Oncofusion

start / graph / hyp-the-serine-arm-has-a-second-chromatin-proximal-menin-input-that-is-not-atf4

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q109-serine-one-carbon

The serine/one-carbon programme has a SECOND, chromatin-proximal menin input that is not ATF4 - which is why PHGDH is menin-selective where CHAC1 is not, and why the arm is absent in KMT2A::MLLT3

also called two-input model of the serine collapse · menin-H3K4me3-PHGDH arm · serine arm is not an ATF4 output

This graph has assumed that a menin inhibitor switches off serine manufacture because it switches off the stress factor ATF4. Two measurements say there must be a second, separate route: a DOT1L inhibitor copies the ATF4 output genes almost exactly but only half-copies the serine genes, and in a different kind of KMT2A leukaemia the ATF4 genes still fall while the serine genes do not move at all. The second route is most likely the menin complex acting directly on the serine genes' promoters, which has already been shown in a different cancer.

The same thing, for a scientist

**The hypothesis.** The serine/one-carbon transcript collapse under a menin inhibitor in KMT2A::AFF1 cells is produced by **two inputs, not one**: (a) the ATF4-activity arm this graph already holds, which is compound-independent and travels across fusion partners, and (b) a **chromatin-proximal, menin-dependent input acting locus-by-locus at the serine genes' promoters**, which is compound-selective and is restricted to KMT2A::AFF1/B-lineage in the lines measured. Input (b) is not speculative in kind: **menin inhibition reduces H3K4me3 at the PHGDH promoter and abrogates de novo serine synthesis in Ewing sarcoma (PMID 29672864)**, a paper cited once in this corpus and never for that result. **Four independent measurements motivate it, all from a panel declared to disk before any value was read [E1].** 1. **Compound-selectivity splits the two arms.** RS4;11 day 7, VTP-minus-EPZ5676 margins: ISR output median **-0.024** (range -0.085 to +0.095); serine arm median **-0.55** (PHGDH -1.72, CBS -1.47, GPT2 -1.03); menin-target control median **-1.018**. Serine vs ISR p = 0.011; serine vs menin-target control **not separable**, p = 0.106. 2. **Fusion partner splits them again.** In MOLM13 (KMT2A::MLLT3) the serine arm does not move (PHGDH -0.20 from a detected baseline of 7.22, ALDH1L2 +0.08, PSPH -0.31) while the ISR arm does (CHAC1 -1.89) and the positive control collapses (MEIS1 -2.96). 3. **Kinetics order the arms.** Day-2 share of the day-7 effect: menin targets 0.49-0.76, serine arm median 0.384, ISR output median 0.239. GPT2 at 0.58 sits with the menin targets. 4. **The arm is not internally uniform, which is what two inputs predict and one does not.** PHGDH -3.24, PSAT1 -2.40 and PSPH -0.48 are three steps of one pathway in one sample; MTHFD2's margin is -0.03 and its published upstream factor is MYC, not ATF4. **Prior probability 0.5.** The compound-selectivity and the MOLM13 null are solid measurements; the *attribution* to a chromatin-proximal input is an inference from one Ewing-sarcoma chromatin result and from the shape of the data, and a **dose artefact remains a complete alternative explanation for finding 1** - EPZ5676 achieves only 29-61% of the VTP effect on the menin-target control, so every margin in the panel could scale with compound engagement. Finding 2 is not explained by a dose artefact, which is why the prior is not lower.

Go deeper
What this hypothesis replaces, and what it leaves standing

It narrows hyp-the-isr-collapse-is-a-fusion-programme-readout-not-a-menin-interface-readout, which concluded from the ISR arm's compound-independence that the collapse reports fusion-programme disruption by any route. That conclusion stands for the ISR arm and does not extend to the serine arm 1. The two entries are compatible: the ISR arm is a fusion-programme readout by any route, the serine arm has an additional menin-selective component, and the reason the earlier entry did not see it is that PHGDH appeared in its data block as one row of a discriminating-arm list and was not followed up.

It supports but complicates hyp-menin-inhibition-collapses-the-serine-asns-programme-and-should-resensitise-kmt2a-r-all-to-asparaginase. The serine measurement replicates on a pre-registered panel. But ASNS - the gene that hypothesis turns on - has a margin of -0.25 and is 86% reproduced by the DOT1L inhibitor, so ASNS is on the compound-independent side of the split, and the asparaginase prediction therefore does not inherit the menin-selectivity this entry is about.

It leaves the ATF4-activity mechanism entirely intact. ATF4's transcript does not move (-0.86 max) and the eIF2alpha module is flat; nothing here disputes that, and input (a) of this hypothesis IS that mechanism.

