Oncofusion

start / graph / target-ng2-cspg4-the-fusion-driven-antigen-that-normal-vasculature-kills

Targetmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q138-kmt2a-aff1-surfaceome

NG2/CSPG4 is the one antigen in the whole surfaceome that KMT2A::AFF1 itself drives - and it is a pericyte and vascular protein, at 124.5 nTPM in normal blood vessel and 246 TPM in tibial artery

also called NG2 · CSPG4 as a target · MCSP · 7.1 antigen

NG2 is the only surface protein this leukaemia makes because of its fusion gene rather than because of the cell type it came from, which is exactly the kind of target that a leukaemia cannot escape by changing cell type. The problem is that the same protein coats the cells wrapped around blood vessels throughout the body, so a therapy aimed at it would also be aimed at the child's blood vessels.

The same thing, for a scientist

Across 2,404 surface proteins, **CSPG4/NG2 is the strongest fusion-enrichment signal in the screen**: 55.0x a 13-line B-cell-precursor leukaemia comparator (30.9 nTPM in SEM+RS4;11 against 0.56) and 3.05x the KMT2A-rearranged-other-partner arm [E1]. That is not a correlation this beat discovered - **NG2 is an epigenetically regulated direct target gene of MLL-AF4**, shown in Blood in 2024 [E2], and it is the immunophenotypic surrogate for 11q23/KMT2A rearrangement going back to the 7.1 monoclonal in 1996 [E6]. It is therefore precisely the fusion-driven, escape-resistant antigen this project has been looking for. **And the normal-tissue axis kills it.** HPA consensus puts CSPG4 at **124.5 nTPM in blood vessel**, with 31 of 51 normal tissues above 5 nTPM; GTEx v8 independently puts it at **246.21 TPM in tibial artery** with a whole-body median of 10.9 TPM [E3, E4]. Its normal-tissue profile does **not** track the leukocyte marker PTPRC (r = -0.08), so this is genuine vascular parenchyma and not leukocytes inside the tissue [E1]. The founding 1996 report - *'not expressed on the cell surface of normal hematopoietic cells'* [E5] - is a claim about **blood**, and it has been read for thirty years as though it were a claim about the body.

Go deeper
The numbers, both arms and both atlases
measurementvalue
SEM (KMT2A::AFF1)24.2 nTPM
RS4;11 (KMT2A::AFF1)37.6 nTPM
MV4-11 (KMT2A::AFF1, monocytic)13.9 nTPM
B-precursor leukaemia comparator, 13 lines0.56 nTPM
KMT2A-rearranged, other partner, 5 lines10.14 nTPM
all 1,206 HPA cell lines10.46 nTPM
ratio, B-arm over B-precursor comparator55.0x
ratio, B-arm over other KMT2A partners3.05x
HPA normal blood vessel124.5 nTPM
HPA normal bone marrow0.0 nTPM
HPA normal tissues above 5 nTPM31 of 51
GTEx v8 Artery_Tibial246.21 TPM
GTEx v8 whole-body median10.9 TPM
correlation of normal-tissue profile with PTPRC-0.08

The 55x against B-precursor ALL and the 0.0 in bone marrow are why NG2 works as a diagnostic flag. The 124.5 in blood vessel and 246 in artery are why it is a hard therapeutic target, and the two facts have never been put in the same table in this graph before 134.

The 1996 sentence that has been doing more work than it can bear

The founding paper's title says NG2 'is not expressed on the cell surface of normal hematopoietic cells' 5. That is true, it is measured, and it is the basis of NG2's diagnostic value. It is also the entire normal-tissue evidence base most NG2-as-a-target arguments rest on - and its scope is the haematopoietic compartment, which is the one compartment an anti-leukaemia therapy is allowed to damage.

The same paper notes NG2 was found on cell lines derived from non-haematopoietic tissues. NG2 is the melanoma chondroitin sulfate proteoglycan; it is a canonical pericyte marker with 332 PubMed records intersecting NG2 and pericyte; and this graph already holds cell-type-ng2-periarteriolar-cell for the normal marrow cell that carries it 7. 'Absent from normal blood' is not 'safe'.

