In Vivo CAR-T for Autoimmune Diseases: The Potential of T-Cell-Targeted AAV Gene Delivery

Abstract

Chimeric antigen receptor T-cell (CAR-T) therapy is emerging as a potentially transformative treatment for severe autoimmune disease. Autologous CD19-directed CAR-T studies have produced profound B-cell depletion and sustained drug-free remission in refractory systemic lupus erythematosus, idiopathic inflammatory myositis, systemic sclerosis, and related disorders. Importantly, remission can persist after CAR-T contraction and B-cell reconstitution, supporting the concept that autoimmune therapy may require a finite immune reset rather than permanent immune surveillance. ¹^,²^,⁵^,⁶ Conventional ex vivo CAR-T nevertheless remains therapeutically intensive, requiring leukapheresis, cell manufacturing, lymphodepletion, and infusion of a premanufactured cellular product. Recent severe inflammatory events in larger autoimmune CAR-T programs further emphasize the need to define the appropriate magnitude and kinetics of cellular activation rather than assuming that maximal expansion is always desirable. ⁸^,⁹^,³⁵ In parallel, in vivo CAR-T platforms are beginning to generate therapeutic cells directly within patients. Targeted mRNA/LNP delivery has shown early human activity in autoimmune disease, while targeted lentiviral vectors have now generated clinically active CAR-T cells in both oncology and neurologic autoimmune disorders. These advances create a persistence spectrum ranging from transient RNA expression to durable integrated gene transfer. We propose that T-cell-targeted adeno-associated virus (AAV) warrants investigation as a mechanistically distinct platform because recombinant AAV genomes remain predominantly episomal and may therefore support DNA-based CAR expression without intentional permanent integration. In proliferating T cells, episomal dilution could potentially limit expression over time. ¹⁷^,¹⁸ This review examines the clinical evidence for CAR-T-mediated immune reset, evolving safety considerations, recent mRNA/LNP and lentiviral in vivo CAR-T results, and the biological rationale, engineering requirements, limitations, and manufacturing considerations for CD3-targeted AAV. The central hypothesis is that the optimal autoimmune CAR-T platform may ultimately be defined not by maximal persistence, but by the ability to generate the right number of CAR-T cells at the right rate and for the right duration.

1. CAR-T Therapy for Autoimmune Disease: From Cancer Treatment to Immune Reset

CAR-T therapy was developed to exploit the cytotoxic capacity of genetically engineered T cells against refractory malignancies, particularly B-cell cancers. The success of CD19-directed CAR-T established that engineered T cells can produce profound depletion of CD19-positive cell populations, suggesting a parallel strategy for autoimmune diseases in which pathogenic B cells participate through autoantibody production, antigen presentation, cytokine secretion, and interactions with autoreactive T cells. Conventional B-cell-directed therapies such as rituximab can be effective but often require repeated treatment and do not uniformly eliminate tissue-resident or deeply pathogenic B-cell populations. ⁷ CAR-T therefore offers a qualitatively different intervention: deep depletion of the B-cell compartment followed by immune reconstitution.

The first compelling clinical evidence came from refractory systemic lupus erythematosus (SLE). Mackensen and colleagues treated five patients with autologous CD19 CAR-T after lymphodepleting chemotherapy and observed rapid CAR-T expansion, complete peripheral B-cell depletion, improvement in clinical manifestations, and marked reduction in disease-associated serologic abnormalities. ¹ They subsequently reported 15 patients with severe autoimmune disease, including eight with SLE, three with idiopathic inflammatory myositis, and four with systemic sclerosis. After fludarabine/cyclophosphamide conditioning and a single CD19 CAR-T infusion, all eight patients with SLE achieved DORIS remission, all three patients with myositis achieved major clinical responses, and the systemic sclerosis cohort showed meaningful reductions in disease activity; immunosuppressive therapy was discontinued in all patients. ² Additional case reports and allogeneic studies have extended this concept across systemic sclerosis, antisynthetase syndrome, and severe inflammatory myopathies.

