Lymphoid organoids as the safety case for edited B cell therapy

What the work claims

The PlanB2CureHIV program project, led by James Voss at Scripps Research and funded as P01 HL183482 (about $3.48 million for the consortium in the current award year, running 2025-09-05 to 2030-06-30), proposes that a functional cure for HIV can be built by inserting an antibody transgene cassette into the native immunoglobulin heavy-chain locus (IgH) of a patient's B cells, so the cells express a broadly neutralizing antibody as their own antigen receptor, expand under vaccination, and settle into long-lived plasma cell and memory states that secrete the antibody for years.¹

Two delivery routes are being developed in parallel: an ex vivo product, where the patient's B cells are edited, expanded, and returned before vaccination; and an in vivo system, where the editing machinery is administered directly and reprograms B cells inside the body.¹ The organoid component, Project 4 led by Lisa Wagar, claims a supporting but load-bearing role: immune organoids derived from murine, non-human primate, and human lymphoid tissue, which recapitulate aspects of germinal center reactions, can be used to test the edited cells, rank immunogen pairings, and establish the predictive value of the organoid model across species before anyone is dosed.²

This is an active grant, not a results paper. The claims are programmatic; the evidence inside the abstracts is preliminary, and one preliminary result is a genuine failure, which makes the record more rather than less credible.

How it works

Broadly neutralizing antibodies against HIV are rare because they demand unusual B cell precursors and extensive affinity maturation, which is why no vaccine has elicited them. The consortium's workaround is to skip recruitment of rare precursors entirely: take ordinary peripheral B cells and rewrite them. The cassette is inserted into the IgH locus downstream of the endogenous variable-region exons, so the engineered antibody is expressed using the cell's own regulatory elements and constant-region genes. In wild-type immunocompetent mice, reprogrammed cells have been expanded and differentiated into memory and long-lived plasma cells by vaccination, with boostable VRC01-class responses reported by the program.¹

Safety engineering shapes the editing chemistry. Because double-strand-break editing is a poor fit for a therapy that edits immune cells in vivo, one project pursues double-strand-break-free insertion: prime-editing-assisted insertion genome editing (PASSIGE), CRISPR-free recombineering, and retron-mediated donor amplification packaged as an all-RNA system.¹

The organoid layer exists because edited cells are useless if they cannot be guided into a germinal center reaction. Wagar's project cultures lymphoid-tissue organoids from three species, introduces the engineered B cells, and reads out whether they enter germinal-center-like reactions and mature toward durable memory phenotypes. Human validation uses tissue from the Last Gift cohort, people living with HIV on antiretroviral therapy with terminal illnesses who donate their bodies to research; the organoids are meant to discriminate preventive from therapeutic versions of the cell product.² The mouse and non-human primate work runs with Tabby Therapeutics, an industry partner specializing in B cell engineering.²

Where a skeptic should push

The most load-bearing assumption is that an organoid's germinal-center-like reaction predicts what edited B cells will do in a human lymph node for the next twenty years. Nothing in the record demonstrates that predictive validity; the grant explicitly proposes to establish it. Organoid germinal centers reproduce some anatomy and kinetics, but they lack the vascular, neural, and trafficking context of a living node, and durability of antibody titers is precisely the property that only long time horizons reveal.

The record itself supplies the stress test. The macaque preliminary data are the strongest caution: ex vivo edited macaque B cells were infused back into autologous donors and produced the VRC01 antibody, but titers were short-lived and further vaccination did not raise them, a failure to form durable memory in the species closest to humans.¹ Mouse memory, which works, is doing a lot of rhetorical work here; Project 3 exists because the mouse result did not transfer. The skeptic should also note that every dollar figure and timeline in this article comes from grant abstracts, that no organoid readout data are yet reported, and that the ambitious all-RNA in vivo editing system is a plan, not a demonstrated delivery route.

What is genuinely demonstrated: mouse reprogramming by independent groups, the macaque failure mode, and a governance structure unusual enough to be worth naming. The administrative core has scheduled an INTERACT meeting, a formal pre-investigational consultation, with the FDA at the end of year two, and commits in advance to folding the agency's direction into the research plan for years three through five.²

When the organoid carries the safety case

For platform access and governance of computing and engineering on living tissue, this grant is a preview of how living-model systems will acquire authority. The organoid here is not a research curiosity; it is being positioned as the evidence layer that a regulator will consult before allowing a therapy that permanently rewrites part of a person's adaptive immune system. If the Wagar organoids validate, the platform that produced the validation inherits de facto gatekeeping power: a vendor or core facility whose lymphoid organoid protocol is the one the FDA has seen becomes the price of entry for every competing therapy in the class. We have seen this shape before in organoid reference standards; here it is being negotiated with the regulator in advance, which is more honest and more durable than governance by default.

The consent architecture deserves equal attention. The human validation tissue comes from the Last Gift cohort, terminal donors who consented to research in the broadest sense. Their lymphoid tissue will be used to test a therapy they will never receive, in a platform whose results may set the safety bar for in vivo immune editing for everyone else. That is a defensible use, but it concentrates interpretive authority over terminal-donor material in a consortium that includes a commercial partner, Tabby Therapeutics, whose specialization is exactly the manufacturing step being validated.² The threat is not misconduct; it is that provenance, validation, and commercialization become the same pipeline, with no independent organoid benchmark standing beside the regulatory one.

The opportunity is a template. Break the therapy into an editing step, a tissue-model validation step, and a regulatory checkpoint designed in from year two, and you get a working model for how closed-loop systems trained on living neural tissue should be governed: not by post-hoc review of a finished artifact, but by a pre-committed evidence ladder that the platform must climb in public. The ethics of computing on living tissue will be set by whoever owns the rungs.

The bottom line

As established results: B cells can be reprogrammed at the IgH locus to secrete a chosen antibody in mice, and in macaques the first attempt failed to produce durable titers, which is the correct null to build on. As hypothesis: lymphoid organoids from human donor tissue can predict which edited-cell products will mature durably in people, and double-strand-break-free editing can be delivered in vivo safely. What would confirm the claim is cross-species organoid data that retrospectively predicts the macaque failure and prospectively flags a product that then succeeds in primates; what would break it is another macaque-scale failure that the organoids blessed in advance. Either way, the more consequential output may be procedural: a regulatory consultation scheduled before the results exist, with an organoid platform standing in the middle. That is a governance design worth watching closely, and worth copying before someone copies it first.