Immune Engineering

Established Immune engineering is the proposal that the immune system should be reprogrammed rather than suppressed, and its strongest evidence comes from turning a cancer therapy against autoimmune disease. Chimeric antigen receptor T cells — CAR-T — were built to kill B-cell cancers; the discovery of the decade is that deep, temporary elimination of B cells can reset an autoimmune disease that has resisted every drug. Frontier In the Erlangen series led by Georg Schett, a single infusion of CD19 CAR-T produced drug-free remission in every treated patient across lupus, systemic sclerosis and inflammatory myositis, with the earliest patients now years out and no relapse reported through 2025. Frontier That is the most striking result autoimmune therapeutics has produced in a generation, and it rests on a few dozen patients treated without a control arm.

Established The subject splits into a proven capability and four unproven extensions, and the honest account keeps them apart. Proven: autologous CD19 CAR-T works in refractory B-cell autoimmunity, at least short-term. Unproven, in ascending order of speculation: that the remissions are durable at five to ten years; that in-body (in-vivo) engineering can replace the manufacturing plant; that off-the-shelf (allogeneic) cells can match autologous ones; and that engineered regulatory T cells can induce antigen-specific tolerance. Frontier Each extension would change the field’s scale by an order of magnitude, and none is yet demonstrated in the way the core result is.

Established Two hard constraints govern whether any of this reaches patients, and neither is immunological. The first is manufacturing: an autologous product is grown one patient at a time over three to four weeks, and the closest measured analogue — the first approved gene-edited cell therapy — had infused 39 people worldwide roughly 22 months after approval. Established The second is the risk ledger: lymphodepletion chemotherapy, months of infection-prone B-cell aplasia, and an FDA boxed warning on secondary T-cell cancers are a defensible price in a dying cancer patient and a much higher bar in a young person with treatable lupus. Frontier Whether immune engineering becomes a platform or stays a heroic niche depends on collapsing both constraints, not on further proof that the biology works.

Established A note on sourcing. This brief was commissioned in September 2026 from the Institute’s research base. Reading-list entries without links are cited from the bibliographic record rather than re-fetched, and claims are dated no later than early 2026 unless carried by a linked source.

Established The platform was proven in cancer first, and the cancer record sets both the capability and the safety vocabulary. The first CAR-T product, tisagenlecleucel, was approved by the FDA in August 2017 for relapsed paediatric B-cell acute lymphoblastic leukaemia; six autologous CD19 or BCMA products followed. Established The two signature toxicities were named there: cytokine release syndrome, a systemic inflammatory response gradable one to four, and immune effector cell-associated neurotoxicity syndrome, or ICANS. Established List prices ran to several hundred thousand dollars per infusion, and the product is manufactured from the patient’s own apheresed T cells — the fact that later becomes the bottleneck.

Frontier The autoimmune result began with five lupus patients and is the spine of this subject. Andreas Mackensen, Georg Schett and colleagues at Erlangen reported in Nature Medicine in 2022 that five patients with refractory systemic lupus erythematosus received a single infusion of autologous CD19 CAR-T after standard lymphodepletion; all five entered drug-free remission, anti-double-stranded-DNA autoantibodies became undetectable, and B cells returned after roughly three to four months as a naive, re-set repertoire rather than the autoreactive one that had been removed. Frontier Cytokine release syndrome, where it occurred, was grade 1; no high-grade ICANS was reported. Established The mechanistic claim — that deep tissue-level B-cell depletion resets the autoreactive compartment in a way that repeated antibody drugs like rituximab do not — is the central hypothesis of the entire autoimmune programme.