A particularly important observation is that durable remission does not appear to require permanent B-cell aplasia. In the Müller cohort, B-cell aplasia lasted a mean of approximately 112 days, after which B cells reconstituted in most patients while clinical remission persisted. ² These findings support the concept of an immune reset in which sufficiently deep elimination of pathogenic B-cell clones disrupts the autoreactive network and permits reconstitution of a less pathogenic immune repertoire. ⁶^,¹⁰ This distinction provides the biological foundation for considering CAR-T platforms that are potent enough to achieve deep depletion but are not necessarily designed for lifelong CAR expression.

2. Ex Vivo CAR-T for Autoimmune Disease: Highly Effective but Therapeutically Intensive

The efficacy of ex vivo CAR-T should be considered together with the complexity of the treatment process. Conventional autologous CAR-T requires patient selection, leukapheresis, T-cell isolation and activation, genetic modification, ex vivo expansion, product characterization and release testing, lymphodepleting chemotherapy, infusion, and subsequent in vivo expansion. This model provides important advantages because the cellular product can be characterized before administration and its dose, phenotype, viability, and potency can be controlled. At the same time, it introduces substantial logistical, manufacturing, and clinical complexity.

Ex vivo activation and expansion also influence the biology of the administered product. Patients receive a defined population of already engineered cells capable of immediate antigen recognition and further rapid expansion in vivo. In aggressive malignancy, these kinetics may be advantageous because rapid cytotoxic activity and prolonged surveillance are desirable. In autoimmune disease, however, the therapeutic target is usually not an independently proliferating tumor mass but a pathogenic immune-cell compartment that may only need to be depleted deeply enough to permit immune reconstitution. The optimal cell dose, expansion rate, and duration of persistence may therefore differ between oncology and autoimmune disease. ⁶^,¹⁰

Lymphodepletion is another defining component of most ex vivo protocols. Fludarabine and cyclophosphamide reduce endogenous lymphocytes and increase the availability of homeostatic cytokines, creating a favorable environment for expansion of the infused CAR-T product. ¹^,² However, lymphodepletion also causes transient immunosuppression, cytopenias, and susceptibility to infection. These risks may be acceptable in patients with life-threatening refractory cancer but may carry a different benefit-risk balance as CAR-T moves earlier into chronic autoimmune disease. Importantly, early academic autoimmune CAR-T studies should not be characterized as intrinsically unsafe: in the 15-patient Müller series, toxicity was generally manageable, with predominantly low-grade CRS and only one grade 1 ICANS event. ² The appropriate conclusion is therefore that ex vivo CAR-T can produce extraordinary efficacy with acceptable toxicity in selected patients, but remains an intensive therapy whose optimal conditioning, dose, expansion kinetics, and persistence are not yet defined for autoimmune indications.

3. Autoimmune CAR-T Safety Events and the Importance of Therapeutic Intensity

The expansion of CAR-T into autoimmune disease encountered an important safety development in August 2026, when Novartis temporarily paused multiple autoimmune and neurologic studies of rapcabtagene autoleucel after three fatal cases of immune effector cell-associated hemophagocytic syndrome (IEC-HS). Bristol Myers Squibb separately paused autoimmune studies of zola-cel after inflammatory events reported as transient and reversible. ³⁵ These events warrant careful attention but do not establish that ex vivo CAR-T is broadly unsafe for autoimmune disease or identify a single causal mechanism. Publicly available information has not demonstrated that the events were caused specifically by lymphodepletion, cell dose, manufacturing, CAR architecture, expansion kinetics, or patient characteristics.

CRS and IEC-HS illustrate why cellular kinetics matter. CRS reflects systemic inflammatory signaling after immune-effector-cell activation, whereas IEC-HS is a severe hyperinflammatory syndrome characterized by macrophage activation, cytopenias, coagulopathy, liver dysfunction, hyperferritinemia, and related manifestations. Standardized definitions and grading systems have improved recognition of these toxicities. ⁸^,⁹ The recent safety events therefore reinforce a broader principle: the objective in autoimmune disease should not be maximal immune-effector activation by default, but sufficient biological intensity to achieve immune reset while maintaining an acceptable therapeutic index. This framing provides a strong rationale for technologies that can alter how CAR-T cells are generated, how rapidly they appear, and how long they persist.

