CAR cell therapies
An antigen to recognise, a binder that recognises it, a costimulatory domain that sets the tempo (4-1BB slow and durable, CD28 fast and sharp), a vector that installs the receptor, and the cell that carries it, taken from the patient or from a donor. The grid below is target against costimulatory domain: 13 by 2 from 35 medicines, 5 combinations approved, 11 in development, 0 tried and stopped, 23 untried in this corpus.
Antibody-drug conjugatesRadioligandsCAR cell therapiesTCR-T cell therapiesBispecific antibodiesDegraders and molecular gluesSmall-molecule inhibitorsNaked antibodiesCytokinesCancer vaccinesOncolytic virusesOther cell therapies
Approved 5In development 11Tried and stopped 0Recorded, state unclear 0Untried here 2313 target rows x 2 costimulatory domain columns. Click a cell to filter the table.
Every tried combination
16 combinations
| Medicines | Records | Stopped, and why | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| CD19 | Costimulatory domain not recorded | Approved | 1 | 5 | 0 | - | - | |||||
| CD19 | 4-1BB | Approved | 6 | 0 | 0 | scFv | - | |||||
| CD19 | CD28 | Approved | 2 | 0 | 0 | scFv | - | |||||
| BCMA | Costimulatory domain not recorded | In development | 0 | 2 | 0 | - | - | |||||
| BCMA | 4-1BB | Approved | 4 | 1 | 0 | D-domainVHH (single domain)scFv | - | |||||
| CD19 x CD20 | Costimulatory domain not recorded | In development | 0 | 3 | 0 | - | - | - | ||||
| CD19 x BCMA | Costimulatory domain not recorded | In development | 0 | 2 | 0 | - | - | - | - | |||
| B7-H3 | Costimulatory domain not recorded | In development | 0 | 1 | 0 | - | - | - | ||||
| CD22 | Costimulatory domain not recorded | In development | 0 | 1 | 0 | - | - | |||||
| CD3 | Costimulatory domain not recorded | In development | 0 | 1 | 0 | - | - | |||||
| CD30 | Costimulatory domain not recorded | In development | 0 | 1 | 0 | - | - | - | - | |||
| CD70 | Costimulatory domain not recorded | In development | 0 | 1 | 0 | - | - | - | ||||
| Claudin 18.2 | Costimulatory domain not recorded | Approved | 1 | 0 | 0 | scFv | - | - | - | |||
| DLL3 | Costimulatory domain not recorded | In development | 0 | 1 | 0 | - | - | - | - | |||
| GPRC5D | Costimulatory domain not recorded | In development | 0 | 1 | 0 | - | - | - | - | |||
| TROP2 | Costimulatory domain not recorded | In development | 0 | 1 | 0 | - | - | - | - |
Stopped, and why
0 medicines recorded as stopped; 0 recorded reasons, including stopped studies of medicines still in developmentNo stopped study, withdrawn approval or negative result is recorded for a medicine in this format.
Proposed, not yet tried
12 proposals from the open pipeline, scored and searched (docs/OPEN-PIPELINE.md); hypotheses, not records of a medicineClinical evidence 9Preclinical 2No public evidence 1
| Score | Target | Costimulatory domain | Status | Rationale | Evidence | Now in the corpus |
|---|---|---|---|---|---|---|
| 73 | Claudin 18.2 | CD28 | No public evidence | Claudin 18.2 CAR-T with CD28 costimulation. Claudin 18.2 is validated by the first approved solid-tumour CAR-T (satricabtagene-autoleucel), which uses 4-1BB costimulation, and CD28 is validated as a costimulatory domain in approved CD19 CAR-Ts (axicabtagene-ciloleucel). No corpus Claudin 18.2 CAR-T uses CD28, so the cell is untried. Solid-tumour CAR-Ts need to expand quickly against an immunosuppressive microenvironment, which favours CD28's kinetics; the risk is stronger on-target gastric mucosal toxicity from the same rapid activation.Caveat: The 4-1BB construct already causes gastric mucosal injury; a CD28 version would need to show that faster expansion buys response without worse mucosal toxicity.Proposed by OnCo open pipeline, searched 2026-09-23. | Untried | |
| 73 | MET | 4-1BB | Clinical evidence | MET CAR-T with 4-1BB costimulation. MET is validated by an approved ADC (telisotuzumab-vedotin) and kinase inhibitors (capmatinib), and 4-1BB by most approved CAR-Ts (tisagenlecleucel). No corpus CAR-T targets MET. MET is expressed on hepatocytes and many epithelia, so a systemic MET CAR-T risks liver toxicity; intratumoural or transient mRNA CAR delivery is the design that has been tried.Caveat: Hepatocyte MET expression is the obstacle; unless a transient or locally delivered product proves safe, this stays preclinical.Proposed by OnCo open pipeline, searched 2026-09-23. | Untried | |
