SELE (E-selectin) - the vascular-niche chemoresistance receptor with a clinical antagonist, zero KMT2A papers, and a paediatric trial that a company restructuring closed

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.

• THE THERAPEUTIC INDEX CLAIM IS THE WHOLE APPEAL AND IT IS ONE CLAUSE OF ONE ABSTRACT. E1 reports that blockade protects endogenous HSCs while sensitising blasts - the opposite of the CXCR4 problem. That is the reason to pursue this target and it rests on a single unreplicated statement whose figures were not read.
• E-SELECTIN IS AN INFLAMMATORY RESPONSE MOLECULE. Blocking endothelial leukocyte adhesion in a neutropenic infant on intensive chemotherapy impairs neutrophil recruitment to infection sites, and infection is a leading cause of death in infant ALL therapy. No paediatric safety database for this agent exists beyond a single terminated patient.
• THE LIGAND, NOT THE RECEPTOR, MAY BE ABSENT IN THIS LINEAGE. If KMT2A::AFF1 B-precursor blasts do not carry the sialofucosylated glycans, the target is simply not engaged and the whole mechanism is a myeloid one. This is a failure mode that a mouse AML model cannot reveal and that one flow experiment would.
• THE SUPPLY ROUTE IS CLOSED. Both terminated uproleselan trials stopped for commercial reasons, one of them the only protocol that would have admitted a one-year-old. A preclinical programme built on this agent must confirm material availability before it starts.

1. Do KMT2A-rearranged B-ALL blasts bind E-selectin at all? The mechanism in E1 is glycan-dependent and the sialofucosylated ligands are best characterised on myeloid cells. One flow-cytometry experiment with an E-selectin-Ig chimera on infant KMT2A::AFF1 samples would settle whether this target is even available in the index disease, and it is the cheapest decisive experiment in this beat.
2. Are the fucosyltransferases and sialyltransferases that build E-selectin ligands (FUT4, FUT7, ST3GAL4, ST3GAL6) under KMT2A-fusion control? That would convert a general AML niche mechanism into a fusion-specific one and is answerable from existing KMT2A-r expression data.
3. Is uproleselan obtainable for academic preclinical work now that its sponsor has deprioritised it? The scientific case is intact and the commercial one is not, which is a specific and answerable practical question.

1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   In AML mouse models, blasts release inflammatory mediators that upregulate endothelial niche E-selectin; oncogenesis-associated cell-surface glycosylation changes enhance blast binding to E-selectin and promote pro-survival AKT/NF-kB signalling; in vivo the blasts with highest E-selectin binding potential are 12-fold more likely to survive chemotherapy and are the main contributors to relapse; and absence of E-selectin in Sele-/- hosts or blockade with the small-molecule mimetic GMI-1271/uproleselan inhibits this niche-mediated pro-survival signalling, dampens blast regeneration, strongly synergises with chemotherapy, doubles the duration of mouse survival over chemotherapy alone, and protects endogenous HSCs.

   Endothelial E-selectin inhibition improves acute myeloid leukaemia therapy by disrupting vascular niche-mediated chemoresistance · Barbier V, Erbani J, Fiveash C, et al. · Nat Commun 2020
   PMID 32341362DOI
   ▸What this evidence cannot show

   AML mouse models generally - the abstract does not name the driving lesion, so whether any arm is KMT2A-rearranged is unknown and must not be assumed (rule 73 in its strictest form: the system is unstated). The 12-fold figure is a between-fraction comparison of binding potential, not a treatment effect. 'Protects endogenous HSC' is the most therapeutically important clause in the abstract and is the one most in need of the figures, which were not opened (rule 80). Read from the abstract.

