A KRAS-targeted bispecific T cell engager promotes immunity against colorectal solid tumor.

Introduction
The Kirsten rat sarcoma viral oncogene homolog (KRAS) encodes the guanosine triphosphate (GTP)ase KRAS, a key protein involved in the RAS/MAPK pathway. KRAS functions as a GTP/guanosine diphosphate (GDP) binary switch, regulating signal transduction from activated membrane receptors to intracellular molecules, and is involved in differential downstream gene expression.1,2,3,4,5 Mutations in KRAS keep the protein at the “ON” state, constantly bound to GTP, resulting in uncontrollable cell growth. KRAS mutations are detected in approximately 25% of tumors, making them one of the most common mutated oncogenes.6 Importantly, this 25% includes many highly fatal cancers such as colorectal cancer (CRC), non-small cell lung cancer (NSCLC), and pancreatic ductal adenocarcinoma (PDAC), with 75% of amino acid substitutions occurring at the G12 codon position in KRAS.7,8,9,10,11,12 Common mutations of the various KRAS mutants include G12C, G12D, and G12V, with frequencies varying between ethnic groups.7,13,14,15,16 Multiple studies highlight the KRAS G12 mutation as a critical “hotspot” for HLA class I (HLA-I)-restricted T cell epitopes. Several HLA-I-restricted T cell epitopes have been identified, including those associated with HLA-A∗11:01 (G12D, G12V), HLA-A∗03:01 (G12V), HLA-B∗07:02 (G12R), HLA-C∗01:02 (G12V), HLA-C∗08:01 (G12D), and HLA-C∗05:01 (G∗12D).17,18,19,20,21,22,23,24,25,26 KRAS lacks a binding pocket, and in its activated state, it binds tightly to GTP. As a result, blocking KRAS with small molecules has been challenging in the past 40 years despite its high frequency of detection in cancer. Two inhibitors have been developed to treat NSCLC patients with KRASG12C: Sotorasib (AMG150) from Amgen and Adagrasib (MRTX 849) from Mirati.12,27,28,29 Despite this breakthrough, the development of targeted drugs against KRAS mutations still requires considerable effort, as not all KRAS mutants exhibit a pocket structure for small molecule binding, as seen with KRASG12C.28
Recent advances in immunology and cancer research have significantly accelerated the development of new cancer immunotherapies. Based on the mechanism of action, T cell-based cancer immunotherapies can be mainly categorized into two major classes: (1) therapies targeting immunosuppressive factors such as immune checkpoint inhibitors (ICIs), and (2) therapies promoting immunostimulatory pathways, including chimeric antigen receptor (CAR) T cells and bispecific T cell engaging antibodies (bsAbs).30,31,32 ICIs block key immunoresponsive molecules such as programmed cell death-1 (PD-1) and its ligand (PD-L1) to activate cytotoxic T cells.33,34 This approach soon proved effective against different types of cancers, especially solid tumors. Despite early success and great advancement in cancer treatment, the response rates of ICIs remain limited due to immune evasion and/or tumor microenvironment changes.33 On the other hand, CAR-T cells function by expressing a structure composed of an extracellular antigen-recognition domain and intracellular T cell signaling domains. They target tumor-associated antigens (TAAs) to mediate cytotoxicity.35 An advantage of CAR-T cells is the use of patient-originated T cells, which helps reduce the risk of rejection and autoimmune responses. However, the preparation of T cells is time-consuming and complicated.36
In addition to these two powerful approaches, recent years have marked the rise of bsAbs as a prominent tool for fighting cancer. BsAbs focus on redirecting T cells against target cells through the unique recognition of tumor-specific antigens (TSAs) or TAAs on cancer cells and cell surface molecules on T cells. Importantly, TSAs can be presented as peptides on the tumor cell surface in major histocompatibility complex (MHC) proteins, which interact with the T cell receptor (TCR) to stimulate an antitumor response. The TCR exhibits an antibody-like structure and thus, theoretically, can be incorporated into antibodies. Indeed, recent results from the engineering of TCR-mimicking antibodies have proven the viability of this approach,37,38 and have provided greater potential for the development of T cell-redirected bsAbs. This T cell engaging receptor (TCER) class set its position as intermediate between engineered CAR-T cells and bsAbs and emerged as a promising anti-cancer tool. With their unique feature of specific recognition of TSAs, TCERs bridge T cells to cancer cells, thereby enhancing T cell-mediated killing efficiency. Here, we adapted a published TCR structure used in T cell therapy18,22 and engineered TCERs to redirect primary CD3+ T cells to colorectal solid tumor cells carrying KRASG12V TSAs. Using validated cell-based assays, we evaluated their efficiency, specificity, and stability in vitro.

Results

In silico evaluation of published pHLA-A∗11:01/KRASG12V-targeting TCR
This study aimed to engineer a bispecific TCER capable of simultaneously interacting and recruiting cell types of interest, including T cells and target cells expressing specific human leukocyte antigens (HLAs) and presenting peptides with disease-associated mutations. Being involved in various cellular pathways, KRAS mutations are detected in various cancer types. In particular, the G12V mutation appears to be one of the most common mutations in the RAS family, including KRAS, HRAS, and NRAS, which share identical sequences in the first 86 residues at their N terminus.7,13 Although KRAS is recognized as a major contributor to tumor development, its structure makes it challenging to develop adequate small molecules as potential treatments. On the other hand, peptides carrying relevant KRAS mutations can be presented on certain HLAs and recognized by TCRs, making TCERs a promising approach for targeting these tumors. Other research groups have identified TCR structures capable of recognizing and binding to HLA-A∗11:01 displaying KRASG12V peptide.18,22 Lu et al. have identified two TCR structures, while Bear et al. have identified another with high specificity against pHLA-A∗02:01 and pHLA-A∗03:02. Therefore, we first reevaluated these TCRs in silico using publicly available workflows. We then focused on the structure that demonstrated the highest binding efficiency in our evaluation. Such predictions were useful only for prioritization and had to be cross-evaluated for reliability. Using three different bioinformatics tools, including Alphafold 3.0,39 TCR dock,40 and TCR model 2.3,41 we generated and predicted the binding efficiency between TCR candidates and the HLA-A∗11:01/KRASG12V peptide complex. Among the three TCRs tested, we found TCR1−2C TCR3−2E from Lu et al. had the best scoring outcomes (Table 1). In addition, both TCR1−2C and TCR3−2E have been carefully characterized in vitro, with TCR1−2C also evaluated in mice, making it the favored candidate for our initial trial.

