Niche components considered and not written: what the search actually returned for E-selectin, CAR cells, osteoclasts, marrow macrophages and the rest
also called marrow niche rejection ledger · niche beat rejected candidates
Before writing an entry for each part of the bone-marrow niche, every candidate was searched in PubMed and the results counted on both sides. This records the ones deliberately not written and why, so nobody spends a day rediscovering that there is not a single paper connecting this leukaemia to E-selectin.
The same thing, for a scientist
A rejection ledger for the marrow-niche beat of cycle 40, following `ana-cell-identity-rejected-candidates` and `ana-histone-mark-audit-rejected-candidates`. Every candidate was scored by a PubMed eutils intersection with **both halves counted**: disease half `(KMT2A OR MLL-rearranged OR MLL-AF4 OR MLL-AF9)[Title/Abstract]` = **2,919** records, and each niche half counted separately so a small intersection could be distinguished from a small literature [E1]. Fourteen entries were written. **Rejected outright** for essentially no KMT2A-specific literature: **E-selectin/selectin (0 against a selectin literature of 8,919+)**, **CXCL12-abundant reticular / reticular cell as a named entity (0 against 1,963)**, **bone-marrow macrophage / nurse cell (0)**, **Schwann cell (1)**. **Rejected as not niche-relevant despite non-zero counts**: **osteoclast (5, none about an osteoclastic niche in this disease)**, **regulatory T cell (4)**, **angiopoietin/TIE2 (3, none niche-mechanistic)**, **osteopontin/SPP1 (4, of which the one KMT2A-specific record is an abstract-less Letter)**. **Deferred**: the CNS niche, already covered elsewhere and now with a 2026 primary paper; and the immune microenvironment. One methodological finding is recorded because it changed an answer: **PubMed matches a multi-word Title/Abstract term as an unstemmed phrase**, so `extracellular vesicle[tiab]` returns 2 records and misses the very paper this beat cites, while adding the plural returns 5 [E2].
Method, and the rule this beat followed
The standing instruction: a small intersection means nothing unless both halves are counted. All counts were run against NCBI eutils esearch on 2026-09-19 with [Title/Abstract] restriction and no other filters. Every non-zero intersection under 20 records had its titles inspected before a disposition was assigned. The PubMed MCP search_articles tool was not used for any negative conclusion, because it ANDs every keyword and returns zero on topics with large literatures.
Disease half: (KMT2A[tiab] OR MLL-rearranged[tiab] OR MLL-AF4[tiab] OR MLL-AF9[tiab]) = 2,919.
Rejected outright: essentially no literature
| candidate | niche-half records | intersection | disposition |
|---|---|---|---|
| E-selectin / SELE / selectin / uproleselan | 8,919 (E-selectin/SELE alone) | 0 | rejected |
| CXCL12-abundant reticular / reticular cell (as a named entity) | 1,963 | 0 | rejected as a term |
| bone marrow macrophage / nurse cell | not separately counted; both terms common | 0 | rejected |
| Schwann cell | not separately counted | 1 | rejected |
E-selectin is the strongest rejection here and deserves to be stated positively rather than as an absence. Against an E-selectin literature of 8,919 records, and with uproleselan a clinically developed E-selectin antagonist with a substantial AML programme, the intersection with a 2,919-record disease literature is exactly zero. Vascular-niche E-selectin protection of AML stem cells is a well-known mechanism in the general AML literature; nobody has asked whether it operates in KMT2A-rearranged disease. This is a genuine, clean, pharmacologically actionable gap, and it is the single best target this ledger identifies for a future beat.
CAR cell needs care. The term itself returns zero in this disease - and the two records matching CAR cells[tiab] are chimeric antigen receptor T-cell papers, which is a naming collision worth knowing about. An entry was nonetheless written as cell-type-car-cell, because the cell has KMT2A-specific evidence under other labels (Prx1+ mesenchymal cells supplying the CXCL12 that maintains MLL::AF9 LSCs; LEPR+ perivascular cells as the required SCF source). The rejection is of the term, not of the biology, and the entry says so.
