Conventional chimaeric antigen receptors, or CARs, are modular molecules. An extracellular antigen-binding domain, typically a single-chain variable fragment derived from an antibody, is stitched to a hinge, a transmembrane segment and one or more intracellular signalling domains, most famously the CD3-zeta chain paired with costimulatory modules such as CD28 or 4-1BB. Each module has been drawn from natural immune proteins, and the interfaces between them have largely been inherited rather than designed. The new work argues that this inheritance is precisely where much of the variability in clinical performance originates, because the sequence details at these junctions govern how the receptor folds, traffics, clusters and transmits signals once it encounters its target antigen.

The researchers approached the problem by generating large libraries of receptor variants and interrogating them at scale, linking sequence differences to measurable functional outcomes such as antigen binding, surface expression, T-cell activation, cytokine release and tumour-cell killing. This high-throughput strategy allowed the team to move beyond anecdotal comparisons of a handful of clinically used constructs and instead build a comprehensive map of the sequence-function landscape. In such a map, positions that tolerate substitution appear as flat terrain, while positions where a single residue change dramatically alters signalling or expression stand out as peaks and valleys of functional importance.

A central insight from the analysis concerns the hinge and transmembrane regions, segments that have often been treated as inert spacers in receptor design. The data indicate that these regions are anything but passive. Specific sequence motifs influence the propensity of receptors to dimerise or oligomerise in the membrane before antigen engagement, a property with direct consequences for tonic signalling, a phenomenon in which CAR T-cells become chronically activated in the absence of tumour and progressively exhaust themselves. Receptors that maintain an appropriate quiescent resting state, while still clustering productively upon antigen binding, sustained function far better in repeated-stimulation assays, a laboratory proxy for the serial killing that CAR T-cells must perform in patients.

Structural modelling added a second layer of interpretation. Using computational predictions of receptor geometry, the team examined how the length and rigidity of the extracellular spacer determine whether the antigen-binding domain can physically reach its epitope on the target cell, and how the relative orientation of the binding domain and the membrane affects the mechanical force transmitted through the receptor when a T-cell engages a tumour cell. These considerations matter because T-cell activation is not purely chemical; it is also a mechanical event, and receptors that present the binding domain at an unfavourable angle or distance can fail to trigger despite binding antigen effectively. The study connects these geometric parameters to specific sequence choices, giving designers a vocabulary for tuning reach and orientation deliberately.

The intracellular signalling architecture received equally close scrutiny. Rather than simply comparing CD28-based with 4-1BB-based receptors, the analysis resolved the contribution of individual motifs within these domains, including the number and spacing of phosphorylatable tyrosine residues and the docking sites they create for downstream adaptor proteins. Subtle changes in the order or copy number of these motifs shifted the balance of signalling outputs, altering the kinetics of calcium flux, the production of interleukin-2 versus interferon-gamma, and the durability of the cytotoxic response. This granularity suggests that signalling domains can be treated as tunable modules in their own right, rather than as fixed cassettes copied wholesale from natural receptors.

Perhaps the most forward-looking element of the work is its embrace of de novo design. Armed with the sequence and structural rules extracted from the library screens, the researchers constructed chimaeric antigen receptors that do not simply recombine natural domains but incorporate engineered elements designed to satisfy the identified determinants. These purpose-built receptors were evaluated against established clinical-grade constructs in head-to-head comparisons, and the results support the central claim of the study: adherence to the derived design rules yields receptors with strong antigen sensitivity, low tonic signalling and robust anti-tumour activity. In other words, the rules are not merely descriptive of existing receptors; they are prescriptive for building new ones.

The implications for the clinic are considerable. CAR T-cell therapy has produced remarkable remissions in leukaemias and lymphomas, but it continues to face obstacles including relapse through antigen loss, severe cytokine-mediated toxicity, poor performance against solid tumours, and the substantial cost of manufacturing patient-specific products. A rational design framework addresses several of these challenges at once. Receptors tuned for lower tonic signalling may produce longer-lived, less exhausted cell products, potentially improving persistence and reducing the need for repeated infusions. Receptors engineered for precise antigen sensitivity could narrow the therapeutic window, attacking tumours aggressively while sparing healthy tissue that expresses the target at low levels. And because the design principles are antigen-agnostic, they could accelerate the development of receptors against the many solid-tumour targets that have resisted conventional constructs.

The study also carries a broader message for the field of cell engineering. As synthetic biology matures, the limiting factor in designing cellular therapies is increasingly the understanding of molecular grammar: how the parts of an engineered receptor communicate through sequence, structure and membrane context. The approach demonstrated here, combining deep mutational scanning-style libraries, quantitative functional readouts and structural prediction, offers a template that extends beyond CARs to other synthetic receptors, including those under development for autoimmune disease, fibrosis and regenerative medicine. Receptors built on these principles could, in principle, be specified computationally, screened rapidly and optimised iteratively, compressing development timelines that currently stretch over years.

Caveats remain, as they do in any early-stage engineering discipline. Laboratory assays of receptor function, however rigorous, imperfectly predict behaviour in the complex immunological environment of a patient, where antigen density, suppressive microenvironments and T-cell fitness all shape outcomes. The design rules derived from one antigen system may require recalibration for another, and clinical validation will ultimately determine whether the gains observed in vitro translate into safer and more effective therapies. Nevertheless, the study marks a conceptual shift: the chimaeric antigen receptor, long an assemblage of borrowed biological parts, is becoming a genuinely engineerable molecule, with its sequence and structural determinants laid open for inspection. For a field that has achieved so much with empirically assembled receptors, the prospect of designing them deliberately is a compelling next chapter.

Subject of Research: Sequence and structural determinants governing the efficacy of de novo designed chimaeric antigen receptors for T-cell therapy

Article Title: Sequence and structural determinants of efficacious de novo chimaeric antigen receptors

Article References: Sequence and structural determinants of efficacious de novo chimaeric antigen receptors. (n.d.). https://doi.org/10.1038/s41551-026-01790-9

Image Credits: AI Generated

DOI: 10.1038/s41551-026-01790-9

Keywords: chimaeric antigen receptor, CAR T-cell therapy, de novo protein design, tonic signalling, immunotherapy, synthetic biology, T-cell activation, antigen binding, transmembrane domains, cancer immunotherapy, receptor engineering, Nature Biomedical Engineering