GFI1 — the repressor that LSD1 inhibitors actually work through, and the scan hit whose literature points the opposite way
also called GFI1 · Growth Factor Independence 1 · ZNF163 · Q99684
GFI1 is a switch-off protein that decides when a young blood cell stops renewing itself and becomes a mature one. It matters here for an unexpected reason: the leukemia drugs called LSD1 inhibitors seem to work not by blocking an enzyme but by prising LSD1 off GFI1 — which is the same kind of trick this project is trying to pull with menin.
The same thing, for a scientist
GFI1 (UniProt Q99684, 422 aa) is a SNAG-domain zinc-finger transcriptional repressor central to myeloid and lymphoid differentiation. It entered this graph through the coordinator's unbiased DepMap scan at rank 19 of 4,589 for KMT2A-rearranged selectivity (difference -0.312; -0.558 vs -0.247; pan-essentiality -0.021, so a gene almost no cell line needs) [E1]. The most important thing the literature says about GFI1 is **not** that it is a dependency. It is that GFI1's SNAG domain is the docking site for LSD1, that LSD1 inhibitors induce differentiation in AML with MLL translocations by disrupting that **protein-protein interaction** rather than by blocking demethylase activity, and that the complex requires HMG20B to stay on chromatin [E2][E3]. Separately, GFI1 dosage modulates MLL::AF9 leukemia: low GFI1 enhances oxidative phosphorylation via a FOXO1-MYC axis [E4], the GFI1-36N variant raises CDK4/CDK6 and proliferation in MLL-AF9 cells and makes them more palbociclib-sensitive [E5], and **low** GFI1 is linked to accelerated AML and inferior prognosis while enforced GFI1 expression impedes leukemic growth [E6]. That last point puts the literature in direct tension with the scan's knockout-based signal, and the entry says so rather than papering over it.
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.
I have never read one of these pictures before
A protein is one long chain of amino acids that folds up. The ribbon traces that chain from one end to the other; it is not a wire, it is a simplification of where the backbone runs. Drag to rotate, scroll to zoom.
Domains are the stretches of chain that do a particular job, and they are what the coloured blocks above the picture mark out. Click one and the view flies to it. In this project the important question about any domain is usually whether the fusion still has it — the labels say lost where it does not.
Anything in yellow sticks is a bound small molecule, which is usually the drug.
How this gene got into the graph, and why the scan signal is the weakest of the four
GFI1 was surfaced by a scan, not by reading. It ranked 19 of 4,589 in the coordinator's unbiased comparison of 11 KMT2A-rearranged against 43 other leukemia lines 1.
Of the four novel hits the scan produced, GFI1's numbers are the weakest: a difference of only -0.312, a mean gene effect of -0.558 in the rearranged lines — which is a modest depletion, not a strong dependency — and a pan-essentiality of -0.021, meaning essentially no cell line in DepMap needs GFI1 at all. It also sits at rank 19, immediately above RRM2, a ribosomal-class contaminant with the same difference to four decimal places.
So the scan's contribution here is a reason to look, and nothing more. What the looking found turned out to be more interesting than the scan.
The finding that matters: GFI1 is an interface, and someone has already drugged it1 deeper
LSD1 (KDM1A) inhibitors induce molecular and morphologic differentiation of blasts in AML patients harbouring MLL gene translocations 3. The mechanism is not what the drug class name implies.
LSD1 has 'a critical scaffolding function at genomic sites occupied by the SNAG domain transcription repressor GFI1', and the inhibitors 'block both enzymatic and scaffolding activities, in the latter case by disrupting the protein:protein interaction of GFI1 with LSD1' 3. The earlier paper from the same group titles the point directly: enhancer activation by pharmacologic displacement of LSD1 from GFI1 induces differentiation in AML 2.
Mass spectrometry then showed that LSD1 inhibition also disrupts the LSD1–HMG20B interaction; HMG20B co-locates genome-wide with GFI1 and LSD1, the strongest HMG20B binding coinciding with the strongest GFI1 and LSD1 binding; HMG20B depletion induces leukemia cell differentiation; and HMG20B is required for GFI1's repressor activity by stabilising LSD1 on chromatin at GFI1 sites, through its coiled-coil domain 3.
This is the menin argument, made by a different group about a different complex, and it has already reached patients. A transcriptional repressor with no ligandable pocket is rendered tractable by a small molecule that evicts its obligate partner. That is the precise logic on which this project rests, and finding an independent instance of it in the same disease is the most useful thing this beat produced.