Reagents exist, and one of them is being developed to attack a normal tissue on purpose

CSPG4-directed CAR constructs are not hypothetical: 36 PubMed records intersect CSPG4 or NG2 with chimeric antigen receptor, spanning melanoma, soft-tissue sarcoma, chondrosarcoma, glioblastoma, anaplastic thyroid cancer and triple-negative breast cancer 8. In MLL-rearranged B-ALL specifically, in vivo NG2 blockade with chondroitinase-ABC or an anti-NG2 monoclonal reduced engraftment in patient-derived xenografts and was proposed as an addition to induction therapy 9.

The sharpest confirmation of the toxicity concern comes from an unexpected direction: CSPG4.CAR-T cells are being developed to remodel extracellular matrix in Duchenne cardiomyopathy - that is, deliberately directed at a normal, non-malignant, CSPG4-bearing stromal compartment 10. A target whose CAR is a candidate therapy for a non-cancer tissue is a target with a live on-target/off-tumour problem.

What this entry does NOT say

It does not say NG2-directed therapy is impossible. Antibody-based NG2 blockade at doses that reduce engraftment is a different proposition from a CAR-T that lyses every NG2-positive cell it meets, and the xenograft work used the former 9. Local delivery, logic gating against a B-lineage antigen, dose-limited antibody-drug conjugates and transient constructs are all unexplored here.

It says that the normal-tissue axis has not been applied to NG2 in this disease, that when it is applied NG2 does not pass, and that the reassurance most often cited - the 1996 haematopoietic negative - does not cover the compartment at risk.

Expression is not surface presentation. Every expression number in this entry is mRNA. It says nothing about how much protein reaches the membrane, whether the epitope is accessible to an antibody, or whether the protein is shed. A cell line is not a patient, and the Human Protein Atlas consensus tissue panel is adult tissue while the population of concern is infants. For NG2 specifically the RNA caveat is weaker than usual, because surface NG2 protein on blasts is measured directly by unpermeabilised flow cytometry in the diagnostic literature 6; the caveat that remains fully in force is on the normal-tissue side, where only mRNA was measured here.

Facts and figures
genegene-cspg4
uniprotQ6UVK1
ensemblENSG00000173546
length AA2322
encoded by a genetrue
antigen classfusion-driven (direct MLL-AF4 target gene)
screen verdictF1 surface PASS, F2 expressed PASS, F3 fusion-enriched PASS (55.0x), F4 normal tissue KILLED (blood vessel 124.5 nTPM)
b precursor comparator nTPM0.56
other kmt2a partner nTPM10.14
normal blood vessel nTPM124.5
normal bone marrow nTPM0
n normal tissues over 5 nTPM31
lymphoid profile r vs PTPRC-0.082
index arm nTPM
MV4-1113.9MV4-11: 13.9SEM24.2SEM: 24.2RS4;1137.6RS4;11: 37.6
index arm nTPM — the numbers behind the chart
{
  "SEM": 24.2,
  "RS4;11": 37.6,
  "MV4-11": 13.9
}
gtex v8 peak
{
  "tissue": "Artery_Tibial",
  "tpm": 246.21
}
existing reagents
[
  "monoclonal antibody 7.1",
  "chondroitinase-ABC",
  "anti-NG2 monoclonal (PDX, PMID 30635633)",
  "CSPG4.CAR-T / CAR-NK / CAR-CIK in solid tumours"
]
How this could bite us back
  • NG2 marks pericytes throughout the body, and pericytes maintain vascular integrity. Ablating them in an infant is not a manageable on-target cost in the way B-cell aplasia is; there is no NG2 equivalent of immunoglobulin replacement.
  • NG2 is expressed by oligodendrocyte precursor cells and NG2 is implicated in CNS infiltration in this very disease. A CNS-penetrant NG2-directed agent would meet NG2-positive oligodendrocyte precursors in a developing infant brain.
  • NG2 is ABSENT from the candidate stem-cell populations in t(4;11) ALL and 11q23 AML - CD34+CD19-CD117+ and CD34+CD38- - so an NG2-directed therapy would spare exactly the fraction thought to initiate relapse. This is the strongest single argument against NG2 monotherapy and it is thirty years old.
  • NG2 expression varies greatly between centres in multi-centre studies, so a trial stratified on NG2 positivity would be stratified partly on flow-cytometry technique.
  • If NG2 drives glucocorticoid resistance, NG2-directed therapy selects for NG2-negative disease, and the graph's own hypothesis that NG2 tracks the transcriptional programme rather than the rearrangement predicts that the programme can re-emerge without the antigen.
Still open
  1. Is pericyte NG2 accessible to a circulating CAR-T or antibody? Pericytes sit abluminal to the endothelium, and if they are behind an intact barrier the vascular expression may be a paper toxicity rather than a real one. No experiment in this graph addresses it and it is the single question that decides whether NG2 is targetable.
  2. What fraction of NG2 on a KMT2A::AFF1 blast is surface versus intracellular at diagnosis, before any antigen-directed pressure? This graph already asks the same question for CD19 and it is unanswered for both.
  3. Does an NG2-directed agent have a therapeutic window in which blast NG2 density exceeds pericyte NG2 density? Quantitative antibody-binding-capacity on paired blasts and pericytes would settle it and has not been done.
  4. Since NG2 regulates NR3C1 and glucocorticoid response, does NG2-directed therapy RESTORE glucocorticoid sensitivity in the surviving fraction, making the combination more than additive?
Every source behind this entry (10)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    In a surfaceome-wide screen, CSPG4 shows the largest fusion-enrichment of any surface protein: 30.9 nTPM mean across SEM and RS4;11 against 0.56 in a 13-line B-cell-precursor comparator (55.0x) and 10.14 in the KMT2A-rearranged-other-partner arm (3.05x); its normal-tissue profile does not correlate with PTPRC (r = -0.08).

