Oncofusion

start / graph / gene-prom1

Genestrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q45-immunophenotype

PROM1 / CD133 - a direct MLL-AF4 target switched on by an intragenic H3K79me2/3 enhancer, and the one surface target in this disease with a published human-relevant kill

also called PROM1 · CD133 · AC133 · PROML1 · Prominin-1 · Antigen AC133 · O43490

CD133 is the marker most often used to label cancer stem cells, and in this leukaemia it is switched on directly by the fusion protein through the same chemical mark the drug pinometostat was built to erase. It is also a cautionary tale: a CAR-T cell aimed at CD133 wiped out the patient's normal bone marrow along with the leukaemia.

The same thing, for a scientist

PROM1 (UniProt **O43490**, PROM1_HUMAN, **865 aa**, single reviewed human record, synonym PROML1) encodes the pentaspan transmembrane glycoprotein **CD133 / AC133 / Prominin-1**. This graph names it in 21 entries and has never written it, despite carrying `drug-pinometostat`, `gene-dot1l` and a whole H3K79 layer that PROM1 is the cleanest surface read-out of. **The regulation is the most completely worked-out of any marker in this beat.** Godfrey 2021 showed **aberrant PROM1/CD133 expression is essential for leukaemic cell growth and is mediated by direct binding of MLL-AF4**. Activation runs through an **intragenic H3K79me2/3 enhancer element (KEE)** that drives enhancer-promoter interactions between *PROM1* and the neighbouring gene *TAPT1*, a dual-locus arrangement reflected in a strong expression correlation in leukaemia. In cells that do not express CD133, the *PROM1* locus is **repressed by PRC2 binding**, with reduced *TAPT1* expression partly from loss of interaction with the *PROM1* locus [E1]. Earlier work had already shown AF4 promotes CD133 transcription [E2]. **So PROM1 is a surface marker whose ON state is H3K79-dependent and whose OFF state is Polycomb-dependent.** That is a direct, testable prediction: **DOT1L inhibition should lower surface CD133 in MLL-AF4 cells**, and this graph's the PRC2 and H3K79/DOT1L beats both have a stake in it. It also means CD133 is a better candidate than NG2 for the question of whether a marker tracks the *programme* rather than the lesion, because its programme dependency is mapped to a specific mark at a specific element. **The therapeutic answer is already in.** Bueno 2019 titled the result plainly: **CD133-directed CAR T-cells for MLL leukaemia produce on-target, off-tumour myeloablative toxicity** [E3]. CD133 is on normal haematopoietic stem cells. A CAR against it is a conditioning regimen. This is the single most useful negative result in this beat and it is not currently in the graph.

Go deeper
Why CD133 matters to the menin and DOT1L layers of this graph

If CD133 surface expression is driven by an intragenic H3K79me2/3 element bound by MLL-AF4 1, then CD133 is a flow-readable output of the exact chromatin axis this graph is built around. Three consequences:

  1. A cheap pharmacodynamic assay. Surface CD133 by flow is a same-day, single-cell, clinically routine measurement. If it tracks H3K79me2/3 at the KEE, it is a far easier pharmacodynamic marker for a DOT1L or menin inhibitor than a marrow ChIP.
  2. A prediction that distinguishes the two drug classes. DOT1L inhibition removes H3K79me2/3 directly; menin inhibition displaces the fusion from chromatin and should remove it indirectly. If CD133 falls under both, it is reading the programme; if it falls only under pinometostat, it is reading the mark.
  3. A PRC2 link. The OFF state is PRC2-bound 1, and this graph's the PRC2 beat concluded that EZH2 is the fusion's collaborator rather than its antagonist. A locus where MLL-AF4 activation and PRC2 repression are the two alternatives is a concrete place to test that.

A caution from the resistance literature. CD133 appears among the markers in work modelling acquired resistance to DOT1L inhibition in KMT2A-rearranged ALL, which reports adaptive potential rather than clean marker loss - so 'CD133 falls under pinometostat' may be true early and not durable.

