Oncofusion

start / graph / gene-spi1

Proteinmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.king

SPI1 / PU.1 — a myeloid dependency that a KMT2A panel mistakes for its own

also called SPI1

PU.1 is one of the master switches that makes a blood cell myeloid. Cells that have taken that path need it. Our disease's cell lines are mostly myeloid, so PU.1 looked like a dependency of the disease when it is a dependency of the cell type.

The same thing, for a scientist

SPI1 ranks 43 of 17,728 in the genome-wide KMT2A-selectivity scan, which is what put it on the candidate list. The rank is an artefact of panel composition: SPI1 scores **-1.82 SD** below the other-leukaemia arm in KMT2A-rearranged MYELOID lines and **+0.92 SD** in KMT2A-rearranged B-lineage lines, and on a line-normalised lineage-consistency statistic it sits at the **95.52nd percentile** of random genes — the opposite tail. It is a myeloid-identity dependency. That does not make it unimportant: PU.1-dependent enhancers are the mechanism through which LSD1 inhibition acts in AML.

Go deeper
Where PU.1 does matter here

LSD1 inhibition in AML works by recommissioning PU.1- and C/EBP-alpha-dependent enhancers, which places these factors on the causal path of a drug class this graph already tracks under target-lsd1-kdm1a-kmt2a-r and which has been taken into KMT2A-rearranged patients.

So the correct reading is not that PU.1 is irrelevant to KMT2A-rearranged leukaemia. It is that PU.1 belongs to the myeloid programme the fusion operates within, not to the fusion. For KMT2A-rearranged AML — a large share of the disease — that distinction does not reduce its therapeutic interest. For infant KMT2A::AFF1 B-ALL, which is what this project is principally about, it does.

