Credit: S.Harris/Springer Nature Limited

Engineered immune cells called CAR Ts that can seek out and destroy B cells have transformed the treatment of some blood cancers, and are raising hopes for possible cures for autoimmune diseases. But these living cell therapies — made by harvesting cells from patients, laboriously growing and engineering them in a lab to express a chimeric antigen receptor (CAR), and then re-infusing them into patients — are hard to make, difficult to deliver to patients and expensive. Could gene therapy tools that can turn immune cells into CAR carriers, directly in the body, be the solution?

“The beauty of what we're doing — and we have this pasted on the wall — is that it is faster, better and cheaper,” says Phil Johnson, CEO of Interius BioTherapeutics, one of the most advanced companies in this space.

In vivo CAR T generation is done with off-the-shelf reprogramming cassettes, he explains, making it weeks faster than current manufacturing processes for ex vivo CAR Ts. The cost of goods for these cassettes might be as low as US$5,000 per dose, he estimates, making them a fraction of the cost of a living cell therapy. And they could be better, for two key reasons too. Ex vivo CAR T recipients need to be treated with harsh chemotherapy regimens ahead of cell therapy transfusions to make room for the engineered immune cells, but lymphodepletion cannot be used during the generation of in vivo CAR Ts and so patients retain intact immune systems. And because the souped up in vivo immune cells are less exhausted than their ex vivo counterparts, they could be more active too. “There's a real, theoretical possibility that those cells will simply be better cancer fighters,” says Johnson.

These expectations are founded on emerging data from up and coming gene therapy vectors, the delivery vehicles that will carry the code for a CAR selectively to a subset of immune cells. Interius, Umoja and Kelonia have embraced engineered lentiviral vectors that make permanent changes in immune cells. Capstan, Myeloid Therapeutics, Orbital and Orna have instead prioritized lipid nanoparticles (LNPs) paired with RNA molecules, to induce transient surges of CAR expression in targeted cells. Larger pharmaceutical companies are on the case as well, with Sanofi disclosing last year that it had three in vivo CAR T programmes in preclinical development.

The success of these therapies will depend not just on the vectors, but also on which immune cells they target, what CAR payloads they carry, and what disease settings they are used in. After years of tinkering with design decisions and plans, multiple candidates are in and entering the clinic (Table 1). “By 2026 we'll have learned a lot,” says Adrian Bot, CSO of Capstan. “Half a dozen trials will have read out by then. It's going to be tremendously exciting.”

Table 1 | In vivo CAR immune cells in and approaching the clinic

Drug name

Company

Vector (cell-targeting mechanism)

Therapeutic payload

Lead indication

Planned phase I start

INT2104

Interius

Lentivirus (CD7 scFv)

CD20 CAR

B cell cancers

2024

INT2106

Interius

Lentivirus (CD7 scFv)

CD19 CAR

Autoimmune

2025

UB-VV111

Umoja/Abbvie

Lentivirus (CD3 scFv, CD80 and CD58)

CD19 CAR; RACR

B cell cancers

2024

UB-VV400/410

Umoja/IASO

Lentivirus (CD3 scFv, CD80 and CD58)

CD22 CAR; RACR

B cell cancers

2024

UB-VV300/310

Umoja

Lentivirus (CD3 scFv, CD80 and CD58)

CD20 CAR

NHL/Autoimmune

2026

KLN-1010

Kelonia

Lentivirus (CD3 antibody)

BCMA CAR

Multiple myeloma

2025

Discontinued

Sana

Lentivirus

Various CARs

Discontinued

Discontinued

CPTX2309

Capstan

LNP (CD8 antibody)

CD19 CAR (mRNA)

Autoimmune diseases

"Near future"

Undisclosed

Orbital

LNP (undisclosed cell-targeting moiety)

CD19 CAR

Autoimmune diseases

"Near future"

ORN-145

Orna

LNP (no cell-targeting moiety)

CD19 CAR (circular RNA)

B cell cancers

Undisclosed

ORN-252

Orna

LNP (no cell-targeting moiety)

CD19 CAR (circular RNA)

