Oncofusion

start / graph / gene-bmi1

Genestrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q56-homeodomain

BMI1 - the self-renewal gene that KMT2A fusions can do without, because HOXA9 covers for it at the same locus

also called PCGF4 · RNF51 · Polycomb complex protein BMI-1 · Polycomb group RING finger protein 4

BMI1 keeps blood stem cells able to renew themselves, and it does it largely by silencing a tumour-suppressor gene called CDKN2A that would otherwise stop cells dividing. In KMT2A-fusion leukemia the fusion's own output, HOXA9, silences the same gene - so the fusion can transform cells even when BMI1 is completely absent, which is not true of other leukemia fusions.

The same thing, for a scientist

BMI1 is the PCGF4 subunit of canonical PRC1 and the founding polycomb regulator of haematopoietic and leukaemic stem-cell self-renewal [E1]. Its main effector in this context is repression of the *CDKN2A/CDKN2B* locus, which encodes p16INK4a and p14ARF and is the endogenous brake on CDK4 and CDK6 - so BMI1, *CDKN2A* and CDK6 are one circuit rather than three genes (`mechanism-bmi1-cdkn2a-cdk6-axis`). **The KMT2A-specific result is a dissociation, and it is the reason this entry exists:** AML1-ETO and PLZF-RARα trigger senescence in *Bmi1*-null cells, while **MLL-AF9, which drives *Hoxa7* and *Hoxa9*, readily transforms *Bmi1*-null cells** - because Hoxa9 regulates the same *p16Ink4a/p19Arf* locus and can substitute [E2]. A second group reached a partly opposing conclusion in the same year and the same fusion: *Bmi1* was dispensable for in vitro immortalisation of MLL-AF9 GMPs but **indispensable for leukaemia in vivo**, and deleting *Ink4a/Arf* only partially rescued [E3]. Both are murine MLL-AF9. Live DepMap adds a hard bound on the in vitro half: BMI1 knockout has a mean Chronos effect of **+0.0585** across 1,258 lines and **-0.0103** across 57 leukaemia lines, with 1.4% of lines below -0.5 [E4].

What it looks like
loading AF-P35226-F1…
AlphaFold AF-P35226-F1 ↗
colourshape

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.

very high (90+)confident (70–90)low (50–70)very low (<50)
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.

Go deeper
The dissociation, which is a real argument about whether BMI1 is a target here

Smith et al. asked whether leukaemia-associated fusions that differ in their ability to activate Hox could initiate leukaemia without Bmi1 2. The answer split cleanly by Hox status:

fusionactivates Hox?transforms Bmi1-null cells?
AML1-ETOnono - triggers senescence
PLZF-RARαnono - triggers senescence
MLL-AF9yes (Hoxa7, Hoxa9)yes, readily

The mechanism is that Hoxa9 regulates the p16Ink4a/p19Arf locus as Bmi1 does, and Hoxa9 could overcome the senescence AML1-ETO induced. MLL-AF9 could not initiate leukaemia in Bmi1(-/-)Hoxa9(-/-) mice, which have further compromised HSC function, but either gene alone restored it in the double-null background - which is as direct a statement of functional substitution as this literature contains.

Consequence for this project. A BMI1-directed intervention is, in KMT2A-rearranged disease specifically, aimed at a node the fusion has a documented bypass for. The bypass is HOXA9, which is the fusion's own headline output. That is a strong prior against BMI1 as a single target here, and it comes from a paper whose index fusion is MLL-AF9.

It also makes a prediction nobody has tested. If HOXA9 and BMI1 are substitutable at CDKN2A, then a menin inhibitor - which collapses the HOXA programme - should make KMT2A-rearranged cells newly dependent on BMI1. That is a synthetic-lethal proposal with an explicit mechanism, and it is written up in hyp-menin-inhibition-creates-a-bmi1-dependency.

The paper that partly disagrees, and how to hold both

Yuan et al., also 2011, also murine MLL-AF9 3:

  • Bmi1-null GMPs were immortalised in vitro by MLL-AF9 - consistent with 2 and with the DepMap zero 4.
  • But Bmi1-null cells showed enhanced differentiation, retained fewer LSCs, and Bmi1 was indispensable for the development of leukaemia in vivo.
  • Deleting both Ink4a and Arf only partially restored leukaemogenic capacity, so the requirement is not wholly the CDKN2A axis.
  • Re-expressing Bmi1 in immortalised Bmi1(-/-)Ink4a-Arf(-/-) GMPs failed to restore most deregulated genes or in vivo activity - the authors conclude Bmi1 is needed for the faithful reprogramming of progenitors into LSCs, a one-way step rather than a maintained state.

The two papers are not straightforwardly contradictory and the honest reading is this: BMI1 is dispensable for MLL-AF9 transformation and proliferation in vitro (both papers, plus DepMap), and its requirement appears at the in vivo LSC-establishment step, where 2 finds Hoxa9 can substitute and 3 finds it cannot fully. The difference may be the cell of origin (GMP versus unfractionated), the Ink4a/Arf genotype, or the readout.

What is common to both, and is the load-bearing statement: whatever BMI1 does for a KMT2A fusion, it is not a proliferation requirement of the transformed cell.

The dependency bound, and why it is not the refutation it looks like

Fetched live on 2026-09-20 4: BMI1's mean Chronos gene effect is +0.0585 across 1,258 lines - a positive number, meaning knockout is on average very slightly favourable - and -0.0103 across 57 leukaemia lines. 1.4% of lines fall at or below -0.5 and 0.24% at or below -1.0. This graph's own genome-wide scan ranks BMI1 5,450 of 17,728 for KMT2A-rearranged selectivity.