What it does not prove
It does not prove the menin strategy will work, and it should not be used that way. LSD1 inhibitors are not curative in KMT2A-rearranged AML, the differentiation they induce has not translated into the kind of results menin inhibitors have produced, and the graph's existing target-lsd1-kdm1a-kmt2a-r entry holds the clinical record. The argument this supports is narrower and more useful: interface disruption of an undruggable repressor is a real mechanism that works in cells and in patients' blasts, not a theoretical hope.
GFI1 dosage in MLL::AF9 leukemia, and the direction problem
The Essen/Muenster group has built a dose-dependent picture of GFI1 in MLL::AF9 leukemia:
- Lower GFI1 expression enhances oxidative phosphorylation via upregulation of the FOXO1–MYC axis in haematopoietic progenitors and MLL::AF9-induced leukemic cells, suggesting the therapeutic move in GFI1-low leukemia is to target OXPHOS and glutamine metabolism 4.
- The germline GFI1-36N variant — prevalence about 7% in healthy Caucasians and 15% in AML patients — raises PRMT5, CDK4 and CDK6 protein in murine MLL-AF9 leukemic cells, speeds proliferation, and makes those cells more susceptible to palbociclib 5.
- Reduced GFI1 levels in AML patients and mouse models link to accelerated AML development and inferior prognosis, and enforced GFI1 expression inhibits growth and colony formation of leukemic lines, impedes expansion of pre-leukemic cells, and drives myeloid differentiation with increased apoptosis 6.
Here is the problem. The scan's signal is that deleting GFI1 hurts KMT2A-rearranged cells more than other leukemia cells. The literature's consistent direction is that less GFI1 makes myeloid leukemia worse, and more GFI1 kills it. Those are not compatible readings of the same gene without extra structure — a non-monotonic dose response, a distinction between complete loss and graded reduction, or a lineage difference between the B-lineage-heavy DepMap panel and the myeloid mouse models.
The honest conclusion is that the scan and the literature disagree about GFI1, and the literature is the stronger evidence. The entry exists because of the LSD1 interface, not because the scan found a dependency.
The link back to CCND3, which the scan did not know it was making
The GFI1-36N work reports that this variant raises CDK4 and CDK6 in MLL-AF9 leukemic cells and increases their susceptibility to palbociclib 5. The scan independently put CCND3 at rank 4 and GFI1 at rank 19 with no knowledge of either literature.
That is a small, real cross-link: a germline variant enriched in AML patients that pushes MLL-AF9 cells into exactly the CDK4/6-dependent state that the CCND3 hit and target-cdk4-6-inhibition-kmt2a-r describe. If CDK4/6 inhibition is ever revisited in this disease, GFI1-36N genotype is a biomarker hypothesis that already has pre-clinical support and would cost a genotyping assay to test.
| uniprot | Q99684 |
|---|---|
| gene symbol | GFI1 |
| length AA | 422 |
| alphafold ID | AF-Q99684-F1 |
| cytoband | 1p22.1 |
| accession resolution | Resolved by fetching https://rest.uniprot.org/uniprotkb/Q99684.json on 2026-09-18 and reading back the primary gene name GFI1 (synonym listed: ZNF163); UniProtKB ID GFI1_HUMAN; sequence length 422 aa, matching the value stored here. Not resolved by a gene_exact: search, which matches synonyms. |
| domains note | GFI1 has an N-terminal SNAG domain that is the LSD1 docking site and a C-terminal zinc-finger array, but no residue boundaries were resolved from a primary source in this cycle, so none are asserted. The graph's rule is that an unverified boundary is left out rather than guessed, because a domain with a wrong or null coordinate is invisible to every figure. |
| role | SNAG-domain zinc-finger transcriptional repressor governing myeloid and lymphoid differentiation and HSC self-renewal; the chromatin docking partner whose interaction with LSD1 is what LSD1 inhibitors disrupt |
| druggability | GFI1 itself is a zinc-finger repressor with no asserted pocket. The tractable feature is not GFI1 but the GFI1:LSD1 interface, which clinical-stage LSD1 inhibitors already disrupt. That makes GFI1 an interface partner rather than a target in its own right. |
| entered graph via | ana-depmap-kmt2a-selective-scan (cycle 25), rank 19 of 4,589; no prior entry in this graph |
