Oncofusion

start / graph / gene-spp1

Genemoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q33-niche-protein-layer

SPP1 (osteopontin) - the endosteal matrix protein whose knockdown prolongs survival in MLL-AF9 mice, and the one KMT2A::AFF1-specific paper about it cannot be read

also called osteopontin · OPN · BNSP · bone sialoprotein 1 · ETA-1

Osteopontin is a sticky protein in the bone lining that leukaemia cells make for themselves as well as meet in the niche. Knocking it down made mice with MLL-AF9 leukaemia live longer, and the single paper that measured it in the exact infant leukaemia this project studies has no abstract and cannot be read.

The same thing, for a scientist

SPP1/osteopontin (UniProt P10451, 314 aa, secreted, RGD cell-attachment site at 159-161) is the canonical endosteal matricellular niche protein. It is in this graph because of a specific unfinished task: `ana-marrow-niche-rejected-candidates` deferred SPP1 in cycle 40 'pending full text', because the one comparator-controlled KMT2A::AFF1 record is indexed as a Letter with no abstract. **That deferral is now resolved in both halves.** The unreadable paper is confirmed unreadable by all available routes [E3], and is recorded as a title-only claim at weak confidence per this project's known-but-unreadable standard. But the deferral's premise - that SPP1 needed that paper to justify an entry - was wrong: a separate record in the same intersection is an MLL-AF9 in vivo experiment. Knockdown of *Opn* significantly prolonged survival of mice with MLL-AF9-induced AML and markedly reduced tumour burden; LSCs from knockdown groups were fewer and functionally impaired by immunophenotype, colony formation and limiting dilution; OPN prevents LSCs from undergoing apoptosis and cell-cycle arrest; and repression of OPN in human AML lines reproduced the phenotype [E2]. Notably the OPN in that experiment is **autocrine, made by the leukaemic cell**, not supplied by the endosteum.

What it looks like
SPP1314 aa
SignSignal peptide · residues 1–16Osteopontin (mature chain)Osteopontin (mature chain) · residues 17–314Disordered regionDisordered region · residues 41–290RGD cell attachment site · residues 159–1610100200300

No three-dimensional model is available for this protein, so the map above is the whole picture. That is itself informative: AlphaFold does not model chains this long, and the parts that have been crystallised are held under the entries for the individual domains.

Go deeper
Closing the cycle-40 deferral, and why it was deferred on the wrong criterion

ana-marrow-niche-rejected-candidates recorded SPP1 as 'deferred pending full text' with the reasoning that the one genuinely KMT2A-specific, comparator-controlled record - 'Osteopontin-c is overexpressed in KMT2A-AFF1 positive pediatric B-cell lymphoblastic leukemia when compared to those with ETV6-RUNX1' - is a Letter indexed with no abstract, and that this project requires a citation to be read.

Both halves checked here. First, the paper is genuinely unreadable: efetch returns no AbstractText, and the Europe PMC core record returns hitCount 1 with pmcid absent, so there is no fullTextXML route to try - the two-route test this project requires before writing 'unreadable' has been run and both routes failed 3. It is recorded as a title-only claim at weak confidence, and nothing is written about what the paper probably says.

Second, and more useful: the deferral rested on the wrong record. The same four-record intersection contains an MLL-AF9 loss-of-function experiment in vivo 2, which needs no full text because it has an abstract and which is stronger evidence for an entry than a title. The general lesson is the one in this project's standing rules about gap-markers: a deferral names a blocker, and the blocker should be re-read before the deferral is inherited, because the deferral can be true and irrelevant at the same time.

The direction of the biology is not the niche one

Osteopontin is filed in the niche literature as an endosteal matricellular constraint on HSC pool size. The MLL-AF9 result is not that: the abstract is explicit that the question was 'AML cell autocrined OPN', and the perturbation was knockdown in the leukaemic cell 2. So the mechanism on the table for KMT2A-fusion disease is a leukaemia-cell-autonomous secreted factor that keeps LSCs out of apoptosis and cell-cycle arrest, which is a different intervention target from a stromal matrix protein. This distinction is the same one gene-vcam1 records for the counter-receptor problem and gene-cxcl12 records for the ligand-source problem: for every secreted or surface molecule in this layer, which cell makes it is the load-bearing unknown, and it is unresolved in nearly every case.