For most genes that would close the question. Here it does not, for three reasons that are all in the published record rather than invented to save the hypothesis:

  1. Both primary papers already report the in vitro null. 2 says MLL-AF9 readily transformed Bmi1-null cells; 3 says Bmi1 was not essential to the immortalization of GMPs in vitro. DepMap is agreeing with the literature, not contradicting it.
  2. Self-renewal is the phenotype and a 21-day fitness screen cannot see it. Every positive BMI1 result in leukaemia is a transplantation or LSC-frequency result.
  3. Paralogue structure. BMI1 is PCGF4, one of six PCGF proteins; PCGF1 (in PRC1.1) is separately central to this disease and this graph carries protein-prc1-1-complex and resistance-prc1-1-loss-menin-inhibitor-nup98 for it. A single-PCGF knockout is buffered.

But the bound is still worth having, and it is the entry's most actionable number: it says that any BMI1 experiment in this disease must have an in vivo or serial-replating readout, because an in vitro proliferation assay has a 1.4%-of-lines prior of showing anything.

Where the normal biology sets the ceiling

Lessard and Sauvageau established the constraint that has never gone away 1: Bmi-1 regulates the proliferative activity of both normal and leukaemic stem and progenitor cells, and Bmi-1-null leukaemic cells arrest, differentiate and fail to transplant - but the same paper's normal arm shows the identical requirement in normal stem and progenitor cells.

So BMI1's therapeutic index in leukaemia is set by the normal HSC compartment, in an infant, and the founding paper measured both sides of it in one experiment. This is the classic shape of a self-renewal target and it is why the class has been hard: the selectivity has to come from somewhere other than the gene.

In KMT2A-rearranged disease specifically, 2 says it does not come from the fusion either, because the fusion has a bypass.