Design and production of TCER structures
The design of TCER structures aim to retain the potency and sensitivity of the anti-CD3 bispecific T cell engager (BiTE) molecule while preserving the high affinity of sTCR for the pHLA-A∗11:01/KRASG12V complex. In the proposed mechanism of action, the TCER structures serve as a bridge to recruit both cell types and activate T cells (Figure 1A). Although only a few TCER structures have been published42,43,44 at the time of this study, the robustness of bispecific antibody development highlights several key features that enhance the functional efficiency and specificity of TCER. In this study, we engineered three main structures, TCER01, TCER02, and TCER03 (Figure 1B). Our structures inherited key features from different TCERs and BiTE structures that have been either approved or under evaluation by the US Food and Drug Administration (FDA), including Tebentafusp (Kimmtrak), Talartamab (Imdelltra), and Abbv-184.42,43,44,45 Specifically, KRASG12V-targeted TCR alpha and beta chains were connected via a flexible [G4S]5 linker, similar to the linker between the anti-CD3 variable heavy and light chains. In the TCER03 structure, both TCR and anti-CD3 sequences were linked via a short [G4S] linker, making this structure a single-chain TCR-mimicking antibody. The directly related structure, TCER01, retains all features of TCER03 but also includes the fragment crystallizable region (Fc region) in a single-chain form. The addition of an Fc domain has been reported to increase the half-life of the TCER in patient serum.46,47 In addition, we also incorporated certain features from previous studies to optimize Fc regions, including the use of the Glofitamab-based hinge region DKTHTCPPCPAPEAAGGP, with triple silent FcR mutations L234A, L235A, and P329G (LALA-PG).48,49,50,51 In addition, two interchain disulfide bonds were generated by introducing R292C and V302C mutations in both Fc chains. Doing so helps keep both chains relatively connected, similar to the Fc structures in the Talartamab.45 In addition, to strengthen the binding between the two Fc chains, we introduced the Y407T (knob) – into – T366Y (hole) mutations on the relevant chains. We applied the single-chain Fc approach (sFc) for simplicity of production (Figure 1B).52 The features in the Fc regions remained the same in TCER02, except that the TCR and anti-CD3 regions were now separated by Fc regions. As controls, we engineered a TCR-containing antibody without the anti-CD3 domain and another structure that only contained anti-CD3 scFv linked to the single-chain Fc (Figure 1B). For the production and purification of TCERs, please refer to the Supplementary Methods section. The purified proteins were evaluated using SDS-PAGE. High-performance liquid chromatography (HPLC) was used to ensure high purity with limited dimer or multimer formation (Figures S1D–S1H). Using Alphafold 3.0, we also predicted the structure of our TCER in spatial relation to the HLA-A∗11:01/KRASG12V complex and CD3 complex. In our attempt, the TCR region faced toward the pHLA region while anti-CD3 shows interaction with the CD3 complex (Figures S1A–S1C).

TCER binding affinity in vitro
The TCERs binding affinity against CD3e and the HLA-A∗11:01/KRASG12V complex was assessed by flow cytometry. Since SW620 CRC cells do not express HLA-A∗11:01, we engineered them to constitutively express HLA-A∗11:01 transgene and present the native KRASG12V peptide. In brief, cells expressing relevant targets were incubated with TCERs in a dose-dependent manner, and the interaction was detected by antibody staining against the Fc region of TCERs. An exception was the case of TCER03, which lacked the Fc domain and fused with a DYKDDDDK (FLAG) epitope tag at its C terminus. The binding of TCER03 was assessed using anti-FLAG antibody. TCER01, TCER02, and TCER03 showed high binding affinity against CD3e with TCER01 showing a dissociation constant (Kd) of 2.117 nM and TCER02 binding with a Kd value of 2.849 nM. Furthermore, TCER03 showed a weaker binding affinity at a Kd of 12.25 nM. All these three TCERs showed weaker binding against CD3e compared with CD3-Fc control (Kd = 0.1363 nM) (Figures 1C and 1D, and S1H, S2A, and S2B). This observation agrees with a similar approach in Tebentafusp, where the affinity of anti-CD3-bearing TCER was weaker than that of the parental anti-CD3 antibody.42 This modification facilitates the binding of TCER to both cell types, particularly cancer target cells.
Next, we evaluated the binding of TCERs against the HLA-A∗11:01/KRASG12V complex using two different approaches: (1) biolayer interferometry (BLI) to measure the affinities of the TCERs, and (2) on-cell binding assessment with SW620 cells expressing the pHLA complex. For BLI, the HLA-A∗11:01/KRASG12V complex was immobilized on biosensors and incubated with a concentration range of TCERs. TCER01 and TCER02 exhibited high-affinity binding with Kd values of 1.09 nM and 1.24 nM, respectively, comparable to the control structure where single-chain TCR12.1 was fused to the Fc region (Kd = 1.61 nM). In contrast, TCER03 displayed a slightly lower binding affinity to the pHLA complex, with a Kd of 5.67 nM (Figure 1E). Consistently, the binding was also confirmed on SW620 cells expressing HLA-A∗11:01 and presenting the KRASG12V peptide, TCER01 exhibited stronger binding with a Kd of 1.762 nM compared with TCER02 (Kd = 14.82 nM), while the TCR-Fc construct bound to the complex with a Kd of 2.176 nM (Figures 1F and 1G, S1G, S2C and S2D). The interaction of TCER03 exhibited a weaker binding signal, likely reflecting the antibody’s poor stability, which is addressed in a later section (Figures 1F and 1G, S2C and S2D).

TCERs promoted T cell-mediated cytotoxicity in vitro
Since TCERs were designed to specifically target the KRASG12V mutation presented by HLA-A∗11:01, we generated SW620 CRC cells expressing HLA-A∗11:01 using the lentiviral vectors encoding HLA-A∗11:01 and eGFP (selection marker) (Figures S3A and S3B). In order to investigate the ability of TCERs to activate T cells and trigger immune responses, TCERs were co-cultured with HLA-A∗11:01-expressing SW620 CRC cells (Figures S3A and S3B) and T cells in a dose-dependent manner. At 48 h post coculture, target cell viability was accessed via ATP quantification using CellTiter-Glo. On the other hand, TCER01 and TCER02 triggered T cell-mediated killing at higher rates and achieved maximum efficiencies of 75.03% and 68.16%, respectively (Figure 2A). This rate is higher than that of TCER03 (34.25%) and samples treated with control structures that carry either TCR (TCR-Fc) or single-chain anti-CD3 (CD3-Fc) (11.23% and 9.65%, respectively). Notably, optimal efficiency was achieved with TCER01 at a dose of 0.0625 mM, which is 16 times lower than the 1.0-μM dose required for TCER02 to achieve similar efficiency. The CD69 expression levels were used as a marker to evaluate T cell activation by TCERs. Our data showed that TCER01 activated T cells at a maximum rate of 61.53% CD69+ during this treatment, which was higher than the 45.73% CD69+ activation observed with TCER02 at 24 h post coculture. Furthermore, T cells co-cultured with target cells and TCER03, TCR-Fc, or CD3-Fc were activated by approximately 9%–12% without a dose-dependent manner (Figure 2B). In the absence of our TCERs, co-culturing with target cells still activated T cells to approximately 4.42% after normalization compared with T cell culture alone.
We then evaluated the release of cytolytic granules from the TCER-activated T cells. Perforin and granzyme B levels were assessed using flow cytometry by blocking protein release for 4 h before evaluation. Similar to CD69, perforin was detected in T cells in TCER01- and TCER02-treated samples at much higher rates (57.42% and 40.18%, respectively) in comparison with other samples as well as mocks at approximately <10% (Figure 2C). On the other hand, unlike perforin levels, granzyme B was detected at lower levels in TCER01- and TCER02-treated samples. However, granzyme B expression followed a similar pattern as co-cultured T cells producing more granzyme B in TCER01-supplemented samples (27.16% and 17.42%, respectively) (Figure 2D). In agreement with activation marker CD69 and the release of cytolytic granules, the level of interferon gamma (IFN-γ) showed a similar trend (Figure 2E). The killing effects of TCER-activated T cells were evaluated for different coculture durations, including 24, 48, and 72 h. Killing by TCER01-activated T cells was detected after 24 h of treatment at 26.72%, with no significant difference observed between 48- and 72-h co-incubation. On the other hand, TCER02 triggered T cell-mediated killing activity starting at 48 h, with a further increase at 72 h. In contrast, TCER03 had no significant effect on T cells, with the highest activity observed at 24 h (Figure S3C).