4. In Vivo CAR-T: Reprogramming T Cells Directly Within the Patient

In vivo CAR-T seeks to deliver CAR genetic information directly to endogenous T cells, replacing individualized ex vivo cell manufacturing with administration of a gene-delivery agent that generates the therapeutic cell population inside the patient. This strategy could eliminate leukapheresis, ex vivo activation and expansion, cryopreservation, individualized release testing, and the waiting period between cell collection and treatment. More fundamentally, it changes the pharmacology of CAR-T: the administered dose is a vector or nanoparticle, while the final CAR-T population emerges from the number and phenotype of initially transduced cells, vector potency, antigen burden, and subsequent antigen-driven expansion. ¹⁰^,¹⁴^,¹⁵

The role of lymphodepletion may therefore differ across in vivo platforms. Conventional conditioning is designed to support expansion of an infused cellular product, whereas an in vivo vector that depends on endogenous T cells requires those cells to remain available for transduction. Clinical proof that lymphodepletion is not universally required has already emerged in oncology: ESO-T01 generated anti-BCMA CAR-T cells in vivo after a single targeted lentiviral-vector dose without leukapheresis, ex vivo manufacturing, or lymphodepletion. ¹² For receptor-targeted AAV, preservation of the circulating T-cell compartment is even more directly linked to the proposed mechanism because those cells constitute the substrate for vector binding and gene transfer.

In vivo CAR-T also introduces new development variables. The therapeutic cells cannot be fully characterized before treatment because they are generated after vector administration. Biodistribution, target-cell selectivity, off-target gene delivery, initial transduction frequency, CAR expression, T-cell phenotype, expansion kinetics, persistence, and inflammatory responses therefore become interdependent attributes of the administered genetic medicine. The leading platforms currently span transient mRNA/LNP delivery, integrating lentiviral vectors, and potentially predominantly episomal DNA vectors such as AAV.

5. mRNA/LNP In Vivo CAR-T: Early Human Proof of Concept in Autoimmune Disease

Targeted mRNA/lipid nanoparticle (LNP) systems deliver CAR-encoding messenger RNA to selected lymphocyte populations, allowing cytoplasmic translation without nuclear entry or genomic integration. Because the RNA and translated CAR protein are degraded, expression is intrinsically transient, creating a potentially controllable platform that can be discontinued by withholding further doses.

STR-P004 provided early human proof of concept in autoimmune disease. Pan and colleagues reported three patients with refractory autoimmune disorders treated with a CD8-targeted LNP carrying anti-CD19 CAR mRNA. CAR-positive T cells appeared rapidly in peripheral blood, peripheral B cells were completely depleted within days, and the lupus nephritis patient showed improvement in urinary protein and several clinical manifestations. No dose-limiting toxicity or ICANS was reported, while early infusions were associated with transient grade 1 CRS. ¹¹ The cohort is very small and the evidence is currently a meeting abstract rather than a full peer-reviewed clinical report, so broad safety conclusions are premature.

The study nevertheless validates a central premise of in vivo CAR-T: a systemically administered genetic medicine can generate functional CAR-T cells directly within autoimmune patients and produce rapid biological depletion of CD19-positive B cells. At the same time, B-cell depletion lasted only several days in the reported patients, and anti-double-stranded DNA antibody levels initially decreased and then rebounded. ¹¹ These observations do not demonstrate that transient CAR expression is inadequate; repeated dosing may prove advantageous and could permit titratable treatment. They do, however, sharpen the question of how long CAR-T activity must persist to produce durable immune reset.