Facts and figures
gene symbolSPI1
depmap scan rank of 1772843rd
depmap scan difference-0.385
lineage consistency percentile95.52
verdictmyeloid-identity dependency, not KMT2A-specific
uniprotP17947
length AA270
gene primarySPI1
accession addedverified against live UniProt 2026-09-20; rules 66/68/69 require accession AND length_aa so verify-accessions.py can fail a swap
subnuclear localizationNONE ASSIGNED - A BOUNDED ABSENCE, and the near-miss is more informative than the absence. PU.1[tiab] against nucleolus, nucleolar, nuclear speckle(s), nuclear matrix and PML body/bodies returns 10 records; all ten titles were read and NONE places PU.1 in a named subnuclear compartment [E_q91_spi1_loc4]. The closest result in that set inverts the question: a critical tyrosine in the NUCLEAR MATRIX TARGETING SIGNAL of RUNX1 is required for subnuclear targeting and for synergistic activation of myeloid promoters, and PU.1 is the partner whose cooperativity is being tested - so PU.1's transcriptional output has been shown to depend on ANOTHER factor's subnuclear address while PU.1's own has never been measured [E_q91_spi1_loc3]. Note also what must NOT be filed here: PU.1 'decompacts stem cell heterochromatin' as a pioneer factor [E_q91_spi1_loc1]. That is a chromatin-state claim about the DNA, not a compartment claim about the protein, and the two are routinely conflated.
localization shown for fusionNO for localisation, in any fusion, in any lineage - nobody has imaged PU.1 protein in a fusion-bearing cell. What exists in a fusion context is functional and occupancy-level, not compartmental: PU.1[tiab] AND MLL[tiab] returns 13 records and "PU.1"[tiab] AND KMT2A[tiab] returns 3 [E_q91_spi1_loc4], and the entry already carries the LSD1/PU.1-enhancer result in MLL-AF9 AML (its E3). Note the alias asymmetry, which is a retrieval fact worth keeping: the PROTEIN NAME outperforms the current symbol threefold - "PU.1"[tiab] AND MLL[tiab] = 13 against SPI1[tiab] AND MLL[tiab] = 4.
z vs other leukaemia
{
  "kmt2a_myeloid": -1.82,
  "kmt2a_b_lineage": 0.92
}
depmap kmt2a selectivity
{
  "chronos_mean_kmt2a_rearranged": -0.8317,
  "chronos_mean_other_leukaemia": -0.4467,
  "difference": -0.385,
  "pan_essentiality_all_lines": -0.0905,
  "selectivity_rank": 43,
  "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
{
  "CONSTITUTIVE": "NUCLEUS, and the record is thinner than the protein's importance implies. UniProt P17947 (fetched live 2026-09-21) carries exactly ONE subcellular location value, 'Nucleus', supported by an ECO:0000255 PROSITE rule (PRU00237, the ETS-domain profile - a sequence rule, not a measurement) plus ONE ECO:0000269, PubMed:33951726, a study of six agammaglobulinemia patients carrying heterozygous SPI1 mutations [E_q91_spi1_loc1]. There is no annotated NLS and no molecule- or isoform-scoped comment. For the master transcription factor of myeloid and B-lymphoid fate, that is a one-paper experimental record.",
  "no_induced_relocalisation_recorded": "None in UniProt and none retrieved. PU.1 activity is regulated by DOSE and by partner choice rather than by leaving the nucleus [E_q91_spi1_loc2]."
}
cell type
{
  "HUMAN_LOSS_OF_FUNCTION_PHENOTYPE_IS_THE_SPECIFIC_ANSWER": "B CELLS AND CONVENTIONAL DENDRITIC CELLS. Six unrelated patients with heterozygous SPI1 mutations - four de novo - present with AGAMMAGLOBULINAEMIA: they LACK CIRCULATING B CELLS and possess FEW CONVENTIONAL DENDRITIC CELLS, and introducing disease-similar SPI1 mutations into human haematopoietic stem and progenitor cells impaired early in vitro lymphopoiesis [E_q91_spi1_loc1]. This is a human, germline, allele-resolved cell-type statement and it is much more specific than 'haematopoietic'.",
  "WHERE_IT_IS_REQUIRED_IN_THIS_DISEASE": "The myeloid half only, and that is this entry's founding point: SPI1 scores -1.82 SD in KMT2A-rearranged MYELOID lines and +0.92 SD in KMT2A-rearranged B-LINEAGE lines (this entry's E1). A dependency that reverses sign between the two halves of the index disease is a lineage property.",
  "canonical_lineages": "Granulocyte-macrophage and monocytic output, B lymphopoiesis, and conventional dendritic cells; the graded-dose mouse work below adds the abnormal myeloid precursor pool [E_q91_spi1_loc2].",
  "not_localised_in_any_of_them": "No image of PU.1 protein in a KMT2A-rearranged blast of either lineage was retrieved."
}
cell state
{
  "DOSE_IS_THE_STATE_VARIABLE": "PU.1 is the textbook case of a transcription factor whose CONCENTRATION, not its presence, sets the state. Mice carrying hypomorphic Sfpi1 alleles that reduce PU.1 to 20% OF NORMAL developed AML, while homozygous AND heterozygous deletions did not; 80% loss produced a PRECANCEROUS state characterised by accumulation of an abnormal precursor pool that retained responsiveness to G-CSF with disruption of the M-CSF and GM-CSF pathways [E_q91_spi1_loc2]. Complete loss and half loss are both non-leukaemogenic and an intermediate dose is transforming - a non-monotonic dose-state relation that no binary knockout would find.",
  "ENHANCER_STATE_UNDER_THERAPY": "LSD1 inhibition acts in AML by RECOMMISSIONING PU.1- and C/EBPalpha-dependent enhancers (this entry's E3) - i.e. the therapeutically relevant variable is the occupancy state of PU.1 sites, again not the protein's compartment.",
  "system_note": "The graded-dose result is MURINE. The agammaglobulinaemia result is human germline. Neither is a KMT2A-rearranged system."
}
localization bounded absence
{
  "date": "2026-09-22",
  "tool": "NCBI esearch db=pubmed",
  "run_by": "Q91",
  "queries": [
    {
      "query": "\"PU.1\"[tiab] AND (nucleolus[tiab] OR nucleolar[tiab] OR \"nuclear speckle\"[tiab] OR \"nuclear speckles\"[tiab] OR \"nuclear matrix\"[tiab] OR \"PML body\"[tiab] OR \"PML bodies\"[tiab])",
      "querytranslation": "\"PU.1\"[Title/Abstract] AND (\"nucleolus\"[Title/Abstract] OR \"nucleolar\"[Title/Abstract] OR \"nuclear speckle\"[Title/Abstract] OR \"nuclear speckles\"[Title/Abstract] OR \"nuclear matrix\"[Title/Abstract] OR \"PML body\"[Title/Abstract] OR \"PML bodies\"[Title/Abstract])",
      "count": 10,
      "titles_read": 10,
      "on_topic": 0,
      "note": "All ten read. Two are HTLV-1 latency reviews, one a thalassaemia miRNA study, one a pluripotency gain-of-function screen, one a zebrafish NPMc+ paper, one a 1993 marker-chromosome FISH paper, one a TLX3/BCL11B enhancer study, one a macrophage AIF-1 promoter study, one a nucleosome-repositioning study in a myeloid line (chromatin, not compartment), and one - PMID 16149049 - is about RUNX1's nuclear matrix targeting signal with PU.1 as the cooperating factor. Zero place PU.1 in a compartment."
    },
    {
      "query": "SPI1[tiab] AND MLL[tiab]",
      "querytranslation": "\"SPI1\"[Title/Abstract] AND \"MLL\"[Title/Abstract]",
      "count": 4
    },
    {
      "query": "\"PU.1\"[tiab] AND MLL[tiab]",
      "querytranslation": "\"PU.1\"[Title/Abstract] AND \"MLL\"[Title/Abstract]",
      "count": 13,
      "note": "the protein name returns 3.25x what the HGNC symbol returns - an alias-recall demonstration on a gene with no retired symbol at all"
    },
    {
      "query": "\"PU.1\"[tiab] AND KMT2A[tiab]",
      "querytranslation": "\"PU.1\"[Title/Abstract] AND \"KMT2A\"[Title/Abstract]",
      "count": 3
    }
  ],
  "reading": "The subnuclear absence is real and is bounded by titles and abstracts only (rule 80). Both the symbol and the protein name were run; SFPI1/Sfpi1 (the murine symbol) and SPI-1 were not separately enumerated, and the murine literature is where the dose work lives.",
  "absence_class": "never-measured"
}
How this could bite us back
  • This entry exists because a genome-wide rank was mistaken for biology. It is recorded as a myeloid-identity dependency so the rank cannot be cited again as KMT2A evidence.
  • The therapeutic implication runs opposite to the usual reading of a high dependency rank: PU.1 is needed for an LSD1 inhibitor to work, so it is a required cofactor of a therapy rather than a target to remove.
Still open
  1. Does SPI1 dependency reappear in KMT2A-rearranged B-ALL under any condition, or is the +0.92 SD a stable property of the lineage? Five lines cannot settle it.
  2. If LSD1 inhibitors work by recommissioning PU.1-dependent enhancers, should their use be restricted to the myeloid half of KMT2A-rearranged disease? No trial has stratified that way.
Every source behind this entry (7)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.analysis