Fc domain improves the stability of TCERs
Next, we investigated the stability of TCERs used in this study. TCER01 and TCER02 shared similar features (TCR-mimicking antibody structure, anti-CD3, and Fc) but differed in their relative spatial configurations (Figure 1B). TCER03 can be considered the mini version of TCER01 in terms of structural design, lacking the Fc domain, which makes the significant difference in efficiency between TCER01 and TCER03 worth investigating. Data from other studies have suggested that the Fc domain can extend the half-life in BiTE molecules, so this feature was carefully considered in our designs.45 To investigate the stability of the TCERs, TCER01-03 was intravenously (i.v.) administered to Swiss mice, and their stability was evaluated by measuring the residual amounts in serum over 30 h using ELISA. Our results showed that TCER01 had a half-life of 26.88 ± 8.44 h in mouse serum, while TCER02 demonstrated greater stability under similar conditions (t1/2 = 58.62 ± 30.16 h) (Figures 2F; S3E; Table 2). In contrast, TCER03 levels were undetectable and indistinguishable from the background signal.
While the serum levels of TCERs were monitored in vivo over time, it is equally crucial to evaluate the functional changes of TCERs in vitro over time. To examine this, TCER01-03 was pre-incubated in culture media at 37°C for various time periods from 0 to 128 h prior to being co-cultured with T cells and target cell lines for an additional 48 h. The stability of these structures was evaluated indirectly via functional reflection, where the amount of IFN-γ releases was measured (Figure S3D). In this experiment, all three TCER molecules pre-incubated for different time ranges triggered the IFN-γ production at different levels. Specifically, prolonged pre-incubation led to lower IFN-γ release, particularly in samples pre-incubated for over 16 h (Figure S3F). Our data indicate that TCER01 exhibits superior bioactivity compared with TCER02 in terms of IFN-γ induction (Figures 2E and S3F). However, the absolute IFN-γ levels may not accurately reflect the impact of the Fc domain on TCER stability as shown in Figure S3E. To address this, we normalized the IFN-γ level to the readout from the corresponding 0-h pre-incubation samples, where the engagers are at their optimal state. For example, samples treated with TCER01 were normalized to those treated with TCER01 at 0 h pre-incubation (100%) and coculture samples in the absence of TCER01 (0%). This normalization provides the alternative assessment of TCER stability in relation to their functional activity. In this analysis, both TCER01 and TCER02 showed a similar trend of activity reduction. Their activities decreased by 10%–15% in the first hour of pre-incubation and 15%–20% in the next 2–4 h. After 8 h, the efficiency was reduced to 33%–39%. With longer pre-incubation, the reduction in activity slowed, and stabilized at 19.05% for TCER01 and 11.34% for TCER02 at 128 h (Figure 2F). In contrast, TCER03 efficiency was significantly reduced to 35.62% only after 1 h of pre-incubation and was almost abolished at 8 h (Figure S3F). The reduced activity of TCER03 suggests that this single-chain molecule, lacking an Fc domain, has insufficient stability. This rapid degradation limited its ability to bind and recruit relevant target cells, resulting in low binding efficiency to target cells expressing pKRASG12V/HLA-A∗11:01 and the weak bioactivity of TCER03 observed in Figures 1E–1G and 2A–2F, respectively. With prolonged stability and superior activation efficiency, TCER01 and TCER02 effectively promote T cell activation and cytotoxicity against CRC cells expressing pKRASG12V/HLA-A∗11:01 (Figures 2G and S3G). Moving forward, we will focus primarily on TCER01 and TCER02 because of their high efficiencies.

Exploring KRASG12V-targeting TCER sensitivity
As a potential therapeutic approach, it is crucial to evaluate the sensitivity of engineered TCERs under various conditions, particularly considering the variability in T cell profiles among patients, which can influence the relative effector-to-target (E:T) cell ratio. To understand the sensitivity of TCER molecules, T cells were incubated with target cells at various E:T ratios including 0:1, 0.1:1, 0.5:1, 1:1, 2:1, 5:1, and 10:1 while maintaining consistent TCER doses and incubation conditions. In our experiments, TCER01-treated samples achieved optimal efficiency starting at an E:T ratio of 2:1 with a cytotoxicity rate of 62.14%. The killing efficiency remained consistent at higher E:T ratios of 5:1 and 10:1 at approximately 73%–75% (Figure S3G). On the other hand, TCER02 showed E:T-dependent killing with higher ratios leading to increased killing outcomes, reaching up to 68.54% at a ratio of 10:1 (Figure S3H). Another factor that we considered was the availability of relevant KRASG12V peptides for presentation on HLA-A∗11:01. The level of the presented peptide is a key factor in determining the robustness of TCER-triggered T cell killing. To investigate this, we pulsed target cells with different amounts of the KRASG12V peptide while keeping the TCER doses and treatment conditions constant. In this approach, TCER01 reached optimal function in samples pulsed with 0.02 mM of the peptides and above, while this value in TCER-treated samples was 0.2 mM (Figure S3I). Both types of engagers showed the best activity in the range of 0.2–20 mM.