    What this evidence cannot show

    Two index lines in the B-arm, one nTPM value per line, no replicates and therefore no error bars. The 55x ratio is inflated by a near-zero denominator (0.56 nTPM) and should be read as 'essentially absent from other B-ALL lines', not as a calibrated fold-change.

    Evidence that argues against this (1)
    • MV4-11, the third KMT2A::AFF1 line, sits at 13.9 nTPM - below both B-ALL lines - and the other-partner arm at 10.14 is not far behind. NG2 is enriched in KMT2A-rearranged leukaemia broadly, not uniquely in AFF1, and the 3.05x partner ratio is the weaker of the two claims. Oncofusion Knowledge Graph 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    NG2 is an epigenetically regulated direct target gene of the MLL-AF4 fusion protein, NG2 expression is a hallmark of MLL-rearranged B-ALL and is minimally expressed in healthy haematopoietic cells, and NG2 underlies glucocorticoid resistance in MLLr B-ALL by regulating NR3C1.

    PMID 39093982DOI
    What this evidence cannot show

    Abstract read in full; figures not read on this beat. MLL-AF4 is the index fusion, so unlike most evidence in this graph this is not partner-imported.

    Evidence that argues against this (1)
    • If NG2 drives glucocorticoid resistance, removing NG2-positive cells with a targeted agent selects for the NG2-negative, glucocorticoid-sensitive fraction - which is a therapeutic argument, but also means the residual disease after NG2-directed therapy is a different disease from the one treated. Blood 2024
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    CSPG4 reaches 124.5 nTPM in normal blood vessel and exceeds 5 nTPM in 31 of 51 normal tissues in the HPA consensus panel, while sitting at 0.0 nTPM in normal bone marrow.

    What this evidence cannot show

    Adult bulk tissue mRNA; infant vasculature is not measured and is actively remodelling. The bone-marrow 0.0 is a bulk value and would not detect a rare NG2-positive periarteriolar population.

    Evidence that argues against this (1)
    • This graph already holds cell-type-ng2-periarteriolar-cell, a normal marrow cell that carries NG2 - so the 0.0 nTPM bulk marrow value is demonstrably not the whole story even within the marrow. Human Protein Atlas 2026
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    GTEx v8 independently places CSPG4 at 246.21 TPM in tibial artery with a whole-body median of 10.9 TPM across 54 tissues.

    What this evidence cannot show

    GTEx v8 is adult post-mortem tissue and contains no bone marrow, so it cannot address the compartment being treated; it is used here only to confirm the vascular signal in a second atlas.

    Evidence that argues against this (1)
    • Two atlases agreeing on a vascular signal does not establish that pericyte-surface NG2 protein is accessible to a circulating therapeutic; an abluminal pericyte behind an endothelial layer may not be reachable. GTEx Portal 2026
  5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    The human NG2 homologue recognised by monoclonal antibody 7.1 is not expressed on the cell surface of normal haematopoietic cells, but is expressed on AML blasts from poor-prognosis patients with 11q23 abnormalities, and was found on cell lines derived from non-haematopoietic tissues.

