Loading
Old version — revision 1
This is a fixed snapshot of In vivo CAR-T therapy, saved by Import as part of the initial corpus import. It is not edited and it is not updated; the article may have changed since.
Edit summary: Initial import of content/in-vivo-car-t.md — the filesystem corpus, unchanged. Not an edit.
A way of making chimeric antigen receptor T cells inside the patient's body, by infusing a targeted viral vector or lipid nanoparticle instead of manufacturing cells in a facility.
In vivo CAR-T therapy puts the chimeric antigen receptor gene into a patient's own T cells while those cells are still inside the patient, using an infused vector that seeks out T cells rather than a factory that grows them. Seven conventional CAR-T products, all directed at CD19 or BCMA, are approved in the United States and Europe; each one collects a person's lymphocytes, transduces them in a facility over several weeks, and reinfuses them after lymphodepleting chemotherapy, at a cost exceeding hundreds of thousands of dollars per patient.1 The in vivo version replaces that chain with a single intravenous dose. As of November 2026 nine patients with multiple myeloma have been described in peer-reviewed phase 1 reports, alongside five lupus patients in a shorter peer-reviewed letter,21 and nothing is approved.
The older route to making CAR-T therapy cheaper runs through donors rather than through delivery. Allogeneic, off-the-shelf products start from a healthy donor's cells and knock out TRAC and B2M with CRISPR-Cas9 or TALENs to prevent graft-versus-host disease and rejection; they still require ex vivo engineering, heavy lymphodepletion, and have shown limited persistence in patients.1 In vivo delivery attacks the problem from the other end, leaving the cells where they are.
Two delivery platforms are in clinical development, and they differ in almost every respect that matters.13
A targeted lentiviral vector is a replication-incompetent HIV-1 derivative whose envelope has been re-engineered. Instead of the broad-tropism vesicular stomatitis virus glycoprotein used for laboratory transduction, the particle displays an antibody fragment or nanobody against a lymphocyte surface marker such as CD3, CD5, CD7 or CD8, so that entry is restricted to the intended lineage. A second layer of specificity comes after entry: the CAR sits under a lineage-restricted synthetic promoter, so a vector that gets into the wrong cell still does not express a receptor there. Further modifications reduce clearance and immune recognition, including the phagocytosis-inhibitory ligand CD47 on the particle surface and deletion of MHC class I from the producer cells.1 Because lentiviral DNA integrates, one dose can give durable CAR expression.
An mRNA-lipid nanoparticle carries the CAR as messenger RNA in a vesicle of the kind developed for the COVID-19 vaccines, decorated with a binder against CD8 or another T-cell marker. The RNA is translated in the cytoplasm and never enters the genome, so expression is transient and repeat dosing is expected.1 This is the same carrier chemistry described in Lipid nanoparticles, applied to lymphocytes rather than to liver, and the same modified-nucleoside and capping conventions that Personalized mRNA cancer vaccines rely on to keep innate immune sensing down.1 Neither platform uses leukapheresis, and neither of the reported trials used lymphodepleting chemotherapy. Adeno-associated vectors, the workhorse of Somatic gene therapy, are little used here: as discussed in AAV vectors, their tropism is broad and hard to restrict to a circulating cell type, which is the whole problem Targeted drug delivery exists to solve.
| Dimension | Targeted lentiviral vector | Targeted mRNA-LNP |
|---|---|---|
| Genomic integration | Yes | No |
| CAR expression | Durable after one dose | Transient |
| Dosing envisaged | Single administration | Repeated |
| Pre-existing immunity | Possible, to viral epitopes | No viral epitopes |
| Distinctive safety concern | Insertional mutagenesis in off-target cells | Non-specific uptake by myeloid cells |
| Published human data | Myeloma, nine patients | Lupus, five patients |
The idea predates the delivery technology by years. In 2017 Matthias Stephan's group showed that polymer nanoparticles carrying CAR DNA, a construct closer to DNA nanotechnology than to a virus, could program circulating T cells in leukaemia-bearing mice.4 In 2018 Christian Buchholz's group used a CD8-targeted lentiviral vector to generate human CD19-CAR T cells in humanised mice, depleting B cells and, in three animals, producing cytokine elevations and tissue changes the authors read as a cytokine release syndrome. That paper is worth reading as the first warning that the toxicity profile travels with the method.5 In 2022 a team at the University of Pennsylvania injected CD5-targeted lipid nanoparticles carrying mRNA for an anti-fibroblast receptor into mice with heart failure, making transient CAR-T cells that reduced cardiac fibrosis; the target was scar tissue rather than a tumour, which is why the paper is usually cited for the platform rather than the indication.6 All of this was in mice.