   ▸Evidence that argues against this (1)

   • A mechanism that depends on oncogenesis-associated glycosylation changes may not transfer between lineages: the E-selectin ligands on a myeloid blast (sialyl-Lewis-x on PSGL1 and CD44 glycoforms) are not necessarily present on a B-precursor blast, and infant KMT2A::AFF1 disease is B-lineage with a characteristically immature immunophenotype. No E-selectin binding measurement exists on any KMT2A-rearranged B-ALL sample. NCBI PubMed 2026
2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   The reviewed human E-selectin record is UniProt P16581 (LYAM2_HUMAN), 610 aa, a type I membrane protein with a C-type lectin domain at 22-139, an EGF-like domain at 140-175 and six Sushi/CCP domains spanning 178-549, extracellular 22-556, transmembrane 557-578, cytoplasmic 579-610.

   UniProtKB P16581 (LYAM2_HUMAN) - E-selectin · UniProtKB 2026
   ▸What this evidence cannot show

   Resolved by gene_exact:SELE AND organism_id:9606 AND reviewed:true, exactly ONE reviewed human record; primary gene name read back as SELE, 610 aa, synonym ELAM1. Fetched 2026-09-22. The two cell-membrane location terms carry ECO:0000269. E-selectin's ligand engagement is carbohydrate-dependent through the C-type lectin domain, so the protein sequence alone does not predict which cells it binds - the glycosylation state of the counterpart cell does, which is the mechanism E1 turns on.

   ▸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
3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   No paper names E-selectin and KMT2A/MLL-rearranged leukaemia together in a title or abstract: (SELE[tiab] OR E-selectin[tiab] OR uproleselan[tiab]) AND (KMT2A[tiab] OR MLL-rearranged[tiab] OR MLL-AF4[tiab] OR MLL-AF9[tiab]) returns 0 records with valid querytranslation and an explicit 'No items found.', against operands of 8,608 for E-selectin[tiab] alone and 2,922 for the disease half.

   NCBI E-utilities esearch, db=pubmed - bounded-absence query with querytranslation readback and operand controls · NCBI PubMed 2026
   ▸What this evidence cannot show

   Rule 67 controls were run through the IDENTICAL query shape with the same disease half: (LEPR[tiab] OR leptin[tiab]) returns 2, (SPP1[tiab] OR osteopontin[tiab]) returns 4, VCAM1[tiab] returns 2 - so the AND composes and this is a real zero. Bounded over titles and abstracts (rule 80): an E-selectin binding assay inside a supplementary figure of a KMT2A paper would be invisible. The name list was SELE, E-selectin and uproleselan; CD62E, ELAM1 and GMI-1271 were NOT searched and are the way this bound could be too tight. Queried 2026-09-22.

   ▸Evidence that argues against this (1)

   • The hyphenation trap measured elsewhere in this beat applies to this family too: VCAM1[tiab] returns 2 records against the same disease half while VCAM-1[tiab] returns 0. 'E-selectin' was searched in its hyphenated form here, which is the common one, but 'E selectin' unhyphenated and 'CD62E' were not, so the zero is bounded by a three-name list and says so. NCBI PubMed 2026
4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.trial_registry

   NCT05569512, a phase 1/2 trial of uproleselan with high-dose busulfan pre-transplant conditioning in chemotherapy-resistant AML with a minimum eligible age of 12 months, was TERMINATED after 1 participant, with whyStopped recorded as notification on 26 July 2024 that GlycoMimetics was terminating contracting following company restructuring, ahead of the study meeting criteria to progress from phase 1 to phase 2.

   A Phase 1/2 Trial of Uproleselan Combined With High Dose Busulfan Pre-Transplant Conditioning in Hematopoietic Stem Cell Transplantation for Patients With Chemotherapy Resistant Acute Myeloid Leukemia · ClinicalTrials.gov 2024
   NCT05569512
   ▸What this evidence cannot show

   Registry record only, fetched from the raw ClinicalTrials.gov v2 API 2026-09-22 because the whyStopped field is not returned by the MCP tool. n=1, so nothing about the drug is learnable from the study itself. The 12-month minimum age makes this the only uproleselan record in the registry that would have admitted an infant, and it is an AML transplant-conditioning protocol rather than an infant ALL protocol.