6. Targeted Lentiviral In Vivo CAR-T: From Oncology to Autoimmune Disease

Engineered lentiviral vectors offer a mechanistically distinct route to in vivo CAR-T because productive transduction results in chromosomal integration of the CAR cassette and therefore can support stable expression through subsequent T-cell proliferation. Achieving selective in vivo delivery requires retargeting the vector toward chosen lymphocyte populations using engineered envelopes, nanobodies, receptor-binding ligands, or related strategies. Preclinical studies established that such vectors can generate CAR-T or CAR-NK cells directly in vivo. ¹⁴^,¹⁵

Clinical evidence first emerged prominently with ESO-T01, an immune-shielded, nanobody-directed lentiviral vector encoding an anti-BCMA CAR. In a phase 1 study of five patients with relapsed or refractory multiple myeloma, a single intravenous dose generated CAR-T cells without leukapheresis, ex vivo manufacturing, or lymphodepleting chemotherapy. Four of five patients achieved objective responses, including three stringent complete responses. ¹² The study also demonstrated that in vivo generation does not inherently eliminate inflammatory toxicity: CRS occurred in four of five patients, including three grade 3 events, and one patient developed grade 1 ICANS. These findings indicate that the kinetics and magnitude of CAR-T generation, antigen burden, and host inflammatory response remain important regardless of whether the cells are manufactured ex vivo or in vivo.

A major advance followed in September 2026, when Cheng and colleagues reported JY231, a lentiviral in vivo CD19 CAR-T approach, in 16 patients with refractory neurologic autoimmune disorders. The cohort included neuromyelitis optica spectrum disorder, MOG antibody-associated disease, generalized myasthenia gravis, and inflammatory myopathy. CAR-T cells were generated in all treated patients, complete B-cell depletion was observed, and preliminary clinical improvement occurred across disease groups; reported early toxicities were generally manageable, with grade 1 CRS in most affected patients and no reported ICANS. ¹³ This study provides direct peer-reviewed evidence that an integrating viral vector can generate clinically active CD19 CAR-T cells in patients with autoimmune disease.

These results materially change the in vivo CAR-T landscape. The field now has human autoimmune evidence for both transient mRNA delivery and integrating lentiviral delivery, so the question is no longer simply whether in vivo CAR-T can work. The more important issue is how delivery-platform properties determine the magnitude, controllability, persistence, and safety of the resulting cellular response. Genomic integration should not be equated with inevitable malignant transformation; modern self-inactivating lentiviral vectors have an improved safety profile and extensive clinical experience. Nevertheless, permanent integration creates a potential insertional-genotoxicity concern and necessitates long-term follow-up. The FDA has required boxed warnings concerning secondary T-cell malignancies after approved BCMA- and CD19-directed autologous CAR-T products, although the contribution of vector integration to individual events remains complex and uncommon. 16

7. CAR-T Persistence as a Therapeutic Design Variable

The clinical success of immune-resetting CAR-T and the emergence of transient and integrating in vivo platforms together suggest that CAR persistence should be treated as a therapeutic design variable rather than an attribute that should automatically be maximized. In oncology, long-term persistence may provide continued surveillance against recurrent malignant cells. In autoimmune disease, durable remission after B-cell reconstitution suggests that a finite interval of deep depletion may be sufficient for some indications. ²^,⁶^,¹⁰

mRNA/LNP and lentiviral vectors occupy distinct regions of this persistence spectrum. With mRNA, CAR expression is governed primarily by RNA and protein turnover and can be renewed through repeat dosing. With lentivirus, integrated DNA can be maintained through cell division and may support prolonged or permanent expression in surviving clones. The STR-P004 and JY231 experiences are therefore informative not because one platform has already been shown superior, but because both can generate biologically active CAR-T cells in autoimmune patients while imposing fundamentally different genetic persistence mechanisms. 11,13

Different autoimmune diseases may ultimately require different therapeutic windows depending on tissue distribution of pathogenic B cells, the contribution of long-lived plasma cells, antigen target, previous therapy, and the kinetics of immune reconstitution. The key principle is that maximal persistence should not automatically be assumed to be optimal. A platform capable of providing an intermediate expression window could therefore be valuable if several days of activity prove too short for some diseases while permanent expression provides more persistence than required.

