Oncofusion

start / graph / gene-mthfd2

Genemoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q109-serine-one-carbon

MTHFD2 (P13995) - the one panel member with a published MLL-AF9 in-vivo survival result, and the serine gene that behaves exactly like an ISR gene

also called methylenetetrahydrofolate dehydrogenase/cyclohydrolase 2, mitochondrial · MTDC_HUMAN · NMDMC

MTHFD2 is a mitochondrial folate enzyme switched on in cancers far more than in normal tissue. It is the only gene in this beat's panel that somebody has already knocked down in an MLL-fusion mouse leukaemia - survival improved - and in our data a DOT1L inhibitor copies the menin inhibitor's effect on it almost exactly, which points at MYC rather than at menin.

The same thing, for a scientist

MTHFD2 (P13995, 350 aa, exactly one reviewed human record, synonym NMDMC) is the mitochondrial methylenetetrahydrofolate dehydrogenase/cyclohydrolase [E3]. **It is the only gene in this beat's pre-registered panel with a published functional result in a KMT2A-fusion model:** MTHFD2 is the most differentially expressed metabolic enzyme in cancer versus normal cells; knockdown decreased growth, induced differentiation and impaired colony formation in primary AML blasts; **in human xenograft and MLL-AF9 mouse leukaemia models MTHFD2 suppression decreased leukaemia burden and prolonged survival**; FLT3-ITD is a biomarker of response; and **MYC regulates MTHFD2** [E4]. **Measured here:** RS4;11 day 7, VTP-50469 **-1.16** from a baseline of 8.65 log2 CPM and EPZ5676 **-1.13** - a margin of **-0.03, the most compound-INDEPENDENT gene in the serine arm and indistinguishable from the ISR-output arm** (CHAC1 -0.09, TRIB3 -0.02) [E1]. Its day-2 share is 24%, again matching the ISR arm rather than the serine arm [E1]. **That fits the published mechanism**: a MYC-driven gene should fall whenever the fusion programme falls, by either route - exactly what `hyp-the-isr-collapse-is-a-fusion-programme-readout-not-a-menin-interface-readout` predicts. MOLM13 -0.29, flat; MV4;11 -1.03 against SGC0946 -0.28 [E1]. DepMap: mean **-0.0282, 1.7% of 1,258 below -0.5, leukaemia -0.1374** - despite the mouse survival result, and that disagreement is the entry's point [E2].

Go deeper
MTHFD2 splits the serine arm in two, and does it with a named mechanism

This beat's finding is that the serine arm carries a menin-selective component the ISR arm does not. MTHFD2 is the member that does not 1:

geneVTP-EPZ marginday-2 sharebehaves like
PHGDH-1.720.33menin target
CBS-1.470.46menin target
GPT2-1.030.58menin target
MTHFD2-0.030.24ISR output
CHAC1-0.090.28ISR output
TRIB3-0.020.13ISR output

And unlike the rest of the panel, MTHFD2 has a published upstream transcription factor that is not ATF4: MYC 4. This graph holds gene-myc and treats MYC as a fusion-programme output, so a MYC-driven gene falling equally under both compounds is the expected result and needs no menin-selective input at all.

The honest reading of the serine arm is therefore that it contains at least two sub-programmes with different upstream inputs, and MTHFD2 is the worked example on the compound-independent side. That is a more defensible statement than "the serine arm is menin-selective", and it is the one the data support.

A mouse survival result on a gene no cell line needs

MTHFD2 suppression prolonged survival in an MLL-AF9 mouse leukaemia model and in human xenografts, and impaired colony formation in primary AML blasts 4. Its DepMap gene effect is -0.0282 with 1.7% of 1,258 lines below -0.5 2.

Those are not in conflict; they measure different things, and the disagreement is instructive for this whole beat. A knockdown that induces differentiation over weeks, in vivo, in a leukaemia-initiating population, is not what a pooled CRISPR proliferation screen in folate-replete culture measures. For a one-carbon enzyme, a flat DepMap score is weak evidence of dispensability - and this entry is the panel's counter-example to its own DepMap section.

