FOXC1 - a mesenchymal forkhead factor derepressed in HOXA-high AML that blocks monocytic differentiation, with a real mechanism and zero published connection to KMT2A rearrangement

This is a prediction, not a measurement. No one has seen this protein in this shape. The colours say how much the software trusts itself, and that is worth reading before anything else — the orange stretches are places the model is effectively saying this part has no fixed shape.

• There is no FOXC1 inhibitor and no obvious pocket - it is a winged-helix DNA-binding transcription factor whose oncogenic function is to STABILISE a repressor complex, so the intervention would have to break a FOXC1-RUNX1 protein-protein interface rather than block an activity.
• The functional dependency of the interface runs through RUNX1, HDAC1 and TLE3 - all of which are required in normal haematopoiesis, and RUNX1 loss is itself leukaemogenic. Degrading the complex is not obviously safer than leaving it.
• FOXC1 knockdown releases enhancers and induces differentiation; in a differentiation-inducing strategy that is the mechanism of differentiation syndrome, which is already the dose-limiting toxicity of the menin inhibitors in this disease.
• The gene runs normal mesenchymal, ocular and skeletal development, and FOXC1 haploinsufficiency causes Axenfeld-Rieger syndrome with glaucoma. An agent that lowers FOXC1 function systemically has a named congenital phenotype to answer for.
• FOXC1 shows no CRISPR fitness effect in any of 1,258 lines - including U-937, which carries the same fusion as the line the mechanism was worked out in. There is no potency read-out to optimise against and the one available assay says nothing happens.

1. Is FOXC1 expressed in KMT2A-rearranged AML or B-ALL at all? Nobody has published the measurement, and it is a qPCR panel across the existing KMT2A-rearranged line collection - the cheapest experiment on this beat and the one that decides whether FOXC1 belongs in this graph as biology or only as a comparator.
2. Does menin inhibition collapse FOXC1 the way it collapses HOXA9/MEIS1? If FOXC1 sits on the HOXA-high programme rather than on the fusion, a menin inhibitor should take it down in any HOXA-high cell, and that would make FOXC1 a pharmacodynamic marker shared across NPM1c, NUP98, CALM-AF10 and KMT2A-rearranged disease.
3. Which PBX paralogue does FOXC1 meet in a myeloid blast, and does it sequester FOXC1 in heterochromatin there? The melanoma result [E6] and the HOXA9 collaboration [E1] point in opposite directions through the same cofactor family, and no leukaemia experiment has ever tested it.
4. Does the RUNX1-to-CEBPA enhancer swap [E2] occur in KMT2A-rearranged cells on differentiation? This graph already carries mechanism-wdr5-maintains-the-cebpa-p30-differentiation-block, so a CEBPA-swap mechanism is live here; whether FOXC1 is the factor holding it shut in KMT2A-rearranged disease is untested.
5. Is the DEK::NUP214 CUT&RUN at the FOXC1 promoter reproduced by any other XPO1-dependent lesion? hyp-xpo1-is-a-shared-address-for-hoxa-driven-leukaemia predicts it should be, and FOXC1 is the one non-HOX locus in that experiment, which makes it the sharpest test of the shared-address idea.

1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   FOXC1 is expressed in at least 20% of human AML cases but not in normal haematopoietic populations; its expression in AML is almost exclusively associated with expression of the HOXA/B locus; FOXC1 contributes to a block in monocyte/macrophage differentiation and enhances clonogenic potential; in vivo FOXC1 collaborates with HOXA9 to significantly accelerate the onset of symptomatic leukaemia; and FOXC1-high human AML cases show reduced morphological monocytic differentiation and inferior survival.

   Frequent Derepression of the Mesenchymal Transcription Factor Gene FOXC1 in Acute Myeloid Leukemia · Somerville TD, Wiseman DH, Spencer GJ, Huang X, Lynch JT, Leong HS, Williams EL, Cheesman E, et al. · Cancer Cell 28:329-342 2015
   PMID 26373280DOI
   ▸What this evidence cannot show

   Abstract fetched from PubMed efetch XML on 2026-09-20 and read in full; the body was not retrieved this cycle, so the AML subtypes carrying FOXC1 expression, the 20% denominator's cohort, and the in vivo model's genotype are NOT asserted here. The in vivo collaboration is with HOXA9 in a murine model; the human arm is expression and outcome association. The 20% figure is 'in silico and other analyses' by the authors' own description and is a cohort-dependent number. Nothing in this abstract names KMT2A, MLL or any fusion, so this paper does not connect FOXC1 to the index disease - the HOXA/B association does that only by inference, and the inference is made here and not by the paper.