Cell type, cell state, localisation, epigenetics
  • Cell type. Normal haematopoietic stem and progenitor cells, neuroepithelium, retinal photoreceptors and multiple epithelia - which is the whole toxicity story 3. On blasts: in KMT2A-rearranged ALL, driven by the fusion 1.
  • Cell state. A stemness-associated state marker rather than a lineage marker; described as a pan-cancer stem cell marker 1.
  • Localisation. Plasma membrane, pentaspan (five transmembrane passes), 865 residues, concentrated in membrane protrusions and cholesterol-rich microdomains 4.
  • Epigenetics. ON via an intragenic H3K79me2/3 enhancer element with MLL-AF4 direct binding; OFF via PRC2 1. Coupled to the neighbouring TAPT1 locus by enhancer-promoter interaction.
Localization, cell state, and why subnuclear is NOT APPLICABLE (Q87, 2026-09-21)

Subcellular. Apical plasma membrane, multi-pass, with microvillus-membrane and ciliary/photoreceptor-outer-segment pools and passage through the ER and ERGIC; shed into extracellular membrane particles in body fluids (UniProt O43490). HPA v25.1 IF: plasma membrane, reliability Approved.

Subnuclear: NOT APPLICABLE. A polytopic membrane glycoprotein has no nuclear pool. Category exclusion, not a bounded absence.

Cell state. CD133 is an operational stem/progenitor sorting marker; in this disease its expression is fusion-driven through an intragenic H3K79me2/3 domain, so the state answer and the fusion answer coincide.

Shown for fusion. MLL-AF4 for regulation; nobody for localisation - the one fusion-scoped localisation hit was retrieved, title-checked and rejected.