Note also the FLT3-ITD biomarker 4. That is the second time in this beat a serine/one-carbon dependency has resolved to FLT3-ITD rather than to a fusion, the first being PHGDH in FLT3-ITD AML (PMID 33436370). MOLM13, the line where this beat's serine arm does not move, carries FLT3-ITD - so the line most predicted to need MTHFD2 is the one where a menin inhibitor leaves it alone.

Facts and figures
gene symbolMTHFD2
uniprotP13995
length AA350
entry nameMTDC_HUMAN
protein nameBifunctional methylenetetrahydrofolate dehydrogenase/cyclohydrolase, mitochondrial
ensembl gene IDENSG00000065911
ensembl readbackdisplay_name read back as MTHFD2 before use
accession resolutiongene_exact:MTHFD2 AND organism_id:9606 AND reviewed:true returns EXACTLY ONE reviewed row (rule 62 asserted): P13995, MTDC_HUMAN, gene_primary MTHFD2, synonym NMDMC, 350 aa.
localizationMITOCHONDRION, but UNCODED. UniProt P13995's single SUBCELLULAR LOCATION statement carries no ECO evidence code and no citation, so it is a curator assertion rather than a measurement - recorded as uncoded rather than discarded, per rule 91. No localisation experiment on MTHFD2 has been performed in any KMT2A-rearranged cell.
subcellular localizationMITOCHONDRIAL by uncoded UniProt annotation and by universal usage in the one-carbon literature, where MTHFD2 IS the mitochondrial arm of the folate cycle. The annotation is weak; the literature consensus is not. Both are stated so a reader can see the difference.
subnuclear localizationNONE, answered negative rather than left blank: no nuclear pool is annotated or reported for MTHFD2, unlike SHMT1 and SHMT2 in the same panel, which both carry ECO:0000269 Nucleus statements.
cell typeRS4;11 and SEM (KMT2A::AFF1 B-ALL), REH (KMT2A wild type), MOLM13 (KMT2A::MLLT3 AML) and MV4;11 (KMT2A::AFF1 myeloid) measured here. AML cell lines, PRIMARY AML blasts, human xenografts and an MLL-AF9 MOUSE leukaemia model in PMID 27325891 - the only primary-sample and in vivo evidence on any gene in this beat's panel. NO B-lineage KMT2A::AFF1 model has been used functionally.
cell stateMeasured in the drug-treated arresting state at day 7 of continuous menin-inhibitor exposure under uncontrolled medium folate and serine supply. Day-2 share 24%, a late profile matching the ISR-output arm. The state in which MTHFD2 has published function is DIFFERENTIATION: its knockdown induced differentiation in AML cells and impaired colony formation in primary blasts, which is the same phenotypic axis a menin inhibitor acts on - so a menin inhibitor and MTHFD2 loss may be pushing the same lever rather than two.
epigenetic readoutNONE MEASURED at the MTHFD2 locus in any menin- or KMT2A-scoped study. The published upstream regulator is MYC, a transcription factor rather than a chromatin mark, and MYC is itself a fusion-programme output in this graph - so MTHFD2's chromatin state is two steps removed from the fusion complex and has never been assayed. No H3K4me3, H3K79me2, KMT2A or menin ChIP at MTHFD2 exists. Bounded absence.
shown for fusionYES, and it is the only one in this panel: MTHFD2 suppression prolonged survival in an MLL-AF9 (KMT2A::MLLT3) MURINE MYELOID leukaemia model and in human AML xenografts (PMID 27325891). The transcript readout under a menin inhibitor is shown in KMT2A::AFF1 (RS4;11 -1.16, MV4;11 -1.03) and is FLAT in the human KMT2A::MLLT3 line MOLM13 (-0.29) - flat in the human counterpart of the model where the functional result was obtained, which is this entry's central tension. Nothing has been shown functionally in KMT2A::AFF1.
the published upstream factor is MYC not ATF4PMID 27325891 states that MYC regulates MTHFD2 expression. That is the only panel gene with a named upstream factor other than ATF4, and it explains the compound-independence directly: MYC is a fusion-programme output reachable by either the menin or the DOT1L route, so a MYC-driven gene should fall equally under both - which is what is measured.
absence classmeasured-not-analysed
absence class noteThe transcript response was in deposited data and unanalysed. Never-measured: an MTHFD2 dependency or knockdown experiment in any KMT2A::AFF1 B-lineage model, and any chromatin measurement at the locus.
the measurement rs411 d7 log2fc
{
  "vtp50469_330nM": -1.16,
  "epz5676_1uM": -1.13,
  "vtp_minus_epz": -0.032,
  "epz_fraction_of_vtp": 0.97,
  "dmso_baseline_log2cpm": 8.65,
  "detected": true,
  "floor_collision": false,
  "d2_vtp": -0.276,
  "d2_share_of_d7": 0.24,
  "reading": "the most compound-INDEPENDENT gene in the serine arm; indistinguishable from CHAC1 (-0.085) and TRIB3 (-0.024)"
}
molm13 kmt2a mllt3
{
  "vtp": -0.292,
  "epz": -0.198,
  "dmso_baseline_log2cpm": 8.09,
  "detected": true
}
mv411 independent lab
{
  "menin_vtp50469_500nM": -1.031,
  "dot1l_sgc0946_5uM": -0.28,
  "margin": -0.751,
  "dmso_baseline_log2cpm": 5.35,