| subnuclear localization | NUCLEAR MATRIX, IN PUNCTATE SUBNUCLEAR STRUCTURES SHARED WITH ETO/RUNX1T1. A portion of Gfi-1 and of its paralogue Gfi-1B attaches to the nuclear matrix, as ETO/MTG8 does, and Gfi-1 and ETO co-localise to punctate subnuclear structures; Gfi-1 also associates with HDAC1, HDAC2 and HDAC3 when co-expressed (PMID 12874834) [E_q89_gfi1_loc2]. A second, possibly distinct punctate pool: Gfi-1 colocalises with the STAT3 inhibitor PIAS3 in 'a characteristic nuclear dot structure' (PMID 11060035) [E_q89_gfi1_loc3] - PIAS proteins are classically PML-nuclear-body-associated, so this may be a PML body, and NOBODY HAS CO-STAINED PML TO FIND OUT. UniProt's uncoded note merges the two into one phrase, 'Colocalizes with PIAS3 and RUNX1T1 in nuclear dots', which is tidier than the evidence. |
| localization shown for fusion | NO for GFI1's own compartment in any KMT2A fusion, in any lineage. The GFI1 work in this disease is MLL-AF9 (KMT2A::MLLT3) murine dosage and metabolism work plus human LSD1-inhibitor differentiation work in AML with MLL translocations - function and interface, never position. THE ONE FUSION-ADJACENT LOCALISATION FACT IS FROM A DIFFERENT FAMILY: GFI1 co-localises in punctate subnuclear structures with ETO/RUNX1T1, the partner of RUNX1::RUNX1T1 [E_q89_gfi1_loc2]. That is an analogous-family observation (PI dimension 5), not a KMT2A one, and this entry labels it as such rather than letting it stand in for the missing KMT2A measurement. |
domains
[]
essentiality
{
"depmap_common_essential": false,
"chronos_mean_all_lines": -0.0213,
"chronos_mean_kmt2a_rearranged": -0.5583,
"chronos_mean_other_leukemia": -0.2468,
"kmt2a_selectivity_difference": -0.3116,
"scan_rank": 19,
"scan_n_genes": 4589,
"scan_percentile": 0.414,
"empirical_p_vs_scan": 0.005666,
"n_kmt2a_rearranged_lines": 11,
"n_other_leukemia_lines": 43,
"n_all_lines": 1208,
"direction_conflict": "The knockout signal (loss hurts KMT2A-rearranged cells) runs opposite to the published direction (low GFI1 accelerates AML and worsens prognosis; enforced GFI1 impedes leukemic growth). Unresolved."
}interfaces
{
"lsd1_kdm1a": "GFI1's SNAG domain is the docking site for LSD1; LSD1 inhibitors disrupt this protein-protein interaction as well as blocking catalysis, and the displacement activates enhancers and induces differentiation in AML with MLL translocations",
"hmg20b": "HMG20B stabilises LSD1 on chromatin at GFI1 binding sites through its coiled-coil domain and is required for GFI1 repressor activity; its depletion induces leukemia cell differentiation"
}germline variant
{
"variant": "GFI1-36N",
"prevalence_healthy_caucasian_pct": 7,
"prevalence_aml_pct": 15,
"effect_in_mll_af9": "raises PRMT5, CDK4 and CDK6 protein, accelerates proliferation, increases palbociclib susceptibility",
"source_pmid": "36003783"
}localization
{
"subcellular": "NUCLEUS, and unusually well evidenced: UniProt Q99684 fetched 2026-09-21 carries one location value, 'Nucleus', on FIVE separate ECO:0000269 attributions (PubMed:8754800, 11060035, 12874834, 18805794, 20547752) - more independent experimental support than most proteins in this tranche - plus the uncoded note 'Colocalizes with PIAS3 and RUNX1T1 in NUCLEAR DOTS' [E_q89_gfi1_loc1].",
"subnuclear": "NUCLEAR MATRIX AND PUNCTATE SUBNUCLEAR STRUCTURES, and the punctae are shared with a DIFFERENT FUSION FAMILY'S PARTNER. A portion of Gfi-1 and Gfi-1B associates with the NUCLEAR MATRIX, as ETO does, and Gfi-1 and ETO CO-LOCALISE TO PUNCTATE SUBNUCLEAR STRUCTURES [E_q89_gfi1_loc2]. Separately, Gfi-1 and PIAS3 colocalise in 'a characteristic nuclear dot structure' [E_q89_gfi1_loc3]. Whether the ETO dots and the PIAS3 dots are the same compartment has never been tested - nobody has co-stained all three.",
"WHY_THIS_IS_A_CROSS_FAMILY_FINDING": "ETO is MTG8 is RUNX1T1 - the 3' partner of RUNX1::RUNX1T1, the t(8;21) fusion. So GFI1's only named subnuclear compartment is one it shares with the partner protein of a different paediatric leukaemia fusion, and it is a compartment (nuclear matrix, punctate) that ETO is known for. This graph holds fusion-runx1-runx1t1 and localization-nuclear-matrix-mitotic-scaffold-kmt2a; this is a direct bridge between them and it arrived from a 2003 paper nobody in this project had read."
