ITGA4 (CD49d, the alpha-4 of VLA-4) - the niche adhesion subunit with an approved antibody and a 2026 KMT2A-r B-ALL xenograft result
also called CD49D · alpha-4 integrin · VLA-4 alpha chain · integrin alpha-4
ITGA4 is half of the molecular velcro a leukaemia cell uses to stick to the inside of the bone. An antibody against it is already approved for multiple sclerosis, and in 2026 it delayed leukaemia and improved chemotherapy in mice carrying human KMT2A-rearranged B-ALL.
The same thing, for a scientist
ITGA4 (UniProt P13612, 1,032 aa) is the alpha chain of VLA-4 (alpha-4/beta-1), the integrin that binds VCAM1 and the fibronectin CS-1 region [E1]. It is the most actionable molecule in this graph's niche layer because the key experiment is in the right disease: in KMT2A-rearranged B-ALL, ITGA4 and NG2/CSPG4 - itself a direct transcriptional target of KMT2A fusions - form an axis in which NG2 drives proliferation and migration through Rho GTPase activity in an ITGA4-DEPENDENT manner; combined genetic ablation of both delayed onset and prolonged survival in immunodeficient mice; and natalizumab, an FDA/EMA-approved anti-ITGA4 antibody, delayed progression and potentiated standard-of-care chemotherapy in KMT2A-r B-ALL patient-derived xenografts [E2]. UniProt independently records that alpha-4 interacts with CSPG4 through CSPG4's chondroitin sulfate glycosaminoglycan [E1], so the axis has a curated biochemical basis as well as a genetic one. Blocking the VCAM1-VLA-4 axis also attenuated intracerebral invasion in an MLL-AF9 mouse model [E3]. ITGA4 is not a cell-intrinsic DepMap dependency (mean -0.1387 over 1,255 lines, zero lines at or below -1.0) [E4] - again the expected monoculture result for an adhesion receptor whose ligand is on another cell.
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.
Architecture
A large single-pass type I membrane protein: signal peptide 1-33, mature chain 34-1032, extracellular domain 35-977, one transmembrane helix 978-1001 and a 31-residue cytoplasmic tail 1002-1032 carrying the GFFKR motif at 1003-1007, the conserved membrane-proximal integrin-alpha motif that couples to talin/filamin regulation. The ligand-binding head is a seven-bladed beta-propeller built from FG-GAP repeats 1-7 (35-100, 110-177, 185-237, 238-291, 292-351, 355-412, 416-478), with an SG1 motif at 606-616 1. Integrin alpha-4/beta-1 recognises the sequence Q-I-D-S in VCAM1, which is why the receptor and its counter-receptor are one mechanism and not two 1.
The 2026 KMT2A-r B-ALL result, and why it is the strongest thing in the niche layer
Rubio-Gayarre and colleagues report that ITGA4 and NG2 form a key biological axis in KMT2A-rearranged B-ALL - the index lineage, human cells, patient-derived xenografts. NG2 expression promotes proliferation and migration and is associated with Rho GTPase activity in an ITGA4-dependent manner; combined genetic ablation of ITGA4 and NG2 significantly delayed disease onset and prolonged survival in immunodeficient mice; and natalizumab delayed leukaemia progression and potentiated standard-of-care chemotherapy in KMT2A-r B-ALL PDX models 2. NG2/CSPG4 was previously established by the same group as a direct transcriptional target of KMT2A fusions, so this is a chain that runs fusion -> target gene -> integrin-dependent behaviour -> approved drug. This graph already holds mechanism-itga4-niche-adhesion-kmt2a and gene-cspg4; what it lacked was an entry for the integrin itself.
Epigenetics: the partner is fusion-controlled and the integrin is not measured
The epigenetic content of this axis sits on the NG2/CSPG4 side, where the locus is a direct, epigenetically regulated KMT2A-fusion target (see gene-cspg4, which records that H3K79me2 has nonetheless never been measured there). For ITGA4 itself there is no chromatin measurement in a KMT2A-rearranged system. ITGA4[tiab] AND (methylation[tiab] OR H3K4me3[tiab] OR chromatin[tiab]) returns 59 records (2026-09-22), overwhelmingly about ITGA4 promoter methylation as a peripheral-blood biomarker in autoimmune and cardiovascular cohorts, and none is a KMT2A study. Bounded over titles and abstracts (rule 80).