Facts and figures
uniprotP35226
gene symbolBMI1
length AA326
alphafold IDAF-P35226-F1
ensemblENSG00000168283
cytoband10p12.2
protein classPCGF4 subunit of canonical Polycomb Repressive Complex 1; RING-finger protein, non-catalytic partner of the RING1/RNF2 E3 ligases
resolution notegene_exact:BMI1 AND organism_id:9606 AND reviewed:true returned ONE reviewed row, primary gene name BMI1, entry BMI1_HUMAN, 326 aa, synonyms PCGF4 and RNF51, fetched from the UniProt REST API on 2026-09-20.
identifier trapBMI1 (P35226) and CDK6 (Q00534) are BOTH EXACTLY 326 aa, and both are written on this beat, on the same axis, one repressing the inhibitor of the other. verify-accessions.py asserts accession plus length_aa and therefore CANNOT catch a BMI1-for-CDK6 swap. A second trap sits alongside: BMI1's synonym RNF51 is one character from RNF2, which is a DIFFERENT polycomb RING protein already in this graph as gene-rnf2 and already logged in data/symbol-collisions.json for listing BAP1 as a synonym. Three PRC1-adjacent RING symbols - RNF1 (RING1), RNF2 (RING1B), RNF51 (BMI1) - differ by one or two characters and name three different proteins.
domains noteNo domain boundaries asserted; the RING and helix-turn-helix-like regions were not read back from UniProt with residue coordinates this cycle.
kmt2a specific verdictBYPASSED. MLL-AF9 transforms Bmi1-null cells readily because Hoxa7/Hoxa9 regulate the same p16Ink4a/p19Arf locus, whereas non-Hox-activating fusions cannot. BMI1 is therefore a WEAKER single target in KMT2A-rearranged disease than in leukaemia generally - the opposite of what the self-renewal literature would suggest if read without the fusion comparison.
subnuclear localizationpericentromeric heterochromatin, retained through mitosis (PMID 9722603), with recruitment requiring EED/EZH2 and assembly requiring DNMT1 (PMID 16314526). The 'polycomb body' framing should be used carefully: CLEM shows BMI1 immunogold density inside a fluorescent PcG body is the SAME as in heterochromatin outside it, so the focus is a local accumulation of heterochromatin rather than a BMI1-concentrating compartment (PMID 21818415).
shown for fusionNO. Measured 2026-09-21 with NCBI esearch: (BMI1[tiab] OR Bmi-1[tiab] OR PCGF4[tiab]) AND (MLL[tiab] OR KMT2A[tiab] OR MLL-AF4[tiab] OR MLL-AF9[tiab] OR mixed lineage leukemia[tiab]) AND (localiz*[tiab] OR localis*[tiab] OR subcellular[tiab] OR nuclear body[tiab] OR nuclear bodies[tiab] OR speckle*[tiab] OR condensat*[tiab] OR nucleolar[tiab] OR immunofluorescence[tiab] OR compartment*[tiab]) returns 2 records, both title-checked off-topic (PMID 15254754, an in silico characterisation of ARHGAP23; PMID 22678079, a liposarcoma cell line with 12q amplification). Operands: gene clause 3,926, gene AND fusion 21 - so the BMI1-and-KMT2A literature is real but makes no spatial claim. Positive control through the identical shape with (MLLT1[tiab] OR ENL[tiab]) returns 23. Zero in titles and abstracts; supporting figures not searched (brief rule 80).
epigenetic readoutH2AK119ub - AND BMI1 DOES NOT CATALYSE IT. This is the distinction the entry needed and did not have: BMI1 is PCGF4, the non-catalytic RING-finger partner of canonical PRC1, and the E3 ubiquitin-ligase activity for histone H2A lysine 119 belongs to RING1/RNF2 (RING1B). The founding biochemistry purified hPRC1L, an H2A-SPECIFIC E3 ubiquitin ligase complex composed of several polycomb-group proteins including Ring1, Ring2, BMI1 and HPH, and showed that reducing the expression of these proteins reduces H2A ubiquitination in vivo - so BMI1 is a REQUIRED SUBUNIT of the writer and not the writer [E_q97_epi1]. Independently, Ring1b is identified as the H2A-K119 ubiquitin E3 ligase and H2A ubiquitination is placed on the inactive X as an epigenetic marker of X inactivation, which fixes which subunit carries the catalysis [E_q97_epi2]. IN THE RIGHT LINEAGE, AND THIS IS THE BEST-PLACED MEASUREMENT AVAILABLE: in B-cell lymphoblastic leukaemia, nuclear Vav3 interacts with Bmi1, Ring1b and Ezh2, and the GEF activity of Vav3 is REQUIRED for Bmi1-dependent B-cell progenitor self-renewal, for the protein interaction with Bmi1, for mono-ubiquitination of H2A(K119) and for repression of PRC1 downstream target loci in leukaemic B-cell progenitors; mechanistically Vav3 prevents an Akt-dependent phosphorylation of Bmi1 at S314 [E_q97_epi3]. That is H2AK119ub, Bmi1-dependent, in B-ALL - the correct lineage for this graph - but the driver is BCR-ABL, not a KMT2A fusion. WHAT IS STILL MISSING, AND IT IS THE WHOLE KMT2A ARM: no source read for this entry measures H2AK119ub at CDKN2A, at HOXA, or anywhere else, in a KMT2A-rearranged cell. The entry's own argument is that BMI1 is BYPASSED in KMT2A-rearranged disease (DepMap Chronos mean +0.0585, 1.4% of lines at or below -0.5), and an H2AK119ub measurement is precisely the experiment that would say whether the bypass is at the level of the mark or downstream of it. SUBUNIT REDUNDANCY IS THE CONFOUND: BMI1 is one of six PCGF paralogues, and this graph treats PRC1.1 (PCGF1) as the central polycomb arm in KMT2A-rearranged disease while gene-usp7 and protein-prc1-1-complex carry the H2AK119ub readout for that flavour. A BMI1 knockout that leaves H2AK119ub intact would be evidence of paralogue compensation, not of the mark being irrelevant - which is rule 64's warning applied to the writer side.
domains
[]
essentiality
{
  "source": "DepMap CRISPR (Chronos) via Open Targets GraphQL, retrieved live 2026-09-20",
  "n_all_lines": 1258,
  "chronos_mean_all_lines": 0.0585,
  "chronos_median_all_lines": 0.0738,
  "chronos_mean_leukaemia_lines": -0.0103,
  "n_leukaemia_lines": 57,
  "frac_lines_at_or_below_minus_0_5": 0.0143,
  "frac_lines_at_or_below_minus_1_0": 0.0024,
  "kmt2a_r_selectivity_rank": 5450,
  "kmt2a_r_selectivity_of": 17728,
  "kmt2a_r_difference": -0.0293,
  "interpretation": "no fitness requirement in vitro, consistent with both primary papers; the published requirement is an in vivo LSC-establishment requirement that this assay cannot see"
}
axis partners
[
  "gene-cdkn2a",
  "gene-cdk6",
  "gene-hoxa9",
  "gene-rnf2",