Why 'chromatin-proximal' rather than 'a second transcription factor'

A second transcription factor would do as well as chromatin for findings 1, 3 and 4. Three things favour the chromatin reading, none of them decisive:

  • It is the published mechanism for this exact gene. H3K4me3 at the PHGDH promoter, menin-dependent, with 13C flux abrogation 2. No second transcription factor has been proposed for the serine genes in a menin context.
  • The effect is locus-specific within a pathway. PHGDH -3.24, PSAT1 -2.40, PSPH -0.48, MTHFD2 -1.16, MTHFD1 +0.09 in one sample. A transcription factor acting on a co-regulated cassette produces a cassette response; a promoter-by-promoter chromatin dependency does not have to.
  • The compound-selectivity ordering matches the target programme. The serine arm is statistically indistinguishable from the HOX/MEIS/MEF2C arm on margin (p = 0.106) and distinguishable from the ISR arm (p = 0.011). The HOX/MEIS arm is where this graph's H3K4me3/menin-occupancy evidence already lives.

The honest alternative, stated because it is cheap and would be a better answer: MYC. MTHFD2 is MYC-regulated and compound-independent; if the serine arm's menin-selective members turn out to be MYC-independent while the compound-independent members are MYC-dependent, the whole split is explained without invoking chromatin at the serine loci at all. A MYC degron arm tests this and is in the experiment.