  "note": "the MV4;11 margin disagrees with the RS4;11 margin of -0.03, as it does for ALDH1L2; housekeeping is not flat in that matrix"
}
the measurement array gse63664 every probe
{
  "201761_at": {
    "rs411_dmso_log2_rma": 12.38,
    "rs411_mi22": -0.36,
    "sem_dmso": 12.42,
    "sem_mi22": -0.39,
    "reh_dmso": 11.92,
    "reh_mi22": 0.41
  },
  "234976_x_at": {
    "rs411_dmso_log2_rma": 12,
    "rs411_mi22": -0.48,
    "sem_dmso": 12.21,
    "sem_mi22": -0.45,
    "reh_dmso": 11.21,
    "reh_mi22": 0.7
  },
  "detection_floor_used": "6.0 log2 RMA; both probes far above it in all three lines"
}
essentiality
{
  "depmap_release": null,
  "depmap_release_note": "Rule 83: no release identifier in the Open Targets payload. Fetched 2026-09-22.",
  "depmap_screens": 1258,
  "depmap_tissue_blocks_flattened": 29,
  "chronos_mean": -0.0282,
  "chronos_median": -0.0154,
  "pct_lines_below_minus_0_5": 1.7,
  "leukaemia_screens": 57,
  "leukaemia_mean_gene_effect": -0.1374,
  "depmap_common_essential": "not asserted",
  "the_disagreement_that_matters": "this flat score coexists with an MLL-AF9 mouse survival result and a primary-blast colony phenotype. They are different measurements, and for a folate enzyme in folate-replete pooled culture the CRISPR score is the weaker instrument. This gene is the panel's counter-example to its own DepMap section.",
  "paralogue": "MTHFD2L mean -0.0721, 1.4% below -0.5, and it does not move in any line measured here (baseline 2.23 log2 CPM in RS4;11, at the floor)"
}
How this could bite us back
  • **MTHFD2's mouse survival result is the strongest functional evidence in this beat and it is in the wrong disease.** MLL-AF9 murine myeloid leukaemia plus human AML xenografts, with FLT3-ITD as the response biomarker. Nothing has been done in B-lineage KMT2A::AFF1, and this project's rule 73 exists because that transfer has repeatedly failed here.
  • **MTHFD2 knockdown induces DIFFERENTIATION, which is also what a menin inhibitor does.** Two agents pushing the same phenotypic lever are more likely redundant than synergistic, and differentiation syndrome is already a named toxicity of menin inhibitors. A combination rationale built on additive differentiation pressure should expect additive differentiation syndrome.
  • **The mitochondrial folate pathway is required in normal proliferating tissue and in development.** The paralogue MTHFD1, in the same panel, causes folate-sensitive neural tube defects and a combined immunodeficiency with megaloblastic anaemia when lost - immunodeficiency being a particularly unwelcome addition in a leukaemia patient. MTHFD2 itself has no curated human disease, which is not reassurance.
  • **Do not cite MTHFD2's near-zero compound margin as evidence that the serine arm is not menin-selective.** It is evidence that the serine arm is not uniform: PHGDH, CBS and GPT2 carry the selectivity and MTHFD2 and ASNS do not.
Still open
  1. Is MTHFD2's fall MYC-dependent in KMT2A-rearranged cells? PMID 27325891 puts MYC upstream and this graph treats MYC as a fusion-programme output. A MYC degron or a BET-inhibitor arm alongside the menin and DOT1L arms would test whether MTHFD2's compound-independence is MYC-mediated, and would test the ISR arm at the same time.
  2. Does MTHFD2 suppression work in a KMT2A::AFF1 B-lineage model? The survival result is in MLL-AF9 murine myeloid leukaemia and human AML xenografts. The index disease has never been tested, and rule 73 names exactly this transfer as this project's commonest synthesis failure.
  3. Why does MTHFD2 not fall in MOLM13, the human counterpart of the model where its suppression prolonged survival? MOLM13 carries FLT3-ITD, and FLT3-ITD is the published biomarker of response to MTHFD2 suppression - so the line most predicted to depend on MTHFD2 is the one where a menin inhibitor leaves it alone.
  4. Is there an MTHFD2 inhibitor worth using? The literature runs on knockdown, and ClinicalTrials.gov returns 0 studies for 'MTHFD2 inhibitor', so a 2016 in vivo result has had a decade with no clinical follow-through and the reason is not in the abstracts.
Every source behind this entry (4)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    On a pre-registered panel, MTHFD2 falls 1.16 log2 in RS4;11 at day 7 under VTP-50469 330 nM from a detected baseline of 8.65 log2 CPM and 1.13 log2 under EPZ5676 in the same samples - a VTP-minus-EPZ margin of -0.032, the smallest in the serine arm and inside the ISR-output arm's range of -0.085 to +0.095; its day-2 value is -0.276, 24% of the day-7 effect, matching the ISR-output median of 0.239 rather than the serine arm's 0.384; it is -0.292 in MOLM13 from a detected baseline of 8.09; it is -1.031 against SGC0946 -0.28 in MV4;11; and on the array both GPL570 probes fall in RS4;11 (-0.36 at baseline 12.38, -0.48 at 12.00) and SEM (-0.39, -0.45) and rise in REH (+0.41, +0.70).

