GenemoderateQ33-niche-protein-layer

VCAM1 - the stromal counter-receptor for VLA-4, and in MLL-fusion AML its UPREGULATION is anti-leukaemic, which is the opposite of the expected sign

also called CD106 · INCAM-100 · vascular cell adhesion molecule 1

VCAM1 is the docking protein on marrow blood-vessel and stromal cells that leukaemia cells grip with VLA-4. The surprise is the direction: in mice with MLL-fusion leukaemia, a drug that pushed VCAM1 UP suppressed the disease.

The same thing, for a scientist

VCAM1 (UniProt P19320, 739 aa) is the endothelial and stromal counter-receptor recognised by integrin alpha-4/beta-1 (VLA-4) at the Q-I-D-S sequence [E1]. Its KMT2A-relevant literature is small - two records for VCAM1[tiab] intersected with the KMT2A/MLL half - and both are about direction rather than magnitude. In MLL-fusion AML, IMPDH inhibition suppressed MLL-AF9-driven disease in vivo and the mechanism ran through overactivation of TLR-TRAF6-NF-kB signalling and UPREGULATION of VCAM1, with a TLR1/2 agonist adding to the effect [E2] - i.e. more VCAM1, less leukaemia. In the other, IRF7 loss accelerated MLL-AF9 AML with raised VCAM1 and intracerebral invasion, and blocking VCAM1-VLA-4 delayed progression [E3] - i.e. more VCAM1 signalling, more disease. Those two are not reconcilable as a single 'VCAM1 is protective/permissive' statement, and this entry does not pretend otherwise. VCAM1 is not a cell-intrinsic DepMap dependency (mean -0.0188 over 1,256 lines, zero lines at or below -1.0) [E4].

What it looks like

VCAM1739 aa

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

Architecture, and one thing not to write into a figure

P19320 is a type I membrane protein: signal 1-24, mature chain 25-739, extracellular 25-698, transmembrane 699-720, cytoplasmic 721-739, built from seven Ig-like C2-type domains (25-105, 109-212, 223-309, 312-399, 408-506, 511-595, 600-684) 1. UniProt also carries a processed chain, 'Soluble Vascular Cell Adhesion Molecule-1', annotated as starting at 25 with NO end coordinate - the end value is null. That chain is deliberately omitted from the domains array of this entry, because a domain with a null boundary is invisible to every figure this graph draws and the standing rule is that residue numbers are integers or the domain is left out with a note. The soluble form is real and is a clinical biomarker; its C-terminus is simply not defined in the record.

The two MLL-fusion results point in opposite directions

Anti-leukaemic direction. AMLs with MLL fusions are susceptible to IMPDH inhibitors in vitro; alternate-day dosing suppressed MLL-AF9-driven AML in vivo; and mechanistically IMPDH inhibition induced overactivation of TLR-TRAF6-NF-kB signalling and upregulation of the adhesion molecule VCAM1, which the authors report as contributing to the antileukaemia effect, with a TLR1/2 agonist combining effectively 2.

Pro-leukaemic direction. In MLL-AF9 mouse AML, IRF7 knockout accelerated disease with intracerebral invasion and elevated VCAM1 correlating with high LSC levels, and blocking the VCAM1-VLA-4 axis delayed progression and attenuated intracerebral invasion 3.

Both are murine MLL-AF9. The plainest reading is that VCAM1 on a leukaemic cell (E2, cell-autonomous, induced by a drug) and VCAM1-VLA-4 adhesion between a leukaemic cell and the vessel wall (E3) are different quantities, but neither abstract resolves which compartment expresses the VCAM1 being measured, and that is the experiment this entry is waiting on.