More than fifteen first-in-human phase 1 trials of in vivo CAR platforms were registered within roughly two years of the first, which opened in mid-2023.1 The published human record is much smaller than that and comes almost entirely from one vector. ESO-T01 is a nanobody-directed, immune-shielded lentiviral vector encoding a humanised anti-BCMA receptor, developed by EsoBiotec, now part of AstraZeneca, and tested in investigator-initiated trials in China.1
In the five-patient report, five heavily pretreated men with relapsed or refractory multiple myeloma received a single infusion of 0.2 x 10⁹ transduction units with no leukapheresis, no ex vivo step and no lymphodepletion. There were no dose-limiting toxicities, but every patient had a grade 3 or higher adverse event, and cytokine release syndrome occurred in four, three of them grade 3. One patient developed grade 1 neurotoxicity and died of spinal cord compression from an extramedullary lesion. Four of five responded, three with stringent complete remission, and all four evaluable responders were negative for minimal residual disease at 10⁻⁵ by day 60. Median follow-up was 6.0 months, and the trial was stopped early in 2025.7
A second, separate ESO-T01 cohort of four patients, first reported in The Lancet in July 2025,8 has been followed for up to 15 months. All four developed a two-phase cytokine release syndrome of grades 1 to 3 and grade 3 or worse haematological toxicity; one had grade 1 neurotoxicity. All four responded, two with stringent complete remission. One held that remission for 15 months. Median progression-free survival was 4.0 months, and three patients relapsed or progressed, all with extramedullary disease, and died during follow-up. The authors report no immunogenicity-related or integration-related toxicity, and conclude that the limited persistence they observed means the platform needs further optimisation.9
Durability, not feasibility, is the open questionThe point the two published cohorts agree on is that an in-body CAR can be made and can clear marrow disease. The point they do not settle is whether it lasts. In the cohort with the longest follow-up, median progression-free survival was 4.0 months and three of four patients died after extramedullary relapse; the authors conclude that the limited persistence they observed means the platform needs further optimisation.9 Nine patients across two trials cannot distinguish a platform problem from a disease problem.
The other clinical programmes have reported at conferences rather than in journals. Kelonia's KLN-1010 is a CD3-targeted lentiviral vector carrying a BCMA receptor,1 and its first-in-human cohort was presented as a late-breaking abstract at the 2025 meeting of the American Society of Hematology. In the four patients reported there, CAR-positive cells reached 22 to 85 per cent of circulating T cells by day 15 without lymphodepletion, all four were negative for minimal residual disease within a month, three were in partial rather than complete response, cytokine release syndrome was grade 1 or 2 in three patients, and no neurotoxicity was seen; an expert quoted by The ASCO Post called the safety signal very good but the follow-up still very short.10 The registry record lists planned enrolment of 70 at sites in the United States and Australia.11 Interius's INT2104 and Umoja's UB-VV111, both lentiviral, entered phase 1 in 2024 without published clinical results.1
The strongest argument for the mRNA route is that some diseases may not need a persistent CAR at all. Early clinical data suggest ex vivo CD19 CAR-T cells have therapeutic potential in autoimmune disease, and if what does the work there is a window of deep B-cell depletion, a transient receptor is a feature rather than a defect. Reviewers frame the choice that way: haematological malignancy may require long-term persistence, while autoimmune disease may require only transient depletion, and the delivery vehicle should be matched to the indication.1
A letter in the New England Journal of Medicine in October 2025 reported in vivo CD19 CAR-T therapy in refractory systemic lupus erythematosus, delivered by targeted lipid nanoparticle.2 A 2026 review identifies the product as MagicRNA's HN2301 and counts five lupus patients in that report.1 Capstan's CPTX2309, a CD8-targeted LNP carrying anti-CD19 CAR mRNA, is in a phase 1 trial in Australia that enrols healthy volunteers as well as patients with rheumatoid arthritis or lupus, an unusual design for a cell therapy.12 No results from it have been published.