TCER01 and TCER02 show high specificity for HLA-A∗11:01/KRASG12V
Investigating the specificity of a molecule is a key factor in engineering bispecific T cell engagers. For KRASG12V-targeting TCERs, the specificity needs to be evaluated from different perspectives. These include (1) peptide specificity by evaluating key interaction residues and peptide homologs, (2) HLA specificity by comparing KRASG12V-presenting HLAs, and (3) TCER sensitivity of peptides presented by HLA-A∗11:01. Because the KRASG12V 9-mer peptide is presented by HLA-A∗11:01, peptides of similar length and sequence may also be presented by this HLA molecule, triggering conformational changes relevant for TCER binding and T cell-mediated cytotoxicity. To test this hypothesis, we used NCBI Blast and ExPAsy’s ScanProsite to search for peptides similar to the KRASG12V 9-mer peptide (VVGAVGVGK), allowing up to four mismatches as the cutoff.53,54 This search identified 19 candidates, excluding mutations at residues G12 or G13D on KRAS (Table 3). The reason for this exclusion was based on the original work from Lu et al., where the TCR used in this study was discovered, which has already been addressed in this section.22 In their work, Lu et al. reported the absence of binding to these residues and only detected a weak binding against pKRASG12C/HLA-A∗11:01 tetramer in HEK-293T cells.22 We further predicted the binding of these candidates against HLA-A∗11:01 to assess their potential to form relevant tetramer for TCER recognition. Two independent tools were used for an unbiased approach: NetMHCpan 4.155,56,57 and MHC-I binding prediction.58 The top four candidates identified from this analysis were chosen for further evaluation (Table 3). Among the four candidates (off-target 01–04), only off-target 03 triggered weak T cell-mediated cytotoxicity in TCER01-treated samples. In contrast, no significant cytotoxicity was observed in the other treatments (Figures 3A, 3B, S4A and S4B). Consistently, only off-target 03 displayed weak binding at its highest treatment dose, whereas other off-targets show no binding to the TCERs (Figures 3C, 3D, S4C and S4D), suggesting that TCER01 and TCER02 exhibit high specificity for their target, with limited binding to naturally occurring potential off-target antigens.
Although TCER01 and TCER02 demonstrate high specificity for their intended target with minimal off-target cross-reactivity, it is critical to assess their binding and functional specificity to other highly homologous peptides, including various KRAS mutant peptides. Previously, Lu et al. investigated the binding specificity of TCR 1-2C to tetramer HLA-A∗11:01 in complex with KRAS mutant peptides. In this study, we focus on the functional specificity of TCERs incorporating TCR 1-2C. The target cells SW620 expressing HLA-A∗11:01 were loaded with relevant peptides and co-cultured with T cells and TCERs in a dose-dependent manner. In addition to G12V and G12WT, seven other common KRAS mutations were tested: G12D, G12C, G12A, G12S, G12H, G12R, and G13D. Our data have shown that T cell-mediated cytotoxicity was observed in cells presenting G12D peptides, with a low rate of 14.51% in samples treated with 1 μM and 4 μM TCER01 (Figure 3E). Consistent with this finding, T cell-dependent killing was also detected in G12D samples treated with TCER02, but at a higher rate of 27.19% (Figure S4E). Interestingly, minimal cytotoxicity was observed for the other KRAS mutant peptides tested under similar conditions (Figures 3E and S4E). Overall, TCER01 and TCER02 exhibit high specificity for cells presenting the HLA-A∗11:01/KRAS G12V-9 mer complex, with minimal non-specific activity against most tested KRAS mutant variants, except for the G12D mutation at higher dose.
The binding of the TCR to the KRASG12V/HLA-A∗11:01 complex primarily results from interactions between the TCR and the peptide, as well as between the HLA and the peptide, which together enhance the TCR-HLA interaction.22 Several other HLAs have been reported to present the KRASG12V peptide. Both the original study by Lu et al. and this study focus on HLA-A∗11:01, a prevalent MHC class I allele in populations of Asian descent, found in 48.8% of individuals in Hong Kong, 31.15% of the Indian population, and 23.41% of Asian Americans, according to the Allele Frequency Net Database. Other studies have identified additional HLAs that present KRASG12V, including HLA-A∗02:01 (common in European and Native American populations), HLA-A∗03:01 (prevalent in European and Central Asian populations), HLA-DRB1∗07:01 (found in European populations), and HLA-C∗01:02 (present in Asian and Pacific Islander populations) (Table 4).18,19,22,59,60,61,62,63,64 To investigate whether TCER01 and TCER02 exhibit functionality with these HLAs, we co-cultured SW620 cells expressing the relevant HLA/KRASG12V tetramer with T cells and TCERs in a dose-dependent manner. In addition to the previously mentioned HLAs, we also tested several HLAs not previously reported to present the KRASG12V peptide, including HLA-A∗02:06, HLA-A∗24:02, HLA-B15:01, HLA-B∗40:01, HLA-B∗54:01, and HLA-C∗07:02. T cell-mediated cytotoxicity was subsequently assessed using the CellTiter-Glo assay. In our data, A∗24:02 and B∗15:01 showed cytotoxicity levels of approximately 23.02% and 19.95%, respectively. Cell killing in A∗02:01 and A∗03:01 was detected at a higher TCER01 dose of around 4 μM. In contrast, the other HLAs exhibited little to no detectable cell lysis (Figure 3F). In on-cell binding assays, about 13.24% of A∗24:02-expressing cells showed a positive TCER01 binding signal, with reduced binding rates for cells expressing B∗15:01, A∗03:01, and A∗02:01, consistent with the T cell-mediated cytotoxicity results (Figure 3G). TCER02 displayed a pattern similar to TCER01, with A∗24:02 cells exhibiting robust cell lysis of up to 34.63%, and B∗15:01 samples showing approximately 21.47% lysis. Cell lysis was also observed in A∗03:01 at the highest tested dose of TCER02, while the remaining HLAs demonstrated low cytotoxicity levels (Figure S4F). Consistent with the cytotoxicity results, TCER02 was detected in 23.25% of cells expressing the A∗24:02/KRASG12V tetramer, and in approximately 13.84% of B∗15:01-expressing cells (Figure S4G). Since each above HLA recognizes and presents a different region of KRASG12V peptide (Figure S4H), we used a 25-mer peptide (MTEYKLVVVGAVGVGKSALTIQLIQ) in this evaluation, which covers the regions recognized by all G12V-presenting HLAs,18 unlike the previous approach where we used the shorter nine-residue peptide surrounding the G12V mutation. Regardless of peptide length, TCER01 could recognize the pHLA-A∗11:01/KRASG12V complex and trigger T cell-mediated killing activity (Figure S4I). For the peptides presented by HLA-A∗03:02, we detected a cytotoxicity rate of 20.89% at a high concentration of TCER01 (50 μM). However, no significant T cell-mediated cytotoxicity was observed for HLA-A∗02:01 and HLA-B∗07:02 (Figure S4I). Together, these data suggested the high specificity of TCER01 and moderate specificity of TCER02 in targeting the KRASG12V peptide and indicated the stability of TCER/HLA-A∗11:01/KRASG12V complex requires the involvement of all components, not only limited to the TCR-peptide interaction.
Another aspect of TCER specificity for KRASG12V is identifying its key residues in spatial relation with HLA-A∗11:01 and TCERs. We evaluated TCER specificity against peptide alanine variants at each position of the KRASG12V 9-mer peptide by comparing the amount of IFN-γ release and T cell-mediated cytotoxicity of relevant HLA-A∗11:01 tetramer against TCER01. Among the eight alanine-integrated peptides, V2A and G8A peptides show no significant difference in comparing with the original KRASG12V 9-mer peptide, suggesting these two residues did not interfere with pHLA-A∗11:01 tetramer formation or affect T cell-mediated cytotoxicity. On the other hand, V1A, G3A, and K9A peptides showed a reduction in TCER01 potency. Importantly, V5A, G6A, and V7A abolished TCER01 binding and T cell-mediated killing efficiency on SW620 cells (Figures 3H and S5A). Overall, the positions of critical residues on the presenting KRASG12V peptide are consistent with other highly selective and specific TCR-peptide interactions.65,66