Frontier The follow-up report extended the finding across three diseases and is the strongest single publication in the field. Muller, Schett and colleagues reported in the New England Journal of Medicine in early 2024 on fifteen patients — eight with lupus, three with systemic sclerosis and four with idiopathic inflammatory myopathy — all treated with a single CD19 CAR-T infusion. Frontier Every lupus patient reached DAS-defined remission and stopped all immunosuppressants; the sclerosis and myositis patients showed major disease responses; cytokine release syndrome was low grade, ICANS was minimal, and the deepest toxicities were transient cytopenias and manageable infections. Frontier Median follow-up at publication was on the order of fifteen months, and by 2025 the earliest lupus patients were reported past two to three years in continuing drug-free remission with no relapse — the observation the whole field is now waiting to see either hold or break.

Established The result triggered a wave of company programmes, and their diversity is itself informative. Cabaletta Bio (CABA-201, a fully human CD19 construct), Kyverna Therapeutics (KYV-101), Bristol Myers Squibb and Novartis (rapcabtagene autoleucel) opened autoimmune CAR-T trials across lupus, myositis, systemic sclerosis and myasthenia gravis. Frontier Cartesian Therapeutics took a different route with Descartes-08, an anti-BCMA CAR-T made with messenger RNA rather than a permanent genomic integration, so the cells are transient and no lymphodepletion is required; the mRNA-CAR trades cell durability for a lighter safety profile, and it is the clearest sign the field has not settled on what an immune-engineering product should even be.

Established The safety ledger is the part most often lost in the enthusiasm, and it has three columns. Acute: cytokine release syndrome is milder in autoimmune disease than in cancer, plausibly because the B-cell antigen burden is far lower, but it is not zero and neither is ICANS. Established Subacute: lymphodepletion plus months of B-cell aplasia produce hypogammaglobulinaemia and prolonged cytopenias, and infections in this window are the commonest serious adverse events. Established Long-term: in January 2024 the FDA required a boxed warning on all approved BCMA- and CD19-directed CAR-T products after reports of secondary T-cell malignancies, some carrying the CAR transgene; the absolute rate is low but the insertional-and-clonal mechanism is real. Established The adjacent gene-therapy record makes it concrete: seven of 67 patients given the lentiviral product Skysona developed a haematologic malignancy with vector insertions in MECOM and PRDM16 at 14 to 92 months — a gene-modified cell can become a cancer years after it cures a disease.

Frontier The manufacturing route is being attacked directly, because it is what limits the whole thing to dozens of patients. Kelonia’s KLN-1010 generates CAR-T cells inside the body: in four patients with relapsed or refractory myeloma, all four reached partial response by month 1 and all four were minimal-residual-disease negative, at 13 to 18 days from consent to infusion with no apheresis, no lymphodepletion and no ex-vivo manufacturing. Frontier In autoimmune disease, Capstan Therapeutics began a phase 1 trial of CPTX2309, an in-vivo CAR-T generated by a CD8-targeted lipid nanoparticle carrying mRNA, in June 2025. Speculative If either route holds, the industrial argument that dominates this brief becomes transitional rather than structural, but both are being read from single-digit patient numbers.

Established The off-the-shelf and tolerance routes are older in concept and thinner in evidence. Gene-edited cell therapy predates CRISPR in the clinic: a TALEN-edited allogeneic CD19 CAR-T product was used in infant leukaemia in 2015 and a zinc-finger nuclease administered inside the body in 2019, so allogeneic and in-vivo engineering are not new ideas, only unfinished ones. Frontier Allogeneic (donor-derived) CAR-T removes the manufacturing wait but faces rejection and shorter persistence, both requiring further gene edits to the donor cell. Frontier CAR regulatory T cells aim not to destroy a target but to suppress a specific response: Sangamo’s TX200 for kidney transplant rejection and Quell Therapeutics’ liver-transplant programme are the most advanced, both early-phase with no durable-tolerance readout. Frontier A related design, the chimeric autoantibody receptor (CAAR-T), engineers a cell to kill only the B cells making one pathogenic autoantibody — Cabaletta’s pemphigus programme is the archetype — and it too remains a small clinical effort.

