In vivo CAR-T therapy: latest updates from the 2025 ASH annual meeting

Abstract

Chimeric antigen receptor (CAR)–T cell therapy has significantly improved outcomes in hematologic malignancies but remains constrained by complex ex vivo manufacturing and lymphodepleting chemotherapy. In vivo CAR-T therapy enables direct genetic programming of endogenous T cells through systemic delivery of CAR-encoding vectors, thereby bypassing ex vivo manipulation and simplifying treatment workflows. This strategy has the potential to reduce production timelines, lower costs, and broaden patient accessibility. In addition, in vivo CAR-T approaches may preserve more physiological T-cell states and offer improved scalability. Importantly, however, current in vivo CAR-T platforms differ substantially in delivery modality, persistence, reversibility, safety profile, and translational feasibility. Here, we summarize the latest advances in major preclinical approaches and early clinical studies presented at the 2025 ASH annual meeting, with emphasis on the comparative strengths, limitations, and translational trade-offs of viral vector-based, RNA/lipid nanoparticle-based, non-viral DNA, and gene-writing strategies.

To the editor

Chimeric antigen receptor (CAR)–T cell therapy has achieved remarkable clinical success in hematologic malignancies [1, 2]. However, its broader application is constrained by complex personalized preparation processes and the requirement for lymphodepleting chemotherapy. In vivo CAR-T therapy has emerged as a fundamentally distinct paradigm, which can directly “edit” T cells within the human body, offering potential advantages in scalability, accessibility, and treatment simplicity by enabling direct generation of CAR-T cells within patients [3]. Here, we highlight the latest preclinical and early clinical advances in in vivo CAR-T cell therapy presented at the 2025 ASH annual meeting and provide a platform-based analysis of this rapidly evolving field.

Preclinical investigations

The core challenge of in vivo CAR-T therapy is to generate functional CAR-T cells directly in patients while balancing target-cell specificity, expression durability, redosing feasibility, safety, and manufacturability. No single system simultaneously optimizes target-cell specificity, expression durability, safety, and manufacturability, necessitating indication-specific platform selection. Based on delivery modality and expression characteristics, the preclinical studies presented at ASH 2025 can be broadly grouped into four technological classes: viral vector-based approaches, transient RNA/lipid nanoparticle (LNP) systems, non-viral DNA/minicircle platforms, and gene-editing or gene-writing strategies.

Viral vector–based approaches, particularly engineered lentiviral systems, enable efficient T-cell transduction together with relatively durable CAR expression. IASO Bio reported an engineered lentiviral platform capable of delivering BCMA-, CD19-, and CD20-directed CAR constructs through T-cell-targeted transduction and T-cell-restricted promoter control, with antitumor activity reported in humanized mice and myeloma models [4] (Table 1). Qihan Bio further optimized viral entry using binder-fusogen combinatorics to functionally separate transduction from early T-cell activation, thereby maintaining a more quiescent metabolic state during the transduction phase and supporting antitumor activity in animal models [5] (Table 1). Similarly, Russell et al. showed that a single intravenous administration of the engineered lentiviral vector (LV-169) selectively delivered a functional BCMA-CAR transgene to niche-resident, resting T cells and eliminated disseminated myeloma in major histocompatibility complex(MHC)-deficient NOD scid gamma (NSG) mice [6] (Table 1). While these systems support robust expansion and persistence, they raise concerns regarding insertional mutagenesis, off-target transduction, immunogenicity, and limited redosing potential.

Table 1.