8. AAV as a Potential Intermediate Platform for In Vivo CAR-T

AAV is one of the most extensively developed in vivo gene-delivery platforms. Recombinant AAV is replication deficient and its vector genomes are maintained predominantly as extrachromosomal episomes rather than through programmed chromosomal integration. Low-frequency integration can nevertheless occur, particularly in association with DNA damage; accordingly, AAV should be described as predominantly episomal rather than absolutely non-integrating. ¹⁷^,¹⁸

This genomic behavior may create a persistence mechanism that differs from both mRNA and lentivirus. In non-dividing or slowly dividing tissues, AAV episomes can support prolonged expression. In proliferating cells, however, vector genomes are not duplicated as part of the chromosome and can be progressively diluted as cells divide. ¹⁷^,¹⁸ For a CAR-T application, antigen recognition could initially drive expansion of an AAV-transduced T-cell population while repeated division simultaneously reduces the average episomal vector copy number. In principle, CAR-T activity could therefore be amplified early and become progressively limited later as episomes are diluted.

This model remains a hypothesis and should not be presented as an established feature of AAV CAR-T. The actual duration of CAR expression in proliferating human T cells in vivo has not been determined and would depend on initial vector copy number, promoter activity, T-cell phenotype, antigen burden, proliferative history, episomal stability, and immune selection. Nevertheless, it creates a mechanistically distinct possibility: sustained DNA-based expression without intentional permanent integration and with the potential for proliferation-dependent loss of the genetic template.

AAV is not automatically a better CAR-T vector. Natural AAV serotypes were not optimized for systemic T-cell transduction, and intravenously administered particles can be diverted to non-target tissues, especially the liver. The feasibility of AAV in vivo CAR-T therefore depends on whether capsid engineering can achieve sufficiently selective and productive T-cell transduction at a systemic vector dose low enough to provide a practical safety margin.

9. Engineering AAV for Direct T-Cell Targeting

Natural AAV serotypes differ in receptor usage, intracellular trafficking, and tissue tropism. AAV6 has attracted particular interest in hematopoietic gene transfer and has been widely used to deliver donor templates during genome editing of hematopoietic stem/progenitor cells and primary T cells. ¹⁹^,²⁰ Natural tropism alone, however, is unlikely to provide the selectivity required for systemic T-cell targeting.

The AAV capsid is highly engineerable. Surface-exposed regions of the 60-subunit VP1/VP2/VP3 capsid participate in receptor interaction, antibody recognition, and trafficking, and can be modified to display peptides, antibody fragments, nanobodies, or other receptor-binding ligands. Hamann and colleagues demonstrated this principle by incorporating CD4-specific nanobodies into AAV2 capsids, improving targeting of CD4-positive cells including primary human PBMCs and purified CD4-positive T lymphocytes. ²¹ These studies show that AAV tropism can be converted from natural tissue preference toward receptor-directed cellular delivery.

CD3 is an attractive target for broad T-cell recognition because it is a defining component of the T-cell receptor complex and is widely expressed across mature T-cell populations. Display of a CD3-specific single-chain variable fragment (CD3scFv) on the AAV capsid could therefore combine receptor recognition and vector delivery within the same particle. This approach is biologically more complex than targeting a passive surface marker because CD3 engagement can itself induce signaling. The magnitude of activation is likely to depend on epitope, affinity, avidity, receptor occupancy, ligand density, and the geometry of multivalent display on the capsid.

Our group has developed CD3scFv-modified AAV capsids for direct T-cell targeting and is evaluating this approach for in vivo CAR-T generation. In preclinical studies, selected capsid configurations enhanced productive transduction of primary human T cells and supported CAR expression, providing experimental support for receptor-directed AAV as a T-cell engineering strategy (H. Chen et al., manuscript under review). Because these findings have not yet completed peer review, they should be considered emerging preclinical evidence rather than established clinical validation; the broader biological rationale in this review is therefore supported independently by published studies of AAV capsid retargeting, AAV6-mediated gene delivery, and in vivo CAR-T platforms.

