Oncofusion

start / graph / localization-meis1-nucleus-is-a-sequence-prediction-and-one-isoform-is-cytoplasmic

Mechanismmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q57-localization2

MEIS1's 'nucleus' is a profile prediction, not a measurement - and a homeodomain-less isoform is cytoplasmic and does the opposite job

also called MEIS1 localization · MEIS1-HdL cytoplasmic · MEINOX nuclear retention · TALE factor localization · PROSITE-ProRule PRU00108

The database says MEIS1 is a nuclear protein, but that statement comes from a computer recognising a homeodomain in its sequence, not from anyone looking. A 2026 study found a shorter MEIS1 that lacks that homeodomain, sits in the cytoplasm, and blocks the very differentiation the nuclear form drives.

The same thing, for a scientist

UniProt O00470 annotates MEIS1 `Nucleus` with **ECO:0000255|PROSITE-ProRule:PRU00108** - a sequence-analysis inference from the homeodomain profile, the weakest evidence class in the location vocabulary and the only one in this beat that is not a measurement at all [E1]. The measured picture is more interesting. Mouse Meis1 produces **MEIS1-FL** (homeodomain-containing, nuclear, promotes ATOH1 degradation and granule-cell-progenitor differentiation) and **MEIS1-HdL** (homeodomain-less, **cytoplasmic**, inhibits ATOH1 degradation via CUL3/COP9 signalosome and blocks differentiation) - functionally antagonistic isoforms in distinct compartments [E2]. Nuclear Meis1 is also stage-selective in vivo: it is nuclear in the dermal papilla specifically in the anagen phase of the hair cycle [E3]. And the textbook role of MEINOX proteins in retaining PBX in the nucleus does not hold everywhere: **PBX1 is mostly cytoplasmic in epithelial endometrial cells despite nuclear MEIS1** [E4]. No study localises MEIS1 protein in a KMT2A-rearranged cell [E5].

Go deeper
What ECO:0000255|PROSITE-ProRule means, and why it matters here

The complete UniProt subcellular-location comment for O00470, fetched 2026-09-20:

{"location": {"value": "Nucleus",
  "evidences": [{"evidenceCode": "ECO:0000255",
                "source": "PROSITE-ProRule", "id": "PRU00108"}]}}

ECO:0000255 is 'match to sequence model evidence used in automatic assertion'. PRU00108 is the homeodomain ProRule. The chain is: this protein contains a homeodomain; homeodomain proteins are nuclear; therefore this protein is nuclear. It is a reasonable inference and it is not an observation of MEIS1.

This beat has now found four distinct non-measurement annotation classes in the graph's core genes:

geneannotationwhat it actually is
MEIS1ECO:0000255 PROSITE-ProRulesequence profile prediction
DOT1L (cytoplasm)ECO:0000250by similarity from mouse
ASH2LECO:0000305curator inference
MLLT1/ENLno code at allunattributed

All four render identically in the flat text a bulk import reads. Only the JSON evidence array distinguishes them.

The isoform that breaks the prediction

Owa et al. sequenced full-length cDNA from mouse cerebellar granule-cell progenitors and found Meis1 among more than 700 genes with high isoform diversity 2. The two isoforms are:

  • MEIS1-FL - contains the homeodomain, localises to nuclei, promotes ATOH1 protein degradation transcriptionally, drives differentiation.
  • MEIS1-HdL - lacks the homeodomain, localises to the cytoplasm, inhibits CUL3-mediated ATOH1 degradation by enhancing COP9-signalosome binding to CUL3, and inhibits differentiation.

The irony is exact: the feature the PROSITE rule used to predict nuclear localisation is the feature the cytoplasmic isoform lacks. So the annotation is right about one isoform and silently wrong about the other, and the record carries no isoform qualifier.