Preclinical studies of in vivo CAR-T presented at ASH 2025

Platform type	Program (company)	Route of Administration	Disease Model	Diseases	Therapeutic Target	T-cell Targeting Marker	Expression Characteristics	Key Findings	Reference
Viral vector-based	V-206/207/208 (IASO Bio)	Tail vein injection	Humanized mice; myeloma models	RRMM, NHL, AID	CD19/CD20/BCMA	CD7	Persistent	T-cell-specific invasion and promoter regulation enabled in vivo delivery of BCMA/CD19/CD20 CARs with antitumor activity and no obvious toxicity reported in the abstract.	[4]
Viral vector-based	Binder-fusogen combination (Qihan Bio Inc)	Not specified	Humanized mouse models	-	Not specified	CD3/CD7/CD8	Persistent, potentially redosable	Functionally separated viral entry from early T-cell activation to maintain a quiescent transduction state while supporting antitumor activity in animal models.	[5]
Viral vector–based	LV-169 (Vyriad & Imanis Life Sciences)	Intravenous injection	Human multiple myeloma xenograft (OPM-2) in MHC-deficient NSG mouse	MM	BCMA	CD3	Persistent	Directly generated functional BCMA-CAR T cells from resting niche-resident T cells after intravenous administration, bypassing ex vivo manufacturing.	[6]
RNA/LNP-based	Orna panCAR (Orna Therapeutics)	Intravenous injection	Humanized mouse models of lupus	B cell-mediated AID	BCMA	Not specified	Transient, redosable	The delivery system consists of proprietary immunotropic LNP + synthetic circular RNA with scalable potential.	[7]
RNA/LNP-based	tLNP in vivo CAR-T (Aera Therapeutics)	Not specified	Multiple preclinical models	B cell-mediated AID	CD19	CD8	Transient, redosable	Rapidly generated transient anti-CD19 CAR-T cells in vivo, producing reversible B-cell depletion and immune remodeling.	[8]
DNA/minicircle-based	STAR-CAR platform (Velvet Therapeutics)	Intravenous injection	Humanized mouse models	AID	CD19	Not specified	Transient, redosable	Demonstrates proof-of-concept for systemic in vivo immune cell reprogramming and immune tolerance restoration in autoimmune disease models.	[9]
Gene-editing/gene-writing	Gene Writer tLNP (Tessera Therapeutics)	Intravenous injection	Humanized mouse models; NHP	SCD, AID	CD20	Not specific	Potentially persistent	Targeted LNP-mediated RNA delivery enabled in vivo gene editing in hematopoietic stem cells and T cells, supporting durable cell engineering applications.	[10]
Gene-editing/gene-writing	Gene Writer tLNP (Tessera Therapeutics)	Intravenous injection	Humanized xenograft mouse models; CD34 + humanized mouse models; NHP	AID	CD19/CD20	Not specified	Potentially persistent	Targeted LNP delivery of an RNA gene writer enabled in vivo generation of functional CAR-T cells and efficient editing of resting T cells without pre-activation.	[11]

AID: autoimmune disease; MM: multiple myeloma; NHL: non-Hodgkin lymphoma; NHP: non-human primate; RRMM: relapsed/refractory multiple myeloma; SCD: sickle cell disease

RNA/LNP-based delivery systems offer a non-integrating alternative characterized by transient expression and improved reversibility. Orna Therapeutics reported a circular RNA (oRNA^®) panCAR platform targeting CD19 that induces controllable B-cell depletion in animal models, reduced autoantibody titers in a lupus model, and favored recovery of naïve B-cell compartments [7] (Table 1). Aera Therapeutics presented a targeted LNP strategy that rapidly generated transient CAR-T cells in vivo without ex vivo preparation or lymphodepletion, resulting in reversible B-cell depletion and immune remodeling in preclinical models [8] (Table 1). These platforms are particularly attractive for autoimmune indications or settings requiring controllable activity; however, limited persistence may reduce efficacy in malignancies requiring sustained immune pressure.

Non-viral DNA and programmable gene-engineering platforms expand the design space further. Velvet Therapeutics described a minicircle DNA-based STAR-CAR platform that supported proof-of-concept systemic immune cell reprogramming with the potential for repeated administration [9] (Table 1). In parallel, targeted LNP-mediated RNA delivery and gene-writing systems from Tessera Therapeutics enabled in vivo gene editing in hematopoietic stem cells and T cells, as well as generation of functional CAR-T cells without conventional ex vivo manufacturing [10, 11] (Table 1). These approaches are especially notable because they seek to combine the manufacturing simplicity of in vivo delivery with more programmable and potentially durable cell engineering. Nevertheless, questions remain regarding editing precision, biodistribution, ectopic expression in non-target cells, and the regulatory complexity of permanently rewritten cells.

Collectively, these platforms reflect trade-offs rather than a single optimal solution. Viral vectors favor persistence and potency but raise integration risks; RNA/LNP systems enhance reversibility but require repeat dosing; DNA-based platforms improve safety and flexibility but may reduce durability; and gene-writing approaches offer programmability at the cost of increased biological and regulatory complexity. Future progress will depend on improving cell-type specificity, minimizing off-target delivery, controlling CAR expression kinetics, and enabling scalable manufacturing.