    What this evidence cannot show

    One line per platform, single doses, NO dose-response, n=2-3 per arm, day 7 of continuous exposure so arrest is a competing explanation for any coordinated biosynthetic fall. No series records the amino-acid composition of the medium, and this is an amino-acid-response panel. MI-2-2 in the array arm is the WEAK compound, so the REH wild-type contrast has little dynamic range.

    Evidence that argues against this (1)
    • A near-zero margin can be produced by both compounds saturating the locus just as easily as by a shared upstream input, and MTHFD2's fall is only 1.16 log2 - close enough to the 1.0 inclusion threshold that a small dose difference could move it out of the analysis entirely. Its classification as ISR-like rests on a 0.03 difference between two estimates with n=2-3. this repository 2026
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    MTHFD2 DepMap gene effect is mean -0.0282, median -0.0154, with 1.7% of 1,258 screens below -0.5 and a leukaemia mean of -0.1374 over 57 screens, fetched with all 29 tissue blocks flattened and approvedSymbol read back as MTHFD2; the paralogue MTHFD2L is -0.0721 with 1.4% below -0.5.

    What this evidence cannot show

    Rule 83: the payload carries NO release identifier, so depmap_release is recorded as null and absolute values drift quarterly; the cross-gene contrast is the durable part. Leukaemia screens are matched on the payload's own diseaseFromSource string against four substrings and are NOT split by KMT2A status. Rule 84: no median-split proportion is reported - the percentage is against an absolute -0.5 threshold. A CRISPR knockout in folate- and serine-replete medium understates a biosynthetic dependency. This entry's own E4 is a direct counter-example to reading this number as dispensability, and that is said in the body rather than only here.