Facts and figures
gene symbolSPP1
uniprotP10451
length AA314
accession resolutiongene_exact:SPP1 AND organism_id:9606 AND reviewed:true returned TWO reviewed human records: P10451 (primary gene SPP1, osteopontin, 314 aa) FIRST and Q9BX95 (primary gene SGPP1, 441 aa) second. The correct record is first today, so the obvious query is safe today; the rule-62 multiple-record condition nonetheless obtains and ordering should be re-checked rather than assumed. Primary gene name and length read back before selection. Fetched 2026-09-22.
localizationSECRETED, with three ECO:0000269 experimental supports on UniProt P10451 and no other compartment on the record. Functionally it is a matricellular protein: secreted, then bound into the extracellular matrix, so like CXCL12 its active pool is immobilised rather than freely diffusible. 94% of the mature chain (41-290) is intrinsically DISORDERED, which is why osteopontin has no domain map to speak of and why its activity is set by post-translational modification and proteolytic cleavage rather than by folded modules.
subnuclear localizationNOT APPLICABLE BY TOPOLOGY, argued from architecture. A 314-residue secreted, largely disordered matricellular protein with a cleaved signal peptide and no chromatin-binding module. This field does NOT claim nuclear osteopontin was sought and excluded; no such search was run in this beat and a curated record's silence is not evidence.
subnuclear absence classnot-applicable-by-topology
cell typeTWO CANDIDATE SOURCES AND THE EVIDENCE POINTS AT THE UNEXPECTED ONE. The niche-literature source is the OSTEOBLAST and the endosteal matrix, and this graph holds cell-type-osteoblast and concept-endosteal-niche for it. But the only KMT2A-fusion experiment available asked about AML-cell-AUTOCRINE OPN and knocked it down in the leukaemic cell [E2], so the compartment with evidence in this disease is the BLAST itself. No experiment in a KMT2A-rearranged system has tested stromal or osteoblastic OPN.
cell stateAN APOPTOSIS-RESISTANT, CELL-CYCLE-COMPETENT LEUKAEMIA STEM CELL STATE: Opn knockdown reduced LSC number and function by immunophenotype, colony formation and limiting dilution, and the stated mechanism is that OPN prevents LSCs from undergoing apoptosis and cell-cycle arrest [E2]. Note this is the opposite polarity from the classical niche account, in which endosteal osteopontin CONSTRAINS stem-cell pool size and promotes quiescence.
epigenetic readoutNOT MEASURED. No histone-mark, DNA-methylation or menin-occupancy measurement at SPP1 in a KMT2A-rearranged system was found. The searchable literature is very small - (SPP1[tiab] OR osteopontin[tiab]) intersected with the KMT2A/MLL half returns 4 records in total (querytranslation read back, 2026-09-22), all four accounted for in this entry and its sources - so there is little for a chromatin measurement to be hiding in and the bound is correspondingly informative. Bounded over titles and abstracts (rule 80). ABSENCE CLASS: never-measured. One caveat specific to this gene: the KMT2A::AFF1 record concerns the osteopontin-c SPLICE VARIANT [E4], and a splice-variant question is an exon-usage question that chromatin data at the promoter would not answer.
shown for fusionYES for murine MLL-AF9, as a loss-of-function survival benefit with an LSC phenotype, and the OPN in question is autocrine from the leukaemic cell [E2]. TITLE-ONLY for KMT2A::AFF1: one Letter asserts osteopontin-c overexpression in KMT2A-AFF1 versus ETV6-RUNX1 paediatric B-ALL and cannot be read [E3, E4]. NO for any readable KMT2A::AFF1 measurement, NO for any infant cohort, and NO for stromal or osteoblastic OPN in any KMT2A-rearranged system.
intact union membershipABSENT from the 277-partner IntAct REST union of the six core baits (analyses/interactome-union-migrated-to-rest/results.json, generated 2026-09-22), checked on the symbol and on P10451.
closes deferralana-marrow-niche-rejected-candidates recorded SPP1 as 'deferred pending full text' in cycle 40. This entry closes that deferral: the blocking paper is confirmed unreadable by both required routes [E3] and is recorded title-only at weak confidence, and an independent readable MLL-AF9 in vivo result [E2] justifies the entry without it. The earlier beat's niche-relevance judgement is not overturned - E2's OPN is autocrine and leukaemia-cell-derived, so on a strict structural-niche criterion the earlier disposition was defensible.
collision noteNo row in data/symbol-collisions.json for SPP1, osteopontin, SGPP1 or P10451 (registry read 2026-09-22, case-sensitive over d['collisions'], 177 rows). Two cautions measured here: gene_exact:SPP1 returns two reviewed human records (SGPP1 carries SPP1 as a synonym) with the correct one first TODAY; and 'osteopontin' is a better PubMed token than 'SPP1' for this protein's older literature, so a symbol-only query under-counts.
domains
Signal peptide1.000Signal peptide: 1Osteopontin (mature chain)17.0Osteopontin (mature chain): 17Disordered region41.0Disordered region: 41RGD cell attachment site159.0RGD cell attachment site: 159
domains — the numbers behind the chart
[
  {
    "name": "Signal peptide",
    "start": 1,
    "end": 16
  },
  {
    "name": "Osteopontin (mature chain)",
    "start": 17,
    "end": 314
  },
  {
    "name": "Disordered region",
    "start": 41,
    "end": 290
  },
  {
    "name": "RGD cell attachment site",
    "start": 159,
    "end": 161
  }
]
depmap essentiality
{
  "depmap_common_essential": false,
  "tissue_blocks": 29,
  "n_screens": 1258,
  "mean_gene_effect": 0.0782,
  "min_gene_effect": -0.4486,
  "max_gene_effect": 0.4264,
  "n_lines_at_or_below_minus_1": 0,
  "open_targets_data_version": "year 26, month 06",
  "open_targets_api_version": "26.6.3",
  "fetched": "2026-09-22",
  "control": "Live zero-screen control in the same call and code path: MEN1 returned 29 tissue blocks, 1,258 screens, mean gene effect -0.3616, minimum -2.5262, 79 lines at or below -1.0 - so a zero count of dependent lines is a measured zero, not an empty payload with a passing approvedSymbol readback. DepMap release recorded as Open Targets dataVersion year 26 month 06 (rule 83), API 26.6.3, fetched 2026-09-22. No median-split proportion is reported (rule 84). AND THE NULL IS AN ABSENCE BY CONSTRUCTION: DepMap screens cells in monoculture with no stroma and no endothelium, so a niche factor acting in trans cannot score however essential it is in vivo."
}
How this could bite us back
  • Osteopontin is a wound-healing, bone-remodelling and immune-regulatory protein; systemic blockade in a growing infant touches bone turnover and macrophage biology at once, and no agent against it has paediatric exposure.
  • If the operative pool is autocrine and intracellularly routed to the leukaemic cell's own surface, a circulating neutralising antibody may never reach the relevant concentration at the relevant place - a delivery failure that would look like the mechanism being wrong.
  • The classical niche account has osteopontin CONSTRAINING stem-cell pool size. An agent that removes it could in principle expand the normal HSC pool and the leukaemic one together, and the direction in a KMT2A-rearranged patient is not established by any experiment.
Still open
  1. Is the MLL-AF9 OPN requirement autocrine only, or does endosteal osteopontin contribute? The knockdown in E2 was in the leukaemic cell and the stromal arm has never been tested in a KMT2A-fusion model.
  2. What does PMID 32114371 actually show? A library request or an author contact would close a title-only citation in the index fusion - the cheapest unresolved item in this beat.
  3. Why is SPP1 flat in DepMap if the requirement is autocrine? Unlike the other genes in this tranche the monoculture objection does not obviously apply, and the tension is unresolved.
Every source behind this entry (5)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    The reviewed human osteopontin record is UniProt P10451 (OSTP_HUMAN), 314 aa, secreted, with signal peptide 1-16, mature chain 17-314, a long disordered region 41-290 and an RGD cell-attachment site at 159-161; a gene_exact:SPP1 query returns two reviewed human records, P10451 and Q9BX95 (SGPP1, 441 aa).

