Oncofusion

start / graph / gene-cxcr4

Genestrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q33-niche-protein-layer

CXCR4 - named in eighteen entries of this graph with no entry of its own, and the one niche molecule with a drug, a paediatric trial and a direct KMT2A-r experiment

also called CD184 · fusin · LESTR · NPYRL · C-X-C chemokine receptor type 4

CXCR4 is the receiver on a blood cell that reads the marrow's hold-still signal, and blocking it with the licensed drug plerixafor pushes cells out of the bone into the blood. In infant KMT2A-rearranged leukaemia, blocking it made a targeted drug work far better in mice - and no trial has ever given it to an infant with this disease.

The same thing, for a scientist

CXCR4 (UniProt P61073, 352 aa, seven-transmembrane GPCR) is the receptor half of the marrow retention axis and the best-evidenced niche dependency in KMT2A-rearranged disease. Induced Cxcr4 deletion in adult mice collapses the HSC pool [E1]. In primary infant MLL-rearranged ALL, stroma protects blasts differentially relative to non-MLL-rearranged ALL, soluble stromal factors suffice, stroma protects both bulk blasts and LSCs from the FLT3 inhibitor lestaurtinib, and CXCR4 inhibition with AMD3100/plerixafor markedly enhanced lestaurtinib efficacy in a xenograft [E2] - this is the index-lineage, human-primary-sample result the rest of the niche literature lacks. In murine MLL-AF9 a cell-surface CRISPR screen made CXCR4 a top dependency, and deletion eradicated leukaemia WITHOUT impairing homing, with the phenotype tracking to oxidative stress and differentiation rather than adhesion [E3]. MEIS1 drives CXCL12/CXCR4 signalling indirectly by transactivating SYTL1, which controls CXCR4 intracellular trafficking [E4], and in RS4;11 (KMT2A::AFF1) knockdown of MEIS1 or HOXA7/9/10 hampers the axis [E5]. The receptor is NOT a cell-intrinsic dependency in DepMap - mean gene effect +0.0399 over 1,253 lines, zero lines at or below -1.0 [E6] - and that contradiction with the in vivo screen is the most useful single fact in this entry.

What it looks like
CXCR4352 aa
N-teN-terminal extracellular region (chemokine binding, HIV-1 coreceptor) · residues 1–21Extracellular topological domainExtracellular topological domain · residues 1–38TM1 TM1 (helical) · residues 39–63TM2 TM2 (helical) · residues 78–99TM3 TM3 (helical) · residues 111–130CytoCytoplasmic loop 2 (contains the reported NLS 146-RPRK-149) · residues 131–154TM4 TM4 (helical) · residues 155–174TM5 TM5 (helical) · residues 196–216TM6 TM6 (helical) · residues 242–261TM7 TM7 (helical) · residues 283–302Cytoplasmic C-terminal tailCytoplasmic C-terminal tail · residues 303–352DisoDisordered C-terminus · residues 329–3520100200300

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 a compartment UniProt does not annotate

P61073 is a canonical class-A GPCR: extracellular N-terminus 1-38, seven helical transmembrane segments (39-63, 78-99, 111-130, 155-174, 196-216, 242-261, 283-302), a cytoplasmic C-terminus 303-352 that is disordered from 329, and mapped chemokine-binding patches at 94-97, 113-117 and 186-190 [E1b]. UniProt's subcellular location comment lists cell membrane, cell junction, early endosome, late endosome and lysosome - the itinerary of a receptor that is internalised on ligand binding - and no nucleus.

It is nonetheless reported in the nucleus. In renal cell carcinoma, CXCR4 nuclear localisation is enriched in metastases, associates with poor prognosis, and requires a nuclear localisation sequence at residues 146-RPRK-149; mutating it abolishes nuclear CXCR4, and nuclear CXCR4 co-localises and interacts with HIF-1-alpha in a feed-forward loop 7. Residues 146-149 fall inside UniProt's cytoplasmic topological domain 131-154, so the claim is architecturally coherent - and UniProt annotates no NLS on this record. That is the same shape as rule 91's MEN1 finding: a mapped, published, functionally tested localisation signal that the database's feature table does not carry.