Established The delivery reality is the sobering anchor. The nearest measured throughput figure comes from the first approved gene-edited cell therapy, which reported roughly 165 first cell collections, 39 patients actually infused and 25 authorised treatment centres as of 30 September 2025 — a ceiling set by hospital and manufacturing capacity, not by whether the biology works. Frontier Autoimmune diseases affect a far larger population than the rare cancers CAR-T was built for, so a throughput that is merely slow for a rare cancer would be irrelevant to lupus, which is the arithmetic every in-vivo and allogeneic programme is trying to escape.

Frontier The first open question is durability, and the data cannot yet answer it. The reset hypothesis predicts lasting remission because the returning B cells are naive; the alternative is that autoreactive memory or long-lived plasma cells eventually reconstitute the disease. Speculative Long-lived plasma cells in the bone marrow do not express CD19, so a CD19 CAR-T does not touch them — a mechanistic reason to expect that some antibody-driven diseases will relapse and some will not, and nobody can yet say which.

Frontier The second is generalisation beyond B-cell-mediated disease. Lupus, myositis and systemic sclerosis are diseases in which autoantibodies and B-cell help are central, which is why depleting B cells resets them. Speculative Whether the same logic reaches T-cell-driven diseases — multiple sclerosis, type 1 diabetes — is untested, and a CD19 target is the wrong tool for one. Frontier The honest boundary of the proven result is refractory, B-cell-driven, autoantibody-positive autoimmunity; everything past it is hypothesis.

Frontier The third is whether the manufacturing step can be removed without losing the effect. The specific unknown is whether a transient, lipid-nanoparticle-delivered mRNA CAR can achieve the same depth of tissue B-cell depletion as an expanded, persistent, integrated autologous product — depth being the property the reset hypothesis says matters. Frontier A shallower depletion that is easier to deliver may simply be a better rituximab, which is a real advance and a different claim from a reset.

Speculative The fourth is tolerance: can an engineered cell teach the immune system to stand down against a specific antigen rather than blunt it wholesale? CAR-Tregs and CAAR-T both promise what suppression never delivers — a targeted, durable, antigen-specific correction with the rest of immunity intact. Handwave There is no human demonstration of durable engineered tolerance; the transplant CAR-Treg trials are dosing for safety, and the leap from a phase 1 readout to a tolerance that lets a patient stop all immunosuppression is exactly the step where the argument works by assertion.

Established The binding bottleneck is manufacturing, a capacity problem rather than a science problem. An autologous product requires apheresis, a GMP clean-room slot, vector or nanoparticle supply, a two-to-four-week engineering window, release testing and a transplant-capable centre — every one a scheduled physical resource, and the 39 infusions in 22 months figure is the measured expression of how tight the constraint is. Frontier The two routes out — in-vivo generation and off-the-shelf allogeneic cells — are precisely those that would convert a per-patient factory into a distributable drug, which is why they attract disproportionate investment relative to their evidence.

Established The second bottleneck is lymphodepletion, the conditioning chemotherapy that makes room for the engineered cells. Fludarabine and cyclophosphamide are genotoxic, immunosuppressive, and the proximate cause of much of the infection risk; they are tolerated in oncology because the alternative is death. Frontier In autoimmune disease a non-genotoxic conditioning regimen — antibody-based lymphodepletion, or an in-vivo route that needs none at all — would change the risk-benefit calculus more than any improvement in the CAR itself. Frontier Cartesian’s lymphodepletion-free mRNA-CAR and Capstan’s in-vivo nanoparticle are both, in effect, bets that removing the chemotherapy is the highest-value move available.

Frontier The third bottleneck is targeting specificity for the in-vivo route. Delivering a CAR-encoding payload to the right cell inside the body — a CD8 T cell for Capstan, a broader T-cell pool for lentiviral approaches — without transducing the wrong cell is unsolved, and it is the same targeting problem that governs all in-vivo genetic medicine. Speculative An accidental, persistent CAR in an unintended long-lived lineage is exactly the kind of event the secondary-malignancy warning is about.

