An NG2-directed therapy meets pericytes - and unlike B-cell aplasia there is no replacement therapy for a damaged vascular wall
also called NG2 pericyte toxicity
NG2 sits on the cells that wrap around small blood vessels and hold them together. A therapy that kills every NG2-bearing cell would damage those, and there is no equivalent of the antibody infusions that make the loss of normal B cells survivable.
The same thing, for a scientist
CSPG4/NG2 is at 124.5 nTPM in HPA normal blood vessel and 246.21 TPM in GTEx tibial artery, above 5 nTPM in 31 of 51 normal tissues, and its normal-tissue profile does **not** track PTPRC (r = -0.08) - so the signal is vascular parenchyma, not resident leukocytes [E1, E2]. NG2 is a canonical pericyte marker and an oligodendrocyte-precursor marker. **The asymmetry with CD19 is the point.** CD19-directed therapy causes B-cell aplasia, which is survivable because immunoglobulin replacement exists. **There is no replacement therapy for pericyte loss.** Pericytes maintain capillary integrity, blood-brain-barrier function and vascular stability, and an infant is building that vasculature. The corroboration comes from an unexpected direction: **CSPG4-directed CAR-T is being developed to remodel matrix in Duchenne cardiomyopathy** - a deliberate attack on a normal CSPG4-bearing stromal compartment [E3]. A target whose CAR is a candidate therapy *for* a non-cancer tissue has an unretired on-target/off-tumour problem.
What would retire this risk, and what would not
Would not retire it: a clean PDX efficacy study. Immunodeficient mice have murine pericytes, and a human-specific anti-NG2 reagent cannot engage them, so the model is structurally blind to this toxicity 4.
Would retire or quantify it:
- quantitative antibody-binding capacity on primary human blasts versus primary human pericytes from the same donor, establishing whether a therapeutic window exists;
- immunohistochemistry on a paediatric normal-tissue array - the atlases used here are adult;
- evidence that abluminal pericytes are inaccessible to a circulating cell therapy, which would make the vascular expression a paper toxicity.
Modality matters more than target here. The published in-vivo work used chondroitinase-ABC and a blocking monoclonal, not a cytolytic construct 4. Blocking NG2 function to mobilise blasts is a different proposition from lysing every NG2-positive cell, and the risk recorded here applies with full force only to the second.
| kind | toxicity |
|---|---|
| severity | potentially high for a cytolytic modality; low to moderate for a blocking antibody |
| mechanism | on-target/off-tumour engagement of NG2/CSPG4 on pericytes, vascular smooth muscle and oligodendrocyte precursors |
| asymmetry with cd19 | B-cell aplasia is survivable because immunoglobulin replacement exists. There is no replacement therapy for pericyte loss. |
| mitigation | Prefer a blocking or function-modifying modality over a cytolytic one; logic-gate against a B-lineage antigen; quantify antibody-binding capacity on primary blasts versus primary human pericytes to establish a window; run IHC on a PAEDIATRIC normal-tissue array before any cytolytic construct. |
affects
[ "target-ng2-cspg4-the-fusion-driven-antigen-that-normal-vasculature-kills" ]
normal expression
{
"HPA blood vessel nTPM": 124.5,
"GTEx Artery_Tibial TPM": 246.21,
"HPA tissues above 5 nTPM": "31 of 51",
"r vs PTPRC": -0.082
}- This risk is recorded from adult mRNA atlases. If it is quoted to block an NG2 programme it will have outrun its evidence: no protein measurement on human pericytes was made here.
- Over-weighting this risk pushes the field back to CD19, which this disease escapes by lineage switch. A target with a manageable toxicity that the disease escapes may be worse than one with a real toxicity that it cannot.
- The DMD CAR-T observation is inferred from a title. If that construct turns out to be a CSPG4-binder used for delivery rather than for lysis, the corroboration weakens considerably.
- Are abluminal pericytes accessible to a circulating CAR-T? This single question decides whether the risk is real or notional, and nothing in this graph addresses it.
- Has any CSPG4-directed CAR trial in solid tumours reported vascular toxicity? 36 papers exist and none was read in full on this beat; a clinical-safety readout may already exist.
- Does NG2 density on a KMT2A::AFF1 blast exceed pericyte density by enough for an affinity-tuned low-avidity CAR to discriminate? That is the standard remedy for this class of problem and it has not been costed here.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
CSPG4 is at 124.5 nTPM in HPA normal blood vessel, above 5 nTPM in 31 of 51 normal tissues, and its 51-tissue profile correlates with PTPRC at r = -0.08, indicating parenchymal rather than leukocyte origin.
Human Protein Atlas, consensus tissue RNA nTPM across 51 normal tissues (rna_tissue_consensus.tsv); HPA release v25.1, fetched 2026-09-23 · Human Protein Atlas 2026What this evidence cannot show
Adult bulk tissue mRNA. 'Blood vessel' in HPA is a composite tissue and the value does not separate pericytes from smooth muscle or endothelium.
Evidence that argues against this (1)
- mRNA abundance in a vessel wall does not establish that the protein is presented on a surface a circulating T cell can reach; pericytes sit abluminal to the endothelium. Human Protein Atlas 2026
- E2strong 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.
GTEx v8 median gene expression, CSPG4 (ENSG00000173546.7), 54 tissues; Artery_Tibial 246.21 TPM, median 10.9 TPM; fetched 2026-09-23 · GTEx Portal 2026What this evidence cannot show
Adult post-mortem tissue; GTEx has no bone marrow and no paediatric donors.
Evidence that argues against this (1)
- Two adult atlases agreeing tells you nothing about an infant, in whom vascular remodelling is active and pericyte turnover differs. Human Protein Atlas 2026
- E3moderate 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.
CSPG4.CAR-T Cells Modulate Extracellular Matrix Remodeling in DMD Cardiomyopathy. · Int J Mol Sci 2025What this evidence cannot show
Title and journal verified by esummary; abstract not read on this beat, so the inference rests on the title, which is explicit about the indication.
Evidence that argues against this (1)
- A construct engineered for a fibrotic cardiac compartment may have been deliberately tuned for a selectivity that an anti-leukaemia construct would not share, so the two are not directly comparable. Int J Mol Sci 2025
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In vivo NG2 blockade in MLL-rearranged B-ALL PDXs used chondroitinase-ABC or an anti-NG2 blocking monoclonal antibody and produced mobilisation and reduced engraftment, in immunodeficient mice.
NG2 antigen is a therapeutic target for MLL-rearranged B-cell acute lymphoblastic leukemia. · Leukemia 2019What this evidence cannot show
Murine vasculature and murine pericytes; the model cannot report human on-target pericyte toxicity for a human-specific reagent. Abstract read in full.
Evidence that argues against this (1)
- That no pericyte toxicity was reported in these studies is not evidence of safety, because the experiment could not have detected it. Absence of a reported toxicity in a system blind to it is not a negative result. Leukemia 2019
the toxicity that the normal-tissue kill implies, and the asymmetry with B-cell aplasia
the normal NG2-bearing cell type already in this graph