Every clinical candidate targets an antigen already validated by ex vivo products: CD19, CD20, CD22 or BCMA. The blood cancers those antigens define are where the method has been tested, and solid-tumour work has largely shifted to in vivo CAR-macrophages instead, which is a different cell and a different problem.1 Every trial in the 2026 review's tables is a single-arm phase 1 study, so no randomised comparison against an approved ex vivo product exists.1
Payload capacity bounds what else a vector can carry. The enhancements reviewers list as crowded out by it include armouring, logic gating, and the genome and epigenome editing that an ex vivo process can accommodate through Base editing, Prime editing or Epigenome editing.1 Apart from CAR-macrophage products, every current in vivo platform expresses a second-generation receptor.1
Dose control is harder than it looks. With an ex vivo product the infused cell dose is counted before it goes in; with an in vivo vector the number of CAR-positive cells is an outcome of transduction efficiency in that patient, which is why the KLN-1010 cohort spans a fourfold range in CAR-positive fraction.10 The integrating platforms cannot be un-dosed. The transient platforms face the opposite problem: sustaining an effect means repeating the infusion, and no dosing interval has been established in a published trial.1
The economic case is the one most often made and least often tested. Proponents argue that removing apheresis, centralised manufacturing and the return shipment removes the several-week turnaround and the six-figure cost that keeps ex vivo CAR-T confined to specialist centres.1 No in vivo product has a price. Simpler manufacturing has not so far produced a cheaper one-time genetic medicine, which is the problem Gene therapy for aging runs into for a different reason and Access and inequality takes up directly.
How far the intervention can be undone depends entirely on which platform delivered it. An mRNA-LNP CAR decays on its own, because nothing was written into a chromosome; a lentiviral CAR is a permanent insertion in every cell it reached, and there is no route back for those cells.1 That difference, rather than any property of the receptor, is what makes the reversibility of in vivo CAR-T a per-product question.
Reviewers name three risks specific to delivering a vector into a person rather than into a culture dish.1 Off-target transduction of cells that are not T cells raises the prospect of insertional mutagenesis wherever an integrating vector lands, and transduction of tumour cells themselves can mask the target epitope and shield the cancer. Accumulation in organs such as the liver, the default destination for systemically infused particles. And immune reactions, which fall hardest on patients already exposed to the viral epitopes a lentiviral particle presents.
The toxicities familiar from ex vivo therapy have not been avoided. Cytokine release syndrome occurred in four of five patients in the first published cohort and in all four of the second, with grade 3 events in both, and neurotoxicity has been reported at grade 1.79 In 2024 the US Food and Drug Administration required a boxed warning about T-cell malignancies, including CAR-positive lymphoma, on the six approved autologous CD19- and BCMA-directed products, having concluded the risk applied across the class and advising lifelong monitoring for secondary cancers.13 That action concerns ex vivo products and no in vivo case has been reported, but the mechanism it was written about is insertion of a transgene by an integrating vector, and in vivo delivery removes the step at which a product could be characterised before it enters a patient. The resulting detection problem resembles the one set out in Off-target effects in genome editing: the unintended change is rare, scattered, and in cells nobody sequenced.
"In vivo CAR" is wider than "in vivo CAR-T"The first in vivo CAR trial, opened in Australia in mid-2023, used lipid nanoparticles to program macrophages rather than T cells, and CAR-macrophage candidates are now the main in vivo approach to solid tumours.1 Papers and press releases often use the broader term, so a reported patient count may include cells that are not T cells.
At least five large pharmaceutical companies bought into the field between March 2025 and April 2026: AstraZeneca acquired EsoBiotec, AbbVie acquired Capstan, Gilead's Kite unit acquired Interius, Bristol Myers Squibb acquired Orbital Therapeutics, and Eli Lilly, which had agreed in February 2026 to buy Orna Therapeutics, agreed on 20 April 2026 to pay $3.25 billion upfront and up to $7 billion in total for Kelonia.141 Lilly closed that purchase on 25 June 2026, accounting for it as a business combination and recording about $4.7 billion as an acquired in-process research and development asset attributed to KLN-1010.15 Trung Huynh, an analyst at RBC Capital Markets quoted when the Kelonia deal was announced, noted that AbbVie and Bristol Myers Squibb had paid $2.1 billion and $1.5 billion for companies without comparable clinical efficacy data, and judged Lilly's price justified by the strength of Kelonia's data.14
That is a large amount of capital resting on a human record still counted in dozens of patients rather than hundreds, with no randomised data and the longest follow-up in a peer-reviewed report standing at 15 months.9 The most load-bearing uncertainty is not whether a vector can make a CAR-T cell in the body, which the published cohorts settle, but whether a therapy whose dose is set by the patient's own biology can be titrated well enough to be given outside a transplant centre. If it cannot, in vivo CAR-T inherits the access constraint it was designed to remove.
paperVilcot, F., Brugère, C. A., Fuentealba, J., Houot, R. "In vivo CAR therapies: Turning the patient into their own CAR factory." HemaSphere, 2026;10:e70328.↩ ↩2 ↩3 ↩4 ↩5 ↩6 ↩7 ↩8 ↩9 ↩10 ↩11 ↩12 ↩13 ↩14 ↩15 ↩16 ↩17 ↩18 ↩19 ↩20 ↩21 ↩22 ↩23An open-access review that tabulates current LNP-based clinical in vivo CAR trials and other programmes and their stage; the patient counts it reports are drawn from the primary reports and conference abstracts it cites.