Immune responses of TCER01 and TCER02 across cancer cell types and donors
The responses of TCER-activated T cells were further evaluated against different tumor cells, including K562 chronic myeloid leukemia, NCI-H520 (H520) non-small cell lung carcinoma, and SW620 cells. T cell-mediated cytotoxicity for all three cell types was evaluated and compared on various categories, including HLA-A∗11:01 expression, B2M level, and peptide pulsing. In all three cancer cell types, TCER01 and TCER02 triggered the immune responses via IFN-γ production only when HLA-A∗11:01 formed a tetramer with KRASG12V peptide. The presence of KRASG12wt peptide, however, was insufficient to promote IFN-γ production by co-cultured T cells (Figure 4A). Interestingly, when manipulating B2M expression via lentiviral transduction, no significant difference was observed in T cell-mediated cytotoxicity in K562 and SW620 cell lines. However, H520 cells with B2M overexpression showed 49.21% killing in TCER01-treated samples and 51.24% in TCER02-treated samples, approximately 5-fold higher than non-B2M-transgene samples (Figure 4A). These data suggested the important role of the B2M level in determining the efficiency of bispecific engagers.
There is evidence that patients may respond differently to the same immunotherapy.67,68 Therefore, in addition to evaluating various tumor cell lines, we were also interested in exploring how our TCERs interact with T cells from different donors. This information is highly critical for the development of not only bispecific TCERs but also CAR-T engineering, as these therapies were developed based on patients’ specific conditions and their efficacy is strongly dependent on the response profile of these patients.67,68 In this study, we examined the function of TCERs using T cells isolated from different donors. While TCER01 showed highly consistent effects across the five donors through T cell-mediated cytotoxicity, we observed variable efficiency for TCER02, particularly TCER03, with robust activity for donors #03 and #05 (Figure 4B). These data suggested that TCERs exhibit variable efficacy across different donors and patients.

Efficacy of TCER01 and TCER02 in CRC spheroid
Our 2D in vitro experiments demonstrated the high specificity and potency of TCER01 and TCER02 in promoting T cell-mediated cytolysis against CRC cells expressing HLA-A∗11:01/KRASG12V. We are interested in determining whether these TCERs retain their efficacy in more complex systems, such as 3D spheroids or patient-derived organoids (PDOs). Recently, there has been growing enthusiasm for utilizing 3D models to assess the effectiveness of preclinical trial drugs for cancer and other diseases.69,70,71
Due to ethical protocol constraints and challenges in obtaining samples from cancer patients positive with HLA-A∗11:01/KRASG12V, we were unable to use the PDO model in current study. Instead, we created 3D spheroids containing HLA-A∗11:01/KRASG12V-expressing CRC cells and human stromal cells (ASCs, supporting the spheroid formation), which were co-cultured with T cells and TCERs in a dose-dependent manner. We assessed the effects of TCERs in these co-cultured samples using fluorescence-based approach and by measuring IFN-γ levels. Additionally, given that we achieved up to 80% cell killing in 2D samples, we are interested in exploring whether combining our TCERs with commonly used chemotherapies, such as oxaliplatin and irinotecan, could enhance their efficacy.72,73,74,75,76,77 Both drugs were pre-evaluated for their effects on our spheroids, and sublethal optimal doses were selected for combination with TCERs (data not shown). Our results indicate that both engagers required approximately 68–72 h to eliminate CRC cells in the spheroids, with minimal or no impact on the normal cell ASCs. Notably, the addition of oxaliplatin or irinotecan did not enhance the efficacy of TCERs, particularly for TCER01 (Figures 4C–4E and S5B–S5D). TCER monotherapy and TCER combined with oxaliplatin induced a dose-dependent increase in IFN-γ release, correlating with enhanced TCER-mediated cytotoxicity. However, the combination with irinotecan resulted in significantly lower IFN-γ levels compared with other groups. These results indicate that IFN-γ is likely a critical mediator of TCER’s mechanism of action (Figures 4C–4E, and S5B–S5D). These findings suggest that TCER01 and TCER02 can effectively and selectively target CRC cells expressing relevant HLA tetramers while sparing non-target cells.