| gene symbol | ITGA4 |
|---|---|
| uniprot | P13612 |
| length AA | 1032 |
| accession resolution | gene_exact:ITGA4 AND organism_id:9606 AND reviewed:true returned exactly ONE reviewed human record, P13612 / ITA4_HUMAN, primary gene name read back as ITGA4, 1,032 aa, synonym CD49D. Fetched live 2026-09-22. |
| localization | PLASMA MEMBRANE, single-pass type I, 943 of 1,032 residues extracellular. Worth recording exactly: UniProt's only subcellular-location term for P13612 is 'Membrane' carrying ECO:0000255, a SEQUENCE-ANALYSIS inference rather than ECO:0000269 experimental support. For a protein whose surface expression is measured routinely as CD49d in diagnostic flow cytometry, the database's localisation annotation is uncoded - the class rule 91 describes. The functional localisation that matters is polarised and dynamic: integrin engagement clusters VLA-4 at the cell-stroma contact, which is where the Rho GTPase signal in E2 is generated. |
| subnuclear localization | NOT APPLICABLE BY TOPOLOGY, argued from architecture and not from a literature search. ITGA4 is a single-pass type I membrane receptor with a 31-residue cytoplasmic tail and no chromatin-binding module; 91% of the protein is outside the cell, so there is no mechanism for a subnuclear pool. This field does NOT claim that nuclear ITGA4 was sought and excluded - no such search was run in this beat, and UniProt's silence on the point is not evidence, for the reason recorded against E1. The nuclear consequence of the axis is indirect and real: NG2/CSPG4, the ITGA4 partner, is a direct transcriptional target of the KMT2A fusion, so the axis is switched on from the nucleus and executed at the membrane. |
| subnuclear absence class | not-applicable-by-topology |
| cell type | KMT2A-REARRANGED B-ALL BLASTS, in human cells and patient-derived xenografts [E2] - this is the only molecule in this tranche whose principal experiment is in the index lineage. Also murine MLL-AF9 AML cells, where the VCAM1-VLA-4 axis drives intracerebral invasion [E3]. ITGA4/CD49d is expressed on most leukocyte lineages and on HSPCs, which is the source of the drug's toxicity surface (natalizumab carries a PML risk in MS). |
| cell state | AN AGGRESSIVE, MIGRATORY, CHEMORESISTANT STATE. In KMT2A-r B-ALL the ITGA4-dependent Rho GTPase activity downstream of NG2 drives proliferation and migration, and removing it potentiates chemotherapy [E2]; NG2 positivity is independently associated with early relapse and glucocorticoid resistance. In the MLL-AF9 model the VCAM1-VLA-4 state is a CNS-INVASIVE one [E3], which connects to this graph's cell-state-cns-niche-adapted rather than to marrow retention. |
| epigenetic readout | THE AXIS IS EPIGENETICALLY DRIVEN ON THE PARTNER'S SIDE, NOT MEASURED ON THIS ONE. NG2/CSPG4 is an epigenetically regulated DIRECT transcriptional target of KMT2A fusions (see gene-cspg4), so the ITGA4-NG2 axis is a fusion-driven chromatin output executed at the membrane. For ITGA4 itself: ITGA4[tiab] AND (methylation[tiab] OR H3K4me3[tiab] OR chromatin[tiab]) returns 59 records (2026-09-22), dominated by ITGA4 promoter-methylation biomarker studies in autoimmune and cardiovascular cohorts, and none is in a KMT2A-rearranged system. No H3K79me2, H3K4me3 or menin occupancy measurement at ITGA4 in KMT2A-r disease. Bounded over titles and abstracts (rule 80). |
| shown for fusion | YES for KMT2A-r B-ALL as a class, in human cells and PDX, with a drug: NG2-ITGA4 axis, combined genetic ablation, and natalizumab potentiating standard-of-care chemotherapy [E2]. YES for murine MLL-AF9 in the CNS-invasion setting via VCAM1-VLA-4 [E3]. NOT SHOWN for KMT2A::AFF1 specifically - the abstract does not name the PDX fusion partners - and NOT SHOWN in an infant sample or under a menin inhibitor. |
| 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 P13612. Note the contrast: UniProt DOES record an ITGA4-CSPG4 interaction [E1], and CSPG4 is a fusion TARGET rather than a fusion PARTNER - so the physical link runs to the fusion's output, not to the fusion complex. |
