On-target, off-tumour toxicity is harm to healthy tissue that carries the same target a treatment was built to recognise, so accurate recognition injures the tissue as well as the cancer.
On-target, off-tumour toxicity is injury from recognition of the intended target on healthy tissue. It differs from recognition of an unintended molecular target. The predictable, lymphoma-specific consequences of aiming at lineage antigens (B-cell aplasia, low immunoglobulin, and the effects of CD52, CCR4 and CD47 targeting) are described in the record on the cost of aiming at a lineage antigen, which this record does not repeat.
Many candidate cancer targets are normal proteins found at different levels or places in the body, so tumour-associated does not mean tumour-exclusive. A treatment that recognises its intended target wherever it is accessible, such as a chimeric antigen receptor (CAR) T cell, can harm any healthy tissue that displays it, and tissue location, target density, receptor properties and exposure all shape the outcome. This differs from off-target cross-reactivity, in which the receptor recognises an unintended molecule, and from cytokine release syndrome, which is systemic inflammatory activation; the three can occur together. It also differs from payload toxicity in an antibody-drug conjugate: a pooled analysis of ADC trials found the key grade 3 to 4 toxicities grouped by payload and concluded they were likely off-target and related to the payload (Masters et al., 2018). Binding specificity is not tissue selectivity, and experience with an antibody against a target does not automatically carry over to a living cell with a different recognition and signalling system.
Two primary reports show the contrast. A patient with colon cancer metastatic to the lungs and liver received 10 billion autologous T cells expressing a trastuzumab-based CAR against ERBB2 (HER2), the largest number permitted in the first dose-escalation cohort, the day after a lymphodepleting regimen of cyclophosphamide (60 mg/kg daily for 2 days) and fludarabine (25 mg/m2 for 5 days). Within 15 minutes the patient had respiratory distress, chest imaging showed pulmonary oedema, and the patient died 5 days after treatment, with marked rises in interferon-gamma, GM-CSF, tumour necrosis factor-alpha, interleukin-6 and interleukin-10. At autopsy multiple organs showed systemic ischaemia and haemorrhagic microangiopathic injury and the lungs diffuse alveolar damage; histological examination of autopsy tissue found ERBB2 at 3+ in liver metastases and 2+ in lung metastases and much lower levels in normal lung parenchyma, normal liver and normal breast. The authors hypothesised that highly active ERBB2-directed T cells recognised ERBB2 on normal lung cells on first pass through the lung and released inflammatory cytokines that began a cascade ending in multiorgan failure; the autopsy findings were read as initial lung injury followed by days of hypotension and ischaemia, and the authors call the significance of the cells' reactivity against cultured primary cells unclear. This is a single case, with a stated dose and conditioning, whose mechanism is the authors' hypothesis and which gives no incidence for any CAR product (Morgan et al., 2010). In a trial of an affinity-enhanced T-cell receptor against HLA-A*01-restricted MAGE-A3 in myeloma and melanoma, the first two patients developed cardiogenic shock and died within days; no MAGE-A3 expression was detected in heart tissue at autopsy, and beating cardiomyocytes made from induced pluripotent stem cells triggered T-cell killing through recognition of an unrelated peptide from titin. That is off-target recognition of an unintended molecule, not the intended target in the wrong tissue (Linette et al., 2013).
How the question is asked before and during trials. FDA's CAR T-cell development guidance says to assess both undesired targeting of healthy tissue that expresses the intended antigen (on-target/off-tumour) and unintended targeting of other antigens, using examples such as tissue cross-reactivity studies with an antibody carrying the same recognition domain, protein arrays, cytotoxicity and cytokine-release testing on panels of primary cells, cell lines or stem-cell-derived organ systems, and relevant animal models, and encourages a combination of methods; antigen recognition domains can differ in affinity or avidity or recognise a different site, which should be evaluated, and CAR T cells can carry added components such as suicide genes. In practice a readout is a staining intensity or positive fraction, a binding profile, killing or cytokine release by healthy cells at a named time, and finally organ injury in the treated population; stained fixed tissue does not show a living organ's accessibility, an animal model may lack the human target, and a negative result in one panel is not a safety certificate. No universal expression cut-off is given here because no source read gives one. Whether affinity tuning, recognition logic or control switches improve safety for a particular design needs that design's own evidence. The 2026 ASTCT consensus abstract lists on-target, off-tumour toxicity among the toxicities for which it considered definitions and grading; its criteria were not readable.
Showing the technology this term belongs to: CAR-T cell therapy.
Shares TCR-T cell therapy, CAR-T cell therapy.
Shares Cytokine release syndrome (CRS), HER2, CAR-T cell therapy.
Shares TCR-T cell therapy, CAR-T cell therapy.
Every source this page cites, grouped by who published it. Grouping is read from the web address, not assigned by hand.
A source that has moved or gone is reported by the weekly link check. /audit/