Facts and figures

gene symbol	VCAM1
uniprot	P19320
length AA	739
accession resolution	gene_exact:VCAM1 AND organism_id:9606 AND reviewed:true returned exactly ONE reviewed human record, P19320 / VCAM1_HUMAN, primary gene name read back as VCAM1, 739 aa. Fetched live 2026-09-22.
domains omitted	UniProt P19320 carries a processed chain 'Soluble Vascular Cell Adhesion Molecule-1' whose location is start 25, END NULL. It is deliberately absent from the domains array: a residue boundary of null is invisible to every figure in this graph and the standing rule is integers or omission-with-a-note. The soluble form is real and is used as a circulating biomarker; only its C-terminal boundary is undefined in the record.
localization	CELL MEMBRANE plus a SECRETED soluble form, and the evidence codes run the counter-intuitive way. UniProt gives the membrane form 'Cell membrane' with ECO:0000305 (curator inference) and the soluble chain 'Secreted' with ECO:0000269 (experimental) - so on this record the experimentally coded compartment is the shed ectodomain. Biologically VCAM1 is an ENDOTHELIAL and STROMAL surface protein, i.e. it sits on the OTHER cell in this graph's central interaction; the leukaemic side of the pair is ITGA4/ITGB1.
subnuclear localization	NOT APPLICABLE BY TOPOLOGY, on the architecture rather than on a literature search. VCAM1 is a type I membrane protein with a 19-residue cytoplasmic tail, 674 of its 739 residues extracellular, and no chromatin-binding module anywhere in the sequence, so there is no mechanism by which it could occupy a subnuclear compartment. WHAT THIS FIELD DOES NOT CLAIM: it does not claim that nuclear VCAM1 has been looked for and excluded. The UniProt record annotates no nuclear location and no NLS, and per this beat's own CXCR4 case - where UniProt annotates no NLS and a mapped, mutated, functionally tested NLS at 146-RPRK-149 is published - a curated list of accepted locations is not a record of locations sought. No targeted literature search for nuclear VCAM1 was run in this beat, so the honest statement is: not applicable on topology, unsearched as a literature question.
subnuclear absence class	not-applicable-by-topology
cell type	PRINCIPALLY NOT THE BLAST. VCAM1 is predominantly expressed on endothelial cells and marrow stroma - this graph holds cell-type-sinusoidal-endothelial-cell and cell-type-car-cell - and it is the counter-receptor side of the VLA-4 interaction. Both KMT2A-relevant papers measure VCAM1 in a leukaemia model without stating in the abstract which compartment expresses the VCAM1 being measured [E2, E3], and in E2 the induced VCAM1 is plausibly on the leukaemic cell itself. Resolving that is the open question of this entry.
cell state	In the MLL-AF9 IRF7-knockout model, elevated VCAM1 correlated with HIGH LEUKAEMIA STEM CELL levels and a CNS-INVASIVE state, and blocking the axis attenuated both [E3]. In the IMPDH-inhibitor setting, induced VCAM1 accompanied a TLR-NF-kB-activated, SUPPRESSED-leukaemia state [E2]. The two associate VCAM1 with opposite states, which is why this field records both rather than choosing.
epigenetic readout	NOT MEASURED IN THIS DISEASE, and the surrounding biology is transcription-factor-driven rather than chromatin-driven in both available papers: VCAM1 moves downstream of TLR-TRAF6-NF-kB [E2] and of IRF7 loss [E3], neither of which is a histone-mark measurement. No H3K79me2, H3K4me3, H3K27me3 or menin occupancy measurement at VCAM1 in a KMT2A-rearranged system has been reported. Bounded over titles and abstracts (rule 80): VCAM1[tiab] intersected with the KMT2A/MLL disease half returns 2 records total, both cited here, so the chromatin question has no literature to hide in. ABSENCE CLASS: never-measured.
shown for fusion	YES for murine MLL-AF9, twice, IN OPPOSITE DIRECTIONS: VCAM1 upregulation as part of an anti-leukaemic IMPDH-inhibitor mechanism [E2], and VCAM1-VLA-4 blockade as anti-leukaemic in the IRF7-null CNS-invasion setting [E3]. NO for any KMT2A-rearranged B-ALL experiment, NO for KMT2A::AFF1, and NO for any infant sample. The B-lineage ITGA4 work [gene-itga4 E2] does not measure VCAM1.
intact union membership	ABSENT 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 P19320.
collision note	VCAM1 has no row in data/symbol-collisions.json (read 2026-09-22, case-sensitive over d['collisions'], 177 rows). A retrieval trap WAS measured here and is worth the registry: the hyphenated form is a DIFFERENT STRING to PubMed. VCAM1[tiab] intersected with the KMT2A/MLL half returns 2 records; VCAM-1[tiab] against the identical half returns 0 with a valid translation. A beat that searched only the hyphenated protein name - which is how the protein is written in most adhesion papers - would have recorded a clean false zero for this gene.