    In the genome-wide DepMap scan contrasting KMT2A-rearranged against other leukaemia lines, SPI1 ranks 43rd of 17,728 with a difference of -0.385. Splitting the index arm by lineage dissolves it: SPI1 scores -1.82 SD below the other-leukaemia arm in KMT2A-rearranged MYELOID lines and 0.92 SD in KMT2A-rearranged B-lineage lines, i.e. no dependency at all in KMT2A-rearranged B-ALL.

    Lineage-split test of myeloid transcription-factor dependencies in KMT2A-rearranged leukaemia · Oncofusion knowledge graph (internal analysis); DepMap Chronos via Open Targets 2026
    What this evidence cannot show

    Computed by the coordinator on 2026-09-19. Five B-lineage and eight myeloid lines, so each half is small and one line moves a mean. Lineage is assigned from the DepMap disease label rather than an expression signature, and KMT2A-rearranged disease includes genuinely ambiguous-lineage cases. Chronos measures knockout.

    Evidence that argues against this (2)
    • A genome-wide rank from that scan is not evidence about KMT2A, because the scan's index arm is myeloid-enriched (8 AML lines against 5 B-ALL). A myeloid-identity gene rises in that ranking for reasons that have nothing to do with the fusion, which is precisely what the lineage split demonstrates for this gene. The rank is retained here as the observation that needed explaining, not as support. Oncofusion knowledge graph (internal analysis); DepMap Chronos via Open Targets 2026
    • The B-lineage half is five lines and two of them are the field's most heavily passaged KMT2A::AFF1 models, so the absence of dependency rests on a small and culture-adapted set. A dependency that is real but weak in B-lineage disease would be hard to distinguish from zero here. Oncofusion knowledge graph (internal analysis); DepMap Chronos via Open Targets 2026
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.analysis

    Scored against 647 random protein-coding genes on a line-normalised statistic that takes the WEAKER lineage half - max(z_myeloid, z_lymphoid), so a one-sided effect cannot rank - SPI1 falls at the 95.52th percentile. MEN1 and PSIP1 fall at the 0.00th and the pan-essential control ACTB at the 43.59th.