Autoimmune diseases

2026

ORN-328

Orna

LNP (no cell-targeting moiety)

BCMA CAR (circular RNA)

Multiple myeloma

2026

MT-302

Myeloid

LNP (no cell-targeting moiety)a

TROP2 CAR (mRNA)

Epithelial tumours

2023

MT-303

Myeloid

LNP (no cell-targeting moiety)a

GPC3 CAR (mRNA)

Liver cancer

2024

Undisclosed

Carisma/Moderna

LNP (no cell-targeting moiety)

GPC3 CAR (mRNA)

Liver cancer

Undisclosed

These results could transform the way CAR Ts are categorized, he adds, shifting them from cell therapies to gene therapies. “The distinctions between classes of therapies is blurring in front of our eyes,” says Bot.

Lentiviral reprogramming

Years before ex vivo CAR T cell therapies first entered the clinic, researchers were already hard at work retooling lentiviral vectors to enable in vivo reprogramming of immune cells.

For Christian Buchholz, who heads the molecular biotechnology and gene therapy research group at the Paul-Ehrlich-Institut, early inspiration struck 20 years ago. Another team had reported in 2005 that when a cancer-killing measles virus was covered with single-chain antibody fragments, it could be targeted to cancerous immune cells. Buchholz, seeing these findings, wondered whether a similar strategy — engineering a lentivirus to express an antibody-based cell targeting moiety, to target a cell type of choice, and a measles virus glycoprotein, to drive the vector’s uptake into host cells — could help get lentiviral vectors into specific cell types. Sure enough, surface-engineered CD20-targeted vectors were preferentially taken up by B cells, he reported in 2008. CD8-targeted vectors got into T cells in vivo in mouse models as well, he reported in 2012.

But to turn a vector into a drug, a potent payload is needed. “When you go in vivo, you only hit a very tiny fraction of cells,” says Buchholz. “You need to have a therapeutic strategy that will give those cells a selective advantage in vivo to finally get a readout”. With B-cell depleting CAR Ts first entering the clinical in 2010, Buchholz saw an opportunity. A CD8-targeted CD19-CAR-loaded lentiviral vector transduced T cells in vivo in mice and wiped out B cells, he reported in 2018.

Several biotechs have embraced this strategy, with a few tweaks along the way. Each firm has chosen a preferred cell-targeting moiety, the antibody-based binder that defines which subset of cells are targeted. The glycoprotein matters too, acting as the entry key into engaged immune cells. And then there is the therapeutic payload, the CAR that enables the retrained T cells to seek out malignant foes. Layered on top of all that is manufacturing.

“The challenge really has been, how do you turn this into something that can be scaled and transferred into human use,” says Johnson.

His team at Interius has chosen to embed an anti-CD7 antibody fragment into the vector to target CD7+ cells, enabling the targeting of both T cells and NK cells. “We're creating a more diverse, if you will, population of effector cells,” he adds.

“We know that CAR NK cells are extremely potent, so it will be interesting to see how that decision bears out in the clinic,” says Laura Evgin, a CAR T researcher at the University of British Columbia.

Umoja, by contrast, uses a multidomain solution to efficiently target a different subset of T cells. An early iteration of this vector used just an antibody fragment against the canonical CD3 T cell marker, hewing closely to the cell types that make up ex vivo CAR T products. But the reprogramming efficiency was only in the 10–15% range. “Frankly we had to spend a lot of energy retooling it,” says Umoja COO David Fontana. By fusing the CD3 antibody fragment to the co-stimulatory domains of CD80 and CD58, the team ultimately found it could dramatically increase the reprogramming efficiency of the vector.

“We're mimicking as closely as we believe we can the antigen-presenting cell and T cell interaction. We believe that's important to our game here,” says Fontana. The multidomain approach offers a reprogramming efficiency in the 50% range, he adds.

On the glycoprotein side, Buchholz has shown that measles-modified vectors offer ‘near-absolute’ selectivity, with little to no off-tissue editing. Interius and Umoja have each studded their vectors with glycoproteins from other virus families, and these use different mechanisms to gain entry into target cells. This might increase the risk of off-tissue editing, cautions Buchholz.