The KMT2A-rearranged evidence, and it is better than the beat brief assumed

The premise that niche biology has never been measured in KMT2A-rearranged disease specifically is false, and the counter-example is in the index lineage. Sison and colleagues used primary infant MLL-rearranged ALL samples, showed MLL-R ALL is differentially protected by stroma from spontaneous apoptosis relative to non-MLL-R ALL, showed stroma-produced soluble factors alone promote proliferation and cell-cycle entry, showed stroma protects both bulk blasts and the LSC fraction from lestaurtinib, and then showed in a xenograft that AMD3100 (plerixafor) markedly enhanced lestaurtinib efficacy 2. The same group later showed the CXCR4 antagonist POL5551 chemosensitises paediatric ALL 8.

What is genuinely absent is narrower and should be stated that way: no experiment reports a CXCL12/CXCR4 measurement in KMT2A::AFF1 specifically as opposed to MLL-rearranged as a class, no trial of a CXCR4 antagonist has enrolled an infant with leukaemia, and nobody has measured what a menin inhibitor does to this axis (see absence-cxcr4-axis-never-measured-under-menin-inhibition).

DepMap says not essential; an in vivo screen says top dependency. Both are right.

Open Targets/DepMap, release dataVersion 26/06: 29 tissue blocks, 1,253 screens, mean gene effect +0.0399, range -0.3603 to +0.53, and zero lines at or below -1.0 6. In the syngeneic MLL-AF9 mouse, CXCR4 was a top cell-surface dependency and its deletion eradicated leukaemia 3.

The reconciliation is the assay: DepMap screens are monoculture, and a receptor whose ligand is supplied by a stromal cell has nothing to receive in a dish. This is the clearest example in this graph of a dependency class that DepMap cannot see by construction, and it is a caution for every DepMap-based argument about niche genes: a flat DepMap profile for a receptor, adhesion molecule or chemokine is uninformative rather than reassuring. It is also a reason not to quote a median-split position for these genes (rule 84).

Epigenetics: the CXCR4 locus is bivalent in at least one human primary cell type, and unmeasured in this disease

In human endothelial colony-forming cells, pro-angiogenic pathway genes including CXCR4 are held under bivalent H3K27me3/H3K4me3 marks, and an epigenetic-drug combination that resolves the bivalency toward H3K4me3 activates CXCR4 signalling along with VEGFR, WNT, NOTCH and SHH 9. So the locus is a bivalency-controlled, pharmacologically movable gene in at least one primary human cell type.

In a KMT2A-rearranged system nobody has looked: CXCR4[tiab] AND (H3K79[tiab] OR H3K4me3[tiab] OR DOT1L[tiab]) returns 6 records and none is a KMT2A study, and KMT2A[tiab] AND CXCR4[tiab] AND (ChIP[tiab] OR occupancy[tiab] OR promoter[tiab]) returns exactly one record, which is a T-ALL RUNX2 paper. H3K79me2 is the mark this graph's DOT1L arm predicts at every direct KMT2A-fusion target, and CXCR4 - a gene whose function the fusion's own HOXA/MEIS1 output controls 45 - has never had it measured.