    Lineage-consistency null test for KMT2A dependencies: line-normalised, random-gene percentile · Oncofusion knowledge graph (internal analysis); DepMap Chronos via Open Targets 2026
    What this evidence cannot show

    The null is random protein-coding genes, most not expressed in leukaemia, so percentiles are flattered for genuinely leukaemia-relevant genes and this one still does not rank. The statistic is capped by a four-line B-lineage half by construction.

    Evidence that argues against this (1)
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    LSD1 inhibition exerts its antileukemic effect in AML by recommissioning PU.1- and C/EBP-alpha-dependent enhancers, placing SPI1 on the mechanism of action of an epigenetic drug class that has been tested in KMT2A-rearranged disease.

    PMID 29453291DOI
    What this evidence cannot show

    Established in AML models. Whether the same enhancer-recommissioning mechanism operates in KMT2A-rearranged B-lymphoid disease, where PU.1 shows no DepMap dependency, is not addressed by this work.

    Evidence that argues against this (2)
    • A mechanism of action is not a dependency. LSD1 inhibitors acting THROUGH PU.1-dependent enhancers shows PU.1 is required for the drug's effect, which is close to the opposite of PU.1 being a target — the therapy needs PU.1 present and functional in order to work. Reading this entry as support for inhibiting SPI1 would invert it. Blood 2018
    • The clinical record for this class is thin. The first-in-human LSD1 study enrolled 27 patients with relapsed/refractory AML in a cohort enriched for MLL/KMT2A rearrangement, which is a dose-finding result rather than evidence that the enhancer mechanism translates. Nature Genetics 2015
  4. E_q91_spi1_loc1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProtKB P17947 (SPI1_HUMAN, primary gene SPI1, 270 aa, read back live 2026-09-21) carries exactly ONE subcellular location value - 'Nucleus' - with two attributions: ECO:0000255 to the PROSITE rule PRU00237 (the ETS-domain profile, a SEQUENCE rule rather than an experiment) and ECO:0000269 to PubMed:33951726. There is no location note, no annotated nuclear localisation signal, and no isoform- or fragment-scoped comment. The cited experimental paper studied six unrelated agammaglobulinaemic patients each carrying a heterozygous SPI1 mutation, four of them de novo; the patients LACKED CIRCULATING B CELLS and possessed FEW CONVENTIONAL DENDRITIC CELLS, and the paper frames PU.1 as a pioneer factor that controls haematopoietic cell fate by DECOMPACTING STEM CELL HETEROCHROMATIN so that non-pioneer factors can enter otherwise inaccessible sites.

    What this evidence cannot show

    A one-paper experimental record for the master haematopoietic pioneer factor. The paper is a disease-genetics study whose localisation content, where it exists, concerns mutant protein behaviour rather than the compartment of the wild type; this beat read its abstract, not its figures, so whether it contains the microscopy UniProt is pointing at cannot be settled from here.

    Evidence that argues against this (1)
    • The strong reading - 'PU.1 is nuclear, established' - is correct as biology and badly supported as a record, and the two should not be confused. Strip the PROSITE rule, which is an inference from having an ETS domain and would return 'Nucleus' for any ETS protein including ones with cytoplasmic pools, and one experimental citation remains. Compare SBDS in this same tranche, a far less studied protein, whose compartment is pinned by two dedicated imaging studies including one in primary human myeloid progenitors. Importance in the literature and strength of the localisation record are uncorrelated here, which is exactly the pattern rule 89 found at the top of this graph's degree ranking. Orelio C et al. 2009
  5. E_q91_spi1_loc2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In mice, GRADED REDUCTION of PU.1 is leukaemogenic where complete or half loss is not: animals carrying hypomorphic Sfpi1 alleles that reduce PU.1 expression to 20% of normal levels developed acute myeloid leukaemia, UNLIKE mice carrying homozygous or heterozygous deletions of Sfpi1. An 80% loss induced a precancerous state characterised by accumulation of an abnormal precursor pool that retained responsiveness to G-CSF, with disruption of the M-CSF and GM-CSF pathways.