Frontier The fourth is allogeneic persistence. An off-the-shelf donor cell is rejected by the host and tends to persist for weeks rather than months, so it may not achieve the depth of depletion the reset requires. Frontier The engineering answer — knocking out the T-cell receptor and HLA class I to hide the cell — adds gene edits, and each edit adds a manufacturing step and a genotoxicity question, so allogeneic products trade one bottleneck for several smaller ones.

Established This subject consumes capabilities that neighbouring fields build for their own reasons. From delivery science it needs the targeted lipid-nanoparticle and viral-vector chemistry that would make in-vivo engineering reliable; that chemistry is the object of study in Nanomedicine, and the in-vivo CAR programmes are that field’s most direct clinical application. Established From genome engineering it needs the multiplex editing that makes an allogeneic cell invisible and safe, and the vector-manufacturing capacity to supply it, both assessed in Genetic Engineering. Frontier Neither field has immune engineering’s needs as its objective, which is the ordinary condition of an applied discipline and worth stating rather than assuming.

Established Two dependencies are shared with the wider cell-and-gene-therapy enterprise. Non-genotoxic conditioning would lift the throughput ceiling and shorten the infection window, and it is a stem-cell-transplant problem rather than an immunology one. Frontier Companion diagnostics — the biomarkers that say which patient has the B-cell-driven disease this works on — are the same regulated device class that gates the products assessed in Precision Medicine, and immune engineering will need them to avoid treating the patients it cannot help.

Speculative The dependency nobody lists is a predictive theory of relapse. There is no validated way to say in advance which autoimmune disease will stay in remission after a reset and which will reconstitute from CD19-negative plasma cells. Handwave Without it the field cannot tell a durable cure from a long remission until years have passed, and it is pricing and consenting patients as though it already knows.

Frontier The decisive test is whether drug-free remission after a single CD19 CAR-T infusion holds at three to five years in a controlled cohort large enough to measure a relapse rate. Everything the field claims rests on durability, which an uncontrolled series of a few dozen patients with fifteen months of follow-up cannot establish. Frontier A randomised comparison against best available therapy, powered for relapse over five years rather than response at one, would either confirm that immune reset is a cure or reveal it as a long remission — and both answers are worth having, because the reimbursement and consent frameworks being built assume the first.

Frontier The second experiment is a head-to-head of in-vivo against autologous depletion depth. The in-vivo route’s promise is that it removes manufacturing without losing effect, and the measurable proxy is tissue B-cell depletion, not peripheral-blood clearance. Speculative A trial measuring lymph-node and bone-marrow depletion after an in-vivo CAR against an autologous one would settle whether the easier product is a genuine reset or merely a better antibody drug; nothing in the current single-digit datasets does this.

Speculative The third is a functional tolerance readout for a CAR-Treg. A transplant CAR-Treg trial that let recipients withdraw all immunosuppression and stay rejection-free for a defined window would be the first human demonstration that engineered tolerance is real rather than a safety-phase hope. Handwave No such withdrawal endpoint has been reported, and until one is, antigen-specific tolerance remains the field’s most attractive claim and its least evidenced.

Frontier The fourth is a registry of secondary malignancies with the CAR-transgene status of each clone. The boxed warning rests on case reports; a systematic denominator would convert an alarming anecdote into an actuarial number a young lupus patient could weigh. Established That registry is a bookkeeping exercise, not a scientific one, and its absence is a choice.

Established The central engineering system is cell manufacturing, and the design goal is to stop doing it one patient at a time. Closed, automated benchtop systems that shorten vein-to-vein time are the incremental route; they widen the bottleneck without removing it. Frontier The step-change routes are in-vivo generation, which removes manufacturing entirely, and allogeneic production, which converts it into a batch-manufactured inventory. Frontier The field runs all three in parallel because none has yet proven it can carry the clinical result the autologous product established.