Facts and figures
gene symbolCXCR4
uniprotP61073
length AA352
accession resolutiongene_exact:CXCR4 AND organism_id:9606 AND reviewed:true returned exactly ONE reviewed human record (n=1), P61073 / CXCR4_HUMAN, primary gene name read back as CXCR4, length 352 aa. Fetched live 2026-09-22. The record carries no gene-name synonyms, so there is no rule-62 ambiguity here.
localizationPLASMA MEMBRANE AND THE ENDOCYTIC ITINERARY. UniProt P61073 records cell membrane (ECO:0000269 x3 plus ECO:0000305), cell junction, early endosome, late endosome and lysosome, each ECO:0000269 - the expected profile for a GPCR that is internalised and degraded on ligand engagement. The endosomal and lysosomal terms are mechanistically load-bearing rather than incidental here, because the MEIS1 link to this axis runs through SYTL1 control of CXCR4 INTRACELLULAR TRAFFICKING [E4], i.e. through exactly those compartments.
subnuclear localizationNUCLEAR ENTRY IS A REAL PUBLISHED CLAIM; A SUBNUCLEAR COMPARTMENT IS NOT, AND NEITHER HAS BEEN LOOKED FOR IN THIS DISEASE. (1) What is established: CXCR4 nuclear localisation occurs in renal cell carcinoma, is enriched in metastases, requires a mapped NLS at residues 146-RPRK-149 whose mutation abolishes it, and nuclear CXCR4 forms a feed-forward loop with HIF-1-alpha [E7]. Residues 146-149 lie inside UniProt's cytoplasmic topological domain 131-154, so the claim is architecturally coherent - and UniProt annotates NO NLS on P61073, the same database gap rule 91 documents for MEN1. (2) What is absent: no assignment to a subnuclear compartment (nuclear speckle, nucleolus, lamina, chromatin) in any tissue, and no search for nuclear CXCR4 in any KMT2A-rearranged blast. ABSENCE CLASS for the KMT2A-r question: never-measured - no instrument has been pointed at it, and the nuclear-CXCR4 antibody/fractionation work that exists is in solid tumours.
subnuclear absence classnever-measured
cell typeHAEMATOPOIETIC STEM CELLS (Cxcr4 deletion collapses the adult HSC pool, E1; this graph holds cell-type-hsc-bone-marrow) and LEUKAEMIC BLASTS AND LEUKAEMIA STEM CELLS - primary infant MLL-rearranged ALL blasts and the LSC fraction within them, both protected by stroma and both sensitised by AMD3100 [E2], and murine MLL-AF9 LSCs in which CXCR4 is a top dependency [E3]. CXCR4 is expressed far more widely than this (T cells, monocytes, endothelium, neurons), which is the reason the toxicity surface of a CXCR4 antagonist is broad.
cell stateTWO STATES, opposite in sign. In normal haematopoiesis CXCR4 signalling maintains the QUIESCENT HSC state and pool size [E1]. In leukaemia it maintains a CHEMOPROTECTED, STROMA-DEPENDENT blast state - and the MLL-AF9 result shows the state it holds is a LOW-ROS, UNDIFFERENTIATED one, because losing CXCR4 signalling produced oxidative stress and differentiation in vivo rather than a homing defect [E3]. This graph holds cell-state-differentiation-block; the E3 result argues CXCR4 signalling is one of the inputs that sustains it.
epigenetic readoutTHE LOCUS IS BIVALENT AND DRUGGABLE IN ONE PRIMARY HUMAN CELL TYPE, AND UNMEASURED IN THIS DISEASE. In human endothelial colony-forming cells CXCR4 sits among pro-angiogenic genes held under bivalent H3K27me3/H3K4me3, and resolving the bivalency toward H3K4me3 with an epigenetic-drug combination activates CXCR4 signalling [E9]. In a KMT2A-rearranged system there is no measurement: CXCR4[tiab] AND (H3K79[tiab] OR H3K4me3[tiab] OR DOT1L[tiab]) returns 6 records, none KMT2A (the top hits are a diabetic gastroparesis ChIP-seq study and the endothelial-progenitor paper above); KMT2A[tiab] AND CXCR4[tiab] AND (ChIP[tiab] OR occupancy[tiab] OR promoter[tiab]) returns 1 record, a T-ALL RUNX2 paper. H3K79me2 is the mark this graph's DOT1L arm predicts at direct KMT2A-fusion targets and it has never been measured at CXCR4. Bounded over titles and abstracts (rule 80); queried 2026-09-22.