    PMID 8562938
    What this evidence cannot show

    1996 flow cytometry with one monoclonal. The scope of the negative is the haematopoietic compartment and hematopoietic cell lines; the same abstract reports positivity on non-haematopoietic-derived lines, which is the part usually left out when this paper is cited for safety.

    Evidence that argues against this (1)
    • Thirty years of pericyte literature, and this graph's own cell-type-ng2-periarteriolar-cell entry, establish NG2 on normal non-haematopoietic cells. The 1996 negative is correct and narrow. Human Protein Atlas 2026
  6. E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Monoclonal antibody 7.1 against NG2 identifies childhood acute lymphoblastic leukaemias with t(4;11)(q21;q23) or t(11;19)(q23;p13) and MLL gene rearrangement, measured as surface antigen by flow cytometry.

    PMID 8562939
    What this evidence cannot show

    A diagnostic-accuracy design, not a therapeutic one. Surface protein is established for blasts; nothing in it addresses normal-tissue distribution.

    Evidence that argues against this (1)
    • A multi-centre study found NG2's predictive accuracy for MLL rearrangement varies greatly by centre and depends on staff training and standardised sampling, so 'NG2-positive' is not a stable phenotype across laboratories. Leuk Res 2011
  7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    NG2 is a canonical pericyte marker; PubMed returns 332 records for NG2[tiab] AND pericyte[tiab] and 24 for CSPG4[tiab] AND pericyte[tiab] on 2026-09-23.

    What this evidence cannot show

    A record count establishes that a literature exists, not what it concludes; the 332 records were not read. NG2 is also an oligodendrocyte-precursor marker, so some of the 332 concern brain rather than vessels.

    Evidence that argues against this (1)
    • A large literature on a marker in normal cells says nothing about whether those cells are accessible to, or damaged by, a given therapeutic modality. Human Protein Atlas 2026
  8. E8strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    CSPG4/NG2-directed chimeric antigen receptor constructs exist across melanoma, soft-tissue sarcoma, chondrosarcoma, glioblastoma, anaplastic thyroid cancer and triple-negative breast cancer; PubMed returns 36 records for (CSPG4 OR NG2)[tiab] AND "chimeric antigen receptor"[tiab].

    What this evidence cannot show

    Titles were read; the papers were not. None of the 36 is in KMT2A-rearranged leukaemia, so the reagents exist but the indication does not.

    Evidence that argues against this (1)
    • That a CAR exists in six solid tumours and in none of them has reached routine use is itself evidence that the on-target/off-tumour problem with this antigen is not solved. Int J Mol Sci 2025
  9. E9strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In patient-derived xenografts of MLL-rearranged B-ALL, NG2 is crucial for engraftment after intravenous transplantation, and in vivo NG2 blockade with chondroitinase-ABC or an anti-NG2 monoclonal antibody caused significant mobilisation of leukaemic cells; NG2 blockade was proposed as an addition to vincristine/glucocorticoid/asparaginase induction.

    PMID 30635633DOI
    What this evidence cannot show

    Xenograft models in immunodeficient mice, whose vasculature and pericytes are murine; a mouse cannot report human-pericyte on-target toxicity for a human-specific reagent. Abstract read in full.

    Evidence that argues against this (1)
    • Mobilising leukaemic cells out of the niche is the stated mechanism, and mobilisation into the circulation is not obviously desirable in a disease whose hallmark is high circulating blast counts and CNS infiltration. Leukemia 2018
  10. E10moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    CSPG4-directed CAR-T cells are under development to modulate extracellular matrix remodelling in Duchenne muscular dystrophy cardiomyopathy - a normal, non-malignant CSPG4-bearing stromal compartment.

    PMID 40724839
    What this evidence cannot show

    Title and journal verified by esummary; abstract not read on this beat. The inference that this demonstrates normal-tissue targeting is drawn from the title, and the title is explicit about the indication.

    Evidence that argues against this (1)
    • Deliberately targeting a fibrotic compartment in a diseased heart is not the same as incidentally targeting healthy pericytes, and the DMD construct may be engineered for exactly the tissue selectivity a leukaemia construct would need to avoid. Int J Mol Sci 2025
Connected to