| 73 | TROP2 | 4-1BB | Preclinical | TROP2 CAR-T with 4-1BB costimulation. TROP2 carries three approved ADCs (sacituzumab-govitecan, datopotamab-deruxtecan, sacituzumab-tirumotecan) and 4-1BB is the costimulatory domain of most approved CAR-Ts (tisagenlecleucel). The corpus holds a TROP2 CAR-NK in phase 2 (eb-nk-301) but no TROP2 CAR-T. TROP2 is broadly expressed on normal epithelia, so a CAR-T's sustained killing would hit skin and gut; the ADCs tolerate this because payload dose scales with antigen density, but a CAR has no such threshold unless engineered with logic gating.Caveat: Normal-epithelium expression makes an unmodified TROP2 CAR-T dangerous; logic-gated or transient constructs are the only credible designs.Proposed by OnCo open pipeline, searched 2026-09-23. | Untried | |
| 72 | HER2 | 4-1BB | Clinical evidence | HER2 CAR-T with 4-1BB costimulation. HER2 is the corpus's most validated delivery antigen (trastuzumab-deruxtecan, trastuzumab, zanidatamab) and 4-1BB costimulation is validated in most approved CAR-Ts (tisagenlecleucel, ciltacabtagene-autoleucel). No corpus CAR-T targets HER2. HER2 is expressed at low levels on lung and heart, and a high-affinity HER2 CAR caused a fatal respiratory event in 2010; later low-affinity constructs have been tolerated, so the design question is affinity tuning and local (intraventricular, intratumoural) delivery.Caveat: On-target lung and heart toxicity is documented for a high-affinity HER2 CAR; any new construct needs affinity tuning or regional delivery, and paediatric sarcoma or CNS disease is a more plausible first indication than breast cancer.Proposed by OnCo open pipeline, searched 2026-09-23. | Untried | |
| 71 | Somatostatin receptor 2 | 4-1BB | Preclinical | Somatostatin receptor 2 CAR-T with 4-1BB costimulation. SSTR2 is validated by an approved radioligand (lutathera) and 4-1BB by most approved CAR-Ts (tisagenlecleucel). No corpus CAR-T targets SSTR2. SSTR2 is a seven-transmembrane GPCR with a small extracellular surface, which is hard for an scFv to bind; a CAR built on the octreotide peptide as the binder is the plausible design, and SSTR2 on pancreatic islets and pituitary is the safety question.Caveat: Islet and pituitary expression could give endocrine toxicity, and a GPCR is a difficult CAR target; preclinical work on a peptide-based binder would need to come first.Proposed by OnCo open pipeline, searched 2026-09-23. | Untried | |
| 69 | Nectin-4 | 4-1BB | Clinical evidence | Nectin-4 CAR-T with 4-1BB costimulation. Nectin-4 is validated by an approved ADC (enfortumab-vedotin) and 4-1BB by most approved CAR-Ts (tisagenlecleucel). No corpus CAR-T targets Nectin-4. Nectin-4 is dense and fairly homogeneous in urothelial cancer, and its normal expression is mainly skin; the skin toxicity of enfortumab-vedotin shows the antigen is reachable there, so a CAR-T would need affinity tuning to spare keratinocytes.Caveat: Keratinocyte expression predicts skin toxicity worse than enfortumab's; whether affinity tuning can separate tumour from skin is the first preclinical question.Proposed by OnCo open pipeline, searched 2026-09-23. | Untried | |
| 66 | PSMA | 4-1BB | Clinical evidence | PSMA CAR-T with 4-1BB costimulation. PSMA is validated by an approved radioligand (pluvicto) and three phase-3 alpha emitters, and 4-1BB is the costimulatory domain of most approved CAR-Ts (tisagenlecleucel, lisocabtagene-maraleucel). No corpus CAR-T targets PSMA, although the corpus already flags armoured CAR designs for solid tumours (armored-car). PSMA is tumour-restricted enough for a radioligand, but prostate cancer's bone-marrow niche and TGF-beta-rich stroma suppress T cells; armoured constructs with dominant-negative TGF-beta receptors are the design that has reached patients.Caveat: An armoured PSMA CAR-T caused fatal toxicity in an early trial; the balance between enough potency to clear bone metastases and safety is unresolved.Proposed by OnCo open pipeline, searched 2026-09-23. | Untried | |
| 65 | CD123 | 4-1BB | Clinical evidence | CD123 CAR-T with 4-1BB costimulation. CD123 is validated by an approved ADC (pivekimab-sunirine) and an approved toxin fusion (tagraxofusp), and 4-1BB by most approved CAR-Ts (tisagenlecleucel). No corpus CAR-T targets CD123. CD123 marks leukaemic stem cells and BPDCN uniformly, so the antigen is homogeneous; it is also on endothelium and normal progenitors, and capillary leak from CD123-directed agents (tagraxofusp) is the class toxicity to expect.Caveat: Capillary leak and myeloablation are the expected toxicities; like CD33, this works best as a bridge to transplant or with a suicide switch.Proposed by OnCo open pipeline, searched 2026-09-23. | Untried | |
| 65 | CD33 | 4-1BB | Clinical evidence | CD33 CAR-T with 4-1BB costimulation. CD33 is validated by an approved ADC (gemtuzumab-ozogamicin) and 4-1BB by most approved CAR-Ts (tisagenlecleucel). No corpus CAR-T targets CD33, and the transplant-bridge design that would make it safe is already established for CD45 radioimmunotherapy (iomab-b). CD33 is on normal myeloid progenitors, so a persistent CD33 CAR-T causes marrow aplasia; the workable designs are a transient CAR (mRNA), a suicide switch, or CAR-T as a bridge to transplant with CD33-deleted donor stem cells.Caveat: Marrow aplasia is certain with a persistent CD33 CAR; the therapy only makes sense paired with a transplant plan or an off-switch.Proposed by OnCo open pipeline, searched 2026-09-23. | Untried | |