  "protein-prc1-1-complex"
]
localization
{
  "subcellular": "Nucleus, with an experimentally annotated cytoplasmic pool: UniProt P35226 lists Nucleus and Cytoplasm, both at ECO:0000269. No leukaemia localisation exists.",
  "subnuclear": "PERICENTROMERIC HETEROCHROMATIN, clustered into what are called PcG bodies - and the body may be an artefact of fluorescence. BMI1 with RING1 and hPc2 forms discrete foci, the larger ones near centromeres, tightly associated with 1q12 and related pericentromeric sequences, and stably retained through mitosis. Recruitment to those foci is DYNAMIC, needs EED and EZH2, and proper assembly needs DNMT1. But correlative light-electron microscopy finds no enrichment of BMI1 inside a fluorescent PcG body relative to heterochromatin outside it, so the honest statement is that BMI1 is on heterochromatin generally and the foci mark where heterochromatin is locally dense.",
  "how_shown": "immunofluorescence against RING1, BMI1 and hPc2 with a kinetochore marker and FISH to 1q12, through mitosis (PMID 9722603); photobleaching kinetics plus EED, EZH2 and DNMT1 knockdown (PMID 16314526); correlative light-electron microscopy with high-pressure freezing, cryosubstitution and on-section immunogold in BMI1-GFP U-2 OS cells (PMID 21818415).",
  "systems": "human fibroblasts, U-2 OS osteosarcoma and other non-haematopoietic lines; no leukaemia and no KMT2A-rearranged system has ever been used to localise BMI1",
  "node": "localization-chromocentres-pericentromeric-kmt2a-r"
}
cell type
{
  "expressed_in": "Not resolved to a haematopoietic cell type in this entry.",
  "required_in": "NORMAL and leukaemic stem cells in mouse, and - corrected 2026-09-21 - in MLL-AF9 leukaemia too, IN VIVO. Bmi-1 regulates the proliferative activity of normal and leukaemic stem cells [E1]. In MLL-AF9-transformed Bmi1-null murine GMPs, Bmi1 is NOT needed for immortalisation in vitro [E3] but IS INDISPENSABLE for the development of leukaemia in vivo, with fewer LSCs retained and enhanced differentiation in its absence [E_q66_correction1]. The in vitro result is the one this entry had been carrying; the in vivo result is from the same abstract and points the other way.",
  "not_required_in": "MLL-AF9-transformed murine GMPs IN VITRO ONLY - they immortalise without Bmi1 [E3], but they do not make leukaemia in a mouse [E_q66_correction1]. The independent instrument that looked like corroboration is also an in vitro one: DepMap gives BMI1 a POSITIVE mean gene effect of +0.0585 across 1,208 lines [E4], and a genome-scale screen in established cell lines cannot see a requirement that acts at transformation and shows up on transplantation. Two in vitro readouts agreeing is not two independent lines of evidence.",
  "how_shown": "germline Bmi1 knockout with transformation, serial replating, differentiation and senescence readouts in murine myeloid progenitors, PLUS transplantation assays and Ink4a/Arf epistasis (PMIDs 12714970, 21624810, 21527932); genome-scale CRISPR fitness in established lines (DepMap Chronos via Open Targets)",
  "nodes": [
    "cell-type-hsc-bone-marrow",
    "cell-type-gmp"
  ]
}
cell state
{
  "maintains_state": "SELF-RENEWAL, in normal and leukaemic stem cells, partly through silencing of the CDKN2A/Ink4a-Arf locus [E1] - and the KMT2A-fusion exception this entry's title asserts is narrower than the title says. Bmi1-null MLL-AF9 cells show ENHANCED DIFFERENTIATION, retain fewer LSCs and cannot make leukaemia on transplantation, and deleting Ink4a and Arf together only PARTIALLY rescues them [E_q66_correction1]. Re-expressing Bmi1 afterwards does not restore leukaemogenicity either, so the requirement is for faithful reprogramming at transformation rather than for ongoing BMI1 activity.",
  "regulates_state": "Senescence entry in Bmi1-null cells, which MLL fusions escape and non-Hox-activating fusions do not [E2].",
  "state_dependent_localisation": "BMI1's own subnuclear position is state-dependent in a way nobody has followed up: recruitment to polycomb bodies requires EED and EZH2 and assembly requires DNMT1 [E_q66_loc2]. A cell under a dual EZH1/2 inhibitor - the agent this graph's PRC2 arm is built around - is a cell in which BMI1's address has been perturbed, and that has never been measured.",
  "how_shown": "transformation, senescence and CDKN2A readouts in Bmi1-null murine progenitors (PMIDs 12714970, 21624810, 21527932); photobleaching and knockdown of EED, EZH2 and DNMT1 (PMID 16314526)",
  "nodes": [
    "cell-state-stem-like-self-renewal"
  ],
  "title_overstates": "Recorded rather than silently rewritten, following this project's practice that the title is what gets quoted. This entry is titled 'the self-renewal gene that KMT2A fusions can do without'. That is true of immortalisation in vitro and FALSE of leukaemia development in vivo in the same paper (PMID 21527932). The coordinator should decide whether to retitle; the body now carries both halves."
}
How this could bite us back
  • The fusion already has a bypass. MLL-AF9 transforms Bmi1-null cells because Hoxa7/Hoxa9 cover the same locus [E2], so a BMI1-directed agent in KMT2A-rearranged disease is aiming at a node with a documented, fusion-encoded workaround.
  • Normal HSCs need BMI1 as much as leukaemic ones do, and the founding paper measured both in one experiment [E1]. In an infant, the therapeutic index has to come from somewhere other than the gene, and nothing read this cycle says where.
  • No in vitro readout. BMI1 knockout is fitness-neutral in 98.6% of human cancer lines [E4]. Any BMI1 programme has to be optimised against a transplantation endpoint, which is slow, expensive and poorly suited to a medicinal-chemistry cycle.
  • BMI1 is non-catalytic. It is the PCGF4 partner of the RING1/RNF2 E3 ligases, not an enzyme, so a small molecule must disrupt a protein-protein interface or degrade the protein - and degrading a PRC1 subunit removes it from every target locus, not only CDKN2A.
  • Derepressing CDKN2A is the intended effect and is also a senescence programme. Senescent leukaemic cells are alive and secretory, and the senescence-associated secretory phenotype is documented to support neighbouring tumour cells - so the mechanism could produce a reservoir rather than a clearance.