   ▸Evidence that argues against this (1)

   • FOXC1 expression being 'almost exclusively associated with HOXA/B locus expression' is a correlation across a heterogeneous AML cohort, and HOXA/B-high AML has at least four distinct genetic entrances (NPM1c, NUP98 fusions, CALM-AF10, KMT2A rearrangement). An association with the state is not an association with any one of its causes, and this graph's own DepMap lineage split finds no FOXC1 dependency in KMT2A-rearranged lines of either lineage. Broad Institute DepMap; Open Targets Platform; query run by this project 2026-09-20 2026
2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   In the human PICALM::MLLT10 (CALM-AF10) acute monocytic leukaemia line Fujioka and in primary AML blasts, FOXC1 - highly expressed in about 30% of HOXA-high AML - interacts with RUNX1 through the Forkhead and Runt domains respectively, co-occupies primed and active enhancers close to differentiation genes, and stabilises association of RUNX1, HDAC1 and the Groucho repressor TLE3 to limit enhancer activity; FOXC1 knockdown causes loss of those repressors, gain of CEBPA binding, enhancer acetylation, upregulation of nearby genes including KLF2, genome-wide redistribution of RUNX1/TLE3/HDAC1 from enhancers to promoters, and repression of the self-renewal genes MYC and MYB, with morphological differentiation and CD86 induction.

   Enhancer recruitment of transcription repressors RUNX1 and TLE3 by mis-expressed FOXC1 blocks differentiation in acute myeloid leukemia · Simeoni F, Romero-Camarero I, Camera F, Amaral FMR, Sinclair OJ, Papachristou EK, Spencer GJ, Lie-A-Ling M, et al. · Cell Reports 36:109725 2021
   PMID 34551306DOI
   ▸What this evidence cannot show

   Abstract and full text fetched from Europe PMC (PMC8480281, fullTextXML, 113,135 characters after tag-stripping) on 2026-09-20 and read for the experimental system. RULE 73: the functional experiments are in Fujioka, a PICALM::MLLT10 line that additionally carries NRAS, ETV6, TP53 and EZH2 mutations by the authors' own table, plus one primary normal-karyotype AML (BB475) for ChIP. MV4-11 (KMT2A::AFF1) appears in the paper's cell-line resource table but is not the subject of the knockdown, ChIP-seq or RIME arms - so NOTHING in this paper is a KMT2A-rearranged result. The peak numbers quoted (18,745 FOXC1 peaks in Fujioka, 39,941 in BB475) are read from the text and are MACS2 calls under the authors' stated thresholds, not an absolute count of binding sites.

   ▸Evidence that argues against this (1)

   • Fujioka carries TP53 and EZH2 mutations alongside PICALM::MLLT10, so the enhancer-repressor architecture FOXC1 is shown to hold together in that line may be a property of that genotype. The authors themselves anchor the general claim by citing prior work rather than by reproducing the mechanism in a second line, and the primary-AML arm is ChIP occupancy rather than knockdown. Somerville TD et al. 2015
3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   In t(6;9)/DEK::NUP214 AML, a multi-omics comparison of 57 AML primary samples showed a selective response to the XPO1 inhibitors selinexor and eltanexor and a distinct transcriptomic signature characterised by overexpression of FOXC1 and HOX genes; separately, CUT&RUN in t(6;9) cells demonstrated direct binding of DEK::NUP214 to the promoters of FOXC1 and the HOXA/B clusters, and both the expression of these genes and the fusion's binding at their regulatory regions were selectively reduced upon XPO1 inhibition.

   DEK::NUP214 acts as an XPO1-dependent transcriptional activator of essential leukemia genes · Kaya F, Bewicke-Copley F, Miettinen JJ, Casado P, Leddy E, Deniz Ö, Lavallée VP, Philippe C, et al. · Leukemia 39:1526-1531 2025
   PMID 40204893DOI
   ▸What this evidence cannot show

   Abstract fetched from PubMed efetch XML on 2026-09-20 and read in full; the body (PMC12133580) was not retrieved this cycle, so the identity of the t(6;9) cells used for CUT&RUN and the number of FOXC1 peaks are NOT asserted. The 57-sample cohort and the CUT&RUN are SEPARATE experiments on separate material, and the dispatch brief for this beat conflated them - stated here because the conflated version is a stronger claim than the paper supports. DEK::NUP214 is not a KMT2A fusion; this is a second, independent HOXA-high entrance that happens also to switch FOXC1 on, which is the reason FOXC1 is in this graph at all.