Rule 73: mouse cerebellar granule-cell progenitors, in vivo electroporation. Not leukaemia, not blood. But MEIS1 is a core KMT2A-fusion target and a HOXA9 partner, and nobody has asked which MEIS1 isoforms a leukaemic blast makes. If a homeodomain-less cytoplasmic isoform exists there, every MEIS1 western in this field is summing two proteins with opposite functions.

Nuclear MEIS1 is a state, not a constant

In mouse vibrissa follicles, Meis1 is selectively nuclear in the dermal papilla in the anagen phase of the hair cycle, and conditional Meis1 loss retards hair growth through a secreted factor acting on matrix cells 3. That is a cell-cycle-phase-like, tissue-state-dependent nuclear localisation - the same shape as HOXA9's thrombopoietin-driven nuclear import recorded in this beat's HOXA9 entry.

Taken together with 2, the honest summary is: MEIS1's compartment is regulated by isoform and by cell state, and the database records neither.

The MEINOX-retention model has a published exception

The standard account, which this graph carries in gene-pbx3, is that PBC-class proteins (PBX1/2/3) carry both NLS and NES and that MEINOX-class heterodimerisation blocks export and drives nuclear retention. Dintilhac et al. provide an organogenesis counter-example: PBX1 is mostly cytoplasmic in epithelial endometrial cells of the developing female genital tract despite MEIS1 being nuclear in the same tissue, and the control of PBX1 distribution there is cell-cycle-correlated and MEINOX-independent 4.

So MEIS1 nuclear does not entail PBX nuclear. Any inference in this graph that runs from MEIS1 expression to nuclear HOX/PBX/MEIS trimer formation is assuming a step this paper shows can fail.

Facts and figures
genegene-meis1
uniprotO00470
length AA390
compartment subcellularnucleus for the homeodomain-containing full-length protein; CYTOPLASM for the homeodomain-less MEIS1-HdL isoform (mouse cerebellar granule-cell progenitors). The human annotation is a sequence-profile prediction, not a measurement.
compartment subnuclearunpublished - no subnuclear observation of MEIS1 exists in any system read for this beat
state dependencenuclear in mouse dermal papilla specifically in the anagen phase of the hair cycle
partner inference warningPBX1 is mostly cytoplasmic in epithelial endometrial cells despite nuclear MEIS1, so MEINOX nuclear retention of PBC-class partners is not universal
shown for fusionNO - MEIS1 protein has never been localised in a KMT2A-rearranged cell, despite being one of the canonical fusion target genes
patient material datafalse
bounded absence date2026-09-20
uniprot evidence codes
{
  "Nucleus": "ECO:0000255|PROSITE-ProRule:PRU00108 - SEQUENCE MODEL MATCH, automatic assertion; no experimental localisation evidence in the record"
}
isoform dependence
{
  "MEIS1-FL": "homeodomain-containing, nuclear, promotes ATOH1 degradation and differentiation",
  "MEIS1-HdL": "homeodomain-less, cytoplasmic, inhibits CUL3-mediated ATOH1 degradation and inhibits differentiation"
}
methods
[
  "long-read cDNA sequencing plus in vivo electroporation and isoform-resolved imaging",
  "immunohistochemistry on tissue sections",
  "conditional knockout in ex vivo organ culture"
]
species and systems
[
  "mouse cerebellar granule-cell progenitors",
  "mouse vibrissa follicle / dermal papilla",
  "human and mouse developing female genital tract",
  "human UT-7/TPO and murine progenitors (for the HOXA9 import work)"
]
How this could bite us back
  • A MEIS1-directed therapeutic strategy built on the transcription-factor function would not touch a cytoplasmic isoform acting through CUL3 and the COP9 signalosome, and there is currently no assay in this field that would notice one.
  • MEIS1 is required in normal haematopoietic stem cells and in cardiomyocyte cell-cycle control, so systemic MEIS1 inhibition carries a stem-cell and a cardiac liability independent of any localisation question.
  • Because the nuclear annotation is a profile prediction, any computational pipeline that filters 'nuclear proteins' for a KMT2A-fusion interactome will include MEIS1 on no evidence, and would include a cytoplasmic isoform on the same non-evidence.
Still open
  1. Which MEIS1 isoforms does a KMT2A-rearranged blast make? If a homeodomain-less cytoplasmic isoform exists there, every MEIS1 western in this field sums two proteins with opposite functions. Long-read cDNA sequencing of existing samples answers it without a new experiment.
  2. Is MEIS1 nuclear localisation in leukaemic blasts constitutive or regulated? It is stage-dependent in hair follicle and cytokine-regulated for its partner HOXA9, so the constitutive assumption is untested in this disease.
  3. Does nuclear MEIS1 imply nuclear PBX3 in a leukaemic blast? One organogenesis system shows it does not, and this graph's homeodomain layer assumes it does.
  4. Is there any subnuclear structure containing MEIS1? Nothing has ever been reported, and HOXA9 - its closest functional partner - has its own named nuclear bodies.
Every source behind this entry (5)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProt O00470 (human MEIS1, 390 aa) annotates subcellular location as Nucleus with evidence code ECO:0000255|PROSITE-ProRule:PRU00108, i.e. a match to the homeodomain sequence model used in automatic assertion, and carries no experimental localisation evidence and no subnuclear or isoform detail.