  • If BMI1 repression of CDKN2A is what a BMI1 agent relieves, it does nothing in the ~8.5% of infant patients whose CDKN2A locus is deleted outright [E5], and those are plausibly the higher-risk patients.
  • The compartment argument for targeting BMI1 is weaker than the literature's vocabulary suggests. Correlative light-electron microscopy finds BMI1 immunogold density inside a fluorescent 'PcG body' no higher than in heterochromatin outside it (PMID 21818415), so a strategy premised on disrupting a BMI1-containing nuclear body would be premised on a structure that may be an optical artefact of locally dense heterochromatin.
  • Cross-talk with the PRC2 arm, in a direction that confounds combination studies: BMI1's recruitment to polycomb bodies REQUIRES EED and EZH2 (PMID 16314526). A dual EZH1/2 inhibitor therefore moves BMI1 as a side effect, so a BMI1-inhibitor plus EZH-inhibitor combination is not two independent perturbations and any additivity analysis has to account for it.
  • BMI1's pericentromeric complexes are stably retained THROUGH MITOSIS (PMID 9722603). A reversible inhibitor given on an intermittent schedule would have to displace a mark that survives cell division, which is a pharmacokinetic requirement no BMI1 programme has stated.
  • The dispensability result is the decisive one and it is already in this entry: MLL-AF9-transformed murine GMPs immortalise without Bmi1, and DepMap gives BMI1 a positive mean gene effect. Delivering a BMI1 inhibitor to a KMT2A-rearranged blast would be delivering it to the one context where the gene has been shown not to be needed.
  • BMI1 is a subunit of the H2AK119ub writer, not the catalytic centre, so a BMI1-directed agent removes a stimulatory/structural component of an E3 whose catalysis lives in RING1/RNF2 (PMIDs 15386022, 15509584). The pharmacologically cleaner target for the MARK is the ligase, and this graph already holds gene-rnf2 - an intervention aimed at BMI1 to move H2AK119ub is aimed one step off the enzyme.
  • H2AK119ub is a genome-wide repressive mark, so any agent that lowers it de-represses everything polycomb holds down, not only CDKN2A. In B-lymphoid leukaemia the same mark is required for repression of negative regulators of proliferation AND for repression of oncogenic transcription factors (PMID 35650206), i.e. losing it releases both arms at once.
Still open
  1. Does menin inhibition create a BMI1 dependency? If HOXA9 and BMI1 are substitutable at CDKN2A [E2], collapsing the HOXA programme should unmask a BMI1 requirement. This is a clean synthetic-lethal prediction with a named mechanism and it has not been tested; see hyp-menin-inhibition-creates-a-bmi1-dependency.
  2. Is the Hoxa9-for-Bmi1 substitution true in B-lineage KMT2A::AFF1? Both 2011 papers are murine MLL-AF9 myeloid. Infant B-ALL is the index disease and the substitution argument has never been made in it.
  3. Which PCGF does the work in KMT2A-rearranged cells - PCGF4 (BMI1) in canonical PRC1, or PCGF1 in PRC1.1? This graph treats PRC1.1 as central and has never asked whether the two arms are redundant with each other, which would explain the BMI1 DepMap zero without invoking self-renewal at all.
  4. Does BMI1 loss change CDKN2A expression in a KMT2A-rearranged cell where CDKN2A is genomically intact - about 91% of infant patients [E5]? The axis is asserted throughout this literature and the measurement in this disease was not found this cycle.
  5. Does BMI1 occupy pericentromeric heterochromatin in a KMT2A-rearranged blast, and does HOXA9 - which this entry says covers for BMI1 at CDKN2A - occupy the same regions? If the bypass is locus-specific rather than compartment-wide, that is testable by one ChIP pair in an index line.
  6. Is the CLEM result general, or a property of U-2 OS? Every claim in this graph that invokes polycomb bodies rests on fluorescence, and one electron-microscopy study in a leukaemia line would settle whether the compartment exists in the disease at all.
  7. Does a dual EZH1/2 inhibitor displace BMI1 from heterochromatin in leukaemia cells? If it does, part of the phenotype attributed to PRC2 inhibition is a PRC1 relocation.
  8. Which is right for the index disease - the in vitro or the in vivo half of PMID 21527932? In MLL-AF9 they disagree, and this entry's title follows the in vitro half. No Bmi1-null transplantation experiment has been done with a KMT2A::AFF1 B-lineage fusion at all.
  9. Bmi1 re-expression after immortalisation fails to restore leukaemogenicity or most deregulated genes (PMID 21527932). If BMI1 is needed only at the moment of reprogramming, a BMI1 inhibitor given to established disease is aimed at the wrong window - is there any evidence of an ongoing requirement in an established leukaemia?
  10. Is H2AK119ub altered at any locus in a KMT2A-rearranged cell when BMI1 is removed? The entry argues BMI1 is bypassed in this disease and no source read for it measures the mark in this background, so the bypass could sit above or below the mark and nobody knows which.
  11. Does PCGF1/PRC1.1 supply the H2AK119ub that BMI1/PCGF4 does not in KMT2A-rearranged cells? The two PRC1 flavours are written separately in this graph and their marks have never been measured in the same experiment.
  12. Is Bmi1 S314 phosphorylation state different in KMT2A-rearranged versus BCR-ABL B-ALL? PMID 35650206 makes that residue the switch for Bmi1-dependent H2AK119ub in B-lymphoid leukaemia and nobody has looked at it in the index disease.
Every source behind this entry (11)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    In mice, the Polycomb group gene Bmi-1 has a key role in regulating the proliferative activity of normal stem and progenitor cells, and the proliferative potential of leukaemic stem and progenitor cells lacking Bmi-1 is compromised: they undergo proliferation arrest, show signs of differentiation and apoptosis, and fail to transplant the leukaemia; complementation with Bmi-1 completely rescues these defects.