But any possible loss in specificity has to be considered alongside potency and manufacturability, counters Kevin Friedman, CEO of Kelonia Therapeutics. If other glycoproteins offer higher reprogramming rates or fewer non-functional particles, lower doses are achievable, potentially improving a product’s overall safety and efficacy profile. “You really need to have all three pillars — specificity, potency and manufacturing — to be effective,” he adds.

Potency will depend also on the CAR that is used, and which B cell marker it is aimed at. All the usual haematological targets — CD19, CD20, CD22 and BCMA — are contenders.

Beyond the cell-seeking warhead, researchers have spent over a decade optimizing CARs, experimenting with how different domains in the body and tail of the CAR impact its cell-killing activity. For Interius, the properties of the CARs that are used in ex vivo products are likely to be well suited for in vivo lentiviral CAR Ts too. They have as a result embraced a tried and true construct, with a CD8 hinge domain, a CD8 transmembrane domain, a 4-1BB co-stim signalling domain and a CD3ζ signalling tail. “It's a very well established, accepted, efficacious CAR construct,” says Johnson.

Umoja similarly chose validated CARs for its lead programmes. But it has added an extra arrow to the lentiviral quiver, transducing targeted T cells with a second protein called rapamycin activated cytokine receptor (RACR). The small molecule rapamycin turns on this artificial RACR receptor, which in turn delivers IL-15 to reprogrammed cells. “We've demonstrated now very nicely that this not only prolongs the life of those T cells, but that it also generates more engineered CARs in the body in the mice,” says Fontana.

Preclinical data show that these candidates are on track. Interius has reported that a single infusion of its lead product can deplete B cells in both mice and macaques. Umoja has similarly shown that its reprogrammed cells, without RACR, drive B cell depletion in non-human primates.

“It's very compelling data,” says Evgin, of both data sets. “They've demonstrated that they can generate these cells in vivo with a relatively high efficiency, and that those cells are extremely functional and able to deplete B cells in mouse models and in non-human primate models.” Lentiviral reprogrammers don't need to reach extremely high levels of in vivo transduction, she adds, because retrained T cells are subsequently bombarded by antigen stimulation — driving their expansion.

Buchholz is also cautiously optimistic. “It’s good news that reprogramming works in a large animal model. That makes it promising. But of course, non-human primates are not diseased patients,” he says.

Cancer calculus

To get human data, both companies are now enrolling patients into phase I cancer trials.

They face crowded blood cancer landscapes. Not only are ex vivo CAR Ts highly effective, but oncologists have little appetite to repeat the same treatment strategy in patients who do not respond well to a first CAR T. “Our advisors — and these are practicing physicians — said to us we will not retreat patients that have relapsed on a CD19 CAR T with another CAR19 CAR T product,” says Johnson.

His team consequently has aimed its lead programme at CD20, another canonical marker of both B cells and of B cell cancers. The trial is currently recruiting patients.

Umoja is more optimistic about the path for a CD19-targeted in vivo CAR T, advancing a lead product for both CAR-T-naive and CAR-T-exposed patients with blood cancers. But to get around the retreatment issue, it is also advancing a CD22-targeted in vivo CAR T.

Both of Umoja’s lead reprogramming cassettes also carry the RACR gene, providing the opportunity to test the effect of these therapies with and without the rapamycin boost. “For patients who have a partial response, we have a couple options: we can re-dose them with the reprogramming vector, or we can give them rapamycin, stimulating those cells to kill tumours. It just gives us another lever that our competitors don't have,” says Fontana.

With the phase I trials geared towards safety, researchers are watching for a few potential hitches.

For Buchholz, a key issue is the cell-targeting selectivity. “The big question in the clinic will be, is the selectivity of these vectors retained?” he says.

Drug developers and regulators will have to keep a close eye on off-tissue reprogramming. During a clinical trial of Novartis’s now-approved ex vivo CAR T therapy tisagenlecleucel (Kymriah), a manufacturing mishap resulted in a CAR inadvertently being transduced into a leukaemic B cell. This generated a CAR-T-resistant leukaemic clone that led to the patient’s death, a worst-case outcome for off-tissue reprogramming.