}cell type
{
"normal": "HAEMATOPOIETIC PROGENITORS AND BOTH MATURE LINEAGES. GFI1 is a SNAG-domain repressor central to myeloid and lymphoid differentiation, as this entry's summary already states. The PIAS3/STAT3 functional work was done in PRIMARY T CELLS, where Gfi-1 amplified IL-6-dependent T-cell activation [E_q89_gfi1_loc3]; the bounded query GFI1[tiab] AND (\"hematopoietic stem\" OR \"haematopoietic stem\" OR granulocyte OR neutrophil OR GMP)[tiab] returns 137 records on 2026-09-21, a large literature this beat counted and did not enumerate.",
"in_this_disease": "MURINE MLL::AF9 LEUKAEMIA and human AML with MLL translocations. This entry's existing E4 uses in-vitro and ex-vivo murine MLL::AF9 models for the GFI1-dosage metabolism work; E3 reports LSD1 inhibition inducing differentiation of blasts in AML patients harbouring MLL gene translocations. NAMING THE FUSION PRECISELY: MLL-AF9 is KMT2A::MLLT3, NOT the index fusion KMT2A::AFF1, and no GFI1 work exists in KMT2A::AFF1 or in B-lineage infant disease.",
"where_the_localisation_was_measured": "Not stated in the abstracts and not haematopoietic-specific. The nuclear-matrix and ETO co-localisation work used mammalian cells with co-expressed constructs; the PIAS3 dots came from a yeast two-hybrid lead confirmed by coprecipitation in eukaryotic cells."
}cell state
{
"DIFFERENTIATION": "The state GFI1 governs. GFI1 decides when a progenitor stops self-renewing and matures; see cell-state-differentiation-block, the state this disease is defined by, and note the direction - LSD1 inhibitors INDUCE differentiation in MLL-translocated AML by prising LSD1 off GFI1's SNAG domain (this entry's E2, E3), which is a protein-protein-interface manoeuvre and therefore a localisation-adjacent one: it changes what GFI1 is bound to, not how much of it there is.",
"DOSAGE": "A distinct state axis this entry already records: the germline GFI1-36N variant raises protein levels and is enriched in AML patients (E5), while LOW GFI1 protein accelerates AML development (E6). Both directions are pathogenic, which is unusual and is why the entry flags a direction problem.",
"bounded": "No GFI1 cell-state localisation measurement exists - nobody has asked whether GFI1's nuclear-matrix fraction or its punctate pool changes with differentiation state, with LSD1 inhibition, or in a fusion-bearing cell."
}localization bounded absence
{
"date": "2026-09-21",
"tool": "NCBI esearch db=pubmed",
"query": "GFI1[tiab] AND PIAS3[tiab]",
"querytranslation": "\"GFI1\"[Title/Abstract] AND \"PIAS3\"[Title/Abstract]",
"count": 2,
"ids": [
"26951200",
"11060035"
],
"reading": "The 'nuclear dots' half of UniProt's note rests on a two-record literature, one of which is the founding paper. The ETO half is separate and has its own source (PMID 12874834). No study has co-stained GFI1 with PML, SC35 or a nuclear-matrix marker in a haematopoietic cell, and none has imaged GFI1 in a KMT2A-rearranged cell at all.",
"absence_class": "never-measured"
}- GFI1 is required for normal myeloid and lymphoid differentiation and for HSC self-renewal. Anything that suppressed it would be attacking a core haematopoietic regulator in a patient whose haematopoiesis is already failing.
- The direction of benefit is genuinely unclear. One literature says raise GFI1; the scan implies lowering it is damaging to the leukemia. An intervention built on the wrong direction would be worse than useless, and `risk-wrong-strike-is-invisible` is the entry that describes what that failure looks like.
- The GFI1:LSD1 interface needs a third protein, HMG20B, to hold together on chromatin. Two-body models of interfaces have already misled this project once; the menin:KMT2A interface may have its own HMG20B that nobody has looked for.
- LSD1 inhibitors demonstrate that this interface can be drugged and have not cured anybody with KMT2A-rearranged leukemia. That an interface is druggable is a much weaker statement than that drugging it works, and the same gap applies to menin.
- Which is right about direction — the scan's knockout signal or the literature's dose findings? The cheapest discriminating experiment is a titration: graded GFI1 knockdown across a KMT2A::AFF1 and a KMT2A::MLLT3 line, reading proliferation and differentiation, to see whether the response is monotonic.
- Is the GFI1 dependency signal B-lineage or myeloid? Every published functional GFI1 experiment cited here is myeloid, and the scan's rearranged panel is mixed. If the signal is myeloid-only it is not about infant B-lineage disease at all.
- Does the GFI1:LSD1 interface have a measurable Kd and a co-structure that could be compared with the menin:MBM1 interface? A side-by-side comparison of two interfaces that both yielded clinical molecules would be a genuinely useful addition to ana-comparative-interface-druggability-scan.