Rejected as not niche-relevant: non-zero counts that do not survive reading
| candidate | intersection | what the records actually are |
|---|---|---|
| osteoclast / osteoclasts | 5 | zoledronic acid inhibiting MLL-AF9 cell growth; a weightlessness-osteoporosis sequencing paper; a prostate-cancer histone demethylase paper; a 1998 monoblastic line that differentiates to osteoclast-like cells with M-CSF and IL-4; a herbal-medicine review. None concerns an osteoclastic niche in this disease |
| regulatory T cell / Treg | 4 | includes a mouse study of blocking Treg migration to the leukaemic microenvironment; immune microenvironment rather than structural niche |
| angiopoietin / TIE2 | 3 | a CAMK paper, a Baf200 chromatin-remodelling paper, and the ANGPTL2 vesicle paper already written up under cell-type-sinusoidal-endothelial-cell. Note ANGPTL2 is not an angiopoietin despite the name |
| osteopontin / SPP1 | 4 | niche S100A8 in AML; the Ebf2+ MSPC paper (SPP1 among dysregulated niche factors, already cited); a general AML osteopontin-LSC paper; and one genuinely KMT2A-specific record, 'Osteopontin-c is overexpressed in KMT2A-AFF1 positive pediatric B-cell lymphoblastic leukemia when compared to those with ETV6-RUNX1', which PubMed indexes as a Letter with no abstract |
The osteopontin case is the one worth flagging for a future beat rather than dismissing. The title states a real, KMT2A::AFF1-specific, comparator-controlled finding (versus ETV6-RUNX1), osteopontin is a canonical endosteal niche matricellular protein, and osteopontin is independently reported as required for leukaemia stem cell maintenance in AML. It was not written up here because the record carries no abstract, so the only verifiable content is the title, and this project's standing rule is that a citation must be read rather than inferred. Obtaining the full text would likely justify an entry.
The Treg and immune-microenvironment records are deferred rather than rejected: they are a different axis (immune evasion) from the structural niche this beat covers, and the 2026 infant KMT2A::AFF1 CNS paper below shows that axis is now productive.
Deferred, not rejected
1. The CNS niche. Already covered by mechanism-kmt2a-r-tissue-localization and concept-cns-involvement-at-diagnosis, and now with a 2026 primary paper in exactly the right disease: in an immune-competent KMT2A::AFF1 infant B-ALL murine model, leukaemia in the CNS niche shows unique functional and transcriptomic adaptations, suppressed T-cell and macrophage activity indicating niche-specific immune escape, PI3K pathway activation with increased AKT phosphorylation and reduced PTEN driving CNS propagation, and miR-93 as the regulator, upregulated in patient samples and required for CNS engraftment 3. Deferred here only because the CNS is a different anatomical niche from the marrow and has its own entries; it is flagged as high priority.
2. The immune microenvironment of the marrow. T cells, macrophages, NK cells and immune evasion in the marrow compartment. Deferred as a separate beat. ana-cell-identity-rejected-candidates already rejected NK cell and neutrophil as cell-identity candidates; that disposition does not cover them as immune-microenvironment components, in the same way that this beat wrote cell-type-osteoblast after that ledger rejected osteoblast as a cell of origin.
3. Osteopontin/SPP1, pending retrieval of PMID 32114371's full text.
4. Mitochondrial transfer from stroma to blast. The 2024 review 'The impact of mitochondrial transfer on leukemia progression' appeared in the niche intersection. Given that three independent compartments in this beat protect this leukaemia by metabolic means, contact-dependent mitochondrial transfer is a plausible unifying mechanism. Not written because the record found is a review in a non-English-language journal and no KMT2A-specific primary study was identified; a targeted search is warranted.
Method finding worth carrying forward: PubMed does not stem multi-word tiab terms
This changed an answer in this beat and is recorded so it does not cost the next agent a wrong absence claim.
(KMT2A OR MLL-rearranged OR MLL-AF4 OR MLL-AF9)[tiab] AND (extracellular vesicle[tiab] OR exosome[tiab]) returns 2 records, and neither is PMID 33108353, the ANGPTL2 small-extracellular-vesicle paper this beat cites as a primary finding — even though that paper's abstract contains both 'MLL-AF9' and 'extracellular vesicles'. Adding the plural form raises the intersection to 5 and includes it 2.