On-tissue reprogramming might cause cancer too. Last November the FDA started investigating whether ex vivo CAR Ts raise the risk of cancer, spurred on by the concern that the integration of the CAR payload into the T cell’s genome might in rare cases be a cancerous event. Subsequent data has since helped assuage this concern for ex vivo CAR Ts, but has not yet put it entirely to rest.

“The risks associated with ex vivo therapy and in vivo therapy need to be assessed independently,” adds Evgin.

If in vivo CAR Ts can cause secondary cancers, adds Buchholz, it could take years for these to show up.

More immediately, researchers will be watching for cytokine release syndrome (CRS), a form of systemic immune overreaction, and immune cell-associated neurotoxicity syndrome (ICANS), a set of neurologic side effects that also stem from an overabundance of inflammatory cytokines with CAR T treatment. CRS, especially, is part and parcel of CAR T cell therapy, and is noted in the black box warnings of all FDA-approved ex vivo CAR Ts. But the community has established standardized protocols to keep these risks in check. “There's a lot of clinical experience and many tools that can be used to manage those types of toxicities,” says Evgin.

The immunogenicity profile of in vivo CAR Ts also bears consideration, adds Buchholz. Patients who are receiving ex vivo CAR Ts are lymphodepleted before they are treated, and so their immune systems have a limited ability to respond to the engineered cells and the chimeric proteins they carry. By contrast, in vivo CAR Ts cannot be paired with lymphodepletion because this would deplete the cells that need to be modified. But a fully functional immune system might reject either the vector or the CAR it encodes, he cautions.

“I'm very excited that these trials are coming, but there are concerns and we'll see what happens,” says Buchholz.

There could also be early hints of efficacy. “The levels and the depth of B cell depletion will be important for us,” says Fontana.

Beyond the cancer trials, autoimmune applications are close behind. Hopes here are based on preliminary evidence that B cell killing by ex vivo CAR Ts can induce immune reset and possible cure in lupus. Whereas the high costs, lymphodepletion regimens and side effects of ex vivo CAR T are a burden in the oncology setting, they are even more problematic for patients with autoimmune diseases. Interius has therefore prioritized a CD19-targeted in vivo CAR T to enter the clinic in autoimmune diseases in 2025. Umoja and partner IASO Biotherapeutics aim to start testing a CD22-targeted candidate in vivo CAR T in China in patients with autoimmune diseases next year as well.

“We're pretty excited about the autoimmune space, for us and for the whole in vivo field,” says Fontana.

Transient transformations

Not everyone is sold, however, on lentiviral reprogramming. In 2017, just months before Buchholz showed that lentiviral vectors could drive B cell depletion in mice, another team reported a similar breakthrough with a nanoparticle-based vector. The success of mRNA-based COVID-19 vaccines subsequently raised the profile of lipid nanoparticle systems, and biotech’s including Capstan and Orbital jumped on the in vivo CAR T opportunity.

For Bot at Capstan, a big advantage of mRNA-loaded LNPs is their tunability. These therapeutic candidates only transiently retrain T cells to temporarily express a B-cell-killing CAR, he explains, and so dose adjustments and repeat administrations can be used to maximize efficacy while minimizing long-term safety issues. As an added benefit, LNPs are easier to manufacture than lentiviral vectors.

“Cost wise and complexity wise, it's a lot easier to make an LNP and an mRNA versus an entire virus,” says Evgin, who has collaborated with NanoVation on the use of LNPs to generate in vivo CAR Ts.

LNPs nevertheless also need to be optimized for in vivo reprogramming. The LNP vectors that were used in COVID-19 vaccines were by design reactogenic, to maximize the immune response against the viral antigen encoded by the mRNA payload they carried. But re-dosable reprogramming therapeutics need to fly under the immune system’s radar. “Instead of going for a Teflon particle that is very immunogenic and persists a long time in the body, we went with a snowflake type design, so that unless the particle is taken up in a receptor-mediated fashion very rapidly, it just disintegrates,” says Bot. “That's the ideal profile.”