    What this evidence cannot show

    Resolved by gene_exact:SPP1 AND organism_id:9606 AND reviewed:true, which returned TWO reviewed human records - P10451 (primary gene SPP1, 314 aa) and Q9BX95 (primary gene SGPP1, 441 aa, which carries SPP1 as a synonym). The correct record was returned first on 2026-09-22, so the query is safe today, but the multiple-record condition of rule 62 obtains and ordering is not guaranteed to persist. The primary gene name and length were read back before selection. Note also that osteopontin is heavily and variably phosphorylated and glycosylated and is cleaved by thrombin and MMPs, so 'osteopontin' names a family of species; the splice variant in E4 (osteopontin-c) is one of them.

    Evidence that argues against this (1)
    • A curated feature table records what curators accepted, not what was sought and excluded, so no absence claim follows from it. The worked example is in this beat: UniProt P61073 annotates no nuclear localisation signal for CXCR4 while a mapped, mutated, functionally tested NLS at 146-RPRK-149 is published. Bao Y et al. 2019
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Knockdown of Opn significantly prolonged the survival of mice with MLL-AF9 cell-induced AML and markedly reduced tumour burden; LSCs from Opn-knockdown groups showed decreased numbers and impaired function by immunophenotype, colony-forming and limiting dilution assays; Opn prevents LSCs from undergoing apoptosis and cell-cycle arrest; and repression of OPN in human AML cell lines reproduced the phenotype. The question addressed was AML-cell-autocrined OPN.