   ▸Evidence that argues against this (1)

   • A transcriptomic signature and a promoter-binding event do not establish that FOXC1 is functionally required in t(6;9) AML; the paper's functional read-out is drug response to XPO1 inhibition, which collapses the whole signature at once and cannot attribute the effect to FOXC1 rather than to the HOXA/B genes bound in the same experiment. FOXC1 is not a CRISPR dependency in any line. Broad Institute DepMap; Open Targets Platform; query run by this project 2026-09-20 2026
4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   Fetched live on 2026-09-20, DepMap Chronos gene effect for FOXC1 has mean +0.0391 across 1,258 cell lines with 1.27% of lines at or below -0.5; among leukaemia lines the KMT2A-rearranged myeloid half sits at z +0.25 and the B-lineage half at z -0.24 relative to 44 other leukaemia lines, with OCI-AML3 (NPM1c, HOXA-high, KMT2A-wild-type) at z -1.25. FOXC1 is not a fitness dependency in any leukaemia arm tested. MEN1 through the identical query shape returned z -2.41 / -2.43.

   DepMap CRISPR (Chronos) gene effect for FOXC1 (ENSG00000054598) and MEN1 (ENSG00000133895), retrieved via the Open Targets Platform GraphQL API, with the lineage split of analyses/lineage-confound-myeloid-tf-dependencies re-run for this gene · Broad Institute DepMap; Open Targets Platform; query run by this project 2026-09-20 · Database query 2026
   ▸What this evidence cannot show

   A bounded negative with a date, a query route and a positive control run through the identical query shape (rule 67). Seven myeloid and four B-lineage KMT2A-rearranged lines. THE BOUND THAT MATTERS HERE: a CRISPR fitness screen measures 21-day proliferation, and FOXC1's published function is a DIFFERENTIATION BLOCK with reduced clonogenicity on knockdown - a phenotype that a proliferation screen in an already-transformed line can miss entirely. This number bounds FOXC1 as a fitness dependency; it does not bound FOXC1 as a differentiation-block maintainer, and the two are different questions. Ensembl id from HGNC primary (HGNC:3800); approvedSymbol read back as FOXC1.

   ▸Evidence that argues against this (1)

   • FOXC1 knockdown in Fujioka cells produces differentiation, reduced clonogenicity and a reduced S/G2/M fraction - a real functional effect that this CRISPR fitness number does not see, in a line that is not in the KMT2A-rearranged arm. Read together, the two say FOXC1 matters where it is highly expressed and not where it is not, which is an expression-gated dependency and exactly the pattern a pan-line fitness mean averages away. Simeoni F et al. 2021
5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.review

   Counted absence, measured 2026-09-20 with PubMed E-utilities esearch: FOXC1[tiab] returns 977 records and (MLL OR KMT2A)[tiab] returns 9,228, while FOXC1[tiab] AND (MLL OR KMT2A)[tiab] returns exactly 1 - PMID 15492844, 'Human FOX gene family (Review)', Int J Oncol 2004 - which title-checks to zero on-topic records. FOXC1[tiab] AND ('MLL-AF4' OR 'KMT2A::AFF1' OR 't(4;11)')[tiab] returns 0. No paper has named FOXC1 and KMT2A/MLL together in a title or abstract in an on-topic way.

   Human FOX gene family (Review) · Katoh M, Katoh M · International Journal of Oncology 25:1495-1500 2004
   PMID 15492844
   ▸What this evidence cannot show

   BOUNDED TO TITLES AND ABSTRACTS; supporting figures and supplementary tables were NOT searched (rule 80). Both operand counts are given so the zero has a denominator. NOTE ON THE CITATION SHAPE: the citation record here is the FOX-family review that the query returned, with its title, authors, year and venue copied field-by-field from efetch; the query text sits in citation.query_note. It was first written with the query text in the title and authors fields and a PMID beside it, which conflates 'a citation of paper X' with 'a query that returned paper X' and which verify-citations.py --strict correctly flagged. The review is cited as the single returned record and NOT as support for anything it says. The alias list searched on the FOXC1 side was the current symbol only - FKHL7, IRID1, FREAC3, ARA, IGDA and IHG1 were NOT run against the KMT2A clause, and those carry the pre-2000 ophthalmology literature. On the KMT2A side both current and retired symbols were covered by the (MLL OR KMT2A) clause. The likeliest hiding place for a FOXC1-in-KMT2A measurement is a supplementary expression table in an AML subtype paper, and that was not opened.