To target these LNPs to T cells, Capstan conjugated an anti-CD8 antibody to its vectors. The goal is to leave CD4+ T cells untouched, he adds. Whereas CD8+ T cells pack the cell-killing blow, CD4+ helper T cells typically intensify inflammatory activity and CD4+ regulatory T cells dampen it. Reprogramming CD4+ cells could as a result complicate the therapeutic effect, says Bot, and could perhaps also exacerbate the CRS associated with CAR T. “We want to stay away from those. We want to avoid unnecessarily co-opting CD4+ T cells,” says Bot.

Another focus for Capstan was on overhauling the body and tails of the CAR construct. The CARs used in ex vivo and lentiviral in vivo applications are optimized for the long-term persistence of the engineered cells, he explains, but transient reprogrammers need instead a rapid effect. “Developing CARs for this purpose was a key effort for us,” says Bot.

The company expects to provide more details on the make-up of its lead candidate and its activity in preclinical models later this year. “This will be a strong springboard for us to move this programme into clinic,” says Bot.

Capstan will start in autoimmune diseases. If an immune reset can cure lupus, the longer-lasting effect of the ex vivo and lentiviral CAR products may not be needed. And because the mRNA-based products don’t integrate into the genome, they should also be less likely to cause cancer. “From a safety perspective, LNPs are a lot more comforting,” says Evgin.

But there could be tradeoffs. In blood cancers, ex vivo CAR Ts can persist for years after they are infused, and these long-lived cells may play a role in the durability of the anti-cancer response. Will transient CAR Ts have the same staying power? “It'll be really interesting to compare the lentivirus approach versus an mRNA approach in both cancer and autoimmune indications, to try to identify which scenarios we need to have very long-term persistence of the CAR versus a short burst of CAR expression that can mediate therapeutic activity,” says Evgin.

Whereas much of the earliest ex vivo CAR T work was advanced by academic groups, this time biotechs are leading the charge into the clinic, she adds. The researcher community is consequently eager for more public data to figure out how and where to best advance these technologies.

Beyond T cells

Ultimately, researchers working in this space hope that both lentiviral and LNP vectors will be useful for reprogramming more than just T cells. LNPs, especially, are already well suited to reprogramme myeloid cells, which take up these vectors without any cell-targeting moieties at all. Even when LNPs are studded with T-cell targeting antibodies, a good portion of the vector often ends up in myeloid cells, says CEO of Myeloid Therapeutics Daniel Getts. “We can target monocytes, macrophages, dendritic cells and even neutrophils. We can bring the full gamut of those cells into play,” he adds.

Orna has consequently embraced a ‘panCAR’ tactic. Their lead LNPs — without any cell targeting moieties — efficiently reprogramme T cells, NK cells and macrophages in non-human primates, says the company. The firm has prioritized these for the full range of liquid cancer and autoimmunity indications.

For Getts, this misses the bigger limitation of the CAR T modality: lack of activity in solid tumours. This shortfall may in part be because researchers have struggled to find cancer-associated antigens without off-tissue liabilities. But Getts suspects that the real problem is the poor penetration and low survival rate of T cells in the tumour microenvironment. CAR myeloid cells, he says, are the answer. Not only are myeloid cells drawn by chemokines directly into the body of the tumour, but they can live there readily — phagocytosing tumour cells, activating inflammatory pathways and recruiting and training other immune cells in the tumour.

Myeloid Therapeutics already has two in vivo CAR products in the clinic, taking aim at solid tumour cells expressing either TROP2 or GPC3. Carisma and its partner Moderna have also prioritized an LNP-based GPC3 CAR macrophage reprogramming cassette, for liver cancer.

“Without a doubt there's big biological risk here,” says Carisma CSO Michael Klichinsky. Researchers have yet to prove either the efficacy or the approvability of CAR macrophages. But solid tumours account for 90% of the cancer burden, making the possibility of activity in these settings with an in vivo reprogramming agent all the more appealing. “We’re going after a real unmet need,” he adds.