dose-matched DOT1L arm matches the menin arm on PHGDH, the compound-selectivity finding was an artefact, the two-input hypothesis loses its main support, and eleven gene entries become descriptive transcript records. That is the correct outcome if it is what happens, and the experiment is designed so that it is visible rather than arguable.
- **No arm of this experiment touches a patient and none should be read as a step toward one.** Every menin-selective gene it interrogates is dispensable in over 98% of 1,258 DepMap lines and the first two enzymes of the pathway cause a lethal malformation syndrome when lost in the germline.
- **The flux readout could be the most deflating part.** If serine synthesis is unchanged despite a ten-fold PHGDH transcript fall - which PSPH's unmoved transcript makes plausible, since the pathway's exit step does not change - then this whole line of work is a pharmacodynamic marker and not a metabolic vulnerability.
- Which laboratory should run it? The two public series come from independent groups that already grow these lines under both compound classes, and the marginal cost of adding a dose-titration arm to an existing protocol is small. This is a collaboration question rather than a scientific one.
- Does NALM-6 express GPT2, CBS and ALDH1L2 above the detection floor? The control line proposed here is a candidate on the strength of one paper about PSAT1, and its baselines for the rest of the panel are unmeasured. That check costs one qPCR plate and must precede the experiment.
- Should the MYC arm be included from the start? It tests the cheapest alternative to the chromatin attribution, and MTHFD2's MYC dependence plus its compound-independence make it the single most informative extra arm.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset
REH cannot serve as the KMT2A-wild-type control for three of this beat's panel genes: GPT2's single GPL570 probe reads 5.17 log2 RMA in REH against 10.78 in RS4;11, below the declared 6.0 floor; all three CBS probes are below floor in REH (4.84, 3.85, 3.92) and also in RS4;11 (4.83, 4.07, 3.86); and only one of three ALDH1L2 probes is above floor in RS4;11 (6.47) while reading 3.14 in REH. GSE127506+GSE127508, which carry the potent compounds, contain no KMT2A-wild-type line at all. PHGDH, PSAT1 and SHMT2 are well above floor in all three array lines and are usable.
analyses/q109-is-the-serine-collapse-downstream-of-atf4-or-parallel - pre-registered panel over GSE127506+GSE127508, GSE63664 and GSE260456; results.json · this repository 2026What this evidence cannot show
The 6.0 log2 RMA floor is a convention declared in advance, not a measurement, so a gene at 5.9 is not categorically different from one at 6.1 - GPT2's 5.17 and CBS's 4.07 are clearly below, but the rule is a threshold. The array arm uses MI-2-2, the weak compound, so even the usable genes have little dynamic range in this comparison.
Evidence that argues against this (1)
- NALM-6 is proposed as the replacement on the strength of its use in the PSAT1 literature, and its expression of GPT2, CBS and ALDH1L2 has NOT been checked here. Substituting one unverified control line for a measured-unusable one is an improvement only if the substitute's baselines are printed first, and this entry asserts the requirement rather than the fact. Liu Y et al. 2023
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The chromatin and flux readouts proposed here have both been performed in Ewing sarcoma under a menin inhibitor: menin inhibition reduced H3K4me3 enrichment at the PHGDH promoter and completely abrogated de novo serine and glycine biosynthesis as shown by 13C-glucose metabolic tracing, with the serine pathway the most significantly affected pathway by Bru-seq.
Menin regulates the serine biosynthetic pathway in Ewing sarcoma. · Svoboda LK, Teh SSK, Sud S, et al., Lawlor ER · J Pathol 2018What this evidence cannot show
Abstract read this session; full text not retrieved, so the ChIP antibody, promoter coordinates, tracing protocol and replicate counts - all of which a laboratory repeating this would want - were NOT seen. The system is Ewing sarcoma, which carries no KMT2A fusion; this evidence item establishes that the experiment is FEASIBLE and has been done once, not that it will give the same answer here.
Evidence that argues against this (1)
- In that system MI-503 causes loss of menin and MLL1 protein, so their chromatin experiment could not separate an interface-specific effect from a consequence of protein loss. The design proposed here inherits that limitation unless the menin and DOT1L arms are matched on target engagement - which is exactly why the dose-matching is the load-bearing part of this design rather than the ChIP. Jimenez JA et al. 2021
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
NALM-6 is a KMT2A-wild-type B-cell ALL line used alongside REH in a study that eliminated both by downregulating PSAT1, with synergy against cytarabine and activity in a cell-line-derived xenograft - establishing it as an available B-lineage comparator that expresses the serine pathway functionally.
Lycorine eliminates B-cell acute lymphoblastic leukemia cells by targeting PSAT1 through the serine/glycine metabolic pathway. · Liu Y, Du Z, Li T, et al., Chen C · Eur J Pharmacol 2023What this evidence cannot show
Abstract read this session; full text not retrieved. The paper establishes that NALM-6 is killed through PSAT1; it does NOT report NALM-6's expression of GPT2, CBS or ALDH1L2, which is what this design actually needs. So NALM-6 is a candidate, not a validated control, and its baselines must be measured first.
Evidence that argues against this (1)
- That the proposed wild-type control line is itself killed through a serine-pathway gene is an awkward fact for the design: a control line chosen to show the programme is fusion-dependent is a line in which the programme is functionally important. That does not invalidate the regulation contrast - regulation and requirement are different claims - but it means a negative result in NALM-6 will need careful wording. this repository 2026
the dose-matched experiment is the central falsifier of this hypothesis
the decisive number is the PHGDH log2 fold change in the dose-matched DOT1L arm
the ATF4-protein and phospho-eIF2alpha western should be folded into this design rather than run separately
- tests →Pinometostat (EPZ-5676) — the DOT1L inhibitor, and the most instructive failure in this field
the DOT1L arm must be titrated to match the menin arm on MEIS1 and MEF2C - the load-bearing feature no published experiment has
fixed-dose menin arm; VTP-50469 or revumenib