The design objective should therefore not be maximal CD3 binding. An optimized CD3-targeted AAV should maximize productive gene transfer while minimizing unwanted polyclonal activation. Development should evaluate vector binding and internalization together with transduction efficiency, activation markers, proliferation, cytokine release, differentiation state, and exhaustion. The number and position of CD3scFv moieties on the capsid may be especially important because multivalent presentation could produce signaling properties different from those of soluble anti-CD3 antibodies.

Targeted AAV may also generate CAR-T cells with different kinetics from an ex vivo infusion. Vector distribution, receptor binding, internalization, intracellular trafficking, nuclear delivery, vector-genome processing, transcription, translation, and surface CAR expression must occur before antigen recognition. CAR-positive cells should therefore emerge progressively after dosing rather than being delivered as a preformed cellular bolus. Whether this difference reduces inflammatory toxicity is unknown and should be tested directly; the ESO-T01 experience demonstrates that in vivo generation alone does not prevent CRS. 12

10. Potential Therapeutic Advantages of T-Cell-Targeted AAV

If receptor-directed AAV can achieve efficient T-cell transduction at sufficiently low systemic doses, several potential advantages follow. Like other in vivo platforms, it could replace individualized cell manufacturing with a standardized vector product. More specifically, preservation of the endogenous T-cell compartment is integral to the proposed mechanism: circulating T cells are the substrate to be transduced, so broad lymphodepletion immediately before vector administration would be mechanistically counterproductive. This does not exclude all forms of immune modulation, but the platform concept is fundamentally different from ex vivo protocols in which conditioning is used to support an infused cellular product.

Consistent with this development strategy, our ongoing preclinical work with CD3scFv-modified AAV is examining whether receptor-directed capsid engineering can increase functional T-cell transduction and thereby reduce the vector exposure required for CAR-T generation (H. Chen et al., manuscript under review). Until those data are peer reviewed, however, low-dose efficacy and any associated safety advantage should be regarded as hypotheses requiring direct quantitative comparison with untargeted AAV and other in vivo CAR-T platforms.

A second potential advantage is dose reduction through receptor targeting. Conventional systemic AAV gene therapy may require very high vector doses when large tissue compartments must be transduced. High-dose systemic AAV has been associated with transaminitis, thrombocytopenia, complement activation, sinusoidal endothelial injury, and other toxicities; nonhuman-primate studies have demonstrated clinically important liver and vascular injury at doses on the order of 10^14 genome copies/kg and above. ²² For targeted CAR-T, the vector may need to generate only a sufficiently large founding population of CAR-expressing T cells, after which antigen-driven proliferation can amplify the therapeutic population. The most meaningful measure of successful targeting may therefore be how much it reduces the systemic vector dose required to achieve a defined level of B-cell depletion.

A third potential advantage is the combination of DNA-based expression with predominantly episomal persistence. This could provide a distinct relationship between CAR-T expansion and transgene duration: antigen-driven proliferation may initially amplify the number of CAR-positive cells while progressively diluting the episomal templates that sustain expression. If validated, such cell-division-dependent pharmacology could create a biological self-limiting mechanism that differs from both mRNA degradation and stable lentiviral integration.

11. Challenges and Unanswered Questions for AAV-Mediated In Vivo CAR-T

The potential advantages of AAV must be considered alongside established limitations. Pre-existing anti-AAV antibodies are common and can reduce systemic transduction, while vector administration usually induces additional anti-capsid immunity that complicates repeat dosing. ²³^,²⁴ Recent work continues to evaluate alternative capsids, antibody-removal strategies, capsid shielding, and transient immune modulation, including prophylactic abatacept or dasatinib, but routine systemic AAV redosing is not yet a general clinical capability. ²⁵^,²⁶ Whether a single AAV dose will be sufficient for autoimmune immune reset is therefore a critical clinical question.

Off-target biodistribution is another major concern. The liver can act as a substantial sink for systemically administered AAV, creating non-therapeutic exposure and potentially increasing toxicity. An effective T-cell-targeted vector may therefore require both positive targeting toward T cells and detargeting from liver and other non-target tissues. Capsid engineering can be complemented by transcriptional restriction or post-transcriptional detargeting strategies where appropriate.