Facts and figures
uniprotO43490
gene symbolPROM1
length AA865
cd numberCD133
therapeutic statusCD133-directed CAR T-cells in MLL leukaemia: on-target, off-tumour MYELOABLATIVE toxicity (PMID 30778134)
accession resolutiongene_exact:PROM1 AND organism_id:9606 AND reviewed:true, fields include gene_primary and length; one row; primary name read back as PROM1; length 865 asserted for verify-accessions.py.
subnuclear localizationNOT APPLICABLE - PROM1/CD133 is a five-transmembrane apical plasma-membrane glycoprotein with no nuclear pool. Recorded as a category exclusion rather than a bounded absence.
cell stateSTEM-LIKE / PRIMITIVE, as an operational flow-cytometry definition rather than a measured compartment. CD133 is used to sort haematopoietic stem and progenitor cells and is read as a stemness marker in leukaemia; (PROM1[tiab] OR CD133[tiab]) AND ("hematopoietic stem"[tiab] OR "haematopoietic stem"[tiab]) returns 438 records (2026-09-21), which is a literature size and not a validation. This entry already records the regulatory half - PROM1 as a direct MLL-AF4 target driven by an intragenic H3K79me2/3 domain - so for PROM1 the cell-state answer and the fusion answer are the same fact.
shown for fusionMLL-AF4 for REGULATION, nobody for LOCALISATION. PROM1 is a direct MLL-AF4 target with an intragenic H3K79me2/3 domain, which is an occupancy-and-expression result, not a localisation one. The fusion-scoped localisation query returns one record, PMID 25370525 ('Nuclear cupping in the blasts - more to the cup than myeloid'), which is a cytology image note with no abstract and nothing about PROM1's compartment; retrieved, title-checked and rejected. Screened count on topic: 0.
other names
[
  "AC133",
  "Prominin-1",
  "PROML1"
]
regulation
{
  "on_state": "direct MLL-AF4 binding; intragenic H3K79me2/3 enhancer element (KEE); enhancer-promoter interaction with neighbouring TAPT1",
  "off_state": "PRC2 binding at the PROM1 locus",
  "pmid": "32242051",
  "title": "H3K79me2/3 controls enhancer-promoter interactions and activation of the pan-cancer stem cell marker PROM1/CD133 in MLL-AF4 leukemia cells.",
  "authors": "Godfrey L, Crump NT, O'Byrne S, et al.",
  "venue": "Leukemia",
  "year": 2021,
  "title_provenance": "Title/authors/venue backfilled from PubMed esummary on 2026-09-20 because this citation stored a bare PMID with no title, which made it unscoreable by verify-citations.py. BACKFILLING IS NOT VERIFICATION: it confirms the PMID resolves to a real paper, not that this is the right paper for the claim. That still requires reading."
}
depmap kmt2a selectivity
{
  "chronos_mean_kmt2a_rearranged": 0.1011,
  "chronos_mean_other_leukaemia": 0.1002,
  "difference": 0.0009,
  "pan_essentiality_all_lines": 0.1237,
  "selectivity_rank": 9267,
  "rank_denominator": 17728,
  "provenance": "Joined 2026-09-21 from analyses/depmap-kmt2a-selective-scan-full (generated 2026-09-19T10:12:06Z), the project's own genome-wide scan over 11 KMT2A-rearranged and 43 other leukaemia DepMap lines. NOT re-fetched and NOT independent of that scan: every caveat it carries applies here, including that the KMT2A-rearranged arm is a small number of lines drawn from nine individuals, that the rank denominator is 17,728 genes, and that a difference is not a dependency. Added because only 26 of 308 gene entries carried this number, which made any KMT2A-selectivity question answerable for 8% of the corpus.",
  "how_to_read": "A negative difference means MORE required in KMT2A-rearranged lines than in other leukaemia lines. A rank near 17,728 is ANTI-selective. Neither is evidence of a fusion-driven dependency on its own: check pan_essentiality_all_lines before calling anything selective, and note that a paralogue-redundant gene can score near zero for reasons that have nothing to do with this disease - see ana-the-paralogue-redundancy-excuse-is-not-tested-and-my-proxy-decided-the-answer."
}
localization
{
  "subcellular": "APICAL PLASMA MEMBRANE, multi-pass. UniProt O43490 lists Apical cell membrane (multi-pass), Cell projection/microvillus membrane (multi-pass), Cell projection/cilium/photoreceptor outer segment, Endoplasmic reticulum and the ER-Golgi intermediate compartment, with the note that it is found in extracellular membrane particles in cerebrospinal fluid, saliva, seminal fluid and urine. HPA v25.1 IF, reliability Approved, gives a single location: Plasma membrane.",
  "subnuclear": "NOT APPLICABLE. Prominin-1 is a five-transmembrane glycoprotein inserted co-translationally into the ER membrane and trafficked to the apical surface and to membrane protrusions. There is no soluble nuclear species, so there is no subnuclear address to determine. Recorded as a category exclusion, like HLA-A and NRAS in the previous tranche, rather than as an unmeasured gap.",
  "the_topology_is_the_informative_part": "PROM1 is selectively concentrated in PLASMA-MEMBRANE PROTRUSIONS - microvilli, the photoreceptor outer segment, and the membrane particles it is shed in. That is a real subcellular address with structure to it, and it is why CD133 epitope detection is so sensitive to membrane state: a CD133-negative flow result can mean the protein is absent, or that it has been shed, or that the epitope's glycosylation changed.",
  "how_shown": "UniProt's apical and ciliary topology terms are ECO:0000250 (sequence similarity), not experimental; the plasma-membrane call is corroborated independently by HPA antibody immunofluorescence at reliability Approved",
  "not_applicable_reason": "polytopic plasma-membrane glycoprotein, no nuclear pool",
  "node": "localization-plasma-membrane-leukaemic-surface-antigens"
}
How this could bite us back
  • CD133-directed CAR T-cells for MLL leukaemia caused on-target, off-tumour myeloablative toxicity. CD133 is on normal haematopoietic stem cells, so a CD133 CAR is functionally a conditioning regimen, and the patients are infants. This is the clearest published refutation of a surface target in this disease and it should be the first thing anyone proposing a CD133 strategy reads.
  • A CD133-based pharmacodynamic readout for DOT1L or menin inhibition may fail exactly when it matters. Acquired resistance to DOT1L inhibition in KMT2A-rearranged ALL is adaptive rather than binary, so the marker could report drug effect early and lose the signal as resistance develops.
  • The AC133 flow epitope is glycosylation-dependent. A therapy or a stress that changes glycosylation could make the blast look CD133-negative without changing PROM1 expression - which would produce a false escape signal and a false MRD reassurance at the same time.
  • PROM1 is coupled to TAPT1 by enhancer-promoter interaction. Any intervention that shuts the PROM1 locus is also reducing TAPT1, a gene this graph has never considered, and the consequences of that are unknown.
Still open
  1. Does surface CD133 fall under menin inhibition as well as under DOT1L inhibition? If it falls under both it reads the programme; if only under pinometostat it reads the mark, and the two answers point to different pharmacodynamic uses.
  2. Is the intragenic H3K79me2/3 enhancer element at PROM1 present in primary infant KMT2A::AFF1 blasts, or only in the cell lines in which it was mapped?
  3. What does losing TAPT1 do? It is dragged along by any PROM1-directed epigenetic intervention and nothing in this graph has looked at it.
  4. Can a CD133 CAR be made safe by logic-gating against a second antigen? TanCAR CD19/CD133 exists by title and the myeloablation question for it is unanswered here.
Every source behind this entry (6)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In MLL-AF4 leukaemia cells, aberrant PROM1/CD133 expression is essential for leukaemic cell growth and is mediated by direct binding of MLL-AF4. Activation is controlled by an intragenic H3K79me2/3 enhancer element (KEE) that increases enhancer-promoter interactions between PROM1 and the nearby gene TAPT1, and this dual-locus regulation is reflected in a strong correlation of expression in leukaemia. In PROM1/CD133 non-expressing cells the PROM1 locus is repressed by polycomb repressive complex 2 binding, associated with reduced expression of TAPT1 partly due to loss of interactions with the PROM1 locus. MLL-AF4 promotes leukaemogenesis mainly through recruitment of DOT1L and increased H3K79me2/3.