Frontier Allogeneic engineering is a multiplex gene-editing problem: knock out the T-cell receptor to prevent graft-versus-host disease, disrupt HLA class I to slow rejection, and often add a persistence or safety switch. Frontier Each edit is a manufacturing and safety cost, and their accumulation is why allogeneic persistence stays shorter than autologous — the cell is safer to give and easier to reject.

Frontier In-vivo engineering is a delivery problem, and it is the same one that governs all genetic medicine. A payload — mRNA in a targeted lipid nanoparticle, or a genome-integrating lentivirus — must reach the intended T-cell lineage and no other, at a dose that produces enough engineered cells to matter. Speculative The mRNA route is transient and safer but may not persist for a deep reset; the lentiviral route persists but reintroduces the insertional-oncogenesis question. Handwave No in-vivo platform has yet shown it can hit both specificity and sufficient depth at once in autoimmune disease.

Established The nearest neighbour is genetic engineering, because every engineered cell is a genetic-engineering product wearing an immunology label. The vectors, the multiplex edits for allogeneic cells and the manufacturing constraints assessed in Genetic Engineering apply in full, and the pre-CRISPR clinical history of cell engineering — the 2015 TALEN CD19 case and the 2019 in-body zinc-finger administration — is that field’s history as much as this one’s. Frontier The traffic runs both ways: the safety signal that produced the CAR-T boxed warning is the same insertional-and-clonal hazard that field tracks in every integrating therapy.

Frontier Delivery is the second adjacency, and it is the fork that decides the field’s scale. The targeted lipid-nanoparticle chemistry examined in Nanomedicine is the route by which in-vivo CAR generation could replace the manufacturing plant, and the autoimmune in-vivo programmes are that field’s most consequential live application. Frontier If in-vivo delivery to immune cells becomes reliable, the industrial constraints in this brief become transitional; if it does not, they are the steady state.

Frontier Three further adjacencies are load-bearing. The autologous manufacturing chain and companion-diagnostic architecture assessed in Precision Medicine govern who can be treated and how fast, and the two subjects share the same throughput ceiling. Speculative The transplant-tolerance ambition of CAR-Tregs connects to the graft-survival problem in Lab-Grown Organs, where a tolerance that let a recipient keep an organ without lifelong immunosuppression would be transformative. Speculative And the emerging idea of CAR-T directed at senescent cells links this subject to the ageing-intervention programme in Cellular Rejuvenation, where an engineered cell that clears senescent cells is being explored as a route to the same clearance that small molecules attempt.

Established Three institutional facts, not three scientific ones, decide whether immune engineering reaches patients at scale. The first is payment for a one-time, high-cost cell therapy in a chronic disease: health systems pay recurring costs from recurring budgets, and a single-administration product that replaces a lifetime of drugs inverts that logic. Frontier The oncology CAR-T products, at several hundred thousand dollars each, already strain that architecture for rare cancers; applying it to a common autoimmune disease multiplies the strain by the eligible population.

Established The second is the academic-versus-commercial manufacturing split. The Erlangen result was produced with an academic CD19 construct manufactured on-site, not a licensed commercial product, which is how a university hospital could treat a series at all. Frontier Whether the field scales through decentralised point-of-care manufacturing or centralised commercial products is unresolved, and the two imply completely different cost, regulatory and access outcomes.

Established The third is a regulatory pathway for in-vivo cell engineering, which does not yet exist as a category. An in-body product that turns a patient’s own cells into CAR-T is neither a conventional cell therapy nor a conventional drug, and regulators are assessing the first such trials case by case. Frontier A platform approval — regulating the delivery process and design rules rather than each product — is what an in-vivo field would need, and no agency has established one, which is the same gap the n-of-1 genetic medicines face.