    Bmi-1 determines the proliferative capacity of normal and leukaemic stem cells · Lessard J, Sauvageau G · Nature 423:255-260 2003
    PMID 12714970DOI
    What this evidence cannot show

    Abstract read this cycle. MURINE, and the leukaemia is Hoxa9+Meis1a-induced rather than a KMT2A fusion - which matters more than usual here, because the whole question this entry turns on is whether Hoxa9 can substitute for Bmi1. A leukaemia driven BY Hoxa9 is not a neutral test of that. The normal and leukaemic requirements are measured in the same paper and are of similar magnitude, which is the therapeutic-index problem stated at its source.

    Evidence that argues against this (1)
    • A requirement demonstrated in a Hoxa9/Meis1a-driven murine leukaemia does not transfer automatically to a KMT2A fusion, and the paper that tested fusions directly found MLL-AF9 transforms Bmi1-null cells readily. Smith LL et al. 2011
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In murine cells, AML1-ETO and PLZF-RARalpha, which do not activate Hox, trigger senescence in Bmi1(-/-) cells, whereas MLL-AF9, which drives expression of Hoxa7 and Hoxa9, readily transforms Bmi1(-/-) cells; MLL-AF9 could not initiate leukaemia in Bmi1(-/-)Hoxa9(-/-) mice, but either gene alone restored its ability to establish leukaemic stem cells in the double-null background; and Hoxa9, like Bmi1, regulates expression of the p16(Ink4a)/p19(ARF) locus and can overcome senescence induced by AML1-ETO.

    Functional crosstalk between Bmi1 and MLL/Hoxa9 axis in establishment of normal hematopoietic and leukemic stem cells · Smith LL, Yeung J, Zeisig BB, Popov N, Huijbers I, Barnes J, Wilson AJ, Taskesen E, Delwel R, Gil J, Van Lohuizen M, So CW · Cell Stem Cell 8:649-662 2011
    PMID 21624810DOI
    What this evidence cannot show

    Abstract read this cycle; full text not retrieved. MURINE MLL-AF9 MYELOID disease and murine germline knockouts - not infant B-lineage KMT2A::AFF1, and a germline Bmi1 null is a developmental absence rather than an acute loss in an established leukaemia. The substitution argument is made at the p16Ink4a/p19Arf locus; whether Hoxa9 substitutes for BMI1's other polycomb targets is not addressed.

    Evidence that argues against this (1)
    • A second 2011 study in the same fusion reached a different conclusion in vivo: Bmi1 was dispensable for in vitro immortalisation of MLL-AF9-transformed GMPs but indispensable for leukaemia development in vivo, and deleting Ink4a and Arf only partially restored leukaemogenic capacity. Yuan J et al. 2011
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Bmi1(-/-) murine granulocyte/macrophage progenitors transformed with MLL-AF9 were immortalised in vitro, showing Bmi1 is not essential to immortalisation, but showed enhanced differentiation and retained fewer leukaemic stem cells; transplantation showed Bmi1 was indispensable for leukaemia development in vivo; deletion of both Ink4a and Arf only partially restored the leukaemogenic capacity of Bmi1(-/-) LSCs; and complementing immortalised Bmi1(-/-)Ink4a-Arf(-/-) GMPs with Bmi1 failed to restore the majority of deregulated genes or in vivo leukaemogenic activity.

    Bmi1 is essential for leukemic reprogramming of myeloid progenitor cells · Yuan J, Takeuchi M, Negishi M, Oguro H, Ichikawa H, Iwama A · Leukemia 25:1335-1343 2011
    PMID 21527932DOI
    What this evidence cannot show

    Abstract read this cycle. MURINE MLL-AF9 MYELOID disease, granulocyte/macrophage progenitors as the cell of origin, germline Bmi1 knockout. The 'only partially restored' phrasing is the authors' and is not quantified in the abstract, so the size of the Ink4a/Arf-independent component is not known from what was read. The failure of Bmi1 re-expression to rescue is the most interesting claim here and also the hardest to interpret from an abstract.

    Evidence that argues against this (1)
    • The companion 2011 paper found MLL-AF9 readily transformed Bmi1-null cells and that Hoxa9 substitutes for Bmi1 at the p16Ink4a/p19Arf locus, so the in vivo requirement reported here may be conditional on the cell of origin or on Hox dosage rather than general to MLL fusions. Smith LL et al. 2011
  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 BMI1 has mean +0.0585 and median +0.0738 across 1,258 cell lines, and mean -0.0103 across 57 leukaemia-annotated lines, with 1.43% of lines at or below -0.5 and 0.24% at or below -1.0. BMI1 knockout therefore imposes no measurable fitness cost in the great majority of human cancer lines, including leukaemia lines.

    DepMap CRISPR (Chronos) gene effect for BMI1 (ENSG00000168283), retrieved via the Open Targets Platform GraphQL API, depMapEssentiality.screens · 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 and a tool, and bounded ONLY on 21-day proliferation of immortalised lines. Both primary papers already report exactly this result in vitro, so it is confirmation and not refutation. BMI1 is PCGF4, one of six PCGF paralogues, so a single-gene knockout is buffered by construction - and PCGF1, the PRC1.1 subunit this graph treats as central, is a separate gene. Self-renewal, LSC frequency and in vivo leukaemogenesis - the phenotypes where every positive BMI1 result lives - are invisible to this assay. No infant-derived line in the panel.

    Evidence that argues against this (1)
    • This project's genome-wide scan ranks BMI1 5,450 of 17,728 for KMT2A-rearranged selectivity with a difference of -0.0293, so there is no selectivity signal either - the gene is uninteresting to DepMap on both the absolute and the differential axis, which is a stronger negative than either alone. analysis run by this project 2026
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    CDKN2A is deleted in about 8.5% (4/47) of infant KMT2A-rearranged ALL, against 28% (5/18) of infant ALL without a KMT2A rearrangement, so in the great majority of infant KMT2A-rearranged patients the p16INK4a/p14ARF brake is genomically intact and any repression of it must be epigenetic rather than structural.