    PMID 32242051DOI
    What this evidence cannot show

    Abstract read in full from raw efetch XML on 2026-09-20; full text not read. **Cell-line work in MLL-AF4 models**, not primary infant patient material, and the abstract does not name the lines. 'Essential for leukaemic cell growth' is a knockdown/knockout result in those systems. The KEE is defined by H3K79me2/3 occupancy, so 'H3K79me2/3 controls' is partly definitional and the causal step requires the perturbation experiments, which are in the full text and not read here. Nothing in this abstract measures SURFACE CD133 by flow under DOT1L inhibition - the pharmacodynamic prediction this entry draws is a prediction, not a result.

    Evidence that argues against this (1)
    • A published model of acquired resistance to DOT1L inhibition in KMT2A-rearranged ALL reports substantial adaptive potential in these cells, so a marker that falls when H3K79 methylation is removed may recover as resistance develops. A CD133-based pharmacodynamic readout could therefore report drug effect early and lose the signal exactly when it is needed. Schneider P et al. 2023
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    The mixed lineage leukaemia (MLL) fusion-associated gene AF4 promotes CD133 transcription.

    PMID 22337994DOI
    What this evidence cannot show

    **Title only; the abstract was not fetched and the full text was not read in this session.** Authors, venue and DOI were fetched from PubMed efetch on 2026-09-20 and are not from recall, but nothing beyond the title was read. The claim is exactly what the title asserts. Carried because it establishes that the AF4-to-CD133 link predates and is independent of the 2021 chromatin mechanism [E1], which matters for whether that mechanism is a single-laboratory finding. A reader needing a number from this paper must fetch it.

    Evidence that argues against this (1)
    • A title-only citation cannot be checked against the claim it is cited for beyond the words in the title. It is recorded here at moderate confidence under this graph's known-but-unread convention rather than being silently upgraded to support the mechanism in E1. Godfrey L et al. 2021
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    CD133-directed CAR T-cells for MLL leukaemia produced on-target, off-tumour myeloablative toxicity.

    PMID 30778134DOI
    What this evidence cannot show

    **The PubMed record carries no abstract** - this is a correspondence/letter item - so the claim here is exactly the title assertion, fetched from raw efetch XML on 2026-09-20 together with its DOI and PMCID (PMC6756031). The five-route full-text test was not completed on this record in this beat; a PMCID exists, so routes 2-4 remain available and a reader wanting the model system, the degree of myeloablation or whether it was murine or human should run them. **Species and system are not established here.** Carried at strong confidence because the assertion is in the title of a peer-reviewed item from the group that has published most of this disease's CD133 and NG2 biology.

    Evidence that argues against this (1)
    • A dual-targeting design may escape the toxicity: TanCAR T cells targeting CD19 and CD133 have been reported to eliminate MLL leukaemic cells, and an Fc-optimised CD133 antibody has been proposed for NK-mediated killing in B-ALL. Neither is shown here to avoid the myeloablation, and both were found only by title in this beat. Li D et al. 2018
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    PROM1 is UniProt O43490 (PROM1_HUMAN), Prominin-1, also known as Antigen AC133 and Prominin-like protein 1, CD antigen CD133, 865 amino acids, a single reviewed human record with primary gene name PROM1 and gene synonym PROML1.

    What this evidence cannot show

    Fetched 2026-09-20 with `gene_exact:PROM1 AND organism_id:9606 AND reviewed:true`, requesting `gene_primary` and `length` per rules 62 and 65; primary name read back as PROM1; exactly one reviewed row; length 865 asserted for `verify-accessions.py`. **The gene symbol PROM1 and the clinical name CD133 do not share a character**, and PROM2 exists as a paralogue - a loose match could return it.