The cause is that PubMed matches a multi-word [Title/Abstract] term as a phrase and does not apply stemming to it, whereas a single word is stemmed. So nestin, osteoblast, adipocyte and pericyte behave as expected, but extracellular vesicle, bone marrow adipocyte, reticular cell, nurse cell and Schwann cell do not match their plurals.
Rule for future beats: for any multi-word Title/Abstract term, include both the singular and the plural. A beat that writes 'only two papers exist' on an unstemmed phrase has made a searching error, not a finding.
The CNS-niche candidate is no longer deferred, and its citation was a placeholder
This entry deferred a CNS-niche candidate and carried its citation with "authors": "Not verified beyond the PubMed record retrieved 2026-09-19" and no DOI. Both are now filled from the PubMed record, fetched 2026-09-20: Duguid A, Malouf C, Nitsche L, Halsey C, Ottersbach K, Experimental Hematology 2026, DOI 10.1016/j.exphem.2025.105347.
The disposition also changes. The paper is now written up as cell-state-cns-niche-adapted in the cell-state beat, so 'deferred' is stale - the candidate was taken up.
Worth noting what the author field was doing: a placeholder string in the authors slot is invisible to verify-citations.py's author check, which skips non-name strings by design so that consortium and 'see PubMed record' entries do not flood the report. That exemption is correct for its purpose and it means a placeholder can sit indefinitely without being flagged.
| script | ad hoc NCBI eutils esearch queries; full parameters in the Method section |
|---|---|
| status | run |
| run date | 2026-09-19 |
| result summary | 14 entries written; 4 candidates rejected outright for essentially no literature (E-selectin, reticular/CAR cell as a term, bone-marrow macrophage/nurse cell, Schwann cell); 4 rejected as not niche-relevant on reading (osteoclast, Treg, angiopoietin/TIE2, osteopontin/SPP1); 4 deferred (CNS niche, marrow immune microenvironment, osteopontin pending full text, stromal mitochondrial transfer). |
| method finding | PubMed matches a multi-word [Title/Abstract] term as an UNSTEMMED PHRASE. `extracellular vesicle[tiab]` returns 2 records in this disease and misses PMID 33108353 whose abstract says 'extracellular vesicles'; adding the plural returns 5. Rule for future beats: for any multi-word Title/Abstract term, include singular AND plural. |
inputs
[ "NCBI eutils esearch, db=pubmed, [Title/Abstract] field restriction", "disease half: (KMT2A OR MLL-rearranged OR MLL-AF4 OR MLL-AF9)[tiab] = 2,919 records" ]
rejected candidates — the numbers behind the chart
[
{
"candidate": "E-selectin / SELE / uproleselan",
"niche_half": 8919,
"intersection": 0,
"disposition": "rejected outright",
"reason": "Zero records against an 8,919-record E-selectin literature and a 2,919-record disease literature. Vascular-niche E-selectin protection of AML stem cells is established in the general AML literature and uproleselan is a clinically developed antagonist, so this is a clean, pharmacologically actionable gap and the best single target this ledger identifies for a future beat."
},
{
"candidate": "CXCL12-abundant reticular cell / CAR cell (as a NAMED TERM)",
"niche_half": 1963,
"intersection": 0,
"disposition": "rejected as a term, not as biology",
"reason": "The term returns zero in this disease and the two records matching 'CAR cells' are chimeric antigen receptor T-cell papers, a naming collision. An entry was written as cell-type-car-cell anyway because the cell has KMT2A-specific evidence under other labels (Prx1+ mesenchymal CXCL12 maintaining MLL::AF9 LSCs; LEPR+ perivascular cells as required SCF source)."
},
{
"candidate": "bone marrow macrophage / nurse cell",
"niche_half": null,
"intersection": 0,
"disposition": "rejected outright",
"reason": "Zero records. Marrow macrophages are established regulators of the erythroblastic island and of HSC retention in the general niche literature, and have no KMT2A-specific literature at all."
},
{
"candidate": "Schwann cell",
"niche_half": null,
"intersection": 1,
"disposition": "rejected outright",
"reason": "One record, and it concerns axonal regeneration rather than leukaemia. Non-myelinating Schwann cells are a documented HSC niche component in the general literature (TGF-beta activation), which makes the absence here notable but not writable."
