Viscoelastic synthetic antigen-presenting cells for augmenting the potency of cancer therapies

Abstract

The use of synthetic antigen-presenting cells to activate and expand engineered T cells for the treatment of cancers typically results in therapies that are suboptimal in effectiveness and durability. Here we describe a high-throughput microfluidic system for the fabrication of synthetic cells mimicking the viscoelastic and T-cell-activation properties of antigen-presenting cells. Compared with rigid or elastic microspheres, the synthetic viscoelastic T-cell-activating cells (SynVACs) led to substantial enhancements in the expansion of human CD8⁺ T cells and to the suppression of the formation of regulatory T cells. Notably, activating and expanding chimaeric antigen receptor (CAR) T cells with SynVACs led to a CAR-transduction efficiency of approximately 90% and to substantial increases in T memory stem cells. The engineered CAR T cells eliminated tumour cells in a mouse model of human lymphoma, suppressed tumour growth in mice with human ovarian cancer xenografts, persisted for longer periods and reduced tumour-recurrence risk. Our findings underscore the crucial roles of viscoelasticity in T-cell engineering and highlight the utility of SynVACs in cancer therapy.

The early diagnosis of cancer and traditional therapies for it, such as surgery, radiation therapy and chemotherapy, have improved the treatment of various cancers. However, there are side effects, resistance to therapies, uncurable genetic mutations and recurrence. To address these limitations, researchers are exploring immunotherapies such as chimaeric antigen receptor (CAR) T-cell therapy^1–4. Current CAR T-cell therapies have shown remarkable success in treating blood cancers and lymphomas⁵. However, several issues persist, including cancer recurrence and the lack of long-term immunity against cancer. In the context of solid tumours, the immunosuppressive tumour environment and the incompetency of existing CAR T-cell expansion methods present additional challenges that need to be addressed^6–8. Therefore, there is an urgent need for the development of more effective CAR T-cell expansion techniques, which can enhance the cancer-killing capability of these cells and promote long-lasting immunity against cancer. Addressing these challenges will impact cancer treatment, including solid tumours, ultimately improving patient outcomes and quality of life.

T-cell stemness, characterized by the capacity to self-renew and differentiate into multiple T-cell subsets, is a vital feature for sustaining a long-lasting and effective immune response against cancer⁹. In particular, T memory stem cells (TMSCs), a subset of memory T cells with stem-like properties, are becoming increasingly recognized for their critical role in sustaining a durable and effective immune response after adoptive transfer^10–12. TMSCs possess both a self-renewal ability and multipotent capacity to differentiate into various antigen-specific T-cell subsets, rendering them exceptionally potent in generating a robust and persistent immune response¹³. Although cytokines and small molecules have been explored to enhance the generation and maintenance of TMSCs^14,15, the results may vary due to the lack of control of other niche factors, such as the mechanical properties of synthetic antigen-presenting cells (APCs) or matrices.

T-cell activation is a critical first step in the adaptive immune response, as it defends the body following the initial interaction between T lymphocytes and APCs¹⁶. Recently, various approaches of synthetic APCs have been developed to induce ex vivo and in vivo expansion of T cells^17–19, using advanced biomaterials that replicate specific bioactive signals of the APC surface in both two-dimensional (2D)^20,21 and three-dimensional (3D)^22–24 microenvironments. While current ex vivo T-cell stimulation technologies are useful for efficiently enriching and activating antigen-specific T cells, the conditions remain to be optimized. For example, the dominant method for the activation of T cells is the use of anti-CD3/CD28 (cluster of differentiation 3/cluster of differentiation 28) antibody-coated paramagnetic beads such as Dynabeads (Gibco)²⁵. However, T cells activated by Dynabeads generally result in suboptimal cell expansion rates, fewer CD8⁺ cytotoxic T cells and loss of stemness due to the intrinsic differences between Dynabeads and APCs²⁶. In particular, Dynabeads made of polystyrene are stiff (20–40 MPa)²⁷ and have very distinct mechanical properties compared with APCs that are much softer and show viscoelastic behaviour²⁸, potentially compromising the engagement and activation level of T-cell receptors (TCRs)^29,30.