    What this evidence cannot show

    A database record. gene_exact:MEIS1 AND organism_id:9606 AND reviewed:true returned one reviewed record, gene_primary MEIS1, accession O00470, length 390. ECO:0000255 is not a claim that the protein is not nuclear; it is a claim that the record contains no experimental localisation evidence for human MEIS1, which is what this item asserts.

    Evidence that argues against this (1)
    • MEIS1 is a DNA-binding transcription factor whose entire published function is transcriptional, and it is routinely recovered in nuclear extracts and in ChIP experiments. The nuclear assignment is almost certainly right for the full-length protein; the objection is that a graph should cite the ChIP and the fractionation rather than the profile match, and should not silently extend it to isoforms lacking the homeodomain.
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In mouse cerebellar granule-cell progenitors, Meis1 produces two functionally antagonistic isoforms: MEIS1-FL, which contains the homeodomain, localises to nuclei and promotes ATOH1 protein degradation and progenitor differentiation; and MEIS1-HdL, which lacks the homeodomain, localises in the cytoplasm, inhibits CUL3-mediated ATOH1 degradation by enhancing COP9 signalosome binding to CUL3, and inhibits differentiation.

    Meis1 isoform diversity orchestrates neural progenitor differentiation by regulating ATOH1 degradation at distinct subcellular compartments. · Owa T, Adachi T, Shiraishi R, Ichijo K, Ji K, Mizuno M, Suyama K, Nishitani K, Hasegawa I, Sone M, Kawauchi D, Nishioka T · PLoS Biol 2026
    PMID 42441717DOI
    What this evidence cannot show

    RULE 73: mouse cerebellar granule-cell progenitors with in vivo postnatal electroporation - a neural developmental system, not haematopoietic and not leukaemic. The isoform catalogue comes from long-read cDNA sequencing of that tissue, so the existence of MEIS1-HdL in human blood cells is not established by this work. The MEIS1-FL nuclear localisation is cited by the authors to their own previous study rather than shown afresh here. Read: abstract only (PubMed efetch, 2026-09-20).

    Evidence that argues against this (1)
    • Isoform repertoires are tissue-specific, so the existence of a cytoplasmic homeodomain-less MEIS1 in cerebellar progenitors does not establish one in leukaemic blasts, and this entry should not be read as asserting it. What it does establish is that 'MEIS1 is nuclear' is an isoform-dependent statement, which the annotation does not qualify - and that long-read sequencing of MEIS1 in KMT2A-rearranged cells is an unasked question.
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Meis1 is selectively localised in the nucleus of the dermal papilla in the anagen phase of the mouse hair cycle, and conditional Meis1 loss in ex vivo vibrissa follicle organ culture retards hair growth with defective hair-matrix-cell proliferation, partly rescued by culture supernatant from Meis1-sufficient dermal papilla cells.