    Osteopontin is required for the maintenance of leukemia stem cells in acute myeloid leukemia · Zhou J, Chen X, Zhou P, et al. · Biochem Biophys Res Commun 2022
    PMID 35182972DOI
    What this evidence cannot show

    Murine myeloid MLL-AF9 and human AML lines, not infant B-lineage KMT2A::AFF1 (rule 73). The perturbation is knockdown in the leukaemic cell, so this establishes an autocrine requirement and says nothing about endosteal or stromal OPN in this disease. Effect sizes, numbers of mice and the identity of the human lines are not in the abstract, which is all that was read (rule 80). Biochem Biophys Res Commun is a venue with brief peer review and the result has not, as far as this beat found, been independently replicated in a KMT2A-fusion model.

    Evidence that argues against this (1)
    • The cycle-40 marrow-niche beat searched the same intersection and judged osteopontin 'rejected as not niche-relevant / deferred', on the basis that the four records contained only one KMT2A-specific comparator-controlled study and that one was unreadable. That judgement and this entry disagree about the same four records, and the disagreement is recorded rather than quietly overwritten: the earlier beat was applying a niche-relevance criterion, under which an autocrine leukaemic-cell factor genuinely is out of scope. da Fonseca Bastos ACS et al. 2020
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The one KMT2A::AFF1-specific osteopontin record, PMID 32114371, is confirmed unreadable by the routes available here: PubMed efetch returns no AbstractText and the Europe PMC core record returns hitCount 1 with no pmcid, so no fullTextXML route exists.

    PMID 32114371DOI
    What this evidence cannot show

    Checked 2026-09-22 by efetch (publication type Letter, no AbstractText element) and by Europe PMC search with resultType=core, where result[0].pmcid is absent. The publisher route via the DOI was NOT attempted and the article is in a subscription Elsevier journal, so 'unreadable' here means unreadable by the two routes this project requires before the label is applied, not unobtainable by a library request. Per the known-but-unreadable standard, the title's assertion is recorded and nothing is written about what the paper probably shows.

    Evidence that argues against this (1)
    • This project has twice found papers labelled unreadable on an empty abstract that were in fact open-access full text in PMC, because the check requested the field that says so and never read it. The pmcid field WAS read here and is absent, which is the stronger check - but a fifth route, resolving the DOI to the publisher, remains untried and would settle it. da Fonseca Bastos ACS et al. 2020
  4. E4weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    The title of PMID 32114371 asserts that osteopontin-c is overexpressed in KMT2A-AFF1-positive paediatric B-cell lymphoblastic leukaemia compared with ETV6-RUNX1-positive disease.

    PMID 32114371DOI
    What this evidence cannot show

    TITLE-ONLY. The paper has no abstract and no accessible full text [E3], so nothing is known about the cohort size, the assay, the direction or magnitude of the difference, whether the comparison was adjusted, or whether osteopontin-c was measured as transcript or protein. This item exists so that a future cycle knows the record exists and what it claims; it must not be used to support any quantitative or mechanistic statement, and confidence is weak for that reason.

    Evidence that argues against this (1)
    • A title is a claim by the authors and not a result a reader has checked. The comparator, ETV6-RUNX1, is the favourable-risk subtype, so a difference between it and KMT2A-AFF1 may reflect any of the many biological differences between those subtypes rather than anything specific to osteopontin. Zhou J et al. 2022
  5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    SPP1 is not a cell-intrinsic dependency in DepMap: 29 tissue blocks, 1,258 screens, mean gene effect +0.0782, range -0.4486 to 0.4264, zero lines at or below a gene effect of -1.0.

    What this evidence cannot show

    Live zero-screen control in the same call and code path: MEN1 returned 29 tissue blocks, 1,258 screens, mean gene effect -0.3616, minimum -2.5262, 79 lines at or below -1.0 - so a zero count of dependent lines is a measured zero, not an empty payload with a passing approvedSymbol readback. DepMap release recorded as Open Targets dataVersion year 26 month 06 (rule 83), API 26.6.3, fetched 2026-09-22. No median-split proportion is reported (rule 84). AND THE NULL IS AN ABSENCE BY CONSTRUCTION: DepMap screens cells in monoculture with no stroma and no endothelium, so a niche factor acting in trans cannot score however essential it is in vivo. There is an extra wrinkle for SPP1: E2's mechanism is AUTOCRINE, so unlike CXCL12 or KITLG this gene's phenotype is in principle visible to a monoculture screen - and it still does not appear. That tension is not resolved here. Possible readings include that the autocrine requirement is in vivo-specific, that it is specific to the LSC fraction rather than bulk culture, or that the knockdown phenotype in E2 does not generalise across lines.

    Evidence that argues against this (1)
    • The same monoculture assay returns a flat null for CXCR4 while an in vivo syngeneic MLL-AF9 CRISPR screen made CXCR4 a top cell-surface dependency whose deletion eradicated leukaemia - a direct demonstration that a DepMap null for a niche gene is uninformative about in vivo requirement. Ramakrishnan R et al. 2020
Connected to