| drug | Natalizumab, FDA- and EMA-approved anti-ITGA4 monoclonal antibody, repurposing proposed for KMT2A-r B-ALL on the strength of E2. No leukaemia trial: natalizumab[tiab] AND leukemia[tiab] returns 23 records (2026-09-22) and the KMT2A-r one is E2 itself; the rest are largely PML and MS-safety reports. Also used as a mobilising agent with motixafortide in a sickle-cell gene-therapy protocol (NCT05618301). |
| collision note | ITGA4 has no row in data/symbol-collisions.json (read 2026-09-22, case-sensitive over d['collisions'], 177 rows). Two search cautions measured here: the alias 'CD49d' is the flow-cytometry name and papers use it interchangeably, so a symbol-only query under-counts; and 'VLA-4' names the alpha-4/beta-1 DIMER, so a VLA-4 result is evidence about ITGA4 AND ITGB1 jointly and must not be attributed to ITGA4 alone - E3 is exactly that case. |
domains — the numbers behind the chart
[
{
"name": "Signal peptide",
"start": 1,
"end": 33
},
{
"name": "FG-GAP 1",
"start": 35,
"end": 100
},
{
"name": "FG-GAP 2",
"start": 110,
"end": 177
},
{
"name": "FG-GAP 3",
"start": 185,
"end": 237
},
{
"name": "FG-GAP 4",
"start": 238,
"end": 291
},
{
"name": "FG-GAP 5",
"start": 292,
"end": 351
},
{
"name": "FG-GAP 6",
"start": 355,
"end": 412
},
{
"name": "FG-GAP 7",
"start": 416,
"end": 478
},
{
"name": "SG1 motif",
"start": 606,
"end": 616
},
{
"name": "Extracellular topological domain",
"start": 35,
"end": 977
},
{
"name": "Transmembrane (helical)",
"start": 978,
"end": 1001
},
{
"name": "Cytoplasmic tail",
"start": 1002,
"end": 1032
},
{
"name": "GFFKR motif",
"start": 1003,
"end": 1007
}
]depmap essentiality
{
"depmap_common_essential": false,
"tissue_blocks": 29,
"n_screens": 1255,
"mean_gene_effect": -0.1387,
"min_gene_effect": -0.8087,
"max_gene_effect": 0.4295,
"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."
}- Natalizumab causes progressive multifocal leukoencephalopathy in multiple sclerosis by blocking lymphocyte trafficking into the CNS. Giving it to an infant already immunosuppressed by intensive chemotherapy is a different risk calculation from giving it to an ambulant adult with MS, and the PDX experiments in E2 are in immunodeficient mice, which cannot model it at all.
- The drug is a mobiliser. Anti-ITGA4 releases CD34+ cells from marrow - it is in active clinical use for exactly that - so the intervention that potentiates chemotherapy may simultaneously distribute blasts, and the MLL-AF9 data show the VCAM1-VLA-4 axis running to INTRACEREBRAL invasion [E3].
- Blocking one chain of a dimer leaves the other. ITGA4 pairs with beta-1 and beta-7; beta-1 also pairs with alpha-5 to bind fibronectin, so fibronectin adhesion has an alpha-5/beta-1 bypass that anti-alpha-4 does not touch.
- The axis partner is a diagnostic marker. NG2/CSPG4 surface positivity is used to call an 11q23 rearrangement; a therapy that pushes on this axis could move the marker used to identify the disease.
- Do the PDX models in E2 include a KMT2A::AFF1 / t(4;11) case? The index fusion's representation in the natalizumab result is the single most decision-relevant unknown for this lab.
- Does natalizumab disperse disease as well as detaching it? It is used clinically as a mobiliser, and no extramedullary or CNS readout is reported in E2.
- Is ITGA4 surface level itself fusion-dependent, i.e. does it fall under a menin inhibitor, or is only its partner NG2 fusion-driven?
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
The reviewed human ITGA4 record is UniProt P13612 (ITA4_HUMAN), 1,032 aa, a single-pass type I membrane protein with seven FG-GAP repeats, a GFFKR motif at 1003-1007, and documented ligands fibronectin CS-1/CS-5 and VCAM1 (recognising Q-I-D-S in VCAM1); it also records an interaction with CSPG4 through CSPG4's chondroitin sulfate glycosaminoglycan.
UniProtKB P13612 (ITA4_HUMAN) · UniProtKB 2026What this evidence cannot show
Resolved by gene_exact:ITGA4 AND organism_id:9606 AND reviewed:true, exactly ONE reviewed human record; primary gene name read back as ITGA4, length 1,032 aa, synonym CD49D. Fetched 2026-09-22. The single subcellular location term is 'Membrane' with only ECO:0000255 (sequence-analysis inference), not ECO:0000269 - so even the localisation of a protein this well studied is UNCODED in UniProt, which is the class rule 91 warns about.