    PMID 31543346DOI
    What this evidence cannot show

    RULE 73: mouse hair follicle, an ex vivo organ culture - as far from KMT2A-rearranged infant leukaemia as any system in this beat. It is included because it is one of very few in-vivo observations that MEIS1 nuclear localisation is STAGE-DEPENDENT rather than constitutive. 'Selectively localized in the nucleus... in the anagen phase' implies non-nuclear or absent in other phases, which the abstract does not spell out. Read: abstract only (PubMed efetch, 2026-09-20).

    Evidence that argues against this (1)
    • Stage-selective nuclear staining can equally reflect stage-selective EXPRESSION rather than stage-selective localisation; the abstract does not separate them, and without a total-protein comparison across hair-cycle phases the localisation reading is not secure.
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    PBX1 is mostly cytoplasmic in epithelial endometrial cells of the developing female genital tract despite nuclear localisation of MEIS1 in the same tissue, demonstrating control of PBX1 intracellular distribution that is independent of MEINOX proteins and is cell-cycle-correlated.

    PBX1 intracellular localization is independent of MEIS1 in epithelial cells of the developing female genital tract. · Dintilhac A, Bihan R, Guerrier D, Deschamps S, Bougerie H, Watrin T, Bonnec G, Pellerin I · Int J Dev Biol 2005
    PMID 16172981DOI
    What this evidence cannot show

    RULE 73: developing female genital tract, an organogenesis system with no relation to haematopoiesis. The result concerns PBX1 rather than PBX3, which is the PBC-class member this graph carries; PBC paralogues are not interchangeable for localisation. The observation is immunohistochemical co-staining in tissue sections, so 'mostly cytoplasmic' is a qualitative description. Read: abstract only (PubMed efetch, 2026-09-20).

    Evidence that argues against this (1)
    • The MEINOX-retention model is well supported in other systems, including the limb-bud and HOX-trimer literature this graph's homeodomain entries rest on, so this is an exception rather than a refutation. Its value is that it forbids treating 'MEIS1 is nuclear' as sufficient for 'the HOX/PBX/MEIS trimer is nuclear', which is an inference the graph makes implicitly.
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    Bounded absence, measured 2026-09-20: no study located reports MEIS1 protein subcellular or subnuclear localisation in a KMT2A-rearranged cell of any lineage or species.

    What this evidence cannot show

    QUERY: `MEIS1[tiab] AND ("nuclear localization"[tiab] OR subcellular[tiab] OR immunofluorescence[tiab] OR "nuclear import"[tiab])` = 23 records on 2026-09-20, all 20 retrieved titles checked; the systems are hair follicle, cerebellum, limb bud, genital tract, cardiomyocyte, pulmonary artery smooth muscle and ovary, plus one t(4;11) paper (PMID 33924850, nuclear FGFR2 interacting with MLL-AF4) that concerns FGFR2 and not MEIS1 localisation. RULE 67 CONTROL: the query returns 23 through the identical shape, so the instrument returns ones. NAME LIST: MEIS1, Meis1; the previous/alternate designations MEIS and HOXA9-cofactor phrasing were not searched, and the murine gene name Mrg1 (a genuine older alias in some literature) was NOT searched - this is a known weakness of the bound. RULE 80 (added mid-beat): this count is bounded by TITLES AND ABSTRACTS, not by papers. Localisation panels are very often supporting figures whose abstracts never mention them, so a PubMed zero here means 'no paper ANNOUNCES this measurement', not 'no such figure exists'. Supporting figures and supplementary material were NOT searched.

    Evidence that argues against this (1)
    • MEIS1 occupancy in KMT2A-rearranged cells is measured by ChIP-seq, which constrains where MEIS1 is on the genome without constraining where it is in the cell. As with DOT1L and PSIP1 in this beat, the missing measurement is an image, not information - and the specific risk here is isoform summation rather than compartment ignorance.
Connected to