},
{
"candidate": "osteoclast",
"niche_half": 36637,
"intersection": 5,
"disposition": "rejected as not niche-relevant",
"reason": "Five records: zoledronic acid inhibiting MLL-AF9 cell growth, a weightlessness-osteoporosis sequencing study, a prostate-cancer histone demethylase paper, a 1998 monoblastic line differentiating to osteoclast-like cells with M-CSF and IL-4, and a herbal-medicine review. None describes an osteoclastic niche in this disease."
},
{
"candidate": "regulatory T cell / Treg",
"niche_half": null,
"intersection": 4,
"disposition": "rejected here, deferred to an immune-microenvironment beat",
"reason": "Four records including a mouse study of blocking Treg migration to the leukaemic haematopoietic microenvironment. This is the immune microenvironment rather than the structural niche this beat covers."
},
{
"candidate": "angiopoietin / TIE2",
"niche_half": null,
"intersection": 3,
"disposition": "rejected as not niche-relevant",
"reason": "Three records: a CAMK paper, a Baf200 chromatin-remodelling paper, and the ANGPTL2 endothelial vesicle paper already written up in cell-type-sinusoidal-endothelial-cell. ANGPTL2 is not an angiopoietin despite the name, so the ANGPT1/TIE2 axis proper has zero KMT2A-specific literature."
},
{
"candidate": "osteopontin / SPP1",
"niche_half": null,
"intersection": 4,
"disposition": "deferred pending full text",
"reason": "One genuinely KMT2A-specific, comparator-controlled record exists - 'Osteopontin-c is overexpressed in KMT2A-AFF1 positive pediatric B-cell lymphoblastic leukemia when compared to those with ETV6-RUNX1' (PMID 32114371) - but PubMed indexes it as a Letter with NO ABSTRACT, so only the title is verifiable and this project's rule is that a citation must be read. Osteopontin is a canonical endosteal matricellular niche protein and is independently reported as required for AML leukaemia stem cell maintenance, so obtaining the full text would likely justify an entry."
},
{
"candidate": "hypoxic niche protection",
"niche_half": 203310,
"intersection": 17,
"disposition": "rejected because the evidence points the OPPOSITE way",
"reason": "Already rejected by the resistance beat and re-confirmed here by fetching the paper: hypoxia induction SENSITISED KMT2A-rearranged AML to venetoclax plus MI-503, with the combination repressing HIF-1A and its target HDAC9 (PMID 38053201). Written up as concept-marrow-oxygen-gradient stating the direction, not as a protection entry."
},
{
"candidate": "niche protection against menin inhibitors",
"niche_half": 543067,
"intersection": 9,
"disposition": "rejected outright, with the search re-run rather than inherited",
"reason": "Menin-inhibitor half (216 records) x niche/stroma half (543,067) = 9 records, all titles read, none testing niche protection against a menin inhibitor. The category appears only as a heading in a 2026 review. Written up as the largest gap in the beat with a specified experiment, exp-stromal-coculture-menin-inhibitor-protection."
},
{
"candidate": "stromal-to-blast mitochondrial transfer",
"niche_half": null,
"intersection": null,
"disposition": "deferred",
"reason": "A 2024 review on mitochondrial transfer in leukaemia progression appeared in the niche intersection but is a review in a non-English-language journal and no KMT2A-specific primary study was identified. Given that three compartments in this beat protect this leukaemia metabolically, contact-dependent mitochondrial transfer is a plausible unifying mechanism and merits a targeted search."
},
{
"candidate": "CNS niche",
"niche_half": null,
"intersection": null,
"disposition": "deferred to its own beat",
"reason": "Covered by mechanism-kmt2a-r-tissue-localization and concept-cns-involvement-at-diagnosis, and now with a 2026 primary paper in an immune-competent KMT2A::AFF1 infant model (miR-93/PTEN and CNS-specific T-cell exhaustion, PMID 42139054). A different anatomical niche; flagged as high priority."