| 63 | BCMA | CD28 | Clinical evidence | BCMA CAR-T with CD28 costimulation. BCMA is validated by two approved 4-1BB CAR-Ts (idecabtagene-vicleucel, ciltacabtagene-autoleucel) and the CD28 costimulatory domain by two approved CD19 CAR-Ts (axicabtagene-ciloleucel, brexucabtagene-autoleucel). No corpus BCMA CAR-T uses CD28 costimulation, so the cell is untried in the corpus. Every approved BCMA CAR-T uses 4-1BB (idecabtagene-vicleucel, ciltacabtagene-autoleucel); CD28 gives faster, stronger expansion (axicabtagene-ciloleucel) but shorter persistence and more cytokine release, and myeloma relapses are often late, where persistence matters.Caveat: An earlier CD28 BCMA CAR-T was stopped after phase 1; whether that reflected the domain, the binder or the era's manufacturing is the question to settle before repeating it.Proposed by OnCo open pipeline, searched 2026-09-23. | Untried | |
| 61 | CD38 | 4-1BB | Clinical evidence | CD38 CAR-T with 4-1BB costimulation. CD38 is validated by two approved antibodies (daratumumab, isatuximab) and 4-1BB by most approved CAR-Ts (tisagenlecleucel). No corpus CAR-T targets CD38, which would offer an autologous option after BCMA and GPRC5D antigen loss. CD38 is on activated T cells themselves, so a CD38 CAR-T risks fratricide during manufacturing unless CD38 is knocked out; affinity-tuned CARs that spare CD38-low normal cells have been described.Caveat: Fratricide and CD38 expression on NK cells and red-cell precursors are the obstacles; a CD38-knockout T-cell product is the likely requirement.Proposed by OnCo open pipeline, searched 2026-09-23. | Untried | |
| 60 | CD30 | 4-1BB | Clinical evidence | CD30 CAR-T with 4-1BB costimulation. CD30 is validated by an approved ADC (brentuximab-vedotin) and 4-1BB by most approved CAR-Ts (tisagenlecleucel). The corpus holds a CD30 CAR-NK in phase 2 (eb-car30-nk) but no CD30 CAR-T. CD30 is restricted to activated lymphocytes and Reed-Sternberg cells, so normal-tissue toxicity is low; the obstacle is the Hodgkin microenvironment, which excludes T cells, and the small number of patients who fail brentuximab and checkpoint blockade.Caveat: The population that needs it is small and the Hodgkin microenvironment resists T-cell infiltration; a CCR4-armoured design may be required for cells to reach the tumour.Proposed by OnCo open pipeline, searched 2026-09-23. | Untried |
Not placed
1 of 36 medicines in this format (3%)These medicines belong to the format but their target is on no record the engine reads: the targets field is empty, the target lists no such drug, the trials linked to it name no target of its own, and the modality, mechanism and summary text name none the corpus knows. They are listed here rather than placed by guesswork; adding the target to the record puts them in the grid on the next build.
How this grid is read from the records
Each medicine's parts come from its own record: the targets field first, else the target records that list the medicine, else a sibling conjugate that shares its antibody name, else the trials linked to it, else the target names the corpus knows found in its modality, mechanism or summary text. The costimulatory domain, binder, vector and cell source are read from the mechanism and modality text (4-1BB or CD28, scFv or VHH, lentiviral or retroviral, autologous or allogeneic); a part the text does not name is shown as not recorded. Every part carries the fields it was read from and a confidence in the JSON file.
What the states mean, and do not
Approved: a medicine in the cell has an approval row or a regulator's approved entry and is not recorded as withdrawn. In development: a recruiting, active or planned trial, a development-phase record status, or active studies in the ClinicalTrials.gov index, and no approval. Tried and stopped: every medicine in the cell is withdrawn, negative or historic, or its only trial evidence is a terminated, withdrawn or negative study, or its approval was withdrawn. Untried: no medicine in this corpus combines the two parts.
A cell is coloured by its strongest medicine; the table shows each medicine with its own state. The reasons quoted under Stopped, and why are the records' words, including the registry's whyStopped text where the sponsor wrote one, and include stopped studies of medicines that are still in development elsewhere. Nothing here is medical advice.