Facts and figures
statementThe serine/one-carbon transcript collapse under a menin inhibitor in KMT2A::AFF1 cells has TWO inputs: an ATF4-activity arm that is compound-independent and travels across fusion partners, and a chromatin-proximal menin-dependent input acting locus-by-locus at the serine genes' promoters that is compound-selective and, in the lines measured, restricted to KMT2A::AFF1 / B-lineage. The second input is not novel in kind: menin inhibition reduces H3K4me3 at the PHGDH promoter and abrogates de novo serine synthesis in Ewing sarcoma.
origindata-pattern
prior probability0.5
why that priorThe two discriminating measurements are solid - a 0.53 log2 median difference in compound margin with p=0.011, and a 3.04 log2 difference in PHGDH's response between RS4;11 and MOLM13 with a working positive control in both. The ATTRIBUTION to a chromatin-proximal input is an inference from one Ewing-sarcoma chromatin result plus the shape of the data, and a dose artefact is a complete alternative explanation for the compound-selectivity half. The MOLM13 null is not explained by a dose artefact, which keeps the prior at 0.5 rather than below it. MYC is a live alternative for the second input and would be a better answer if it held.
what it would change if trueA menin-selective component in the serine arm means pinometostat and a menin inhibitor are NOT interchangeable on this programme, which matters because this graph's current position - from the ISR arm - is that they are. It would also mean the serine arm is a candidate pharmacodynamic readout specific to menin-interface engagement, which the ISR arm explicitly is not.
what it does NOT licenseAny target nomination. Every menin-selective member of the arm is dispensable in over 98% of 1,258 DepMap lines, PHGDH and PSAT1 loss causes Neu-Laxova syndrome in the germline, and no PHGDH inhibitor has ever been in a human trial.
shown for fusionThe compound-selectivity and the pathway non-uniformity are shown in KMT2A::AFF1 (RS4;11 on RNA-seq, SEM on array, MV4;11 in an independent laboratory). The ABSENCE of the arm is shown in KMT2A::MLLT3 (MOLM13). The chromatin mechanism is shown only in Ewing sarcoma, which carries no KMT2A fusion at all, and that is the hypothesis's largest gap.
absence classmeasured-not-analysed
absence class noteThe compound-selectivity comparison and the MOLM13 arm were both available in data deposited in 2019 and had never been run. The chromatin half is never-measured in this disease: no H3K4me3 ChIP at any serine locus exists in any KMT2A-rearranged cell.
predictions
[
  "A dose of EPZ5676 titrated to match VTP-50469 on MEIS1 still leaves PHGDH, CBS and GPT2 with a menin-selective margin, while CHAC1, TRIB3, ATF5 and MTHFD2 remain matched. This is the central prediction and the hypothesis fails without it.",
  "H3K4me3 at the PHGDH promoter falls under VTP-50469 in RS4;11 and falls LESS, or not at all, under a dose-matched EPZ5676 - while H3K79me2 at the same locus falls under both.",
  "KMT2A and menin ChIP show fusion-complex occupancy at the PHGDH, CBS and GPT2 promoters in RS4;11 and NOT at CHAC1, TRIB3 or ATF5.",
  "In MOLM13, H3K4me3 at the PHGDH promoter does not change under VTP-50469, matching its flat transcript - i.e. the absence of the serine arm in KMT2A::MLLT3 is visible at the chromatin layer and is not a downstream buffering effect.",
  "A MYC degron collapses MTHFD2 and the ISR arm without collapsing PHGDH, CBS and GPT2. If instead it collapses all of them, the second input is MYC and this hypothesis's chromatin attribution is wrong while its two-input structure survives.",
  "Serine and glycine FLUX from 13C-glucose falls in RS4;11 under VTP-50469, as it does in Ewing sarcoma - which would give the mechanism a phenotype it currently lacks, since PSPH is unmoved and no panel gene is a dependency.",
  "In a KMT2A-wild-type B-ALL line that expresses the panel (NALM-6, since REH is below floor for GPT2 and CBS), a potent menin inhibitor does NOT collapse the serine arm."
]
falsifiers
[
  "A DOSE-MATCHED EPZ5676 arm abolishes the serine arm's margin. That kills the compound-selectivity half outright and the hypothesis reduces to the existing single-input model plus a fusion-partner observation.",
  "H3K4me3 at the PHGDH promoter is UNCHANGED under VTP-50469 in RS4;11 while the transcript falls 3.24 log2. The chromatin attribution would then be wrong, and the second input would have to be a factor rather than a mark.",
  "A MYC degron collapses PHGDH, CBS and GPT2 as completely as it collapses MTHFD2. The split would then be MYC-versus-ATF4 rather than chromatin-versus-ATF4, and the entry's specific mechanism is refuted while its two-input structure stands.",
  "A potent menin inhibitor collapses the serine arm in a KMT2A-WILD-TYPE line. That removes the fusion-dependence both inputs rest on and makes the whole programme an off-fusion drug effect.",
  "The MOLM13 null turns out to be FLT3-ITD-dependent - i.e. MOLM13 plus a FLT3 inhibitor reveals the serine arm. Finding 2 would then be about a co-mutation and not about fusion partner or lineage, and PMID 33436370 makes that the leading alternative.",
  "The per-gene SSP magnitudes in PMID 29672864 show PSPH falling as little in Ewing sarcoma as it does here. The 'pathway is not a unit in this disease' argument would then be an artefact of reading a grouped abstract."
]
experiments
[
  "exp-dose-matched-menin-versus-dot1l-arms-with-h3k4me3-chip-at-phgdh-in-rs4-11"
]
analyses
[
  "ana-the-serine-arm-is-not-an-atf4-output-it-is-compound-selective-and-absent-in-kmt2a-mllt3"
]
How this could bite us back
  • **This hypothesis makes the serine arm look more specific and therefore more attractive, and that is precisely the direction in which it should not be read.** Every menin-selective member of the arm - PHGDH (1.1%), CBS (0.4%), GPT2 (0.5%) - is dispensable in over 98% of 1,258 DepMap lines, and the germline phenotypes of the first two enzymes of the pathway are Neu-Laxova syndrome, lethal in utero.
  • **If the arm is KMT2A::AFF1/B-lineage restricted, the addressable population narrows** - and the AML subtype where menin inhibitors are furthest advanced clinically is the one where this beat measured no serine response at all.
  • **A dose-response would refute this entry in a week and has not been done.** Nothing here should be cited as established mechanism, and the entry is graded speculative for that reason rather than as a formality.
Still open
  1. Is the whole compound-selectivity result a dose artefact? Asked by the cycle-58 beat and now again here. It has been the leading alternative explanation for two consecutive cycles and remains unanswered because no public dataset contains a dose-matched arm.
  2. Is MYC the second input rather than chromatin? MTHFD2 is MYC-regulated and compound-independent; nothing has tested whether the menin-selective members are MYC-independent, and a degron experiment would answer it more cheaply than a ChIP.
  3. Does the serine programme have any phenotype in this disease? PSPH is unmoved, no panel gene is a DepMap dependency, and serine flux has never been measured in a KMT2A-rearranged cell. A mechanism with no measured consequence is the weakest part of this hypothesis.
  4. Is the MOLM13 null about fusion partner, lineage or FLT3-ITD? Three confounded variables and one line each. MOLM13 plus a FLT3 inhibitor is the cheapest discriminator and PMID 33436370 makes FLT3-ITD the leading candidate.
Every source behind this entry (3)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    On a panel declared to disk before any expression value was read, at RS4;11 day 7 the per-gene VTP-50469-minus-EPZ5676 margin has median -0.024 for the ISR-output arm and -0.55 for the serine/one-carbon arm (Mann-Whitney two-sided p = 0.01119) while the serine arm is not separable from the menin-target positive control at median -1.018 (p = 0.10629); in MOLM13 the serine arm does not move (PHGDH -0.202 from a detected baseline of 7.22 log2 CPM, ALDH1L2 +0.077, PSPH -0.306, GPT2 -0.213) while CHAC1 falls 1.892 and MEIS1 falls 2.956 in the same samples; the day-2 share of the day-7 effect is 0.49-0.76 for the menin targets, median 0.384 for the serine arm and median 0.239 for the ISR output; and within the serine arm PHGDH (-3.24), PSAT1 (-2.40) and PSPH (-0.48) are three consecutive steps of one pathway with 2.8 log2 between the extremes, while MTHFD2's margin is -0.032 and ASNS's is -0.253.