Evidence that argues against this (1)
- A curated feature table is a summary of what curators accepted, not a statement that other features were sought and excluded, so nothing here licenses an absence claim. The demonstration in this graph is CXCR4: UniProt P61073 annotates no nuclear localisation signal while a published, mutated, functionally tested NLS at 146-RPRK-149 exists. The same caution applies to every UniProt-derived statement in this entry. Bao Y et al. 2019
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In KMT2A-rearranged B-ALL, NG2 and ITGA4 form an axis in which NG2-driven proliferation and migration are associated with Rho GTPase activity in an ITGA4-dependent manner; combined genetic ablation of ITGA4 and NG2 significantly delayed disease onset and prolonged survival in immunodeficient mice; and natalizumab, an FDA/EMA-approved anti-ITGA4 antibody, delayed leukaemia progression and potentiated standard-of-care chemotherapy in KMT2A-r B-ALL patient-derived xenografts.
NG2-ITGA4 axis regulates Rho GTPases and leukemic aggressiveness in KMT2A-r B-ALL and is targetable with natalizumab · Rubio-Gayarre A, Vinyoles M, Tejedor JR, et al. · Blood 2026What this evidence cannot show
Read from the abstract, which does not state which KMT2A fusion partners the PDX models carry, the number of PDX models, or the magnitude of the chemotherapy potentiation. 'KMT2A-r B-ALL' is a class; whether KMT2A::AFF1 is represented is not resolvable from the abstract (rule 80). The single-gene ablation effects are not separated from the combined effect in the abstract, so ITGA4's own contribution is asserted as part of an axis rather than measured alone.
Evidence that argues against this (1)
- Natalizumab is used clinically as a stem-cell MOBILISER in combination with motixafortide (NCT05618301, a sickle-cell gene-therapy mobilisation study), so anti-ITGA4 in a leukaemia releases blasts as well as detaching them. The same mechanism that potentiates chemotherapy here is the mechanism that would disperse disease, and this abstract does not report extramedullary or CNS burden. ClinicalTrials.gov 2026
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
In MLL-AF9-induced mouse AML, IRF7 loss accelerated disease with intracerebral invasion and raised VCAM1, and blocking the VCAM1-VLA-4 axis delayed disease progression and attenuated intracerebral invasion of AML cells.
Loss of IRF7 accelerates acute myeloid leukemia progression and induces VCAM1-VLA-4 mediated intracerebral invasion · Oncogene 2022What this evidence cannot show
Murine myeloid MLL-AF9 with an engineered IRF7 knockout, so the VCAM1 elevation is downstream of a genetic perturbation rather than a property of the fusion, and the setting is CNS invasion rather than marrow retention. VLA-4 is the alpha-4/beta-1 dimer, so this result is as much about ITGB1 as about ITGA4. Author names not transcribed; title, DOI, journal and year taken from the fetched efetch record.
Evidence that argues against this (1)
- The IMPDH-inhibitor study reports the opposite sign for this axis in the same murine MLL-AF9 model: VCAM1 UPREGULATION contributed to an ANTI-leukaemic effect. So a VCAM1-VLA-4 blockade rationale rests on one of two in vivo experiments that disagree about direction, and both are myeloid mouse. Liu X et al. 2023
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
ITGA4 is not a cell-intrinsic dependency in DepMap: 29 tissue blocks, 1,255 screens, mean gene effect -0.1387, range -0.8087 to 0.4295, zero lines at or below a gene effect of -1.0.
Open Targets Platform GraphQL - target.depMapEssentiality for ENSG00000115232 (ITGA4), with meta.dataVersion · Open Targets Platform / DepMap 2026What 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, not a finding: DepMap screens cells in monoculture with no stroma and no counter-receptor-bearing cell, so an adhesion receptor acting in trans cannot score however essential it is in vivo. ITGA4 has the most negative mean of the eight niche genes measured in this tranche and still never reaches -1.0 in any line. The in vivo result [E2] is where the requirement appears, and it is the measurement that should be quoted.
Evidence that argues against this (1)
- The same assay class returns the same uninformative null for CXCR4 (mean +0.0399, zero lines at or below -1.0) while an in vivo syngeneic CRISPR screen in murine MLL-AF9 made CXCR4 a TOP cell-surface dependency whose deletion eradicated leukaemia. A monoculture null for a niche gene is therefore refuted as a statement about requirement, by direct example. Ramakrishnan R et al. 2020
the integrin the mechanism is named after
NG2/CSPG4 interacts with integrin alpha-4 via its chondroitin sulfate glycosaminoglycan; the axis is ITGA4-dependent
- related to →KMT2A-rearranged acute lymphoblastic leukaemia
ITGA4-NG2 axis drives aggressiveness in KMT2A-r B-ALL and is targetable with natalizumab
VCAM1-VLA-4 blockade attenuates intracerebral invasion in MLL-AF9 mouse AML
adhesion arm of the niche, alongside the CXCL12-CXCR4 chemokine arm
integrin alpha-4/beta-1 (VLA-4) recognises Q-I-D-S in VCAM1
one record in the entire literature, published 2026