Biotech is a tough business. Within weeks of each other in 2026, two lentiviral in vivo BCMA CAR-T assets posted almost mirror-image results: one drew a response in all 18 patients treated, with adverse events (AEs) largely confined to Grade 1/2; the other saw its trial stopped early after all five patients developed Grade 3 or higher AEs.
This article examines how EsoBiotec/AstraZeneca's results in March 2026 [1] led many of us to question the viability of in vivo CAR-T in oncology, only for Kelonia Therapeutics' ASCO 2026 readout [2] to suggest the modality is alive and well.
ESO-T01 Multiple Myeloma Results
The Chinese Phase 1 (NCT06791681) of EsoBiotec (owned by AstraZeneca) was published in Nature Medicine in March 2026. ESO-T01 is an in vivo BCMA CAR-T, built on an immune-shielded lentiviral vector (MHC-I KO, high CD47, mutant VSVG, anti-TCR nanobody). The asset requires a single IV dose without the need for lymphodepletion.
Despite preliminary signals of efficacy, the trial was stopped early in 2025; given the toxicity profile, safety is the widely assumed driver, though the publication does not state a reason. Specifically, five heavily pre-treated relapsed/refractory (R/R) multiple myeloma patients were dosed, and 4/5 demonstrated an objective response (80% ORR), three of which a stringent complete response (sCR), with four patients being negative for minimal residual disease (MRD) by day 60.
The toxicity profile was unfavorable: All patients had Grade ≥3 AEs; cytokine release syndrome (CRS) 4/5 (three Grade 3), and one patient death on day 19 (attributed to disease /spinal cord compression but occurred amid Grade 3 CRS).
KLN-1010 Multiple Myeloma Results
The Phase 1 inMMyCAR study (NCT07075185) of Kelonia Therapeutics presented updated data at ASCO 2026. KLN-1010 is an in vivo BCMA CAR-T, built on a modified lentiviral vector via Kelonia's in vivo gene placement system (iGPS, envelope and tropism modifications). Similar to ESO-T01, KLN-1010 also requires a single IV dose without the need for lymphodepletion.
The readout was uniformly positive across efficacy and safety. Specifically, 18 R/R multiple myeloma patients were dosed across three dose levels, with a 100% ORR and MRD-negative bone marrow in all evaluable patients at one month; the first patient dosed remained in a deep, ongoing MRD-negative response beyond 10 months, and of six patients with ≥4 months of follow-up, four were in sCR and two in very good partial response (VGPR), all with ongoing MRD negativity.
The toxicity profile was favorable: CRS occurred in 16/18 patients, all Grade 1-2; one Grade 1 and one Grade 3 ICANS (the Grade 3 event limited to 3 days), with cytopenias and Grade 3-4 infections “notably limited”. No infusion-related reactions were observed following dexamethasone premedication, and the safety review committee approved outpatient infusion.
Analysts and BD professionals have been debating whether Lilly paid too much (~$3.25B upfront plus ~$3.75B in milestones) to acquire Kelonia Therapeutics [3]. Having now seen the readout that led to this deal, the price appears much more reasonable.
These results are on par with the early readouts of Abecma and Carvykti, the approved autologous CAR-T cells for R/R multiple myeloma. ORR is easy to achieve in multiple myeloma and is not a key metric. As such, the main remaining question is around durability; Carvykti has demonstrated an exceptional PFS and OS advantage in this setting, which KLN-1010 will have to at least partially match in order to compete.
Abecma (ide-cel) | Carvykti (cilta-cel) | |
Trial Name | CRB-401 (Phase 1) | CARTITUDE-1 (Phase 1b - US/Global) |
First Major Report | NEJM (May 2019) | ASH Meeting (Dec 2019) |
Number of Patients (N) | 33 | 29 |
Median Prior Lines of Tx | 7 | 5 |
EFFICACY | ||
Overall Response Rate (ORR) | 85% | 100% |
Complete Response (CR/sCR) | 45% | 69% |
Median PFS | 11.8 months | Not Reached (at time of report) |
SAFETY | ||
CRS (All Grades) | 76% | 93% |
CRS (Grade 3 or higher) | 6% | 7% |
Neurotoxicity (All Grades) | 42% | 10% |
Neurotoxicity (Grade 3+) | 3% | 3% |
Why Did the Very Similar Asset ESO-T01 Underperform?