domains

domains — the numbers behind the chart

[
  {
    "name": "Signal peptide",
    "start": 1,
    "end": 24
  },
  {
    "name": "Ig-like C2-type 1",
    "start": 25,
    "end": 105
  },
  {
    "name": "Ig-like C2-type 2",
    "start": 109,
    "end": 212
  },
  {
    "name": "Ig-like C2-type 3",
    "start": 223,
    "end": 309
  },
  {
    "name": "Ig-like C2-type 4",
    "start": 312,
    "end": 399
  },
  {
    "name": "Ig-like C2-type 5",
    "start": 408,
    "end": 506
  },
  {
    "name": "Ig-like C2-type 6",
    "start": 511,
    "end": 595
  },
  {
    "name": "Ig-like C2-type 7",
    "start": 600,
    "end": 684
  },
  {
    "name": "Extracellular topological domain",
    "start": 25,
    "end": 698
  },
  {
    "name": "Transmembrane (helical)",
    "start": 699,
    "end": 720
  },
  {
    "name": "Cytoplasmic tail",
    "start": 721,
    "end": 739
  }
]

depmap essentiality

{
  "depmap_common_essential": false,
  "tissue_blocks": 29,
  "n_screens": 1256,
  "mean_gene_effect": -0.0188,
  "min_gene_effect": -0.4958,
  "max_gene_effect": 0.6526,
  "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 here 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); the statistic is an absolute count of lines at or below gene effect -1.0."
}

How this could bite us back

The sign is not settled. A therapy designed to block VCAM1-VLA-4 adhesion in KMT2A-rearranged disease is being designed against a molecule whose upregulation suppressed MLL-fusion leukaemia in the other available in vivo experiment [E2]. Getting the direction wrong here accelerates the disease.
VCAM1 is the endothelial side of the pair, so systemic blockade hits vascular inflammation biology across the body, and the shed soluble form means a blocking antibody has a circulating decoy.
If the operative mechanism is CNS invasion [E3], then the marrow-retention framing is the wrong one and a mobilising intervention is contraindicated rather than helpful.

Still open

Which cell expresses the VCAM1 that matters - the blast or the vessel? Both MLL-AF9 papers measure VCAM1 without resolving the compartment in their abstracts, and the therapeutic inference inverts depending on the answer.
Is VCAM1 upregulation on a leukaemic cell anti-leukaemic because it triggers TLR-NF-kB-driven differentiation or immune recognition, as E2 implies, rather than because it changes adhesion at all?
Has VCAM1 ever been measured in KMT2A-rearranged B-ALL? The literature is two murine myeloid papers, and the B-lineage ITGA4/natalizumab work does not report its counter-receptor.

Every source behind this entry (4)

