Allogeneic Cell Therapy Is Still Learning What Biology Will Allow

By Arnaud Deladeriere, Ph.D., Cell&Gene Consulting Inc.

Three factors account for allogeneic cell therapies’ quiet shift toward the edge of the industry's consciousness. First, the modality absorbed nearly a decade of difficult clinical results. Second, the conversation turned toward in vivo CAR as the newest entrant. And third, industry's habit of comparing ex vivo to in vivo almost always reduces to in vivo against autologous, which leaves allogeneic outside the frame.

By most estimates, 70% to 80% of eligible patients do not receive an autologous CAR-T.

The ninth Cell&Gene Foundry roundtable revisited allogeneic therapy across T cell, iPSC-derived, and mesenchymal approaches, with a deliberately blunt question: why have so few allogeneic programs reached late-stage development when the promise has been clear for years? The answer that emerged was that the question itself may be outdated and that the path to late-stage success looks different from what the field expected 10 years ago.

New Data Highlights Incremental Design Choices, Not Breakthroughs

The week of the roundtable, Caribou Biosciences reported updated clinical results that changed where the conversation could reasonably begin. In second-line non-Hodgkin lymphoma, its lead program, a CD19-targeted CAR T cell therapy, showed an 82% overall response rate, a 67% complete response rate, and a median progression-free survival of 17.1 months. Those figures sit alongside the approved autologous products generated from a smaller patient number. The internal characterization was that the program is beginning to approach outcomes seen with autologous CAR-T and that reaching this point required several pivots rather than a single linear path.

Caribou's myeloma program, an allogeneic anti-BCMA CAR T cell, followed a similar trajectory. At the 450 million cell dose, it reported a 92% overall response rate, an 83% complete response rate, and 91% MRD negativity, with roughly half of that cohort still in complete response and early indications of durability. The program is also treating patients who have failed prior BCMA-directed therapy, including patients who have failed an autologous CAR-T. That begins to address a long-standing question for the field: whether an off-the-shelf product can recover patients that a personalized product could not.

The discussion attributed these results to a series of design choices rather than any single breakthrough. Much of the roundtable focused on what those choices were and what the field has learned from them.

When Donor Sophistication And Engineering Peak, Manufacturing Steps In

The clearest lesson concerns donor strategy. The lymphoma cohort that generated the strongest durability data used a matched approach: a young donor, a two-allele minimum match, and an age threshold that has narrowed over time and now sits under 30. The principle is not new. The approved allogeneic T-cell therapy from Atara also relies on donor matching, and the underlying logic traces back to decades of transplant practice. The lesson is that the early allogeneic thesis that sufficient gene editing would allow a product to match no one, was too optimistic. Some of the biology has to be respected rather than removed.

Matching is one lever rather than the whole answer. The myeloma program uses a different engineering route, a beta-2-microglobulin knockout paired with a beta-2-microglobulin-HLA-E fusion protein, which reduces the dependence on matching and supports a best-lot strategy with no fixed threshold. Editing and matching act on the same outcome, and the balance between them is modality dependent.

The more uncomfortable observation, raised across the table, is how little the sophisticated donor work has returned. Teams have run single-cell RNA sequencing on donor material before and after manufacturing and interrogated dozens of variables, only to land repeatedly on age and, to a lesser extent, body mass index as the screens that hold. Age is an imperfect correlate, with capable older donors and poor young ones, but it remains the most reliable filter identified to date. The more useful finding is that the manufacturing process appears to influence the final product more than the donor does. Media composition and cytokine selection can shift an older donor profile toward a younger one by the end of the process, which returns a meaningful degree of control to process design rather than donor screening.

This is also where data tooling starts to become genuinely useful. The variables that matter may sit among those that teams have not historically collected, often out of caution. The roundtable noted that correlative repositories, such as the donor-to-outcome data set assembled at the University of Pennsylvania, point toward a problem that can be approached with more variables than a single team can hold at once. This echoes a recurring Foundry theme: meaningful biological signals often emerge only when multiple variables are considered together rather than in isolation.

The group also highlighted the opposite risk: building a product around a single exceptional donor. That may be an attractive story, but it creates both scientific and commercial fragility. Challenging a process against a range of donors and still reaching the target profile is the more defensible position, even when it is the less appealing story to present to investors.

Allogeneic Retains Control

The strongest strategic argument for allogeneic remains control of the final product. This continues a theme the Foundry examined in its discussion of in vivo CAR-T, where the central observation was that control is redistributed when engineering moves from the manufacturing suite into the patient. In an ex vivo allogeneic setting, the drug product is the cell, characterized and released before administration. In an in vivo approach, the administered material is a vector, and the active substance is whatever modified cell results inside a patient whose own T cells have been shaped by multiple lines of prior treatment. The complexity does not disappear. It shifts into the patient's biology, where it is harder to characterize and impossible to release-test.

T-cell fitness is the concrete expression of this. Autologous, in vivo, and bispecific approaches all depend on the patient's T cells being fit enough to perform, and in heavily pretreated patients, they often are not. A donor-derived process can specify that fitness in advance, using young healthy starting material and a process tuned for a consistent product, including attributes such as the CD4-to-CD8 ratio that correlate with both efficacy and toxicity. How an in vivo approach controls that ratio or manages the higher rate of cytokine release that tends to accompany a CD4-skewed product remains unclear. None of this resolves the in vivo question, which clinical data over the next several years will address. It does explain why control of the drug product is the position allogeneic developers should continue to hold.

Autoimmune Targets Could Answer The Lymphodepletion Question

The indication where the allogeneic argument is cleanest is autoimmune disease. The eligible population is far larger than in oncology, and the apheresis and manufacturing capacity required to treat it with an autologous approach does not exist and is unlikely to be built at that scale. This is the supply constraint the Foundry has identified as the real limit on patient access, and it makes an off-the-shelf product close to a precondition for reaching the autoimmune population.

The open question is lymphodepletion. It is essential in oncology, both to create marrow space for expansion and to clear cells that act against the therapy. Whether it is required in autoimmune disease is genuinely unknown, and the early in vivo data suggesting that it may not be is among the more consequential developments for the field. If an immune reset can be achieved with limited or no lymphodepletion, both the patient experience and the access calculation change. Several programs, including one at Kyverna, are running the comparison with and without lymphodepletion that should eventually resolve it.

If the biology and manufacturing are now good enough, autoimmune is where the access problem makes the modality strategically unavoidable.

Key Takeaways

About The Author:

Arnaud Deladeriere, Ph.D., is principal consultant at Cell&Gene Consulting Inc. Previously, he was head of MSAT and manufacturing at Triumvira Immunologics, and before that, manufacturing manager at C3i. He received his Ph.D. in biochemistry from the University of Cambridge.