    Evidence that argues against this (1)
    • MTHFD2 suppression prolonged survival in an MLL-AF9 mouse leukaemia model and in human xenografts and impaired colony formation in primary AML blasts. A flat pooled-CRISPR proliferation score in folate-replete culture is a different measurement from an in vivo differentiation and survival phenotype, and for this gene the two disagree completely. Pikman Y et al. 2016
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProt reviewed human record P13995 (MTDC_HUMAN) has primary gene name MTHFD2 with the synonym NMDMC, length 350 aa, is annotated Mitochondrion with NO evidence code and no citation on that statement, and its only FUNCTION comment states that although its dehydrogenase activity is NAD-specific it can also use NADP at reduced efficiency; no disease is recorded.

    What this evidence cannot show

    Resolved 2026-09-22 with gene_exact:MTHFD2 AND organism_id:9606 AND reviewed:true; exactly one reviewed row, gene_primary and length read back. **The Mitochondrion statement is UNCODED** - no ECO evidence code, no citation - which by this project's own standard makes it a curator assertion rather than a measurement. It is recorded as uncoded rather than discarded, per rule 91's second clause. The FUNCTION comment is unusually thin and does not describe the reaction, so this entry's pathway role comes from the cited literature and not from UniProt.

    Evidence that argues against this (1)
    • An uncoded mitochondrial assignment on the gene this entry calls mitochondrial is a weaker localisation record than SHMT2's five ECO:0000269 statements or ALDH1L2's one, and this beat's mitochondrial-versus-cytosolic framing must not lean on it. The paralogue MTHFD2L (Q9H903) is annotated mitochondrial inner membrane only BY SIMILARITY (ECO:0000250) to a rat protein, which is weaker still. UniProt Consortium 2026
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    MTHFD2 emerged as the top candidate in an analysis seeking metabolic targets in AML and is the most differentially expressed metabolic enzyme in cancer versus normal cells; knockdown of MTHFD2 in AML cells decreased growth, induced differentiation and impaired colony formation in primary AML blasts; in human xenograft and MLL-AF9 mouse leukaemia models MTHFD2 suppression decreased leukaemia burden and prolonged survival; FLT3-ITD is a biomarker of response to MTHFD2 suppression based on primary patient AML data and functional genomic screening; and mechanistically MYC regulates the expression of MTHFD2 and MTHFD2 knockdown suppresses the TCA cycle.

    Targeting MTHFD2 in acute myeloid leukemia. · Pikman Y, Puissant A, Alexe G, et al., Stegmaier K · J Exp Med 2016
    PMID 27325891DOI
    What this evidence cannot show

    Abstract read this session; full text not retrieved, so survival magnitudes, the number of primary blasts and knockdown efficiency were not checked. **The model is MLL-AF9 (KMT2A::MLLT3) MURINE MYELOID leukaemia plus human AML xenografts - not the index B-lineage KMT2A::AFF1 disease**, and murine MLL-AF9 myeloid leukaemia is precisely the system this project's rule 73 names as its commonest over-generalisation. It is nonetheless the only functional KMT2A-fusion result on any gene in this beat's panel. It is a knockdown result, not a pharmacological one: ClinicalTrials.gov returns 0 studies for 'MTHFD2 inhibitor' (2026-09-22).

    Evidence that argues against this (1)
    • MTHFD2's transcript does NOT move in MOLM13, a human KMT2A::MLLT3 line and the human counterpart of the MLL-AF9 model this survival result comes from, falling only 0.29 log2 from a well-detected baseline while MEIS1 falls 2.96 in the same samples. So a menin inhibitor does not reproduce the intervention that prolonged survival in that model - which makes the mouse result an argument for MTHFD2 as an INDEPENDENT target and not an argument that menin inhibition works through MTHFD2. this repository 2026
Connected to