- Is GFI1-36N genotype associated with response to CDK4/6 inhibition in any human dataset? The pre-clinical claim is specific enough to test retrospectively in any trial that banked germline DNA.
- Does HMG20B belong in this graph as its own entry? It is required for the repressor complex and its depletion phenocopies the drug, which makes it a third-party dependency of the kind this project's interface work should care about.
- Are GFI1's ETO punctae and GFI1's PIAS3 dots the same structure, and is either a PML nuclear body? Two co-stains would connect a cross-fusion-family observation to a compartment this graph already tracks; neither has been done in twenty-three years.
- LSD1 inhibitors work in MLL-translocated AML by displacing LSD1 from GFI1's SNAG domain. Does that displacement change GFI1's subnuclear distribution - does GFI1 leave the nuclear matrix, or the punctae - or only its partner list? This is imageable with existing reagents in an existing model.
- GFI1 shares a subnuclear compartment with ETO/RUNX1T1, a different fusion family's partner. Is the nuclear matrix a shared address for repressive fusion machinery across families, and does KMT2A::AFF1 reach it?
- E1weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.dataset
In this project's unbiased scan of 4,589 protein-coding genes, GFI1 ranks 19 by KMT2A-rearranged selectivity (difference -0.312; mean DepMap Chronos -0.558 across 11 KMT2A-rearranged leukemia lines versus -0.247 across 43 other leukemia lines) with a pan-essentiality of -0.021 across 1,208 lines, and had no entry in this graph before this cycle.
DepMap CRISPR (Chronos) gene effect via the Open Targets Platform GraphQL API, 4,589 HGNC protein-coding genes split by Cellosaurus KMT2A gene-fusion annotation; analyses/depmap-kmt2a-selective-scan/ · Broad Institute DepMap; Cellosaurus (SIB); HGNC; Open Targets Platform; analysis run by this project · Databases and derived analysis 2026What this evidence cannot show
Marked weak deliberately. This is the smallest effect of the scan's four novel hits: a mean gene effect of -0.558 is a modest depletion rather than a dependency, and the gene ranks immediately above RRM2, one of the scan's own acknowledged pan-essential contaminants, with an identical difference to four decimal places. Our own unreviewed computation; 11 lines, none infant-derived; 24% of the protein-coding genome; no lineage conditioning. The published literature points the other way on direction, which is recorded in the sections rather than hidden here.
Evidence that argues against this (1)
- Reduced GFI1 levels are linked to accelerated AML development and inferior prognosis, and enforced GFI1 expression inhibits leukemic growth and drives myeloid differentiation — the opposite direction from a knockout-based dependency signal. Hones JM et al. 2017
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Pharmacologic displacement of LSD1 from GFI1 activates enhancers and induces differentiation in acute myeloid leukemia.
Enhancer Activation by Pharmacologic Displacement of LSD1 from GFI1 Induces Differentiation in Acute Myeloid Leukemia · Maiques-Diaz A, Spencer GJ, Lynch JT, Ciceri F, Williams EL, Amaral FMR, Wiseman DH, Harris WJ, Li Y, Sahoo S, Hitchin JR, Mould DP, Fairweather EE, Waszkowycz B, Jordan AM, Smith DL, Somervaille TCP · Cell Reports 22:3641-3659 2018What this evidence cannot show
Read as title and abstract-level summary only; the full text was not retrieved in this cycle. AML generally rather than KMT2A-rearranged specifically in this title, though the companion paper from the same group frames the clinical differentiation response as occurring in patients with MLL translocations [E3]. The mechanism claim — that the therapeutic effect runs through an interaction rather than through demethylase catalysis — is the load-bearing part and should be confirmed from the full text before being used in any grant or design argument.
Evidence that argues against this (1)
- LSD1 inhibitors have not produced durable responses in KMT2A-rearranged leukemia despite inducing differentiation, so demonstrating the interface mechanism is not the same as demonstrating that interface disruption is sufficient therapy. The graph's own target-lsd1-kdm1a-kmt2a-r entry holds the clinical record. Maiques-Diaz A et al. 2022
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Pharmacologic inhibition of LSD1 induces molecular and morphologic differentiation of blast cells in AML patients harbouring MLL gene translocations; LSD1 has a critical scaffolding function at genomic sites occupied by the SNAG-domain transcription repressor GFI1, and inhibitors block both enzymatic and scaffolding activities, in the latter case by disrupting the GFI1-LSD1 protein-protein interaction; LSD1 inhibition also disrupts the HMG20B-LSD1 interaction; HMG20B co-locates with GFI1 and LSD1 genome-wide, its depletion induces leukemia cell differentiation, and it is required for GFI1 repressor activity by stabilising LSD1 on chromatin at GFI1 binding sites via its coiled-coil domain.