    Acute myeloid leukemia induced by graded reduction of a lineage-specific transcription factor, PU.1. · Rosenbauer F, Wagner K, Kutok JL, Iwasaki H et al. · Nat Genet 2004
    PMID 15146183DOI
    What this evidence cannot show

    MURINE, myeloid, and not a KMT2A-rearranged system - rule 73 applies and the first clause of the claim says so. The dose figures are relative to wild type in that allelic series and do not transfer numerically to human blasts.

    Evidence that argues against this (1)
    • This result argues directly against the way PU.1 arrived in this graph. The entry's own founding observation is a DepMap CRISPR rank, and CRISPR knockout is a complete-loss perturbation - precisely the perturbation this paper shows is NOT the transforming one. A screen that can only ask 'is the gene required' cannot see a phenotype that appears at 20% of normal and disappears at 0% and at 50%. So the DepMap rank and the graded-dose biology are measuring different things, and the entry's conclusion that the rank is a lineage artefact survives, but for a second independent reason the entry did not state: the instrument is blind to PU.1's actual disease-relevant dose window. Le Coz C et al. 2021
  6. E_q91_spi1_loc3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    The only subnuclear result retrieved in which PU.1 participates is about its PARTNER's address, not its own: point mutations in the nuclear matrix targeting signal (NMTS) of human RUNX1 that leave nuclear import, DNA binding and C/EBPalpha-dependent synergy intact nonetheless abolish subnuclear targeting, and a critical tyrosine in the NMTS is required both for subnuclear targeting and for activation of myeloid-specific promoters; PU.1 is one of the two lineage factors whose transcriptional cooperativity with RUNX1 is being assayed.

    PMID 16149049DOI
    What this evidence cannot show

    A transfection/reporter study on the M-CSF receptor and granulocyte promoters; PU.1's own distribution is not imaged anywhere in it. Read from the abstract only. It is recorded here because it is the nearest thing in the retrievable literature to a subnuclear statement involving PU.1, and because it sets up an experiment nobody has run.

    Evidence that argues against this (1)
    • The obvious inference - that PU.1 must therefore share RUNX1's nuclear-matrix compartment - does not follow and the paper gives the reason. The NMTS mutants retained C/EBPalpha-dependent synergy while losing the matrix-targeted activity, so subnuclear targeting is separable from partner-dependent activation in the same protein at the same promoters. Cooperativity between two factors is not evidence that they occupy one compartment; it is evidence that they act on one promoter. Recording 'PU.1 is at the nuclear matrix' from this would be the same over-reach as recording a compartment for SUMO1 from a paper about one of its substrates. Rosenbauer F et al. 2004
  7. E_q91_spi1_loc4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    Counted absence, 2026-09-22, querytranslation read back on every line. PU.1 has no published subnuclear compartment: "PU.1"[tiab] against a six-term compartment list returns 10 records, all ten titles read, 0 on topic. PU.1 has never been localised in a KMT2A-rearranged cell: "PU.1"[tiab] AND MLL[tiab] = 13, SPI1[tiab] AND MLL[tiab] = 4, "PU.1"[tiab] AND KMT2A[tiab] = 3, and none of those is a localisation study.

    What this evidence cannot show

    Bounded by titles and abstracts (rule 80). The compartment list covered nucleolus, nucleolar, nuclear speckle(s), nuclear matrix and PML body/bodies; it did NOT cover Cajal bodies, paraspeckles, chromocentres, transcription factories or the nuclear lamina, and PU.1's pioneer activity at heterochromatin makes chromocentres a live unchecked possibility. The murine symbol Sfpi1 was not run.

    Evidence that argues against this (1)
    • The 13-versus-4 split between "PU.1" and SPI1 is the load-bearing part of this measurement and it argues against how the rest of this corpus bounds absences. This gene has NO retired HGNC symbol - the trap the brief warns about for SEPT2 and WDR61 does not apply - and the current symbol still under-retrieves by a factor of 3.25, simply because the field writes the protein name. A counted absence for any protein with a widely used trivial name is therefore not bounded by the symbol alone, and this graph's absence claims are overwhelmingly symbol-keyed. Cusan M et al. 2018
Connected to