CD3 targeting itself must be carefully controlled because excessive receptor engagement could cause cytokine production and broad T-cell activation independent of CAR-antigen recognition. The optimal capsid may therefore be the one that achieves the highest ratio of productive transduction to unwanted signaling rather than the highest binding affinity. In parallel, the phenotype of the initially transduced T cells should be characterized because naive, memory, effector, regulatory, and exhausted populations differ substantially in proliferative capacity, trafficking, persistence, and cytokine production.

AAV genome capacity imposes an additional engineering constraint. The natural genome is approximately 4.7 kb, and packaging efficiency and full-length genome integrity decline as recombinant genomes approach or exceed the native capacity. ²⁷^,²⁸ This is highly relevant to CAR design because promoter, signal peptide, antigen-binding domain, hinge, transmembrane region, costimulatory and signaling domains, regulatory elements, and polyadenylation sequence must fit within a constrained genome. Dual CARs, cytokines, safety switches, and other armored functions may therefore require compact cassettes or multi-vector strategies.

Finally, AAV integration cannot be described as zero risk. Although vector genomes remain predominantly episomal, low-frequency integration can occur. ¹⁷ For a T-cell-targeted application, integration-site analysis and long-term clonal monitoring should be incorporated into translational development, particularly because antigen-driven expansion could amplify individual transduced clones. Together with quantitative biodistribution, neutralizing-antibody testing, cytokine profiling, and dose-response studies, these analyses will be essential for determining whether the theoretical persistence and dose advantages of targeted AAV translate into a meaningful therapeutic profile.

12. AAV Production, Manufacturing, and Translational Considerations

Translation of receptor-targeted AAV will depend on a reproducible relationship among capsid design, vector-genome integrity, manufacturing, product quality, administered dose, T-cell transduction, and biological activity. This is especially important for ligand-displaying capsids because modifications that improve receptor targeting may simultaneously alter capsid-protein expression, assembly, VP stoichiometry, genome packaging, particle stability, purification behavior, or potency. Capsid engineering and manufacturing should therefore be developed as an integrated process rather than as sequential activities.

Multiple rAAV production systems are available. Transient transfection of HEK293 cells remains widely used and provides flexibility during early development, whereas baculovirus/Sf9 systems offer a scalable suspension-manufacturing platform; stable mammalian producer-cell and helper-virus approaches provide additional options. The relative advantages of these systems depend on vector design, scale, yield, product quality, and process economics, and no single platform can be assumed optimal for every engineered capsid.

For a receptor-targeted vector, conventional physical measurements are necessary but not sufficient. Vector-genome titer is essential for dose assignment, yet a preparation can contain the expected number of genomes while showing reduced biological activity if the targeting ligand is improperly folded, present at an inappropriate density, or structurally altered during production and purification. A meaningful potency assay for CD3scFv-AAV should therefore assess receptor-dependent binding followed by productive transduction and CAR expression in relevant human T cells. This is particularly important because small differences in the initial number of transduced cells could be amplified substantially by antigen-driven proliferation.

Downstream processing must also control full, empty, and partially packaged capsids, host-cell proteins, residual nucleic acids, aggregates, and production-system-specific impurities. Chromatographic purification has become increasingly important for scalable rAAV downstream processing. ³³ For ligand-displaying vectors, additional critical quality attributes may include the fraction and stoichiometry of modified capsid proteins, ligand accessibility, receptor-binding activity, and retention of targeting potency after formulation and storage. Genome size and integrity must be considered early because increasingly complex CAR cassettes can approach the practical AAV packaging limit. ²⁷^,²⁸

The manufacturing objective for T-cell-targeted AAV should therefore not be maximal bulk vector yield alone. It should be maximal yield of functionally active, correctly targeted particles per administered dose. If receptor-directed capsid engineering can substantially increase productive T-cell transduction, gains in biological potency and resulting dose reduction may be more important than a simple increase in vector-genome output.