    Evidence that argues against this (1)
    • The AC133 epitope that most CD133 antibodies detect is a glycosylation-dependent epitope, not simply the protein. CD133 protein can be present while the AC133 epitope is not detected, so 'CD133-negative' by flow and 'PROM1 not expressed' are different statements and the accession does not bridge them. UniProt Knowledgebase, reviewed (Swiss-Prot) 2026
  5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProt O43490 places prominin-1 exclusively in membranes - apical cell membrane and microvillus membrane, both multi-pass, the photoreceptor outer segment, the endoplasmic reticulum and the ER-Golgi intermediate compartment - and notes its presence in extracellular membrane particles in body fluids. HPA v25.1 immunofluorescence gives one location, Plasma membrane, at reliability Approved. Neither record contains a nuclear term, so PROM1's subnuclear localisation is NOT APPLICABLE.

    What this evidence cannot show

    Most of UniProt's topology terms for O43490 are ECO:0000250 (inferred by sequence similarity), so the APICAL specificity in particular is an inference rather than a measurement on this protein; only the plasma-membrane call is independently corroborated. Neither source is haematopoietic - the apical and ciliary work is epithelial and retinal - and a CD133 readout is a glycosylation-dependent antibody epitope, so two clones are not interchangeable (rule 74).

    Evidence that argues against this (1)
    • The NOT APPLICABLE call is about the PROTEIN and must not be read as saying the PROM1 LOCUS is irrelevant to nuclear biology in this disease, because that is precisely where the interesting fusion result lives: PROM1 is a direct MLL-AF4 target carrying an intragenic H3K79me2/3 domain. The fusion's relationship with PROM1 is entirely a chromatin one, at a place in the nucleus the protein never visits. The comparator for what a real fusion-scoped chromatin-address result looks like is the Tip60/H2A.Z work at Hoxa9, where the fusion recruits an enzyme to a named locus - PROM1 has a result of the same kind, so an entry that records only 'no subnuclear address' would invert the actual state of knowledge. Yamagata K et al. 2021
  6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    No study reports PROM1/CD133 protein localisation in a KMT2A-fusion context. Measured 2026-09-21: (PROM1[tiab] OR CD133[tiab] OR prominin[tiab]) AND (KMT2A[tiab] OR MLL[tiab] OR "MLL-AF4"[tiab] OR "MLL-AF9"[tiab] OR "MLL-ENL"[tiab] OR "MLL-AF10"[tiab] OR "t(4;11)"[tiab] OR "t(9;11)"[tiab]) AND (localization[tiab] OR localisation[tiab] OR subnuclear[tiab] OR nuclear[tiab] OR nucleolar[tiab] OR speckle[tiab] OR speckles[tiab] OR "PML bodies"[tiab] OR chromatin-bound[tiab] OR cytoplasmic[tiab]) returns 1 record, PMID 25370525, which was retrieved and title-checked and is off topic. Screened count: 0.

    What this evidence cannot show

    Bounded by titles and abstracts only. 'CD133' is a CD number, and rule 69 part 4 warns a CD number can stand for one gene, no gene, several genes or two antigens; here it is one gene, but the epitope is glycosylation-dependent, so the surface-staining literature and the protein-localisation literature are not the same corpus and a query over one does not bound the other. Rule-67 positive control for this query SHAPE: the identical construction returned non-zero for seven of the sixteen genes run in this tranche on the same day - H2AZ1 1, EED 2, CHAF1B 1, CEBPA 3, TIFAB 1, SAMHD1 1, PROM1 1 - so the instrument can return a one. The querytranslation was read back on every run and confirms that the quoted cytogenetic terms expand to "t 4 11"[Title/Abstract] and "t 9 11"[Title/Abstract], not to a UID filter (rule 95).

    Evidence that argues against this (1)
    • This absence carries very little weight against the question anyone would actually ask, and reporting it without that qualification would misrepresent the field. Nobody wants to know which membrane PROM1 sits in in a KMT2A-rearranged blast; they want to know whether the fusion's H3K79me2/3 domain over the PROM1 gene body is what switches the surface marker on - and that has been addressed in a named fusion. The general lesson from the H2A.Z case applies: a fusion-scoped chromatin result and a fusion-scoped localisation result are different questions, and a localisation-term query cannot see the first one at all. Yamagata K et al. 2021
Connected to