Discussion
In this study, we engineered TCERs targeting KRASG12V/HLA-A∗11:01 by incorporating TCR from an adoptive T cell therapy study and anti-CD3 scFv. We demonstrated that TCER01 and TCER02 engaged T cells and induced potent redirected lysis of various tumor cells, including solid colorectal tumor cells. Moreover, TCER01 showed the highest tumor suppression efficacy among TCERs tested in this study. Similar cytotoxicity efficiencies were observed when using both the short 9-mer peptide and the long 25-mer peptide covering the KRASG12V region suggesting that, once imported, the long peptide is processed, and a portion of the peptide, with the desired length and sequence, is recognized as presented by HLA-A∗11:01, sufficient for TCER recognition. This closely mimics how the native protein is processed via a ubiquitin-mediated pathway.78 It is also worth mentioning that native expression of B2M is necessary for the membrane-presenting efficiency of HLA, thus contributing to the efficacy of TCERs.
Overall, our TCERs were highly specific for the KRASG12V/HLA-A∗11:01 complex. Binding and functional assays showed that both TCERs were highly selective for theKRASG12V 9-mer peptide with minimal cross-reactivity to other homologous peptides. Meanwhile, using other HLAs known to present the KRASG12V did not significantly induce T cell-mediated cytotoxicity. Even in the case of HLA-A∗03:01, which shares high similarity in peptide- and TCR-binding regions with HLA-A∗11:01, except in the dual binding region. Specifically, Q156 in HLA-A∗11:01 is positional equivalent to L156 in HLA-A∗03:01. Q156 is involved in the interaction between HLA-A∗11:01 with G3, A4, and V5 of KRASG12V peptide and residues on CDR regions of TCR.22 Future studies are required to confirm whether the hydrophobic and hydrophilic amino acid difference at the 156th residue contributes to the binding affinity of each HLA against TCER01. In addition, HLA-A∗03:01 has been reported to present a 10-mer peptide,18 suggesting its conformation and essential residue might differ from those of HLA-A∗11:01 despite the high similarity between the two protein sequences. In addition, as the intracellular proteins are continuously targeted for proteasomal degradation into short peptides, various naturally occurring peptides can be presented on HLA-A∗11:01. These random tetramers could theoretically initiate non-specific killing by TCER-activated T cells. Therefore, future studies are required to investigate this possibility.
In addition to the specificity of the incorporated TCR region, the structure of the TCER plays an important role in its efficiency. The relative location between the TCR region and anti-CD3 will determine their binding affinity toward target cells. At the same time, the linkers between these components can modulate the spatial distance between the effector and target cells, thus contributing to cell lysis efficiency. The addition of cysteine residues on Fc domains allows the formation of interchain disulfide bonds that work together with the knob-into-hole modification to maintain the stable Fc dimerization. In addition, incorporating the Fc domain helped increase the half-life of TCER01 and TCER02 and ensured sufficient T cell-mediated killing. The absence of the Fc domain in TCER03 may lead to faster degradation due to reduced structural stabilization or increased susceptibility to proteolysis. In contrast, TCER01 and TCER02 likely benefit from the Fc domain’s ability to enhance protein stability and half-life, as commonly observed in Fc-fusion biologics. The Fc domain is well-known for improving protein stability and prolonging serum half-life in Fc-fusion biologics, as demonstrated by AMG 757. This molecule was engineered to maintain the potency and sensitivity of the anti-DLL3 BiTE molecule while extending serum half-life to allow for longer dosing intervals. The prolonged half-life of AMG 757 is achieved by incorporating a stable, effector-functionless Fc domain at the molecule’s carboxy terminus.45 Furthermore, designing a single-chain component linked by flexible linkers simplifies and facilitates the production and purification of TCER without significantly interfering with its functional efficiency. However, there remains significant potential for optimization, such as reducing the molecular size or incorporating options for drug conjugation.
The recent studies have shown that the KRASG12V mutation is the second most frequent KRAS mutation in cancers (24%), with HLA-A∗11:01 being highly prevalent in Chinese and Asian American populations (both over 30%). Additionally, after adjusting for 2020 Census demographics, the population-weighted genotype frequency of HLA-A∗11:01 was 11.2% across the US population. Thus, our HLA-A∗11:01-restricted TCER, targeting the KRASG12V neoantigen, holds potential for TCR-based therapies in a significant proportion of KRASG12V cancer patients, estimated at over 7.2% in Chinese and Asian American populations and 3.36% across the US population.59,60,61 In summary, our work highlights the prominent application of TCR not only in CAR-T cell generation but also in engineering BiTE receptors, offering a valuable tool for cancer suppression and improved antitumor effectiveness. The structural design could inform the future development of therapeutics targeting the KRASG12V mutation and other cancer-specific mutations.
This study provides a compelling and elegant demonstration of the in vitro potential of the TCER01 molecule. The data on its potency, specificity, and stability are robust and establish a strong foundation. However, to bridge these findings toward clinical translation, a thorough in vivo investigation is the essential next step. In future work, evaluation of TCER01 in relevant preclinical models, such as humanized mouse models bearing KRASG12V/HLA-A∗11:01-positive tumors, is needed. Such studies are critical to (1) confirming antitumor efficacy in a complex physiological environment, (2) assessing the pharmacokinetic and pharmacodynamic properties of the molecule, and (3) crucially, identifying potential on-target, off-tumor toxicities or cytokine release syndrome, which represent the most significant safety hurdles for this class of therapeutics. Addressing these points will advance this promising therapeutic candidate to the next stage of development.

Materials and methods

Ethics statement
Human blood samples were collected for research purposes in accordance with the protocols approved by the Ethics Committee of the Medical Genetics Institute (approval number 02/2024/CT-VDTYH) and conducted in accordance with the ethical principles of the Declaration of Helsinki (1964). Written informed consent was obtained from each participant in accordance with the Declaration of Helsinki. All procedures complied with the guidelines and regulations, and the sample collection was performed at the Medical Genetics Institute.

Cloning and plasmid construction
A list of plasmids generated for this study is shown in Table S1. The original plasmids used in this study were obtained from Thermo Fisher (pcDNA 3.1 [+] V79020), BioIntron (pcDNA 3.4 [+]), or Takara Bio (pLVX Puro 632164). All structures were designed in-house, and relevant TCER-expressing plasmids were engineered either through BioIntron, DNA sequencing, or in-house using a Gibson-based approach.79 Fragments containing TCER sequences were amplified using Q5 High Fidelity DNA polymerase (NEB M0491S). These fragments were then inserted into pcDNA 3.1 (+) or pcDNA 3.4 (+) backbones (for TCER structures) or pLVX (for lentivirus expressing plasmids) via Gibson assembly (Thermo Fisher A46628) following standard procedure.79 The assembled fragments were transformed into 10-beta competent cells (NEB C3019I) for TCER plasmids and Stbl3 (Thermo Fisher C737303) for lentivirus plasmids. Engineered plasmids were confirmed using Sanger Sequencing at BioIntron, DNA Sequencing, and Gene Solutions Lab.

Production and purification of TCERs
TCERs were produced and purified either in-house or by BioIntron Biologicals Inc. All TCERs were produced in suspension CHO-S or HEK293F cell lines where 1 mg of plasmid per 1 million cells was transfected using PolyEthylenImine (PEI) 10 kDa (Sigma-Aldrich 765090) in FreeStyle expression media (Gibco 10319322) at 37oC, 70 rpm, and 5% CO2. Proteins were expressed for 7 days with additional nutrient supplementation. The supernatant was harvested by centrifugation and filtered using a 0.45-mm filter. Prior to purification, protein expression was evaluated using SDS-PAGE followed by silver staining or ELISA anti-immunoglobulin (Ig)G horseradish peroxidase (HRP) detection.
In-house purification was performed using the AKTA protein purification system (Cytiva) with a HiTrap MabSelect PrismA column (Cytiva 17549853) following the manufacturer’s instructions. Protein fractions were analyzed using 10% SDS-PAGE followed by silver staining and size-exclusion chromatography. High-yield and high-purity fractions were combined for buffer exchange. TCERs were stored in a buffer containing 0.55 mg/mL glutamic acid, 0.03 mg/mL polysorbate 80, and 28.5 mg/mL sucrose at pH 5.2.

In silico peptide evaluation and protein complex prediction
The peptide homolog identification was performed using NCBI BLAST and Expasy ProScan with the UniProt protein database for Homo sapiens, using a cutoff of four mismatches. Homologous and alanine-scanned peptides were then evaluated with KRASG12V 9-mer peptide for binding affinity with HLA-A∗11:01 using NetMHC Pan 4.1 or MHC-I prediction. The binding models of TCERs with KRASG12V/HLA-A∗11:01 and CD3 were performed using a combination of Python-scripted Alphafold 3.0, TCR model 2.3, and TCR dock. PyMOL was used to evaluate and optimize the model visualization.