HMG20B stabilizes association of LSD1 with GFI1 on chromatin to confer transcription repression and leukemia cell differentiation block · Maiques-Diaz A, Nicosia L, Basma NJ, Romero-Camarero I, Camera F, Spencer GJ, Amaral FMR, Simeoni F, Wingelhofer B, Williamson AJK, Pierce A, Whetton AD, Somervaille TCP · Oncogene 41:4841-4854 2022What this evidence cannot show
Read as abstract only. The statement about differentiation in patients with MLL translocations is the paper's framing sentence, citing the field rather than reporting new clinical data here. Cell lines, chromatin profiling and mass spectrometry; no in vivo or clinical experiment in this paper. GFI1 is the repressor whose partner is displaced, not the drug target itself, so this is evidence about an interface rather than about GFI1 as a dependency.
Evidence that argues against this (1)
- That the GFI1:LSD1 complex needs a third protein, HMG20B, to stay on chromatin is a warning as much as a finding: interfaces that look like a clean two-body problem often are not, and the same may be true of the menin:KMT2A interface this project treats as the handle. Maiques-Diaz A et al. 2018
- E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Dose-dependent GFI1 expression regulates metabolism in haematopoietic progenitor and leukaemic cells: using in vitro and ex vivo murine models of MLL::AF9-induced AML and extracellular flux assays, lower GFI1 expression enhances the oxidative phosphorylation rate via upregulation of the FOXO1-MYC axis, suggesting therapeutic exploitation of GFI1-low leukemia by targeting oxidative phosphorylation and glutamine metabolism.
Dose-dependent expression of GFI1 alters metabolism in the haematopoietic progenitors and MLL::AF9-induced leukaemic cells · Patnana PK, Liu L, Frank D, Nimmagadda SC, Behrens M, Ahmed H, Xie X, Liebmann M, Wei L, Gerdemann A, Thivakaran A, Humpf HU, Klotz L, Dugas M, Varghese J, Trajkovic-Arsic M, Siveke JT, Hanenberg H, Opalka B, Duhrsen U, Reinhardt HC, Guenther U, von Bubnoff N, Khandanpour C · British Journal of Haematology 202:1033-1048 2023What this evidence cannot show
Read as abstract only. Murine MLL::AF9 models in vitro and ex vivo; no human KMT2A-rearranged patient data and no in vivo therapeutic test. The claim is about GFI1 level modulating metabolism, not about GFI1 being required, and it should not be cited as a dependency. Note the therapeutic direction implied is to treat GFI1-LOW leukemia, which again runs opposite to a knockout-based dependency reading.
Evidence that argues against this (1)
- A metabolic shift measured by extracellular flux in murine models is several steps from a clinical vulnerability, and OXPHOS targeting in AML has repeatedly looked promising in models without translating. Lin KH et al. 2019
- E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
The germline GFI1-36N single-nucleotide polymorphism, with a prevalence of about 7% in healthy Caucasians and 15% in AML patients, raises protein levels of PRMT5, CDK4 and CDK6 in murine MLL-AF9 leukemic cells, accelerating their proliferation, and those cells are more susceptible to the CDK4/6 inhibitor palbociclib than GFI1-36S cells.
Presence of the GFI1-36N single nucleotide polymorphism enhances the response of MLL-AF9 leukemic cells to CDK4/6 inhibition · Vorwerk J, Sun K, Frank D, Neumann F, Huve J, Budde PM, Liu L, Xie X, Patnana PK, Ahmed HMM, Opalka B, Lenz G, Jayavelu AK, Khandanpour C · Frontiers in Oncology 12:903691 2022What this evidence cannot show
Read as abstract only, and the PubMed abstract renders with gene and variant names stripped by lost italic markup, so the reading above was reconstructed against the full title as NCBI esummary returns it ('Presence of the GFI1-36N single nucleotide polymorphism enhances the response of MLL-AF9 leukemic cells to CDK4/6 inhibition'). Murine MLL-AF9 cells only; proteomics and immunoblot rather than a clinical correlation; no patient genotype-response data. The prevalence figures are quoted from the abstract and were not independently checked against a population database.
Evidence that argues against this (1)
- A biomarker built on a germline variant that raises CDK4/6 and speeds proliferation has to contend with the finding that CCND3 upregulation is itself associated with palbociclib resistance, so 'more cyclin-CDK' does not reliably mean 'more inhibitor sensitivity'. Ketzer F et al. 2022
- E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Analysis of AML patients and AML mouse models with reduced GFI1 expression shows a direct link between low GFI1 protein level and accelerated AML development and inferior prognosis; upregulating GFI1 in leukemic cell lines inhibits growth and colony formation, impedes expansion of murine pre-leukemic cells, and in a humanised AML model drives myeloid differentiation morphologically and immunophenotypically with increased apoptosis and fewer cKit-positive cells.