}
]entries written
[ "concept-bone-marrow-niche", "concept-endosteal-niche", "concept-perivascular-sinusoidal-niche", "concept-marrow-oxygen-gradient", "cell-type-car-cell", "cell-type-nestin-msc", "cell-type-ng2-periarteriolar-cell", "cell-type-osteoblast", "cell-type-sinusoidal-endothelial-cell", "cell-type-bone-marrow-adipocyte", "cell-type-bone-marrow-sympathetic-nerve", "mechanism-leukaemia-remodels-its-own-niche", "mechanism-cxcl12-cxcr4-marrow-retention", "mechanism-niche-developmental-stage-sets-leukaemia-lineage", "mechanism-niche-chemoprotection-kmt2a", "mechanism-itga4-niche-adhesion-kmt2a", "trial-poe-10-03-plerixafor-paediatric", "ana-cxcr4-directed-trials-kmt2a-infant", "hyp-niche-remodelling-precedes-relapse-kmt2a", "exp-stromal-coculture-menin-inhibitor-protection", "exp-cxcr4-antagonist-antioxidant-rescue", "exp-beta2-blockade-kmt2a-xenograft", "risk-niche-disruption-mobilises-normal-hsc" ]
- Every count here is an upper bound on absence: a Title/Abstract search misses niche components used as methods or controls, and the fields that matter most are often reported in supplementary material.
- The E-selectin zero is the ledger's strongest claim and also its most fragile, because a single unindexed conference abstract or supplementary figure would change it; it should be re-run before anyone builds a grant on it.
- A rejection ledger creates a hazard of its own: future agents may treat 'rejected' as 'settled'. Three of the dispositions here explicitly reverse an earlier ledger's rejection (osteoblast) or reject a term while keeping the biology (CAR cell), which is the pattern to expect rather than the exception.
- Does vascular-niche E-selectin protect KMT2A-rearranged leukaemia stem cells? The intersection is exactly zero against an 8,919-record E-selectin literature, uproleselan exists, and the mechanism is established in general AML. This is the cleanest actionable gap the beat found.
- What does PMID 32114371 actually report about osteopontin-c in KMT2A::AFF1 versus ETV6-RUNX1 B-ALL? The title states a comparator-controlled, genotype-specific finding and the record has no abstract.
- Is there stromal-to-blast mitochondrial transfer in KMT2A-rearranged leukaemia? Three compartments in this beat protect metabolically and mitochondrial transfer would unify them.
- Marrow macrophages and non-myelinating Schwann cells are established niche components with zero and one KMT2A-specific records respectively. Are these genuine absences or simply questions nobody has asked in this genotype?
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
PubMed eutils esearch intersections run on 2026-09-19 with [Title/Abstract] restriction, disease half (KMT2A OR MLL-rearranged OR MLL-AF4 OR MLL-AF9) = 2,919 records: niche = 40; mesenchymal/stromal = 63; endothelial = 20; hypoxia/hypoxic/HIF-1 = 17; CXCR4 = 11; CXCL12/SDF-1 = 7; nestin = 6; perivascular/sinusoidal = 6; endosteal/endosteum = 5; osteoblast/osteoblastic = 3; osteoclast = 5; osteopontin/SPP1 = 4; angiopoietin/TIE2 = 3; VCAM-1/VLA-4/integrin alpha4 = 2; adipocyte = 2; sympathetic/adrenergic = 2; CXCL12-abundant reticular or CAR cells = 2 (both chimeric antigen receptor papers); E-selectin/SELE/selectin/uproleselan = 0; bone marrow macrophage or nurse cell = 0; reticular cell as a named entity = 0; Schwann cell = 1; regulatory T cell/Treg = 4; pericyte/NG2/CSPG4 = 19, nearly all concerning NG2 on the leukaemic blast rather than the niche pericyte. Corresponding niche-half sizes include niche 57,129; perivascular/sinusoidal 57,982; osteoblast 47,395; hypoxia 203,310; CXCR4 17,206; CXCL12/SDF-1 11,423; E-selectin/SELE 8,919; nestin 7,564; endosteal 3,777; pericyte/NG2/CSPG4 13,613; osteoclast 36,637; leptin/LEPR 45,534; reticular/CAR-cell terms 1,963.