shown for fusionYES, and in more than one fusion, which makes this the strongest shown_for_fusion in the niche layer. MLL-rearranged infant B-ALL (primary human samples plus xenograft): stromal protection is differential versus non-MLL-R ALL and AMD3100 potentiates lestaurtinib [E2]. Murine MLL-AF9: top cell-surface dependency by CRISPR screen, deletion eradicates leukaemia, homing intact [E3]. KMT2A::AFF1 (index fusion, RS4;11): the CXCR4/SDF-1 response depends on MEIS1 and HOXA7/9/10 [E5]. NOT SHOWN: any measurement of CXCR4 surface level, signalling or occupancy in a KMT2A::AFF1 patient sample under treatment; nuclear CXCR4 in any leukaemia; and anything at all about this axis under a menin inhibitor [E10].
intact union membershipABSENT from the 277-partner IntAct REST union of the six core baits recorded in analyses/interactome-union-migrated-to-rest/results.json (generated 2026-09-22); checked on both the symbol and the accession P61073. Expected - CXCR4 acts on the fusion program transcriptionally and through MEIS1/SYTL1, not by binding the complex.
clinical exposurePlerixafor (AMD3100) is licensed, and two paediatric leukaemia trials of CXCR4 antagonism exist and BOTH EXCLUDE INFANTS. NCT01319864 (POETIC, Seattle Children's, plerixafor as a chemosensitiser in relapsed paediatric acute leukaemia and MDS) is phase 1, COMPLETED, n=20 actual, eligibility 3 years to 29 years. NCT01655875 (Emory, AMD3100 as a sensitising agent before myeloablative allogeneic transplant in chemotherapy-resistant paediatric acute leukaemia) is phase 1, TERMINATED with whyStopped fetched from the raw ClinicalTrials.gov API as 'lack of patient enrollment', n=1 actual, eligibility 2 years to 22 years. Fetched 2026-09-22 from the v2 API, not the MCP tool, precisely so whyStopped was visible.
collision noteCXCR4 has no row in data/symbol-collisions.json (read 2026-09-22, case-sensitive over d['collisions'], 177 rows) and needs none. Note the second CXCL12 receptor is ACKR3, PREVIOUSLY NAMED CXCR7 - rule 69 applies: a previous official symbol is not an alias, and a search for ACKR3 alone under-counts a literature written as CXCR7. Both tokens return 0 mentions in this corpus.
domains
N-terminal extracellular …1.000N-terminal extracellular region (chemokine binding, HIV-1 coreceptor): 1Extracellular topological…1.000Extracellular topological domain: 1TM1 (helical)39.0TM1 (helical): 39TM2 (helical)78.0TM2 (helical): 78TM3 (helical)111.0TM3 (helical): 111Cytoplasmic loop 2 (conta…131.0Cytoplasmic loop 2 (contains the reported NLS 146-RPRK-149): 131TM4 (helical)155.0TM4 (helical): 155TM5 (helical)196.0TM5 (helical): 196TM6 (helical)242.0TM6 (helical): 242TM7 (helical)283.0TM7 (helical): 283Cytoplasmic C-terminal ta…303.0Cytoplasmic C-terminal tail: 303Disordered C-terminus329.0Disordered C-terminus: 329
domains — the numbers behind the chart
[
  {
    "name": "N-terminal extracellular region (chemokine binding, HIV-1 coreceptor)",
    "start": 1,
    "end": 21
  },
  {
    "name": "Extracellular topological domain",
    "start": 1,
    "end": 38
  },
  {
    "name": "TM1 (helical)",
    "start": 39,
    "end": 63
  },
  {
    "name": "TM2 (helical)",
    "start": 78,
    "end": 99
  },
  {
    "name": "TM3 (helical)",
    "start": 111,
    "end": 130
  },
  {
    "name": "Cytoplasmic loop 2 (contains the reported NLS 146-RPRK-149)",
    "start": 131,
    "end": 154
  },
  {
    "name": "TM4 (helical)",
    "start": 155,
    "end": 174
  },
  {
    "name": "TM5 (helical)",
    "start": 196,
    "end": 216
  },
  {
    "name": "TM6 (helical)",
    "start": 242,
    "end": 261
  },
  {
    "name": "TM7 (helical)",
    "start": 283,
    "end": 302
  },
  {
    "name": "Cytoplasmic C-terminal tail",
    "start": 303,
    "end": 352
  },
  {
    "name": "Disordered C-terminus",
    "start": 329,
    "end": 352
  }
]
depmap essentiality
{
  "depmap_common_essential": false,