Enforced GFI1 expression impedes human and murine leukemic cell growth · Hones JM, Thivakaran A, Botezatu L, Patnana P, Castro SVDC, Al-Matary YS, Schutte J, Fischer KBI, Vassen L, Gorgens A, Duhrsen U, Giebel B, Khandanpour C · Scientific Reports 7:15720 2017What this evidence cannot show
Read as abstract only. AML broadly rather than KMT2A-rearranged specifically. This is the clearest statement of the direction problem: the therapeutic suggestion is to RAISE GFI1 in GFI1-low leukemia, which is the opposite of what a CRISPR-knockout dependency signal would recommend. Both cannot be straightforwardly true, and this entry does not claim to resolve it.
Evidence that argues against this (1)
- Against reading the direction as settled: GFI1 works through an LSD1-scaffolded repressor complex whose disruption also induces differentiation. A repressor whose loss and whose enforced expression both push toward differentiation is behaving non-monotonically, and the resolution may be that what matters is the complex, not the GFI1 level. Maiques-Diaz A et al. 2022
- E_q89_gfi1_loc1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
UniProt Q99684 (GFI1_HUMAN, primary gene GFI1, synonym ZNF163, 422 aa, read back 2026-09-21) carries one subcellular location value, 'Nucleus', supported by five independent ECO:0000269 attributions (PubMed:8754800, 11060035, 12874834, 18805794, 20547752), with the uncoded note 'Colocalizes with PIAS3 and RUNX1T1 in nuclear dots'. No cytoplasmic value, no tissue-specificity comment.
UniProtKB Q99684 (GFI1_HUMAN) SUBCELLULAR LOCATION comment and note, fetched 2026-09-21 · UniProt Consortium · UniProtKB 2026What this evidence cannot show
Statement about the record. The subnuclear content sits entirely in the UNCODED note, so a filter keeping only ECO:0000269 rows would return 'Nucleus' and discard both punctate observations - the same discard pattern rule 91 documents for KMT2A's fragment comments. Five attributions is more than most, but attribution COUNT is not independence: this beat did not check whether the five are from independent groups.
Evidence that argues against this (1)
- UniProt's note compresses two separate experiments into one phrase and the compression is misleading in a way that matters. 'Colocalizes with PIAS3 and RUNX1T1 in nuclear dots' reads as one compartment containing three proteins. It is two papers, seven years and two systems apart: PIAS3 co-localisation from a STAT3-signalling study, ETO/RUNX1T1 co-localisation from a nuclear-matrix and HDAC repression study. PIAS proteins are SUMO ligases classically associated with PML nuclear bodies; ETO is a nuclear-matrix co-repressor. Those are not obviously the same structure, and reading the note as though they were would create a compartment that has never been shown to exist. Nobody has co-stained GFI1, PIAS3, RUNX1T1 and PML together. McGhee L et al. 2003
- E_q89_gfi1_loc2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
GFI1 HAS A NUCLEAR-MATRIX SUBNUCLEAR RECORD, AND IT IS SHARED WITH THE t(8;21) PARTNER. Gfi-1 interacts with ETO (also termed MTG8, the gene RUNX1T1, first identified through its involvement in the t(8;21) translocation of acute myelogenous leukaemia) and related proteins both in vitro and in vivo, and with histone deacetylase proteins in vivo. A PORTION OF Gfi-1 AND Gfi-1B ASSOCIATED WITH THE NUCLEAR MATRIX, as is the case with ETO, and Gfi-1 AND ETO CO-LOCALISE TO PUNCTATE SUBNUCLEAR STRUCTURES. When co-expressed in mammalian cells Gfi-1 associates with HDAC-1, HDAC-2 and HDAC-3.
Gfi-1 attaches to the nuclear matrix, associates with ETO (MTG8) and histone deacetylase proteins, and represses transcription using a TSA-sensitive mechanism. · McGhee L, Bryan J, Elliott L, Grimes HL, Kazanjian A, Davis JN, Meyers S · J Cell Biochem 2003What this evidence cannot show
Co-expression in mammalian cells rather than endogenous protein, and the abstract does not name the cell line, so the system is unstated (rule 73) and this claim does not name one. Not haematopoietic in the imaging experiments, no KMT2A fusion. 'Nuclear matrix' is an operationally defined fraction - what survives extraction - and its status as a structure rather than an artefact of extraction has been contested for decades. Graded moderate on both counts. Abstract read in full; full text not read.