Frontier The defining ethical fact is that immune engineering imports an oncology risk profile into non-fatal disease. Lymphodepletion chemotherapy, months of infection-prone B-cell aplasia, and a boxed warning on secondary T-cell cancers are proportionate in a patient who will otherwise die of leukaemia. Frontier In a 25-year-old whose lupus is controlled, if imperfectly, on immunosuppressants, the same risks demand a far higher standard of evidence for durable benefit — evidence the field does not yet have. Speculative The risk-benefit line obvious in cancer is genuinely contested in autoimmunity, and treating the two as the same problem is the field’s characteristic ethical error.

Established The consent problem is sharpened by irreversibility and unquantified long-term risk. A patient consenting to a CD19 CAR-T for lupus is consenting to a permanent genomic modification of their T cells whose worst-case outcome, a transgene-carrying malignancy, may appear years later at a rate nobody can yet quote. Frontier Consent to an unquantified long-term hazard, in a disease that is not immediately life-threatening, is ethically different from consent in a terminal setting.

Established The distributional question is posed directly by manufacturing. A therapy delivered through a handful of transplant-capable centres, one patient at a time, is in practice available to people who live near them and whose systems will pay. Speculative If the autologous route is the steady state, immune engineering is a therapy for the few; only the in-vivo and allogeneic routes make it a therapy for the many, so the equity question and the manufacturing question are one question.

Speculative The strongest civilizational claim is that autoimmune disease could move from lifelong management to one-time correction. Autoimmune conditions affect a large fraction of the population and consume a correspondingly large share of chronic-care spending; a durable immune reset that generalised across B-cell-driven diseases would be one of the larger shifts in the burden of chronic disease in a century. Frontier The measured result is consistent with that future and nowhere near sufficient to establish it.

Speculative The deeper claim is that immune engineering becomes a general-purpose platform. The same engineered-cell toolkit is being pointed at cancer, autoimmunity, transplant tolerance, chronic infection and, speculatively, the clearance of senescent cells in ageing. Handwave That vision presupposes exactly the two things this brief says are unproven — cheap, distributable manufacturing and durable, specific control — so the platform claim is a coherent extrapolation resting on assertions.

Frontier Declare the tie honestly. For: a single infusion produced drug-free remission in every treated autoimmune patient in the founding series, a result no drug has matched, with the earliest patients still in remission years later. Established Against: the sample is a few dozen, uncontrolled, short-followed, manufacturing-limited, and the risk ledger includes a boxed cancer warning. Frontier The honest verdict is that immune engineering is a real and possibly transformative capability demonstrated at a scale far too small to know whether it is a cure or a spectacular remission, and the next five years of follow-up decide which.

These horizons track results that would change what can honestly be said about engineering the immune system, not predictions of clinical availability.

• 10 yr: Frontier The durability question is answered one way or the other: five-year, controlled follow-up either confirms that CD19 CAR-T produces durable drug-free remission in B-cell autoimmunity or shows a measurable relapse rate. Frontier In-vivo and allogeneic routes report their first efficacy readouts, deciding whether the manufacturing bottleneck is transitional; expect the safety registries to finally attach a denominator to the secondary-malignancy warning.
• 25 yr: Speculative Either in-vivo generation has made engineered-cell therapy a distributable drug rather than a per-patient service, in which case autoimmune disease becomes a candidate for one-time correction at population scale, or autologous manufacturing remains the steady state and the field stays a specialist niche. Speculative CAR-Treg tolerance either produces its first durable immunosuppression-withdrawal result in transplantation or remains a safety-phase promise.
• 50 yr: Speculative If both manufacturing and durable specific control are solved, immune engineering matures into a general platform spanning cancer, autoimmunity, transplantation and chronic infection from one regulatory and manufacturing base. Handwave The alternative endpoint is equally coherent: the platform works brilliantly for a defined set of B-cell diseases and never generalises, and the grander claims are quietly retired.
• 100 / 250+ yr: Handwave At this horizon the interesting question is whether the immune system becomes a programmable, updatable substrate — edited on demand to add, remove or retune responses — rather than a system we occasionally reset. Handwave Every component of that is asserted rather than demonstrated, and it is worth ending on the observation that the field’s hardest present constraint is making one patient’s cells reliably and cheaply, which has been true at every point of its history.