Clinical trials

Clinical translation remains at a very early stage, and the currently available evidence is limited to small phase 1 cohorts reported in conference abstracts. KLN-1010 (inMMyCAR) is an envelope-modified, replication-incompetent lentiviral vector designed to generate anti-BCMA CAR-T cells in vivo. In the first three heavily pretreated patients with relapsed/refractory multiple myeloma (RRMM), KLN-1010 induced peripheral CAR-T expansion and persistence with a memory phenotype. The reported best responses were partial responses that deepened to very good partial responses, with MRD-negative assessments in all three patients. Toxicities were reported as manageable, including grade 2 cytokine release syndrome in two patients, no ICANS, and transient hematologic adverse events [12] (Table 2). In a phase 1 investigator-initiated study of GT801 for relapsed/refractory CD19-positive B-cell lymphoma, two patients with non-Hodgkin lymphoma (NHL) received treatment without prior lymphodepletion. High CAR expression was detected in peripheral blood T cells, and reproducible CAR-T cell expansion observed after each dose. The abstract further stated that no CAR signal was detected in non-T-cell PBMC populations, suggesting limited off-target delivery in this preliminary dataset. By week 4, both patients achieved partial response [13] (Table 2).

Table 2.

Clinical trials of CAR-T presented at ASH 2025

Product	Clinical Trial ID	Clinical Stage	Target	Diseases	Dose Level	Patient Number	BOR	CRS	ICANS	AEs	Grade 3 or Higher Adverse Events	ORR	CAR expansion/persistence data	Lymphodepletion Status	Follow-up Duration	Reference
KLN-1010	NCT07075185	Phase 1	BCMA	RRMM	Dose-escalation; exact doses not disclosed	3	VGPR (3/3)	Grade 2 (n = 2)	None	3	Grade 3 anemia (1 patient), Grade 3/4 neutropenia (transient)	100%	Peak expansion days 13–18; CAR+ cells 22–72% of CD3 + T cells; persistence in blood and bone marrow through ≥ 3 months with memory phenotype	NR	Up to 3 months (ongoing study)	[12]
GT801	NCT07205315	Phase 1	CD19	R/R B cell malignancies (NHL)	Effective at ≥ 0.01 mg/kg	2	PR (2/2)	/	NR	NR	None	100%	> 30-fold CAR-T expansion in vivo; sustained CAR expression over time	None	NR	[13]

CAR-T: chimeric antigen receptor T-cell therapy; BOR: best overall response; CRS: cytokine release syndrome; ICANS: immune effector cell-associated neurotoxicity syndrome; ORR: objective response rate; BCMA: B-cell maturation antigen; NR: not reported; R/R: relapsed/refractory; RRMM: relapsed/refractory multiple myeloma; VGPR: very good partial response

In vivo CAR-T therapy still entails several critical biological and translational safety concerns that warrant close attention, including off-target transduction, ectopic CAR expression in non-T cells, biodistribution, and vector shedding. The current main challenges are how to establish a reliable biological distribution monitoring system, extend follow-up time, formulate re-administration strategies, and integrate controllable systems (such as inducible promoters or safety switches).

Abbreviations

ASH

American Society of Hematology

CAR

Chimeric antigen receptor

BCMA

B-cell maturation antigen

CD

Cluster of differentiation

MHC

Major histocompatibility complex

NSG

NOD scid gamma

CRS

Cytokine release syndrome

ICANS

Immune effector cell–associated neurotoxicity syndrome

LNP

Lipid nanoparticle

MRD

Minimal residual disease

NHP

Non-human primate

PoC

Proof-of-concept

PR

Partial response

R/R

Relapsed/refractory

tLNP

Targeted lipid nanoparticle

NHL

Non-Hodgkin lymphoma

Author contributions

SW and XY drafted this manuscript. XY and XZ revised the manuscript. XZ provided direction and guidance throughout the preparation of the manuscript. All authors read and approved the final manuscript.

Funding

This work was supported by National Natural Science Foundation (No.82570247); Shandong Provincial Natural Science Foundation (ZR2025QC1699); Taishan Scholars Program of Shandong Province (No.tstp20250756).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.