    The landscape of somatic mutations in infant MLL-rearranged acute lymphoblastic leukemias · Andersson AK, Ma J, Wang J, Chen X, Gedman AL, Dang J, Nakitandwe J, Holmfeldt L, Parker M, et al. · Nature Genetics 47:330-337 2015
    PMID 25730765DOI
    What this evidence cannot show

    The FREQUENCY is quoted from this graph's gene-cdkn2a entry, not from the paper - this beat did not read Andersson et al. The PMID was first seen in that sibling entry and was then FETCHED FRESH from PubMed efetch on 2026-09-20, from which this title, DOI, venue and author list were copied, because the brief forbids re-using a sibling entry's citation block. The number itself remains second-hand: treat it as a pointer to gene-cdkn2a. n=47 infant KMT2A-r against n=18 infant non-KMT2A-r, P=0.058, so the comparison is underpowered and the '91% intact' figure carries that uncertainty.

    Evidence that argues against this (1)
    • A deletion frequency is a bulk-sample measurement with no clonality data in any study gene-cdkn2a read, so 'intact in 91%' describes the diagnostic sample and not necessarily the clone that relapses. Andersson AK et al. 2015
  6. E_q66_loc1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    BMI1's subnuclear address is PERICENTROMERIC and it is inherited through mitosis: antibodies to RING1, BMI1 and hPc2 show in several human cell lines that the PcG complex forms a discrete nuclear structure - PcG bodies - with the larger foci generally near centromeres, clustered rather than randomly dispersed, tightly associated with the large pericentromeric heterochromatin region 1q12 on chromosome 1 and with related pericentromeric sequences on other chromosomes, and these heterochromatin-bound PcG complexes remain stably associated THROUGHOUT MITOSIS.

    The human polycomb group complex associates with pericentromeric heterochromatin to form a novel nuclear domain. · Saurin AJ; Shiels C; Williamson J; Satijn DP; Otte AP; Sheer D; Freemont PS · J Cell Biol 1998
    PMID 9722603DOI
    What this evidence cannot show

    Human fibroblasts and cell lines, none of them haematopoietic and none KMT2A-rearranged. The assignment rests on immunofluorescence with a kinetochore marker for reference, so 'near the centromere' is an optical proximity statement at 1998 resolution. The mitotic-retention claim is the strongest part and is the part least often quoted. Read: abstract only. Identifiers copied field by field from the efetch record fetched 2026-09-21, ArticleId anchored per brief rule 81.

    Evidence that argues against this (1)
    • Correlative light-electron microscopy says the PcG body is not a body. In human U-2 OS cells expressing BMI1-GFP, high-pressure freezing with cryosubstitution and on-section immunogold identified fluorescent PcG bodies NOT as distinct nuclear bodies but as nuclear domains enriched in separated heterochromatin fascicles - and crucially the density of immunogold-labelled BMI1 inside a fluorescent 'PcG body' DID NOT DIFFER from the density in heterochromatin fascicles outside one. The fluorescent focus reflects a local accumulation of heterochromatin, not a concentration of BMI1. Smigová J; Juda P; Cmarko D; Raška I 2011
  7. E_q66_loc2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    BMI1's residence in polycomb bodies is dynamic and requires other machinery: there are at least two kinetically different pools of BMI1, a highly dynamic and a less dynamic fraction; PRC2 members EED and EZH2 are essential for BMI1 recruitment to PcG bodies; and the maintenance DNA methyltransferase DNMT1 is necessary for proper PcG body assembly, independently of DNMT-associated histone deacetylase activity.

    Association of BMI1 with polycomb bodies is dynamic and requires PRC2/EZH2 and the maintenance DNA methyltransferase DNMT1. · Hernández-Muñoz I; Taghavi P; Kuijl C; Neefjes J; van Lohuizen M · Mol Cell Biol 2005
    PMID 16314526DOI
    What this evidence cannot show

    Photobleaching kinetics and knockdown in cultured human cell lines, not haematopoietic and not leukaemic. Two kinetic pools is a fitted description of recovery curves, not a count of physical states. The EED/EZH2 and DNMT1 requirements are for RECRUITMENT and ASSEMBLY respectively, which are different quantities, and the paper does not establish that either is required for BMI1's repressive output. Read: abstract only.

    Evidence that argues against this (1)
    • If EZH2 is required to put BMI1 in polycomb bodies, then the compartment is downstream of the very complex this graph's PRC2 arm treats as separable from PRC1 - and a cell treated with an EZH2 or dual EZH1/2 inhibitor would be a cell in which BMI1's subnuclear address has been altered as a side effect. No study measures BMI1 localisation under an EZH2 inhibitor, and no PcG-body measurement in this graph is corrected for it. Smigová J; Juda P; Cmarko D; Raška I 2011
  8. E_q66_correction1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The in vitro and the in vivo halves of the founding Bmi1/MLL-AF9 experiment point OPPOSITE WAYS, and this entry has been carrying only the first: in murine MLL-AF9-transformed Bmi1-null granulocyte-macrophage progenitors, Bmi1 was NOT essential to immortalisation in vitro, but Bmi1-null cells showed ENHANCED DIFFERENTIATION and retained FEWER leukaemia stem cells, a number of genes including potential tumour suppressors were derepressed, and transplantation assays demonstrated that BMI1 WAS INDISPENSABLE FOR THE DEVELOPMENT OF LEUKAEMIA IN VIVO - with deletion of both Ink4a and Arf only PARTIALLY restoring the leukaemogenic capacity of Bmi1-null LSCs, and re-expression of Bmi1 in immortalised Bmi1-null Ink4a-Arf-null GMPs failing to restore either most deregulated genes or in vivo leukaemogenic activity.