• Depends on This subject waits on results other briefs are producing for their own reasons. It consumes multiplex editing and vector manufacturing from Genetic Engineering, targeted in-vivo delivery chemistry from Nanomedicine, and companion-diagnostic architecture and autologous-manufacturing capacity from Precision Medicine. None of those is a result this subject can produce for itself.
• Requires (not on this map) Five non-scientific constraints bind harder than any unsolved experiment. Decentralised, point-of-care manufacturing is what would let a common disease be treated at all, against a measured ceiling of dozens of patients per programme. A reimbursement mechanism to pay once for a therapy that replaces a lifetime of drugs does not exist at scale. GMP viral-vector and lentivirus supply is a chronic shortage that gates every integrating product. Non-genotoxic conditioning would remove the chemotherapy that drives the infection risk. And a platform regulatory pathway for in-vivo cell engineering — regulating the process, not each product — has no precedent any agency has set.
• Enables What this subject supplies is a route to durable immune modification: transplant tolerance without lifelong immunosuppression, antigen-specific correction of autoimmunity, and potentially targeted clearance of unwanted cell populations. No typed enabling edge is recorded, because that contribution is a capability others build on rather than a discrete technology.
• Adjacent The adjacent subjects share components and failure modes. Genetic Engineering owns the editing and the insertional-oncogenesis hazard. Nanomedicine owns the delivery route that decides the field’s scale. Precision Medicine shares the manufacturing ceiling and the companion-diagnostic gate. Lab-Grown Organs would be transformed by durable engineered transplant tolerance, and Cellular Rejuvenation shares the ambition of clearing senescent cells with an engineered agent.

Frontier “CAR-T cures lupus.” The founding evidence is a few dozen uncontrolled patients with a median follow-up of roughly fifteen months; every one reached drug-free remission, which is remarkable, and none of that establishes a cure. Frontier The correct statement is that a single CD19 CAR-T infusion produced drug-free remission in every patient in a small, uncontrolled refractory-autoimmune series, and whether it is permanent is the open question, not a settled one.

Frontier “It is a one-time cure.” One-time describes the administration, not the outcome. Speculative CD19 CAR-T does not deplete CD19-negative long-lived plasma cells, so some antibody-driven diseases could reconstitute; the durability data cannot yet distinguish a cure from a multi-year remission, and the reimbursement models being built assume the first meaning while the biology has shown only the second.

Frontier “In-vivo CAR-T is already here.” In-body CAR generation has produced deep responses in a handful of myeloma patients and entered autoimmune trials — a genuine demonstration of a manufacturing route. Speculative It is not a demonstration of efficacy in autoimmune disease, the datasets are single-digit, and whether a transient nanoparticle-delivered mRNA CAR achieves the depletion depth the reset requires is exactly what has not been shown.

Established “It is safe because autoimmune patients have low tumour burden.” Cytokine release syndrome is indeed milder in autoimmune disease, a real advantage. Established But the infection risk from lymphodepletion and B-cell aplasia is not lower, and the boxed warning on secondary T-cell malignancy applies to every approved CAR-T product; milder acute toxicity is not low long-term risk, and conflating the two understates the ledger for exactly the young, non-terminal patients for whom it matters most.

Established “Approval means patients get it.” The nearest measured analogue had infused 39 people worldwide roughly 22 months after approval, against a far larger eligible population, because autologous manufacturing runs one patient at a time through a few authorised centres. Frontier For a common autoimmune disease the gap between approval and delivery would be wider still, measured in clean-room slots, vector supply and hospital capacity rather than in science — which is why the in-vivo and allogeneic routes, not any further CAR refinement, decide whether immune engineering is a niche or a platform.