    What this evidence cannot show

    One line per platform, single doses, no dose-response, n=2-3 per arm, day 7 so arrest is a competing explanation. Genes below 1 log2 were excluded by a pre-declared rule, so the arm sizes are 9 and 4 and the p-values rest on those n - they are reported for completeness and the per-gene table is the result. MOLM13 differs from RS4;11 in fusion partner, lineage AND FLT3-ITD status simultaneously, so finding 2 cannot be attributed to any one of them. No medium amino-acid composition is recorded in any series.

    Evidence that argues against this (1)
    • The dose artefact is a complete alternative explanation for finding 1: EPZ5676 at 1 uM reproduces only 29-61% of the VTP-50469 effect on MEIS1, MEF2C, PBX3 and HOXA10, so a margin that scales with compound engagement would put the serine arm exactly where it is, and the ISR arm would be the anomaly (two compounds both saturating a steeply responsive locus). Nothing in any public dataset separates these, and a dose-matched arm is prediction 1 of this hypothesis for that reason. this repository 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In Ewing sarcoma, menin inhibition diminished expression of serine biosynthetic pathway genes, REDUCED H3K4me3 ENRICHMENT AT THE PHGDH PROMOTER, and completely abrogated de novo serine and glycine biosynthesis as shown by 13C-glucose tracing; the serine pathway was the pathway most significantly affected by MI-503 in Bru-seq; and PHGDH inhibition reduced proliferation, viability and tumour growth in vivo.

    Menin regulates the serine biosynthetic pathway in Ewing sarcoma. · Svoboda LK, Teh SSK, Sud S, et al., Lawlor ER · J Pathol 2018
    PMID 29672864DOI
    What this evidence cannot show

    Abstract read this session; full text not retrieved, so the H3K4me3 magnitude, antibody and replicate count were not checked. **The system is Ewing sarcoma, which carries NO KMT2A fusion** - MLL1/KMT2A and menin are overexpressed wild-type proteins there - so this is the mechanism's existence proof in a different lesion class, not in the index disease. MI-503 causes loss of menin and MLL1 PROTEIN in that system, so a chromatin change downstream of protein loss cannot be separated from an interface-specific effect. No H3K4me3 ChIP at the PHGDH promoter exists in any KMT2A-rearranged cell.

    Evidence that argues against this (1)
    • The same senior author's follow-up attributes the SSP loss to loss of ATF4 and states that SSP downregulation 'is indicative of broader inhibition of ATF4-dependent transcription'. So the group that measured the H3K4me3 change subsequently framed the mechanism as ATF4-mediated, and reading their 2018 chromatin result as evidence of an ATF4-INDEPENDENT route is this hypothesis's inference and not the authors' claim. Jimenez JA et al. 2021
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    The genes that fall hardest under a menin inhibitor in this panel are not dependencies and the gene that is a dependency does not fall: PHGDH -3.24 with 1.1% of 1,258 DepMap screens below -0.5, ALDH1L2 -3.40 with 0.0%, CBS -4.07 with 0.4%, CTH -1.57 with 0.0% and a positive mean gene effect, PSAT1 -2.40 with 8.2% - against MTHFD1 at +0.09 under the drug with 28.8% of screens below -0.5 and a leukaemia mean of -0.744, deeper in leukaemia than MEN1 (-0.5459) or DOT1L (-0.3931) in the same fetch.

    What this evidence cannot show

    Rule 83: no release identifier in this payload, so depmap_release is null. Leukaemia screens are matched on the payload's own diseaseFromSource string and are NOT split by KMT2A status, so MTHFD1's -0.744 is a lineage figure and not a fusion figure. Rule 84: the percentages are against an absolute -0.5 threshold, not a median split. A CRISPR knockout in nutrient-replete medium understates every biosynthetic dependency here, and MTHFD2 is this panel's own counter-example - flat in DepMap and yet its suppression prolonged survival in an MLL-AF9 mouse model. MEN1's shallower figure reflects a SELECTIVE dependency, so 'deeper than MEN1' is a statement about depth in leukaemia and not about importance.

    Evidence that argues against this (1)
    • This item is the reason the hypothesis carries no target nomination, and it also weakens the hypothesis itself: if none of the menin-selective genes matters to the cell, the second input this entry proposes is an input to a programme with no demonstrated consequence. A mechanism with no phenotype is a smaller finding than it looks, and the flux measurement that would give it one has never been done in this disease. Pikman Y et al. 2016
Connected to