  "tissue_blocks": 29,
  "n_screens": 1253,
  "mean_gene_effect": 0.0399,
  "min_gene_effect": -0.3603,
  "max_gene_effect": 0.53,
  "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 run in the same call and the same code path: MEN1 returned 29 tissue blocks, 1,258 screens, mean gene effect -0.3616, minimum -2.5262 and 79 lines at or below -1.0. So the zero count of dependent lines is a measured zero and not a silent empty payload with a passing approvedSymbol readback. DepMap release recorded as Open Targets dataVersion year 26 month 06 (rule 83: the payload carries no DepMap release id; this is the OT version), 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 a gene-effect of -1.0.",
  "interpretation": "Flat and slightly positive in vitro while a top dependency in an in vivo syngeneic MLL-AF9 screen [E3]. The discrepancy is the assay, not the gene, and it generalises: DepMap cannot express a dependency on a signal supplied by another cell."
}
How this could bite us back
  • MOBILISATION IS DOUBLE-EDGED. The licensed use of plerixafor is to push stem cells out of marrow into blood. Doing that to a leukaemia releases blasts into circulation; whether that sensitises them or seeds extramedullary and CNS disease is not established in KMT2A-rearranged disease, and this graph separately holds a CNS-adapted cell state (cell-state-cns-niche-adapted) and an intracerebral-invasion result driven by VCAM1-VLA-4 (gene-vcam1 E3).
  • THE TARGET IS REQUIRED BY THE PATIENT'S OWN STEM CELLS. E1 is both the reason the axis matters and the toxicity argument: induced Cxcr4 deletion collapses the adult HSC pool and increases myelotoxic sensitivity. An infant receiving intensive chemotherapy has no marrow reserve to spare.
  • THE MECHANISM MAY NOT BE THE ONE THE STRATEGY ASSUMES. In MLL-AF9, Cxcr4 loss eradicated leukaemia with homing INTACT and a redox/differentiation phenotype [E3]. If that is the operative mechanism, a short antagonist pulse timed to chemotherapy - the mobilisation logic - is the wrong schedule, and sustained signalling blockade is required.
  • ONE TRIAL IN THIS SPACE ALREADY FAILED FOR NON-BIOLOGICAL REASONS. NCT01655875 terminated at n=1 for 'lack of patient enrollment'. In a disease of ~1 infant per 500,000 births a mechanistically attractive combination can still be unrunnable, and a proposal here must state its accrual route.
  • RESISTANCE ROUTE ALREADY VISIBLE: the second CXCL12 receptor, ACKR3 (formerly CXCR7), is not blocked by plerixafor and is entirely unmeasured in this disease (0 mentions of either token in this corpus). A CXCR4-only blockade has an unexamined bypass.
Still open
  1. Does a menin inhibitor lower CXCR4 on a KMT2A::AFF1 blast? MEIS1 and HOXA9 collapse under menin inhibition and both sit upstream of this axis [E4, E5], so the prediction is sharp and the measurement has never been reported [E10]. The RNA-seq to answer it is already deposited.
  2. Is nuclear CXCR4 present in KMT2A-rearranged blasts, and does the 146-RPRK-149 NLS [E7] matter in a haematopoietic cell? Nobody has looked in any leukaemia.
  3. Which of the two mechanisms - retention/adhesion or redox/differentiation - drives the chemosensitisation seen with AMD3100 in MLL-R ALL [E2]? E3 argues against the retention reading in the myeloid model, and the two predict different combination partners.
  4. Would a CXCR4 antagonist be safe in an infant? Both existing paediatric trials set a floor of 2-3 years, and an infant's marrow is still the primary site of a recently completed developmental transition from fetal liver.
Every source behind this entry (11)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Induced deletion of CXCR4 in adult mice severely reduces HSC numbers and increases sensitivity to myelotoxic injury without impairing expansion of more mature progenitors, and most HSCs contact CXCL12-abundant reticular cells.