The cleanest comparison, ESO-T01 versus KLN-1010, has target, disease, vector class, and administration essentially constant, which points to a product-specific cause rather than a flaw intrinsic to in vivo CAR-T as a modality. A reasonable hypothesis is that ESO-T01's toxicity was vector- and innate-immune-driven rather than CAR-driven: one patient developed Grade 3 CRS within two hours of infusion, far too fast for CAR-T expansion and tumor engagement, and the trial showed a biphasic cytokine pattern, an early innate response to the viral vector followed by a later CAR-mediated phase. The most telling evidence is that the same CAR construct (PRG1801) was previously clinically validated ex vivo in 34 patients, which isolates the systemic lentiviral delivery and immune-shielding as the driver.
Premedication and trial-execution differences likely compounded this. ESO-T01 used steroids and tocilizumab reactively, whereas Kelonia applied dexamethasone prophylaxis (eliminating infusion reactions and possibly holding CRS to Grade 1-2); prophylaxis is well-established against the CAR-driven CRS phase but weakest against the hyperacute innate phase tied to ESO-T01's specific vector, and an investigator-led China study may have run a less aggressive supportive-care protocol, possibly under-premedicating deliberately to demonstrate in vivo CAR-T generation. The halt is best read as a real, product-specific safety signal in this particular shielded-vector design rather than evidence that the class is unsafe, given that two contemporaneous in vivo programs (Kelonia, Legend [4]) posted clean profiles.
How Does this Fit Within the Broader Landscape?
The field is now in the very exciting place of having an asset with the potency and curative potential of the CAR-T technology, but one that does not require the highly toxic lymphodepleting step of autologous CAR-T cells, and one that is likely to be soon administered entirely in the outpatient setting. This will eliminate the need for hospitalization for large numbers of patients, and make the therapy accessible in smaller settings, unlike autologous CAR-Ts which are still restricted to specialized centers.
Importantly, the theoretical concern that heavily pre-treated cancer patients will not have enough healthy T-cells to produce CAR-Ts has now been proven surmountable in the clinic, enabling more allogeneic and in vivo CAR-T approaches to be tested.
Despite this fantastic progress, potential headwinds remain; first, the lentiviral risk of insertional mutagenesis remains, which can lead to secondary cancer risk later on. However, this will be tempered by the lack of chemotherapy-driven lymphodepletion, which has also been shown to be a driver of secondary cancers. Second, in multiple myeloma in particular, BCMA is now a target accessible via multiple modalities. As such, the number of patients in the R/R setting who are not BCMA-exposed is constantly shrinking in developed markets. If in vivo CAR-T assets are to be clinically successful, they will have to be tested in earlier settings with more BCMA-naive patients.
Conclusion
These results put viral, integrating in vivo CAR-Ts much further ahead of their transient (i.e. LNP-mediated) CAR-T competition. The main reason transient assets were created is to avoid the risks of secondary cancers later in life. After the Kelonia and Legend readouts, transient assets will have to circumnavigate the issue of re-dosing in order to demonstrate efficacy on par with the integrating assets.
What does this mean for the poster child of commercially successful CAR-T assets, J&J’s Carvykti? Until KLN-1010 demonstrates substantial durability in large numbers of patients in a later stage trial, Carvykti is safe. However, if that happens, Carvykti will face extreme competition on the grounds of ease of administration and no need for lymphodepletion. Its many years of first-mover advantage, however, continues to reduce the numbers of BCMA-naive patients eligible for KLN-1010.
By Sakis Paliouras, PhD
Sources