    Bmi1 is essential for leukemic reprogramming of myeloid progenitor cells. · Yuan J; Takeuchi M; Negishi M; Oguro H; Ichikawa H; Iwama A · Leukemia 2011
    PMID 21527932DOI
    What this evidence cannot show

    Murine MLL-AF9 myeloid leukaemia throughout: species, lineage and fusion partner all differ from infant B-lineage KMT2A::AFF1, and the claim is written with the system first per brief rule 73. Read: abstract only, and the abstract is unambiguous on the in vivo result. This entry's E3 already cites this PMID and quotes only its in vitro sentence; the present item records the rest of the same abstract. The failure of Bmi1 re-expression to rescue is the most interesting and least quoted part - it says the requirement is for faithful REPROGRAMMING at the time of transformation, not for ongoing BMI1 activity, which is a different and much harder thing for a drug to exploit.

    Evidence that argues against this (1)
    • The Ink4a/Arf half of this entry's model survives only partially even here: deleting both Ink4a and Arf only PARTIALLY restored leukaemogenic capacity in Bmi1-null LSCs. So 'HOXA9 covers for BMI1 at the CDKN2A locus' cannot be the whole account, because removing the locus outright does not fully substitute for BMI1 either. Independently, the concurrent comparison-arm work in this graph records BMI1 silencing or pharmacological inhibition impairing CALM-AF10 leukaemia, which makes BMI1 a shared polycomb dependency across two HOXA-driven lesions rather than a KMT2A-specific bypass. Oncofusion knowledge graph, data/graph/nodes/calm-af10/mechanism-calm-af10-bmi1-prc1-dependence.json 2026
  9. E_q97_epi1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    hPRC1L, a histone-H2A-specific E3 ubiquitin ligase complex composed of several polycomb-group proteins including Ring1, Ring2, BMI1 and HPH, was purified and functionally characterised; reducing the expression of these polycomb proteins decreases histone H2A ubiquitination in vivo, and H2A ubiquitination is linked to polycomb silencing.

    PMID 15386022
    What this evidence cannot show

    Abstract fetched and read in full 2026-09-21; full text not read. The founding paper for H2A ubiquitination as a polycomb mark, but it is a biochemical and cell-biological characterisation in HeLa-derived material - not a haematopoietic system and not a leukaemia. BMI1 appears here as one named component of the complex; the paper does not attribute catalysis to it. IDENTIFIER NOTE: this is PMID 15386022, which differs by one digit from 15385962; the record was fetched and the title read back programmatically rather than recalled.

    Evidence that argues against this (1)
    • A complex-level knockdown that lowers H2Aub does not tell you which subunit is required for what, and this graph has been burned by exactly that inference (rule 64). The specific catalytic assignment to RING1B comes from a different paper [E_q97_epi2], and BMI1's contribution in that framework is stimulatory and structural. An entry that reads this paper as BMI1 ubiquitinates H2A would be wrong in a way the abstract does not correct. J Biol Chem 2004
  10. E_q97_epi2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Ring1b is the H2A-K119 ubiquitin E3 ligase and is enriched on the inactive X chromosome in female trophoblast stem cells and differentiating cells, supporting H2A ubiquitination as an epigenetic marker of the inactive X and linking it to the initiation of X inactivation.

    PMID 15509584
    What this evidence cannot show

    Abstract fetched and read in full 2026-09-21; full text not read. Murine trophoblast stem cells and X inactivation - as far from infant B-ALL as a chromatin system gets. Cited here for one thing only: it is where the catalytic subunit assignment for H2AK119ub sits, which is what licenses this entry to say BMI1 is a required subunit of the writer rather than the writer.

    Evidence that argues against this (1)
    • A catalytic assignment made in one developmental system does not guarantee the same division of labour in a leukaemic blast, where PCGF paralogue composition differs. This graph's own PRC1.1 arm runs through PCGF1 and RNF2, not BMI1, so the relevant E3 in KMT2A-rearranged cells may be assembled differently. Nature 2004
  11. E_q97_epi3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In B-cell lymphoblastic leukaemia, nuclear Vav3 interacts with BCR-ABL, Rac and the polycomb proteins Bmi1, Ring1b and Ezh2, and the GEF activity of Vav3 is required for Bmi1-dependent B-cell progenitor self-renewal, for interaction with Bmi1, for mono-ubiquitination of H2A(K119) and for repression of PRC1 downstream target loci in leukaemic B-cell progenitors.

    PMID 35650206
    What this evidence cannot show

    Abstract fetched and read in full 2026-09-21; full text not read. B-LINEAGE, which is the right lineage - but the driver is BCR-ABL, NOT a KMT2A fusion (rule 73), and this graph holds fusion-bcr-abl1 as a comparison arm rather than an index lesion. The H2AK119ub readout is reported as a requirement downstream of Vav3 GEF activity, so it is Vav3-conditional Bmi1 activity that is measured, not baseline Bmi1-dependent H2AK119ub. No KMT2A-rearranged arm is reported.

    Evidence that argues against this (1)
    • The mechanism routes Bmi1 activity through an Akt-dependent phosphorylation at S314 that Vav3 prevents. If BMI1's H2AK119ub output in B-lymphoid leukaemia is gated by a signalling input, then the DepMap near-zero this entry records for BMI1 could reflect cell lines in which that gate is closed rather than genuine dispensability - and equally, a BMI1 inhibitor would act on a target whose activity is already conditionally regulated by kinase signalling the cell can adjust. Nat Commun 2022
Connected to