The viscoelasticity of the extracellular matrix and of surrounding cells plays a notable role in shaping cellular behaviour and function³¹. Viscoelasticity reflects the ability of a substance to resist deformation and return to its original shape after being subjected to stress. Many cell types have been shown to respond to alterations in the viscoelastic properties of their environment³². However, whether the viscoelastic property of synthetic APCs regulates T-cell activation and expansion remains unexplored³³. We postulate that the stiffness and viscoelasticity of APCs play a crucial role in regulating TCR activation and T-cell differentiation. To test this possibility, we developed a scalable technology to produce synthetic viscoelastic activating cells (SynVACs) that mimic the stiffness (kilopascal level) and viscoelasticity of native APCs and demonstrated the robust effects of SynVACs on T-cell expansion and as a potential therapy of lymphoma and solid tumours, in comparison with purely elastic beads and the Dynabeads—a clinically used product as a benchmark.

Results

Fabrication of viscoelastic microbeads by using a high-throughput microfluidic device

Based on the current limitations of existing technologies, our goal was to develop cutting-edge technology to enhance the stemness, CAR transduction rate, tumour-killing efficiency and in vivo persistence of CAR T cells for cancer immunotherapy. Therefore, we designed SynVACs as engineering APCs with fine-tuned mechanical properties (viscoelasticity and stiffness) and surface ligands/antibodies to achieve the desired T-cell activation (Fig. 1a). Alginate is an inert material with excellent biocompatibility, tunable mechanical properties and well-controlled surface chemistry for molecular conjugation, which makes it an ideal candidate for the material base of the technology^34,35. We fabricated alginate-based SynVACs using a high-throughput microfluidic device via a pH-induced internal gelation method. The scanning electron microscopy (SEM) image reveals the channel design of the microfluidic device, specifically the crosslinking area for microbead formation (Fig. 1b). At the T-junction, an alginate solution with pH-responsive calcium meets an acidic oil phase to induce particle formation and ionic crosslinking. Figure 1c and Supplementary Video 1 show the formation of alginate beads.

Fig. 1 |. Development of SynVACs to mimic the viscoelastic properties of native APCs using a microfluidic device.

a, SynVACs use a biomaterial-based approach, using ionically crosslinked alginate networks to create a tunable viscoelastic system that mimics the mechanical properties of native APCs. b, SEM image of the microfluidic device at the crosslinking area. Scale bar, 100 μm. c, SynVACs are fabricated based on a pH-induced internal gelation method. Scale bar, 50 μm. d, Representative images of antibody-coated elastic beads, and SynVACs compared with Dynabeads. Scale bars, 15 μm. e, The size of microbeads is controlled by using a microfluidic device with various channel widths. The data points denote the mean values, and the error bars denote the standard deviations of five biological replicates. f, The size distribution of elastic beads and SynVACs. g, The size of microbeads is controlled by using a microfluidic device with various flow rates. The data points denote the mean values, and the error bars denote the standard deviations of five biological replicates. h, Quantification of cell viability (determined by a LIVE/DEAD staining kit) at day 14 by co-culturing elastic beads and SynVACs with Jurkat T cells. i, Production and collection rates of SynVACs per microfluidic device. j, Stability of GFP-labelled SynVACs over time in RPMI media with 10% FBS (n = 3). k, Left: the strategy for removing SynVACs from T cells after co-culture involved a single centrifugation step (600 g, 5 min). Right: quantitation of removal efficiency (n = 5). In h–k, data represent mean ± standard deviation of three or four biological replicates. In e, g, i and k, significance was determined by a two-tailed, unpaired t-test. NS, not significant. In h, significance was determined by a one-way ANOVA and Tukey’s multiple comparison test.