    PMID 17174120DOI
    What this evidence cannot show

    Mouse, normal haematopoiesis. This is the load-bearing normal-tissue result and therefore also the principal toxicity argument against CXCR4 blockade: the axis a therapy would break is the one that maintains the patient's own stem cell pool.

    Evidence that argues against this (1)
    • Mouse, and normal HSC maintenance rather than leukaemia. An induced deletion that reduces the HSC pool and increases myelotoxic sensitivity describes a physiological requirement, and it cuts against therapeutic CXCR4 blockade as much as for it: the same dependency exists in the normal stem cells an infant needs to keep.
  2. E1bestablished confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    The reviewed human CXCR4 record is UniProt P61073 (CXCR4_HUMAN), 352 aa, a seven-transmembrane receptor with cell membrane, cell junction, early endosome, late endosome and lysosome subcellular locations and no annotated nuclear localisation signal.

    What this evidence cannot show

    Resolved by gene_exact:CXCR4 AND organism_id:9606 AND reviewed:true, which returned exactly ONE record; primary gene name read back as CXCR4, length 352 aa, fetched 2026-09-22. The record carries no synonyms. Four of the five location terms carry ECO:0000269 and the cell-membrane term additionally carries ECO:0000305 (curator inference); per rule 91 an evidence code means a curator saw an experiment, not that the annotation is complete - and the missing NLS (E7) proves that here.

    Evidence that argues against this (1)
    • THIS ENTRY CONTRADICTS ITSELF AND THE CONTRADICTION IS INFORMATIVE. UniProt annotates no nuclear localisation for P61073, while E7 of this same entry cites nuclear CXCR4 in renal cell carcinoma with a mapped nuclear localisation sequence at 146-RPRK-149. A curated subcellular-location list is a summary of what curators accepted, not a statement that other locations were looked for and excluded, and 'no annotated nuclear localisation' must not be reported as evidence of nuclear absence.
  3. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In primary infant MLL-rearranged ALL, stroma differentially protects blasts from spontaneous apoptosis relative to non-MLL-rearranged ALL, stroma-produced soluble factors alone promote proliferation and cell-cycle entry, stroma protects both bulk blasts and LSCs from the FLT3 inhibitor lestaurtinib, and CXCR4 inhibition with AMD3100 (plerixafor) markedly enhanced lestaurtinib efficacy in an MLL-R ALL xenograft.

    PMID 23294096DOI
    What this evidence cannot show

    MLL-rearranged as a CLASS, not KMT2A::AFF1 specifically - the abstract does not break the primary samples down by partner gene, so how much of this is t(4;11) is unknown from the abstract (rule 80: figures not read). The potentiated agent is lestaurtinib, a FLT3 inhibitor that was not adopted, so the combination as tested is not a current regimen. Read from the abstract.

    Evidence that argues against this (1)
    • The companion murine MLL-AF9 study points the other way on MECHANISM: Cxcr4 deletion eradicated leukaemia with marrow homing INTACT and a phenotype of oxidative stress and differentiation, and CXCL12 itself was dispensable. If that is the operative mechanism then the stromal protection reversed by AMD3100 here is not a retention or adhesion effect, and the mobilisation rationale for scheduling a CXCR4 antagonist around chemotherapy does not follow. Ramakrishnan R et al. 2020
  4. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In a syngeneic murine MLL-AF9 AML model a cell-surface CRISPR-Cas9 screen identified CXCR4 as a top regulator of leukaemia growth and survival; Cxcr4 deletion eradicated leukaemia in vivo WITHOUT impairing marrow homing, CXCL12 was dispensable, mutant-receptor experiments showed signalling is what is required, and loss of CXCR4 signalling caused oxidative stress and differentiation in vivo.

    CXCR4 Signaling Has a CXCL12-Independent Essential Role in Murine MLL-AF9-Driven Acute Myeloid Leukemia · Ramakrishnan R, Pena-Martinez P, Agarwal P, et al. · Cell Rep 2020
    PMID 32460032DOI
    What this evidence cannot show

    Murine myeloid MLL-AF9, not infant B-lineage KMT2A::AFF1 (rule 73). This result WEAKENS the retention/adhesion reading of the axis: homing was intact and the phenotype was redox and differentiation, so 'CXCR4 antagonist mobilises the blast out of its niche' is not the mechanism this paper supports.