Evidence that argues against this (1)
- The nuclear matrix is the weakest link in this record and pretending otherwise would be dishonest. The fraction is defined by resistance to high-salt extraction and nuclease digestion, and a long-standing objection holds that the residual network is created by the extraction rather than revealed by it - so 'attaches to the nuclear matrix' may mean no more than 'is tightly bound to something insoluble', which for a DNA-binding repressor bound to chromatin is close to tautological. What survives that objection is the CO-LOCALISATION, which is an imaging result independent of fractionation: Gfi-1 and ETO in the same punctate structures is a positional statement that no extraction artefact produces. So this entry rests the subnuclear claim on the punctae and treats the matrix association as corroborating rather than load-bearing. This graph's own localization-nuclear-matrix-mitotic-scaffold-kmt2a entry carries the same exposure and should be read with the same caution. Rödel B et al. 2000
- E_q89_gfi1_loc3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Gfi-1 and PIAS3 CO-LOCALISE IN A CHARACTERISTIC NUCLEAR DOT STRUCTURE. The interaction was found in a yeast two-hybrid screen and confirmed by coprecipitation from eukaryotic cells; functionally, PIAS3 inhibits STAT3-mediated transcription, Gfi-1 overcomes the PIAS3 block and significantly enhances STAT3-mediated activation, and in PRIMARY T CELLS Gfi-1 amplified IL-6-dependent T-cell activation.
The zinc finger protein Gfi-1 can enhance STAT3 signaling by interacting with the STAT3 inhibitor PIAS3. · Rödel B, Tavassoli K, Karsunky H, Schmidt T, Bachmann M, Schaper F, Heinrich P, Shuai K, Elsässer HP, Möröy T · EMBO J 2000What this evidence cannot show
The co-localisation is in transfected eukaryotic cells; the primary-T-cell experiment is the FUNCTIONAL arm, not the imaging arm, so the dots have not been seen in a primary haematopoietic cell. No marker was co-stained to identify the dots. Abstract read in full, full text not read. 2000, and this beat found only two records in the whole GFI1-PIAS3 literature.
Evidence that argues against this (1)
- A 'characteristic nuclear dot structure' containing a PIAS protein has an obvious candidate identity that the paper never tests - PML nuclear bodies, where PIAS1 and PIAS3 are canonical residents and where SUMO-dependent sequestration of transcription factors is the standard mechanism. If the dots are PML bodies then GFI1's subnuclear record joins a compartment this graph already tracks (localization-pml-nuclear-bodies-pmlrara), and the STAT3 result acquires a mechanism: sequestration rather than titration. If they are not, the observation stands alone and means little. Either way the entry should not assert PML identity, because the co-stain was never done - and the cost of that omission is that a twenty-six-year-old observation still cannot be connected to anything. UniProt Consortium 2026
KDM1A is held by the GFI1 SNAG domain - a fusion-independent recruitment route that could explain the MLL-AF9-locus dependency without any fusion contact
A portion of GFI1 and GFI1B attaches to the NUCLEAR MATRIX as ETO does, and GFI1 and ETO co-localise to punctate subnuclear structures (PMID 12874834) - GFI1's only named subnuclear compartment (Q89 2026-09-21)
- related to →RUNX1::RUNX1T1 (AML1-ETO), t(8;21)(q22;q22)
CROSS-FAMILY (PI dimension 5): GFI1's subnuclear partner ETO/MTG8 is RUNX1T1, the 3' partner of t(8;21); GFI1's only compartment is one it shares with a different fusion family's partner protein (Q89 2026-09-21)
- related to →The differentiation block — what it actually is, operationally, in KMT2A::AFF1 leukaemia
GFI1 governs the progenitor-to-mature transition, and LSD1 inhibitors release the block in MLL-translocated AML by displacing LSD1 from GFI1's SNAG domain - an interface manoeuvre, never imaged (Q89 2026-09-21)
rank 19 of 4,589 — the weakest of the four novel hits (difference -0.312, pan-essentiality -0.021)
GFI1's N-terminal SNAG domain is the LSD1 docking site
GFI1 repressor activity, scaffolded by LSD1 and stabilised by HMG20B, confers the leukemia cell differentiation block
- related to →CDK4/6 inhibition in KMT2A-rearranged leukemia — the class was already tried, and the trial never reported
GFI1-36N raises CDK4/CDK6 in MLL-AF9 cells and increases palbociclib susceptibility — a biomarker hypothesis linking two of the scan's hits
- ← related toKMT2A::MLLT3 (MLL-AF9), t(9;11)(p21.3;q23.3)
GFI1 dosage modulates metabolism and proliferation in murine MLL::AF9 leukemia models