Octet biolayer interferometry
BLI experiments were conducted using an Octet HTX instrument (ForteBio) with anti-GST antibody biosensors for GST-tagged ligands and His-tagged analytes at 1,000 rpm and 25°C. Concentrated TCERs and HLA tetramers were diluted in BLI reaction buffer (25 mM HEPES, pH 7.4, 150 mM NaCl, 0.1 mg/mL bovine serum albumin, and 0.01% Tween 20). GST-fused HLA-A∗11:01/KRASG12V 9-mer was captured on GST biosensors from a 2 μg/mL solution in PBT, followed by a 180-s quench step with 100 μg/mL biotin. After equilibration in PBT, the loaded biosensors were dipped for 600 s into wells containing serial 3-fold dilutions of TCERs and then transferred back to assay buffer for 600 s of dissociation. Sensorgram binding response raw data were reference-subtracted and fitted to a 1:1 binding model using ForteBio’s Data Analysis Software 9.0. Three replicates were measured for each construct, and the average and coefficient of variation were calculated. Binding constants (Kd) were determined by fitting the wavelength shifts in the steady-state regions using a single-site binding model.

Cell-based binding assay
For cell binding assays, target cells were incubated with relevant TCERs for 1 h in defined media and then incubated with a secondary antibody for assessment using flow cytometry.

Cell lines and culturing
All cell lines were cultured at 37°C, 5% CO2. Human embryonic kidney cells for lentivirus production (HEK293T, Takara Bio 632273), colorectal carcinoma (SW620, ATCC CCL-627), and lung cancer cell line (NCI-H520, ATCC HTB-182) were adapted and grown in DMEM media (Thermo Fisher 11995081) supplemented with 10% heat-inactivated fetal bovine serum (FBS) (Thermo Fisher A5256701). Cells awakened from frozen stocks were cultured to the three-four generations prior to the experiments. All cell cultures were used at passage numbers below 15.

Lentivirus production and transduction
HEK293T cells (passage number <15) were seeded at a density of 0.5 × 106 cells per cm2 in six-well plates prior to transfection in DMEM supplemented with 10% FBS and without antibiotic for approximately 18–24 h to reach confluency of 75%–80%. Plasmids used for lentivirus production, including the lentivirus transfer plasmid, Gag/Pol, Rev, and RSV envelope plasmid, were mixed at a molar ratio of 8:6:6:1 following previously established work, respectively.80
At 48–72 h post-transfection, the cell supernatant was collected and filtered through a 0.45-mm PES filter. The flow through containing lentivirus was used for transduction into the target cells. Polybrene (Sigma TR-1003-G) was added to the cell culture at the final 8 mg/mL concentration to increase transduction efficiency. Transduced cells were centrifuged at 800 × g, 37°C for 1 h. At 24 h post-transduction, cell media containing the virus was removed and replaced with fresh complete media supplemented with puromycin (Thermo Fisher A1113803) at the final concentration of 10.0, 5.0, and 5.0 mg/mL for HEK293T, H520, and SW620 cells, respectively.

Transient transfection
Cells designated for transient transfection were seeded in cell media with 5%–10% FBS without antibiotic for 24–36 h before transfection. For PEI-based transfection, cells were cultured in antibiotic-free media, briefly washed with pre-warmed 1x PBS (Cytiva SH30258.02) and placed in 80% volume of Opti-MEM (Thermo Fisher 11058021). Plasmid DNA (pDNA) was mixed with PEI at a ratio of 1:3 wt/wt in Opti-MEM for 20–25 min at 37oC before being added dropwise into the cell culture. For Lipofectamine P3000-based transfection, plasmids were prepared according to the manufacturer’s instruction (Thermo Fisher L3000015) and incubated with the cells for 6 h before being replaced with fresh complete media.

Cell-based functional assays
The peptide-pulsed cells were washed once in 1x PBS and once in AIM-V containing 5% FBS. Effector cells were cultured in AIM-V media supplemented with 5% FBS for at least 24 h and checked for viability. The cells were then co-cultured with effector cells in the presence of the relevant TCERs. The supernatant was collected for the IFN-release ELISA assay (Biolegend 430104). Simultaneously, T cells were collected for T cell activation assays as described above, and target cells were subjected to luminescence-based cytotoxicity assay using CellTiter-Glo (Promega G9242). Both functional assays were performed in accordance with the manufacturer’s instructions.

Peripheral blood mononuclear cell collection and T cell isolation
Peripheral blood mononuclear cells (PBMCs) from donors’ blood samples were isolated using Lymphoprep (StemCell Technologies 07861) following the manufacturer’s instructions with minor modifications. In brief, diluted blood samples were layered on top of Lymphoprep and centrifuged at 1,000 × g at room temperature for 30 min with both brake and acceleration off. Isolated PBMCs were cultured in AIM-V media (Thermo Fisher 12055083) supplemented with 5% FBS at 37°C, 5% CO2 and used within 3 days of culturing.
T cell isolation from total PBMCs was performed using the EasySep Human T cell isolation kit (StemCell Technologies 17951) following the manufacturer’s instruction. Briefly, lymphocytes were resuspended at a density of 5 × 107 cells/mL in a volume of up to 2 mL and supplemented with an isolation cocktail at a ratio of 5/100 (vol/vol). The mixture was incubated at room temperature for 5 min, followed by the addition of Rapishpere at the ratio of 5/100 (vol/vol). The solution was mixed with a wash medium (1x PBS containing 2% FBS and 1 mM EDTA without Ca2+, Mg2+) to a volume of 2.5 mL. The non-CD3+ cells were then captured using magnet, leaving the CD3+ cells in supernatant. T cells were then collected and resuspended in the AIM-V medium.

Peptides
For a complete list of the peptides used in this study, please refer to Table S2. All peptides were synthesized by Genscript Inc with a purity of >95%. Peptides were resuspended in appropriate solvent following the manufacturer’s certificate of analysis (CoA) at the concentration of 20 mM and stored as aliquots at −20oC. Each peptide batch was subjected to no more than two freeze-thaw cycles.

Peptide pulsing and co-culturing
Prior to the experiment, K562, H520, HEK293T, and SW620 cells expressing the HLA of interest were isolated from ongoing culture and washed twice with pre-warmed PBS. Target cells were pre-labeled with CFSE (Invitrogen 65-0850-84) at a final concentration of 0.125 mM in PBS, equivalent to an 80,000× dilution. Staining was performed at 37°C for 5 min in the dark and was immediately stopped by incubating with five volumes of cold FBS at 4°C for 5 min. The cells were then washed with five volumes of pre-chilled PBS up to 3–5 times to reduce the background.
For peptide pulsing, labeled cells were seeded at 6.25 × 104 cells per cm2 in AIM-V media containing 0%–2% FBS and relevant peptides at the specified concentrations for 18–24 h at 37°C, 5% CO2. Pulsed cells were subsequently washed once with PBS and once with AIM-V supplemented with 5% FBS. The cells were resuspended and seeded at the density of 2 × 105 cells/mL.
For coculture assays, effector cells (PBMC or T cells) were grown in AIM-V medium with 5% FBS and mixed with target cells at specified effector: target (E:T) ratios in the presence of relevant TCERs. Co-culturing was done for 24–72 h before analysis.