    Evidence that argues against this (1)
    • Murine MLL-AF9 AML, not human KMT2A::AFF1 infant B-ALL - a different fusion, lineage, species and age. And the finding is explicitly CXCL12-INDEPENDENT, which undercuts the ligand-receptor framing this beat was dispatched to build: whatever CXCR4 is doing here, the marrow-retention story is not obviously it, since deletion eradicated leukaemia WITHOUT impairing marrow homing.
  5. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    MEIS1 promotes leukaemic homing and engraftment by transactivating synaptotagmin-like 1 (SYTL1), which determines intracellular trafficking of CXCR4 and thereby activates CXCL12/CXCR4-dependent migration; replacing SYTL1 in MEIS1-deficient cells restored migration and engraftment.

    PMID 27018596DOI
    What this evidence cannot show

    Murine HOXA9/MEIS1 myeloid leukaemogenesis models, not a KMT2A fusion model and not B-lineage (rule 73). The link to the fusion is indirect but not speculative: MEIS1 is a canonical, menin-dependent output of the KMT2A fusion program, so this paper places CXCR4 function downstream of that program via a named intermediate. Read from the abstract.

    Evidence that argues against this (1)
    • The mechanism is trafficking, not expression - MEIS1 acts through SYTL1 on where CXCR4 goes inside the cell - so this supports a MEIS1-CXCR4 connection without supporting the claim that CXCR4 levels are a fusion-driven output. It is also a different model system from the index disease.
  6. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In RS4;11 (human MLL-AF4 / KMT2A::AFF1 B-precursor line), separate knockdown of MEIS1, HOXA7, HOXA9 or HOXA10 each hampered the CXCR4/SDF-1 axis - reduced migration toward an SDF-1 gradient and loss of SDF-1-augmented proliferation - with similar homing but slowed marrow colonisation in NOD/SCID mice.

    PMID 21518888DOI
    What this evidence cannot show

    The abstract does not say whether CXCR4 surface protein fell or whether only the response fell, and does not give effect sizes. Also note the concordance with E3: homing similar, colonisation slowed - both papers separate homing from retention/expansion, which is the distinction a mobilisation strategy depends on.

    Evidence that argues against this (1)
    • A single human line (RS4;11), knockdown rather than knockout, in-vitro migration. MEIS1 and HOXA knockdown reduce proliferation and engraftment generally in KMT2A-rearranged cells, so the impaired SDF-1 response is compatible with a general fitness defect rather than a specific axis effect.
  7. E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    CXCR4 is not a cell-intrinsic dependency in DepMap: 29 tissue blocks, 1,253 screens, mean gene effect +0.0399, range -0.3603 to +0.53, zero lines at or below a gene effect of -1.0.

    What this evidence cannot show

    Live zero-screen control run in the same call and the same code path: MEN1 returned 29 tissue blocks, 1,258 screens, mean gene effect -0.3616, minimum -2.5262 and 79 lines at or below -1.0. So the zero count of dependent lines is a measured zero and not a silent empty payload with a passing approvedSymbol readback. DepMap release recorded as Open Targets dataVersion year 26 month 06 (rule 83: the payload carries no DepMap release id; this is the OT version), 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 a gene-effect of -1.0. The result must not be read as evidence against CXCR4 as a target: DepMap is monoculture and this receptor's ligand comes from another cell, so the assay cannot express the phenotype. The in vivo syngeneic screen in E3 made the same gene a top dependency. A flat DepMap profile for a niche receptor is uninformative, not reassuring.

    Evidence that argues against this (1)
    • A flat DepMap profile is contradicted as a statement about requirement by the in vivo syngeneic screen in which the same gene was a TOP cell-surface dependency and its deletion eradicated leukaemia. Both measurements are correct; only the in vitro one is generalisable to monoculture. Ramakrishnan R et al. 2020
  8. E7moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    CXCR4 nuclear localisation occurs in renal cell carcinoma tissue, is enriched in metastases and associated with poor prognosis, requires a nuclear localisation sequence at residues 146-RPRK-149 whose mutation abolishes it, and nuclear CXCR4 interacts with HIF-1-alpha in a feed-forward loop in which nuclear HIF-1-alpha in turn promotes CXCR4 transcription.