13. Expert Perspective: Right Cell, Right Dose, Right Duration

The clinical development of CAR-T for autoimmune disease is moving from a proof-of-concept phase toward a platform-design problem. Ex vivo CD19 CAR-T has shown that profound B-cell depletion can produce sustained drug-free remission and that remission can persist after B-cell reconstitution. ¹^,² Targeted mRNA/LNP and lentiviral vectors have now shown that CAR-T cells can also be generated directly within autoimmune patients. ¹¹^,¹³ At the same time, the recent rap-cel safety events and the grade 3 CRS observed with ESO-T01 remind the field that in vivo engineering does not abolish inflammatory risk. ¹²^,³⁵

Future comparisons among platforms should therefore focus less on simple labels such as viral versus non-viral and more on the clinically relevant variables that define the therapeutic window: which T-cell populations are modified, how many CAR-T cells are generated, how rapidly they appear, how strongly they activate, how long CAR expression persists, how completely pathogenic B cells are depleted, how rapidly healthy B cells recover, and whether durable remission survives CAR-T contraction.

Within this framework, T-cell-targeted AAV is best viewed as a testable hypothesis rather than a proven replacement for mRNA/LNP, lentiviral vectors, or conventional ex vivo CAR-T. Its scientific rationale rests on the combination of receptor-directed delivery, the possibility of substantially lowering systemic vector dose, predominantly episomal DNA persistence, and the potential for proliferation-dependent loss of CAR expression. Its limitations—pre-existing immunity, redosing barriers, off-target biodistribution, packaging capacity, low-frequency integration, and high-dose toxicity—must be evaluated with equal rigor.

The CD3scFv-AAV approach discussed here is being evaluated experimentally by our group, including in work currently under peer review (H. Chen et al., manuscript under review). Its inclusion in this perspective is intended to define a testable delivery and persistence model, not to imply that clinical superiority over mRNA/LNP, lentiviral vectors, or ex vivo CAR-T has been established.

The central development principle can be summarized as “right cell, right dose, right duration”. Autoimmune therapy may not require the largest possible CAR-T population or the longest possible persistence. It may require enough CAR-T activity, generated with appropriate kinetics, to achieve deep depletion of the pathogenic immune compartment and permit durable immune reconstitution.

14. Conclusions

CAR-T therapy has established a powerful new principle for severe autoimmune disease: profound B-cell depletion can induce clinical responses that persist after CAR-T contraction and B-cell reconstitution, supporting a finite immune-reset model rather than lifelong suppression. Emerging in vivo platforms now extend this principle beyond individualized cell manufacturing. Targeted mRNA/LNP provides transient, repeatable CAR expression, whereas targeted lentiviral vectors can generate durable CAR-T populations through genomic integration; both have now shown human activity in autoimmune disease. ¹¹^,¹³

AAV introduces a mechanistically distinct possibility. Recombinant AAV genomes are predominantly episomal and may therefore support DNA-based CAR expression without intentional permanent integration, while proliferation of antigen-stimulated T cells could progressively dilute the vector genome. Whether this produces a therapeutically useful intermediate window remains unproven. Success will depend on efficient receptor-directed T-cell transduction at low systemic dose, controlled CD3 signaling, acceptable biodistribution and immunogenicity, reliable manufacturing, and a CAR expression interval appropriate for immune reset.

The next generation of autoimmune CAR-T may therefore be defined not by which platform produces the most persistent engineered T cells, but by which platform generates the “right number of CAR-T cells, at the right rate, for the right length of time”.

Correspondence

Haifeng Chen, Ph.D., Virovek, Inc., Houston, Texas, USA.

Declaration of Interests

H.C. is founder and an equity holder of Virovek, Inc. and AAVivo, Inc., and is involved in the development of T-cell-targeted AAV and AAV-mediated in vivo CAR-T technologies discussed in this article. H.C. is also an inventor on patent applications related to T-cell-targeted AAV technology. The CD3scFv-AAV preclinical work cited in the text as “manuscript under review” has not yet completed peer review. Published evidence and proposed mechanisms are distinguished throughout the article.