Flow cytometry and cell-based staining
For a complete list of the antibodies used, please see Table S3. For cell surface immunofluorescence staining, the cells were washed three times and resuspended with pre-chilled 1x PBS and stained with relevant antibodies at 4°C for 30 min in the dark. Before analysis, the cells were washed three times with pre-chilled 1x PBS. For perforin and granzyme B staining, T cells were incubated with protein transport inhibitor, Monensin (Thermo Fisher 00-4505-51) for 4–6 h at 37°C, 5% CO2. Cells were then permeabilized using intracellular stain perm wash buffer (Biolegend 421002) for 45 min and washed twice before incubation with the relevant antibodies. Flow cytometry was performed on an Agilent NovoFlow cytometer.
For the gating strategy, the FSC and SSC were used to define the single-cell population. After gating on singlets, live cell types of interest with specific markers were defined on plots with relevant stains, including CD3-PE Cy7 for T cells, DYKDDDDK-APC, and HLA-A APC for SW620. Markers of T cell activation and cytokine granule production were evaluated using plots for CD69 APC, perforin-AF700, and granzyme B PerCP by comparing mock and experimental samples under defined conditions.

Spheroid formation and drug administration
To evaluate the cytotoxic effects of T cell engagers (TCERs) alone and in combination with chemotherapy, 3D CRC spheroids were generated. Oxaliplatin (MedChem Express, HY-17371) and irinotecan (MedChem Express, HY-16562) were employed as chemotherapeutic agents alongside TCERs (TCER01 and TCER02). Spheroids were created by co-culturing SW620 CRC cells (ATCC, CCL-227) and human adipose-derived stromal cells (ASCs, Cellosaurus CVCL_4W37) at a 2:1 ratio, totaling 5,000 cells, in ultra-low attachment 96-well round-bottom plates (Corning Costar, 7007). The cells were suspended in 200 μL of AIM-V medium (Gibco, 12055091) supplemented with 5% FBS (Gibco, 10437028) and 2 μM KRASG12V 9-mer peptide to promote TCER-specific antigen presentation on SW620 cells. SW620 cells and ASCs were stably transduced with green fluorescent protein (GFP) and red fluorescent protein (RFP), respectively, to enable visualization of tumor-stromal interactions. The plates were centrifuged at 350 × g for 5 min and incubated at 37°C with 5% CO2 for 48 h to form compact spheroids.
Spheroids were co-cultured with 1 × 105 T cells and exposed to single agents (TCER01, TCER02, oxaliplatin, or irinotecan) or their combinations. For single TCER treatments, TCER01 and TCER02 were diluted 4-fold across a 10-point range (0.015–60 nM). For single chemotherapy treatments, oxaliplatin (0.2–200 μM) and irinotecan (0.1–100 μM) were diluted 10-fold. In combination studies, TCERs (0.00015–60 μM) were combined with fixed sublethal doses of oxaliplatin (10 μM) or irinotecan (100 μM). Controls consisted of spheroids alone, spheroids with T cells, T cells alone, and T cells with oxaliplatin or irinotecan. After 72 h, spheroid morphology and GFP fluorescence were evaluated using an EVOS M5000 Imaging System (Thermo Fisher Scientific). GFP fluorescence intensity was quantified with ImageJ software (NIH, v1.53). Supernatants were collected for IFN-γ measurement via ELISA assay, as previously described.

Pharmacokinetics study
For studying the pharmacokinetics (PK), six male Swiss albino mice were randomly assigned to each experimental TCER group (TCER01, TCER02, TCER03), with three mice in the PBS control group. Mice were injected i.v. through the tail vein with 2 mg kg−1 drug or PBS. Blood samples were collected in at 30 min, 90 min, 3 h, 6 h, 24 h, and 30 h. Samples were centrifuged and serum fractions were stored at −20°C for later ELISA. Animal experimentation was approved by the Animal Ethics Committee at the University of Science in Ho Chi Minh City, Vietnam (protocol number 562/KHTN-ACUCUS).
For ELISA, samples (100 μL) diluted in PBS and dilution series of known concentrations were added to immuno 96-well plates and incubated for 2 h at room temperature (RT). The wells were washed three times in in PBS +0.05% Tween 20 (PBST), 300 μL per well. TCER01 and TCER02 constructs were detected using anti-Fc peroxidase antibody (Thermo Fisher Scientific, A18817) diluted 1:5,000 in 1% BSA in PBST, while TCER03 was detected using anti-DYKDDDDK TAG conjugated HRP (1:1,000, Thermo Fisher Scientific, A01428100). The wells were washed four times with PBS and developed with TMB (eBioscience, 00-4201-56). Each sample was measured in triplicate, and data were analyzed using GraphPad Prism v.8 and Phoenix WinNolin 8.5.2.4 using noncompartmental analysis with model 201.

Statistical analysis
T cell-mediated cytotoxicity was evaluated using the CellTiter-Glo assay, in which the amount of ATP from live cells was used to determine live cell density. The percentage of specific cell lysis was calculated using the following Equation 1:
For IFN-γ release ELISA, readouts were normalized to those of co-cultured samples without TCERs. Flow cytometry signals were normalized to T cells only, target cells only, or co-cultured samples without TCERs. Error bars represent the standard error of the mean (SEM). For multiple comparisons to the same control (in the donor’s evaluation experiment), we used one-way ANOVA, followed by Dunnett’s test. For multiple pairwise comparisons (in tumor cells assessment experiment), we applied one-way ANOVA coupled with Tukey’s honestly significant difference (HSD) test. At least three biological samples and three technical replicates were analyzed for each condition.

Data availability
Datasets supporting the conclusions of this study are available in the Source Data file. The relevant materials will be shared upon request.

Acknowledgments
We are grateful to Phuong-Diem Tran Thi, Le-Son Tran, Trung-Quan Nguyen, Nhu-Xuan Duong, and Minh-Vien Le Thi at Gene Solutions for providing the materials used in this study.

Author contributions
Conceptualization: N.H., C.T.N., and H.-N.N. Methodology: N.H., T.-M.N.T., N.T.B., and C.T.N. Investigation: N.H., T.-M.N.T., N.T.B., and C.T.N. Visualization: N.H. and T.-M.N.T. Project administration: N.H., C.T.N., and H.-N.N. Supervision: N.H., C.T.N., and H.-N.N. Writing – original draft: N.H. Writing – review & editing: N.H., T.-M.N.T., N.T.B., C.T.N., and H.-N.N.

Declaration of interests
All authors are employees of Gene Solutions Inc. The author, H.-N.N. holds equity in Gene Solutions. We confirm that this does not alter our adherence to journal policies on sharing data and materials.