paperWang, Q., Xiao, Z. X., Zheng, X. et al. "In Vivo CD19 CAR T-Cell Therapy for Refractory Systemic Lupus Erythematosus." New England Journal of Medicine, 2025;393(15):1542-1544.↩ ↩2Published as correspondence rather than a full trial report, and indexed without an abstract.
paperBot, A., Scharenberg, A., Friedman, K. et al. "In vivo chimeric antigen receptor (CAR)-T cell therapy." Nature Reviews Drug Discovery, 2026;25:116-137.↩Several authors work at companies developing the platforms the review surveys.
paperSmith, T. T., Stephan, S. B., Moffett, H. F. et al. "In situ programming of leukaemia-specific T cells using synthetic DNA nanocarriers." Nature Nanotechnology, 2017.↩Mouse leukaemia models only; the nanoparticles carried DNA rather than mRNA.
paperPfeiffer, A., Thalheimer, F. B., Hartmann, S. et al. "In vivo generation of human CD19-CAR T cells results in B-cell depletion and signs of cytokine release syndrome." EMBO Molecular Medicine, 2018;10:e9158.↩Humanised mice; CAR induction and B-cell depletion occurred in 7 of 10 animals in one experiment, and three showed cytokine elevations.
paperRurik, J. G., Tombácz, I., Yadegari, A. et al. "CAR T cells produced in vivo to treat cardiac injury." Science, 2022.↩A mouse heart failure model; the receptor targeted fibroblast activation protein, not a tumour antigen, and expression was deliberately transient.
paperAn, N., Wang, D., Zhang, P., Zhang, J. et al. "In vivo generation of anti-BCMA CAR-T cells in relapsed or refractory multiple myeloma: a phase 1 study." Nature Medicine, 2026;32(4):1257-1266.↩ ↩2Five patients at a single dose level, median follow-up 6.0 months; the trial was stopped early in 2025 and the paper reports safety as its primary endpoint.
paperXu, J., Liu, L., Parone, P. et al. "In-vivo B-cell maturation antigen CAR T-cell therapy for relapsed or refractory multiple myeloma." The Lancet, 2025;406(10500):228-231. https://pubmed.ncbi.nlm.nih.gov/40617243/ ↩
paperXu, J., Liu, L., Xie, W., Su, L. et al. "Extended follow-up of in vivo BCMA CAR-T therapy in relapsed/refractory multiple myeloma." Nature Medicine, 2026.↩ ↩2 ↩3 ↩4A separate four-patient ESO-T01 cohort, maximum follow-up 15 months; the authors state that the limited persistence observed requires further optimisation of the platform.
newsThe ASCO Post. "Early Findings From First Human Study of In Vivo CAR T in Myeloma." ASH Meeting Highlights supplement, 25 February 2026.↩ ↩2Reports a late-breaking conference abstract, not a peer-reviewed paper, and quotes an expert noting that follow-up is very short.
trialClinicalTrials.gov. NCT07075185, "A Study to Evaluate a Novel Gene Therapy in Patients With Relapsed and Refractory Multiple Myeloma." Kelonia Therapeutics, phase 1, started 16 July 2025.↩The registry record gives planned enrolment and sites; it carries no results.
trialClinicalTrials.gov. NCT06917742, "A Study of Intravenous CPTX2309 in Healthy Participants and Participants With Moderate to Severe Rheumatoid Arthritis (RA) or Systemic Lupus Erythematosus (SLE)." Capstan Therapeutics, phase 1, started 9 April 2025. ↩
regulatorU.S. Food and Drug Administration. "FDA Requires Boxed Warning for T cell Malignancies Following Treatment with BCMA-Directed or CD19-Directed Autologous Chimeric Antigen Receptor (CAR) T cell Immunotherapies." 2024.↩Applies to the six autologous products approved at the time, all of them manufactured outside the body.
newsFidler, B. "Lilly boosts 'in vivo' cell therapy capabilities with Kelonia buyout." BioPharma Dive, 20 April 2026.↩ ↩2Reports the deal terms and the preceding acquisitions, and attributes the comparison of prices paid to an RBC analyst.
statementEli Lilly and Company. Form 10-Q for the quarterly period ended 30 June 2026, Note 4.↩A company filing: it establishes the closing date and the accounting treatment, not any clinical claim about KLN-1010.