We first optimized the device and fabrication parameters to achieve monodispersed elastic or viscoelastic microparticles with defined sizes. We made viscoelastic alginate beads using low molecular weight (MW) of alginate (75 kDa) and calcium-based ionic crosslinking. In this Article, we used clinical-grade sodium alginate. Alginate is widely acknowledged as a non-antigenic material, showing excellent biocompatibility, particularly in vitro, where it finds extensive utilization. In addition, our previous studies have demonstrated no obvious antigenicity when alginate is used in the form of microparticles or macroscopic hydrogels in vivo^18,19 In its unmodified state, alginate shows little to no cell binding, offering an inert background conducive to the incorporation of precisely defined biological signals. The maximum stiffness achievable by alginate gels may be below 300 kPa³⁶, which is sufficient to cover the dynamic range of native APCs. In this study, we engineered alginate beads with mechanical properties mimicking APCs (kilopascal level). Dynabeads made of polystyrene have a stiffness of up to 20–40 MPa^27,37. As Dynabeads have been optimized with CD3/CD28 antibody coating for T-cell activation and expansion in clinical settings, we use Dynabeads as a benchmark for functional comparison of expanded T cells. To fabricate elastic beads, we used higher-molecular-weight (120 kDa) alginate to first generate ionically crosslinked microbeads then convert them into covalently crosslinked microbeads via carbodiimide-based chemistry. Figure 1d displays the resultant viscoelastic beads and elastic beads, showing transparency compared with rigid polystyrene Dynabeads. The diameter of the microbeads can be adjusted by modifying the channel width of the microfluidic device (Fig. 1e). As the use of alginate with varying molecular weights could lead to changes in the solution’s viscosity, subsequently causing differences in the bead size produced, we used cellulose to counterbalance potential viscosity changes. Carboxy methyl cellulose (CMC) can be easily washed out of microbeads afterwards (Supplementary Fig. 1)^38,39 without interferences with mechanical property and antibody modification process. In this study, we made viscoelastic and elastic beads with diameters ranging from 7 μm to 9 μm (Fig. 1f), mirroring the size of dendritic cells. Furthermore, the bead size can be fine-tuned within a narrow range by adjusting the oil phase flow rate within the channel (Fig. 1g). Our experiments indicate that neither viscoelastic nor elastic beads affect cell viability when co-cultured with human Jurkat T cells, suggesting that the fabricated beads are biocompatible with immune cells (Fig. 1h). A single microfluidic chip can produce approximately 13 million microbeads within an hour, and around 11 million beads can be collected, with losses potentially due to multiple centrifugation steps during the collection process (Fig. 1i). This process can be easily scaled up with a parallel microfluidic system for the high-throughput production of viscoelastic beads, which is critical for their prospective clinical application.

To assess the long-term stability of beads in culture, we conjugated green fluorescence protein (GFP) onto SynVACs during fabrication and maintained the microbeads in RPMI (Roswell Park Memorial Institute) culture media (with 10% fetal bovine serum (FBS)) or HEPES (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid) buffer at room temperature or 37 °C for 15 days, the typical period for human CAR T-cell expansion ex vivo (Fig. 1j and Supplementary Fig. 2). The evaluation of the mean fluorescence intensity revealed no notable changes during this period, indicating that SynVACs were stable in solution. To remove these alginate microbeads from the culture, the beads can be quickly dissolved within a few minutes by introducing chelating agents such as EDTA (ethylenediamine tetraacetic acid) or citrate (Fig. 1j and Supplementary Fig. 2). As an alternative approach, SynVACs can be readily separated from cells by using a physical centrifugation method, resulting in a removal rate of alginate beads (lower in density than cells) greater than 97% (Fig. 1k and Supplementary Fig. 3). A higher removal rate can be achieved through multiple centrifugation processes. Consistent with the stability data, the amount of residual alginate in the supernatant of SynVAC and T cell co-culture was less than 50 ng ml⁻¹ (Supplementary Fig. 4). This low level of alginate had no detrimental effects on T cells in our experiments, similar to previous in vitro⁴⁰ and in vivo¹⁹ studies.