E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
The reviewed human VCAM1 record is UniProt P19320 (VCAM1_HUMAN), 739 aa, a type I membrane protein with seven Ig-like C2-type domains, a soluble processed form, and recognition by integrin alpha-4/beta-1 at the Q-I-D-S sequence.
UniProtKB P19320 (VCAM1_HUMAN) · UniProtKB 2026
What this evidence cannot show
Resolved by gene_exact:VCAM1 AND organism_id:9606 AND reviewed:true, exactly ONE reviewed human record; primary gene name read back as VCAM1, 739 aa, no synonyms on the record. Fetched 2026-09-22. The membrane-form location comment carries only ECO:0000305 (curator inference); the SECRETED soluble form carries ECO:0000269. So the experimentally coded location on this record is the soluble chain, not the membrane protein - the inverse of what a reader would assume.
Evidence that argues against this (1)
- Nothing in a curated feature table licenses an absence claim, because the list records what curators accepted rather than what was sought and excluded. The worked example in this graph is CXCR4, where UniProt annotates no nuclear localisation signal and a mapped, mutated, functionally tested NLS at 146-RPRK-149 is nonetheless published. Read every UniProt-derived statement in this entry as a floor. Bao Y et al. 2019
E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
AMLs with MLL fusions are susceptible to IMPDH inhibitors in vitro; alternate-day administration suppressed MLL-AF9-driven AML in vivo without devastating immune function; and the mechanism involved overactivation of TLR-TRAF6-NF-kB signalling and upregulation of the adhesion molecule VCAM1, which contributed to the antileukaemia effect, with a TLR1/2 agonist combining effectively.
IMPDH inhibition activates TLR-VCAM1 pathway and suppresses the development of MLL-fusion leukemia · Liu X, Sato N, Yabushita T, et al. · EMBO Mol Med 2023
PMID 36453131 DOI
What this evidence cannot show
Murine MLL-AF9 myeloid, not infant B-lineage KMT2A::AFF1 (rule 73). The abstract does not say which cell type upregulates VCAM1, nor how the contribution of VCAM1 to the antileukaemia effect was isolated from the rest of TLR-NF-kB activation. Read from the abstract; the figures were not opened (rule 80). The direction of this result is the reason it is in this entry - it contradicts the intuitive 'adhesion molecule = protection' model.
Evidence that argues against this (1)
- The IRF7 study reports the opposite sign in the same MLL-AF9 model: elevated VCAM1 correlated with higher LSC levels and accelerated disease, and blocking VCAM1-VLA-4 delayed progression. Oncogene 2022
E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
In MLL-AF9 mouse AML, IRF7 knockout accelerated disease with intracerebral invasion, raised VCAM1 in a manner correlating with high leukaemia stem cell levels, and blocking the VCAM1-VLA-4 axis delayed disease progression and attenuated intracerebral invasion.
Loss of IRF7 accelerates acute myeloid leukemia progression and induces VCAM1-VLA-4 mediated intracerebral invasion · Oncogene 2022
PMID 35256780 DOI
What this evidence cannot show
Murine myeloid MLL-AF9 with an engineered IRF7 knockout: the VCAM1 elevation is downstream of that knockout, not a demonstrated property of the fusion. The readout is CNS invasion rather than marrow retention. The blocking reagent and whether it targeted VCAM1 or VLA-4 is not specified in the abstract. Author names not transcribed; title, DOI, journal and year from the fetched record.
Evidence that argues against this (1)
- The IMPDH study reports VCAM1 UPREGULATION as part of an ANTI-leukaemic mechanism in the same MLL-AF9 model. Liu X et al. 2023
E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
VCAM1 is not a cell-intrinsic dependency in DepMap: 29 tissue blocks, 1,256 screens, mean gene effect -0.0188, range -0.4958 to 0.6526, zero lines at or below a gene effect of -1.0.
Open Targets Platform GraphQL - target.depMapEssentiality for ENSG00000162692 (VCAM1), 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 here 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); the statistic is an absolute count of lines at or below gene effect -1.0. AND THIS NULL IS AN ABSENCE BY CONSTRUCTION AND SHOULD NOT BE REPORTED AS INFORMATIVE. For VCAM1 the monoculture objection is stronger than for any receptor in this tranche: the cell that expresses the VCAM1 that matters is the STROMAL or ENDOTHELIAL cell, which is not present in a DepMap screen at all. A DepMap profile for VCAM1 says nothing whatever about niche adhesion, and the figure is recorded here only so that nobody re-derives it and mistakes it for one.
Evidence that argues against this (1)
- The same monoculture assay returns an equally flat null for CXCR4 while an in vivo syngeneic MLL-AF9 CRISPR screen made CXCR4 a top 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

← bindsITGA4 (CD49d, the alpha-4 of VLA-4) - the niche adhesion subunit with an approved antibody and a 2026 KMT2A-r B-ALL xenograft result
integrin alpha-4/beta-1 (VLA-4) recognises Q-I-D-S in VCAM1
expressed in →Bone-marrow sinusoidal endothelial cell — the vessel wall that keeps stem cells alive, and posts vesicles that accelerate MLL-AF9 leukaemia
VCAM1 is predominantly an endothelial surface adhesion molecule
part of →The bone-marrow niche — the address a blood stem cell lives at, which 26 entries in this graph invoke and none had defined
stromal/endothelial counter-receptor of the niche adhesion arm
part of →ITGA4 (VLA-4) and NG2 — the niche-adhesion axis of KMT2A-rearranged B-ALL, and the one approved drug that hits it
the VCAM1 half of VLA-4/VCAM-1 adhesion
related to →The CNS-adapted state — leukaemia-propagating cells reprogrammed by the central nervous system niche, and worse at restarting systemic disease afterwards
VCAM1-VLA-4 blockade attenuates intracerebral invasion in MLL-AF9 mouse AML