    PMID 30177838DOI
    What this evidence cannot show

    Renal cell carcinoma, not leukaemia - this entry cites it to answer the question 'is nuclear CXCR4 a real claim' and the answer is yes, in RCC, with an NLS mapped and mutated. It says nothing about leukaemia. No SUBNUCLEAR compartment (speckle, nucleolus, lamina) is assigned in this paper or anywhere else, and nuclear CXCR4 has never been looked for in a KMT2A-rearranged blast. Read from the abstract.

    Evidence that argues against this (1)
    • UniProt's curated feature table for P61073 annotates NO nuclear localisation signal and NO nuclear compartment, so the database and this paper disagree. Per rule 91 that is an argument about annotation completeness rather than about the experiment, but a reader relying on UniProt alone would conclude nuclear CXCR4 does not exist. UniProtKB 2026
  9. E8moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    POL5551, a Protein Epitope Mimetic CXCR4 antagonist, is potent, mobilises haematopoietic stem cells, and enhances sensitivity to chemotherapy in paediatric ALL.

    PMID 26360610DOI
    What this evidence cannot show

    Paediatric ALL broadly; the abstract as fetched does not establish how much of the effect is in MLL-rearranged samples specifically, and POL5551 has not progressed to a paediatric leukaemia trial. Oncotarget as a venue warrants more than usual caution about follow-up. Read from the abstract.

    Evidence that argues against this (1)
    • PRECLINICAL AND UNDEVELOPED. POL5551 was reported in Oncotarget in 2015 and no clinical programme in paediatric ALL followed it, which is itself evidence about how the field judged the result. Mobilising blasts out of a protective niche is also double-edged - the same manoeuvre that sensitises to chemotherapy releases leukaemic cells into the circulation - and no trial has shown that trade is favourable in infants.
  10. E9moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In human endothelial colony-forming cells, pro-angiogenic pathways including CXCR4 are repressed by bivalent H3K27me3/H3K4me3 marks, and an epigenetic-drug combination that resolves the bivalency toward H3K4me3 simultaneously activates CXCR4, VEGFR, WNT, NOTCH and SHH signalling and improves vascular repair.

    PMID 29033304DOI
    What this evidence cannot show

    Human endothelial colony-forming cells, not haematopoietic and not leukaemic. Cited here for one narrow purpose: that the CXCR4 locus is bivalency-controlled and pharmacologically movable in a primary human cell type, which makes the absence of any chromatin measurement at CXCR4 in KMT2A-rearranged blasts a real gap rather than an idle one. Author names deliberately not transcribed; only the fetched title, DOI, journal and year are stored.

    Evidence that argues against this (1)
    • Human endothelial colony-forming cells, not haematopoietic cells and not leukaemia. A bivalent H3K27me3/H3K4me3 configuration at CXCR4 in endothelial progenitors says nothing about how CXCR4 is regulated in a KMT2A-rearranged blast, and the epigenetic-drug combination used to resolve the bivalency is not a therapy for this disease.
  11. E10strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    No paper names menin and CXCR4 together in its title or abstract: menin[tiab] AND CXCR4[tiab] returns 0 records, with valid query translation and an explicit 'No items found.', against operands of 1,264 (menin) and 17,210 (CXCR4).

    What this evidence cannot show

    Rule 67 positive control through the IDENTICAL query shape: menin[tiab] AND HOXA9[tiab] returns 33, so the AND composes and the zero is not an instrument artefact. querytranslation read back as "menin"[Title/Abstract] AND "CXCR4"[Title/Abstract]; one tag per term (rule 94), no parenthesised group. Bounded over titles and abstracts only (rule 80) - a CXCR4 row inside a menin-inhibitor RNA-seq supplementary table would not be visible to this query, which is exactly why the matching absence entry is classed measured-not-analysed rather than never-measured. Queried 2026-09-22.

    Evidence that argues against this (1)
    • A TITLE-AND-ABSTRACT QUERY IS NOT A SEARCH OF THE LITERATURE. menin[tiab] AND CXCR4[tiab] returning zero bounds what appears in titles and abstracts; it misses full text, methods, supplementary tables and conference abstracts, which is exactly where a subsidiary connection between a drug target and a niche receptor would appear. The honest claim is that no paper is ABOUT both, not that no connection has been reported.
Connected to