Modulation and characterization of the mechanical properties of SynVACs and elastic beads

In this work, four types of alginate microbead with defined viscoelasticity and stiffness were fabricated, including: (1) V1, low-stiffness viscoelastic beads; (2) V2, high-stiffness viscoelastic beads; (3) E1, low-stiffness elastic beads; and (4) E2, high-stiffness elastic beads. The formulation and parameters for each type of microbead are shown in Supplementary Table 1. For viscoelastic beads in the SynVAC group (V1 and V2), we used lower-molecular-weight alginate polymer (75 kDa) to achieve fast stress relaxation (high viscoelasticity) and varied calcium concentrations (15 mM and 50 mM) to ionically crosslink alginate with low (15 kPa) and high (25 kPa) stiffness, while maintaining a half-stress relaxation time around 10 s (Fig. 2a–c). As shown in Supplementary Fig. 5, calcium concentration between 10 mM and 50 mM induced an increase in alginate gel stiffness but had a negligible effect on the relaxation time. To fabricate beads with elastic properties (E1 and E2), we generated microbeads with higher-molecular-weight (120 kDa) alginate polymer using the same method, crosslinked them further with varying concentrations of adipic acid dihydrazide (AAD) and subsequently removed calcium-based crosslinking with sodium citrate. This process of converting viscoelastic crosslinks to elastic covalent bonds enables microbeads with tunable stiffness and slow stress relaxation (Fig. 2a–c and Supplementary Fig. 6). With this slow stress relaxation and covalent crosslinking, stiffness is the dominant mechanical property, and modulation of the stiffness with AAD (between 5 and 25 mM) had no substantial effect on the viscoelastic property (half-stress relaxation time of around 1,000 s) (Supplementary Fig. 7). The stiffness of SynVACs, elastic beads and native APCs was also directly measured using the atomic force microscope (AFM) indentation method, showing consistent results as 2D gels of the same formulation (Fig. 2b and Supplementary Fig. 6a). In addition, the variation of the stiffness of the fabricated microbeads was similar to 2D gel, showing the stable fabrication process (Supplementary Fig. 6b). Moreover, the SynVACs and elastic beads used in the AFM assessments were pre-conjugated with antibodies. This ensures that the measurement results reflect the properties of the actual microbeads used in T-cell activation experiments. Among all the groups fabricated, V1, with low stiffness and high viscoelastic properties (15 kPa, T_1/2 = 10 s), possesses properties most similar to that of mouse dendritic cells with a stiffness of 1–11 kPa and stress relaxation time of 1–5 s (ref. 41). In addition, our side-by-side characterization of stress relaxation of SynVAC beads and the human native APCs (monocytes) by using AFM showed similar results (Supplementary Fig. 8), while elastic beads did not show any stress relaxation.

Fig. 2 |. Characterization of SynVACs in terms of stiffness, viscoelasticity and ligand density.

a, An alginate gel can be ionically or covalently crosslinked to provide viscoelastic or elastic properties, respectively. The molecular weight of the alginate polymer can control the viscoelasticity of crosslinked alginate networks. As the molecular weight increases, a dense network with high physical entanglement and overlap of the polymer chain forms, resulting in an alginate gel with high stiffness and low viscoelasticity, and vice versa. The covalently crosslinked network is elastic as the covalent crosslinks preserve the memory of the initial state. The SynVACs can be transformed into elastic beads by covalent crosslinking and subsequent Ca²⁺ removal. b, Quantitative assessment of the compressive modulus of SynVACs, elastic beads and human APCs conducted using AFM indentation technique (n = 12). c, Quantification of the timescale at which an initially applied stress is relaxed to half its original value (n = 7). The stress relaxation time reflects the viscoelastic property of the gels. d, Shear storage and shear loss modulus as a function of frequency for ionically and covalently crosslinked gels. e, Schematic of antibody conjugation on SynVACs and elastic beads using the tetrazine–TCO ligation method. f, Representative flow cytometry histograms showing the fluorescence intensity of SynVACs and elastic beads expressing various levels of anti-CD3 densities (n = 3). g, Effect of antibody density on human T-cell activation by the analysis of an early T-cell activation marker CD69. SynVACs coated with different amounts of antibodies (anti-CD3/anti-CD28 with a ratio of 1:1) were co-cultured with Jurkat cells (human T lymphocyte cell line) for 18 h, and the percentage of CD69⁺ cells was analysed by flow cytometry. The data points denote the mean values, and the error bars denote the standard deviations of three biological replicates. h, Confocal microscopy images of individually labelled SynVAC, showing a distinct fluorescence signal approximately 100 nm thick around the equatorial section, indicative of surface-conjugated antibodies. Scale bar, 2 μm. In b and c, the data points denote the mean values, and the error bars denote the standard deviations of 12 or 5 biological replicates. In b and c, significance was determined by one-way ANOVA with Tukey’s multiple comparison test. In g, significance was determined by a two-tailed, unpaired t-test.