   ▸Evidence that argues against this (1)

   • An absence in titles and abstracts is not an absence in the literature. FOXC1 is 'almost exclusively associated with HOXA/B locus expression' in AML, and KMT2A-rearranged AML is HOXA-high, so a FOXC1-high KMT2A-rearranged case almost certainly exists inside published expression data - it has simply never been the subject of a paper. The correct reading of this zero is that nobody has ASKED the question, not that the answer is no. Somerville TD et al. 2015
6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   In A7 melanoma cells with elevated nuclear filamin A, FOXC1 is unable to activate transcription and is partitioned into an HP1alpha, heterochromatin-rich region of the nucleus, and this inhibition is mediated through an interaction between FOXC1 and the homeodomain protein PBX1a; UniProt additionally records FOXC1 colocalising with PITX2 isoform 3 at subnuclear chromatin regions.

   FOXC1 transcriptional regulatory activity is impaired by PBX1 in a filamin A-mediated manner · Berry FB, O'Neill MA, Coca-Prados M, Walter MA · Molecular and Cellular Biology 25:1415-1424 2005
   PMID 15684392DOI
   ▸What this evidence cannot show

   Abstract fetched from PubMed efetch XML on 2026-09-20 and read in full; body not retrieved. A7 MELANOMA CELLS and human non-pigmented ciliary epithelial cells - no haematopoietic system anywhere in this work, and the disease context is Axenfeld-Rieger syndrome, not leukaemia. The PITX2 colocalisation is taken from the UniProt Q12948 subcellular-location note (PubMed:16449236 cited there) and that paper was NOT read this cycle, so it is recorded as a database statement and not as a read result. Whether any of this happens in a myeloid blast is entirely unknown.

   ▸Evidence that argues against this (1)

   • The leukaemia literature runs the other way: FOXC1 collaborates with HOXA9, and HOXA9 works through PBX-family TALE cofactors, so a PBX interaction that INACTIVATES FOXC1 by sequestering it in heterochromatin sits awkwardly beside a PBX-dependent HOXA9 partner that COOPERATES with it. Either the melanoma mechanism does not operate in myeloid cells, or the PBX paralogue differs - this graph's own symbol-collisions registry records that PBX3 and not PBX1 is the rate-limiting cofactor in the HOXA9 axis. Unresolved, and recorded as unresolved. Somerville TD et al. 2015

• ← regulatesDEK::NUP214 is the sixth lesion on the XPO1 address - and the only one where the drug response was measured in primary patient cells

  DEK::NUP214 binds the FOXC1 promoter by CUT&RUN; binding and expression fall under XPO1 inhibition

• co-occurs with →HOXA9 — required for survival of established MLL-rearranged leukemia, and dispensable for starting it

  FOXC1 collaborates with HOXA9 in vivo to accelerate leukaemia onset; FOXC1 expression in AML is almost exclusive to HOXA/B-expressing cases

• binds →RUNX1 — activated by KMT2A::AFF1 and then bound by the reciprocal fusion, in the one published two-oncoprotein circuit in t(4;11)

  Forkhead domain to Runt domain; FOXC1 stabilises RUNX1/HDAC1/TLE3 at enhancers

• causes →The differentiation block — what it actually is, operationally, in KMT2A::AFF1 leukaemia

  maintains the monocyte/macrophage differentiation block; knockdown releases it

• expressed in →PICALM::MLLT10 (CALM-AF10), t(10;11)(p13;q14) - the DOT1L-dependent HOXA fusion that contains no KMT2A

  highest FOXC1 expression in the tested AML line panel was Fujioka/P31-FUJ, a PICALM::MLLT10 line, where the whole mechanism was established

• part of →HOXA-high versus HOXA-low KMT2A-rearranged leukemia — the split that breaks the one-mechanism story

  FOXC1 derepression is essentially confined to HOXA/B-high AML

• supports →Chromatin-bound XPO1 as a shared, druggable address for HOXA-driven leukaemia - and the objection that it may be a general chromatin cofactor rather than an address at all

  FOXC1 is the one non-HOX locus in the DEK::NUP214 CUT&RUN and is lost under XPO1 inhibition with the HOXA/B clusters

• inhibits →CEBPA — ranks 25th on the KMT2A scan and sits at the median of random genes

  FOXC1 presence excludes CEBPA from enhancers; knockdown gains CEBPA binding and acetylation

• related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)

  COUNTED ABSENCE: FOXC1 and MLL/KMT2A title-check to zero joint records against operands of 977 and 9,228

• related to →PBX3 - the TALE cofactor the field says is non-redundant, and the experiment that says HOXA9 is the redundant one

  unresolved tension - a PBX1a interaction sequesters FOXC1 in HP1alpha heterochromatin in melanoma while FOXC1 cooperates with PBX-dependent HOXA9 in AML

• related to →Selinexor (XPO1 inhibition) — a real paediatric response rate attached to an unproven mechanism in KMT2A-rearranged disease

  FOXC1 expression falls under selinexor/eltanexor in t(6;9) cells, making it a candidate pharmacodynamic read-out rather than a resistance route