NCBI E-utilities esearch counts for the marrow-niche beat, both halves of every intersection, 2026-09-19 · NCBI E-utilities · NCBI Entrez 2026What this evidence cannot show
Title/Abstract restriction misses work that used a niche component only as a method or control, which is common. The disease half uses four terms and misses records that name only a rarer fusion partner. Counts drift as PubMed is updated. Multi-word terms are matched as unstemmed phrases (see E2), so singular-only queries undercount.
Evidence that argues against this (1)
- A 40-record intersection for 'niche' is not a small literature for this purpose, and the beat's own reading of those 40 titles found at least twelve substantial primary findings - a caution against treating any of these counts as a measure of how much biology exists. The counts guide where to look; they do not measure importance. The cell-identity ledger for cycle 39 made the same point and it applies here: a count is not a finding. Oncofusion Knowledge Graph et al. 2026
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
PubMed matches a multi-word [Title/Abstract] term as an unstemmed phrase: the intersection of the disease half with (extracellular vesicle OR exosome)[Title/Abstract] returns 2 records and does NOT include PMID 33108353, whose abstract contains both 'MLL-AF9' and 'extracellular vesicles'; adding the plural forms raises the intersection to 5 and includes it. Verified by direct query and by confirming that 33108353[uid] AND (extracellular vesicle OR exosome)[Title/Abstract] returns 0 while 33108353[uid] AND MLL-AF9[Title/Abstract] returns 1.
NCBI E-utilities esearch: singular-versus-plural behaviour of multi-word Title/Abstract terms, demonstrated on PMID 33108353, 2026-09-19 · NCBI E-utilities · NCBI Entrez 2026What this evidence cannot show
Demonstrated on one term pair; PubMed's automatic term mapping is not fully documented publicly and its behaviour can change. The safe practice is to include both forms rather than to rely on any model of the indexer.
Evidence that argues against this (1)
- Including plurals is a partial fix, not a complete one: it does not address hyphenation variants (E-selectin versus E selectin), British and American spellings (haematopoietic, tumour), abbreviations, or terms that appear only in the full text. A multi-term OR block remains an approximation and an absence claim built on one is an upper bound. NCBI E-utilities 2026
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Using an immune-competent KMT2A::AFF1-positive infant B-ALL murine model, leukaemia in the central nervous system niche shows unique functional and transcriptomic adaptations with suppressed T-cell and macrophage activity indicating niche-specific immune escape; PI3K pathway activation is a key driver of CNS leukaemia propagation, with increased AKT phosphorylation and reduced PTEN expression and miR-93 emerging as a regulator; miR-93 upregulation is confirmed in patient samples, suggesting use as a biomarker of CNS involvement, and miR-93 knockdown impairs CNS leukaemia engraftment.
miR-93-mediated PTEN suppression and CNS-specific T cell exhaustion shape the leukemia niche in infant KMT2A::AFF1+ B-ALL · Duguid A, Malouf C, Leah T, Nitsche L, Barrett NA, Smith OP, Halsey C, Ottersbach K · Cell Reports 2026What this evidence cannot show
2026 publication, unreplicated. Murine immune-competent KMT2A::AFF1 model, which is the right genotype but a mouse. The patient-sample miR-93 confirmation is an association whose cohort size is not given in the abstract. Cited in this ledger to justify DEFERRING the CNS niche to a dedicated beat rather than to support a marrow-niche claim. Read: abstract only.
Evidence that argues against this (1)
- The CNS niche in this disease is not solely an immune-escape story and a beat built only on that framing would miss the metabolic one: murine work shows the CNS niche metabolically reprograms KMT2A-AFF1 cells, with CNS-adapted cells upregulating lipid and fatty-acid metabolism and losing systemic repopulation capacity, as already recorded in `mechanism-kmt2a-r-tissue-localization`. Read: abstract only. Duguid A et al. 2026
the enumeration behind this beat's entry list, with both halves of every intersection counted
executes the previous ledger's deferred item 1, the bone-marrow mesenchymal and neural niche
defers the CNS niche, now with a 2026 primary paper in an immune-competent KMT2A::AFF1 infant model
closes the cycle-40 'deferred pending full text' disposition for SPP1
converts the cycle-40 flag 'best single target this ledger identifies' into a written entry; the zero was re-measured and holds
re-measures and extends the cycle-40 rejection ledger's counts
successor rejection ledger for the protein layer of the same beat