Somatic mutations can generate neoantigens that are exclusive to cancer cells and absent in normal tissues, making them attractive targets for cancer immunotherapy [[1], [2], [3], [4]]. While most somatic mutations are random passenger mutations unique to individual patients, a subset recurs at critical amino acid positions involved in tumorigenesis, suggesting their potential for broad therapeutic applicability across cancer types and patient populations [[5], [6], [7]]. Recently, adoptive T cell transfer using genetically engineered T cells expressing T cell receptors (TCRs) specific for neoantigens derived from driver gene mutations has shown clinical promise in targeted cancer immunotherapy [8,9]. However, despite the clinical significance of adoptive transfer TCR-T cell therapy, its application faces inherent challenges, primarily stemming from the low functional avidity of naturally occurring TCRs and the risk of cross-reactivity with off-target peptides expressed on healthy tissues [[10], [11], [12], [13], [14], [15], [16], [17]].

The subtlety of a single amino acid alteration within a neoantigen peptide often results in weak immunogenicity, given the extreme scarcity of naturally occurring or induced high-affinity neoantigen-specific T cells [3,18]. To address this, the in vitro engineering of high-affinity TCRs using phage, yeast, or mammalian display systems has attracted considerable interest. Phage and yeast TCR display systems offer a notable advantage in the generation of larger TCR libraries, which may facilitate the identification of TCRs with exceptionally high-affinity. However, both platforms always require the TCRs to be reformatted into single-chain fragments (scTCRs) for effective display. Due to the absence of mammalian protein-folding systems, only a fraction of the scTCRs are able to fold properly, leading to complicated modifications to increase the folding efficiency [[19], [20], [21], [22]].

Despite clinical studies suggesting that therapeutic benefits may be achieved when utilizing higher affinity TCRs in immunotherapy, those with excessive high-affinity carry unpredictable cross-reactivity risks, potentially leading to severe toxicity [[23], [24], [25], [26]]. In two notable examples, high-affinity TCRs targeting MAGE-A3 and MAGE-A12 caused fatal neurological toxicity and cardiovascular toxicity due to off-target recognition of self-peptides [25,26]. In addition, TCRs engineered with supraphysiological affinity (<1 μM) may paradoxically impair T cell function by inducing T cell dysfunction/exhaustion [[27], [28], [29]]. Therefore, it is crucial to achieve an optimal balance between TCR affinity, functional avidity, and specificity for the development of safe and effective TCR-T therapies.

The functional avidity of TCRs—defined as their capacity in responding to antigen presented on cells—depends not only on binding affinity but also on dynamic molecular interactions, TCR clustering, and TCR-CD3 signaling [12,[30], [31], [32]]. This complexity highlights the limitations of traditional binding-affinity based screening methods and underscores the importance of functional screening platforms that recapitulate physiological T cell activation. T cell line display systems, which enable the proper folding and surface expression of TCRs along with their complete activation signals, have emerged as a promising strategy for identifying TCRs with enhanced functional avidity [[33], [34], [35], [36], [37]]. By directly screening for T cell activation, these systems bypass the assessment of binding affinity and facilitate the identification of TCRs with optimal therapeutic potential.

In this study, we developed a TCR engineering platform to enhance the functional avidity of an HLA-A*11:01-restricted KRASG12D-specific TCR KT5 [38] while avoiding reactivity to the wild-type (WT) peptide (Fig. 1). We constructed two libraries, one targeting the CDR3α region and the other the CDR3β region [8,[39], [40], [41]], and screened them using an iterative selection strategy that included positive selection with KRASG12D-pulsed antigen-presenting cells (APCs) and negative selection with KRASWT-pulsed APCs. From these libraries, we identified six TCR variants with enhanced functional avidity for the KRASG12D peptide and with no reactivity for the KRASWT peptide. In particular, CDR3α variants exhibited modest improvements in both functional avidity and specificity, while CDR3β variants showed significant avidity enhancement and potent cytotoxic activity, albeit with variable cross-reactivity profiles. Collectively, this work presents a precision engineering approach to the generation of neoantigen-specific TCRs that balance therapeutic efficacy and safety.