The history of cancer immunotherapy is a history of successive partial solutions. Each decade's advance addressed the principal failure of the one before — and exposed the next constraint.
The pattern is not a criticism of the science. It is what progress in complex biology looks like: each generation of tools does something the prior generation could not, reveals what was previously invisible, and creates new problems worth solving. The current generation of T-cell engagers is no exception.
The early idea
In the 1890s, surgeon William Coley observed tumour regression in patients who developed post-surgical bacterial infections. He inferred that immune activation — triggered non-specifically by microbial products — could suppress tumour growth, and began administering bacterial preparations directly to patients. The data were inconsistent and the mechanism opaque by today's standards. But the core observation has survived every subsequent era: the immune system can, under the right conditions, recognise and kill cancer cells. Every major immunotherapy advance since has been, in some sense, an attempt to make that observation reproducible and controllable.
The monoclonal era
Hybridoma technology, developed by Köhler and Milstein in 1975, gave the field its first precision tool: antibodies manufacturable at scale, targeting defined antigens with specified specificity. By the late 1990s and 2000s, this had translated into the first therapeutic antibodies with meaningful clinical impact — rituximab in B-cell lymphoma, trastuzumab in HER2-positive breast cancer. These were genuine advances. They also defined the limits of the approach.
Therapeutic antibodies direct innate immune mechanisms — complement-dependent cytotoxicity and antibody-dependent cellular cytotoxicity. These are potent but not adaptive. Tumours under sustained antibody pressure acquire resistance through antigen downregulation, target internalisation, or complement-evasion. The mechanism was precise. The durability was not.
The checkpoint revolution
In 2011, ipilimumab — an anti-CTLA-4 antibody — became the first therapy to demonstrate durable benefit in metastatic melanoma by releasing an inhibitory signal on T cells rather than directing antibody-mediated killing. Pembrolizumab established the PD-1 axis as the dominant checkpoint target in solid tumours, with approvals across multiple histologies following from 2014 onward.
Checkpoint inhibitors did not direct T cells at tumour cells. They removed the constraints that prevented pre-existing T cells from acting. In patients with checkpoint-sensitive tumours — high mutational burden, existing immune infiltration — the responses were, in some cases, exceptional and sustained. The limitation became clear quickly: checkpoint inhibition only works where tumour-reactive T cells are already present and only failing because of inhibitory signalling. In tumours that have excluded T cells entirely — through stromal barriers, immunosuppressive microenvironments, or low mutational burden — releasing the brake produces no response. A substantial fraction of solid tumour oncology remains outside the reach of checkpoint monotherapy.
Cell therapy
CAR-T cell therapy addressed the exclusion problem through a different mechanism. Rather than releasing a brake, CAR-T engineering gives T cells new receptor instructions at the DNA level — directing them at specific tumour surface antigens regardless of their prior activation history. Following the first FDA approvals in 2017, CAR-T demonstrated remarkable depth of response in haematological malignancies: complete responses in patients with no remaining standard-of-care options in B-cell ALL and certain lymphomas.
The structural constraints are real. Patient-specific manufacturing requires leukapheresis, ex vivo engineering, and weeks of production time — a window some patients cannot wait through. And in antigen-heterogeneous diseases, escape under CAR-T pressure follows the same evolutionary logic as escape from antibody therapy: tumour cells that lose or reduce target expression are not killed, and they repopulate.
T-cell engagers and the off-the-shelf era
Bispecific T-cell engagers — molecules that simultaneously bind CD3 on T cells and a surface antigen on tumour cells — removed the manufacturing constraint. The molecule bridges T cell and tumour cell directly, forming a cytolytic synapse from the patient's existing T-cell repertoire. No leukapheresis. Off-the-shelf availability. Monthly outpatient dosing. Blinatumomab established the principle in B-cell ALL from 2014. More recently, bispecifics and trispecifics have demonstrated meaningful clinical activity in multiple myeloma, with FDA approvals following.
The format advance is real. The underlying architectural vulnerability is not resolved. A T-cell engager targeting a single antigen imposes the same selective pressure as a CAR-T or a therapeutic antibody targeting that antigen: tumour cells that downregulate the target survive. The delivery mechanism changed. The biology of escape did not.
What comes next
The field's next structural challenge is not finding better single targets — the validated antigen landscape in haematological malignancies is well-characterised. The challenge is designing molecules for which escape through any single mutational event is insufficient. That requires multispecific architectures with independent killing mechanisms at each arm, selected in functional assays that measure the readouts that determine clinical performance: T-cell activation, killing kinetics, cytokine behaviour, and activity across the antigen-expression variance present in real patient tumours.
Each era of immunotherapy solved the problem the prior era revealed. Coley showed the immune system could be engaged. Monoclonals showed it could be targeted. Checkpoints showed it could be released. CAR-T showed it could be reprogrammed. T-cell engagers showed it could be redirected without manufacturing. The problem each of those solutions now shares is the same: tumour evolution. Designing molecules that stay ahead of that evolution is the problem the current generation of multispecific platforms is built to address.