   ▸Evidence that argues against this (1)

   • A single-patient terminated study is thin evidence that a paediatric route to this mechanism is genuinely available: the study also had to compete for accrual in chemotherapy-resistant paediatric AML, and the CXCR4 precedent in this beat (NCT01655875, terminated at n=1 for lack of enrollment) shows how that usually ends. ClinicalTrials.gov 2014
5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.trial_registry

   NCT05054543, a phase 3 randomised double-blinded bridging trial of uproleselan with chemotherapy in Chinese patients with relapsed/refractory AML (n=140, Apollomics Inc.), was TERMINATED with whyStopped recorded as: in light of the strategic realignment of the company's development focus, it has been determined to discontinue the further development of Uproleselan Injection.

   A Phase 3 Randomized, Double-Blinded Bridging Trial to Evaluate the Efficacy of Uproleselan Administered With Chemotherapy Versus Chemotherapy Alone in Chinese Patients With Relapsed/Refractory Acute Myeloid Leukemia · ClinicalTrials.gov 2026
   NCT05054543
   ▸What this evidence cannot show

   Registry record only, fetched from the raw v2 API 2026-09-22. A commercial discontinuation statement is not an efficacy result and this entry does not read it as one - but it is also not nothing: sponsors sometimes deprioritise after seeing unblinded data, and the registry string cannot distinguish that from a portfolio decision.

   ▸Evidence that argues against this (1)

   • The NCI-sponsored randomised phase 2/3 in older adults, NCT03701308, remains ACTIVE_NOT_RECRUITING with 267 participants, so the mechanism is not clinically dead and a definitive randomised readout may still appear. A commercial termination in one geography is weaker evidence about the drug than an active cooperative-group trial is. ClinicalTrials.gov 2026
6. E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   SELE is not a cell-intrinsic dependency in DepMap: 29 tissue blocks, 1,227 screens, mean gene effect +0.0615, range -0.6674 to 0.5847, zero lines at or below a gene effect of -1.0.

   Open Targets Platform GraphQL - target.depMapEssentiality for ENSG00000007908 (SELE), with meta.dataVersion · Open Targets Platform / DepMap 2026
   ▸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. SELE has the lowest screen count of the eight genes in this tranche, 1,227 against 1,258 for MEN1 in the same call, which is normal gene-specific variation in DepMap and not a sign of mixed releases (rule 83). The construction problem is total here: E-selectin is expressed by the ENDOTHELIUM, which is absent from every DepMap screen, so this figure is recorded only to stop anyone re-deriving it and mistaking it for a finding.

   ▸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

• expressed in →Bone-marrow sinusoidal endothelial cell — the vessel wall that keeps stem cells alive, and posts vesicles that accelerate MLL-AF9 leukaemia

  inducible endothelial adhesion molecule of the vascular niche

• part of →The perivascular and sinusoidal niche — where the functional evidence actually sits, and where the marrow is genuinely hypoxic

  vascular-niche chemoresistance receptor

• related to →Niche-mediated chemoprotection in KMT2A-rearranged leukaemia — real for cytarabine, unsupported for menin inhibitors, and the one hypoxia experiment points the other way

  E-selectin-binding blasts are 12-fold more likely to survive chemotherapy in AML models; never tested in KMT2A-rearranged disease

• related to →Niche components considered and not written: what the search actually returned for E-selectin, CAR cells, osteoclasts, marrow macrophages and the rest

  converts the cycle-40 flag 'best single target this ledger identifies' into a written entry; the zero was re-measured and holds

• related to →Every niche-disrupting drug strips the patient's own stem cells of the same shelter — the bite-back that applies to the whole beat

  unlike CXCR4 antagonism, E-selectin blockade is reported to protect endogenous HSCs

• ← supportsNiche biology HAS been measured in KMT2A-rearranged disease - but in the index fusion itself there is essentially one marrow-niche molecular experiment, from 2011

  the only true zero in the census

• ← supportsNiche proteins enumerated and NOT written in this tranche, and four symbol traps measured while doing it - three of them new to the registry

  measures the SELE substring collision recorded in that entry