Further rheological tests were performed to evaluate the viscoelasticity of V1 and E1 beads. Frequency sweep analysis showed that the storage modulus of viscoelastic gels and elastic gels were similar; however, the loss modulus of viscoelastic gels (Supplementary Fig. 9a(i)) was 50 times higher than that of the elastic gel (Supplementary Fig. 9b(i)). Shear-thinning of viscoelastic gels was 10 times more than the elastic gels (Supplementary Fig. 9a(ii), b(ii)). Strain sweep tests also revealed that the viscoelastic gels showed a >20-fold higher loss modulus (Supplementary Fig. 9a(iii)) compared with the elastic gels (Supplementary Fig. 9b(iii)). Consistently, the frequency-dependent and strain-dependent loss factor (tan δ), which represents the ratio of the energy lost to the energy stored in a material during the deformation, revealed higher viscoelasticity for the viscoelastic gels than elastic gels (Fig. 2d and Supplementary Fig. 9c).

Surface modification of SynVACs and elastic beads with T-cell activation signals

To provide T-cell activation signals, SynVACs and elastic beads were modified by conjugating signalling molecules onto these alginate microbeads based on bioorthogonal chemistry for (1) polyclonal T-cell activation using activating antibodies (anti-CD3 and anti-CD28) and (2) antigen-specific CAR T-cell enrichment using mesothelin (MSLN) protein and anti-CD28 (Fig. 2e). Tetrazine–trans-cyclooctene (TCO) clicks reaction occurs rapidly and selectively, resulting in a high conjugation rate and stable covalent linkage between the antibodies and SynVACs⁴². In addition, using tetrazine–TCO click chemistry with a short polyethylene glycol (PEG) linker for antibody conjugation to SynVACs ensures that the ligand’s functionality is directly influenced by the underlying material properties.

To evaluate the specific ligand density in different formulations quantitatively, we constructed a series of stable SynVACs and elastic beads that differ only in their amount of flourescein isothiocyanate (FITC)-labelled anti-CD3 conjugation as we varied the amount of the tetrazine-antibodies used in the formulation. The different level of conjugated antibody density was first visualized by immunofluorescent images as shown in Supplementary Fig. 10. To quantify the number of antibodies on each bead, we used flow cytometry analysis based on the calibration beads with known antibody densities (Fig. 2f and Supplementary Fig. 11). The antibody density or in other means the ligand spacing has been proven to be critical for T-cell activation as it directly affects the interactions between TCRs and their corresponding ligands on APCs^43,44. Our data showed that an antibody density higher than 10^5.9 on SynVACs or elastic beads ensures efficient activation of mouse primary T cells (Supplementary Fig. 12) and human Jurkat cells (Fig. 2g), as evidenced by the flow cytometry analysis of CD69, an early activation marker. This result aligns with previously published data indicating that a ligand spacing of less than 50 nm greatly enhances T-cell activation on 2D surface⁴⁵. We also included Dynabeads as a direct benchmark in our experiments. We estimated the antibody density on Dynabeads using the same method. Our findings indicate the presence of two distinct antibody types on the Dynabeads’ surface, with a molar ratio close to 3:1 (Supplementary Fig. 13), and the total antibody count on the Dynabeads’ surface is approximately 278,958 ± 2,790 (≈10^5.4). As we used a 1:1 cell-to-bead ratio in all conditions, this value is equivalent to the dose of stimulatory antibodies presented to the cells (dose of antibody per cell). With a diameter of 4.5 μm for the Dynabeads, the antibody spacing is calculated to be around 15.9 ± 0.1 nm, similar to our beads (Supplementary Table 1). To further substantiate our results, we assessed early T-cell activation using SynVACs (V1), elastic beads (E1) and Dynabeads in parallel experiments with Jurkat NFAT-zsGreen reporter cells (Supplementary Fig. 14). The activation levels recorded at the 18 h interval are in agreement with our flow cytometry analysis, reinforcing the validity of our approach and reflecting consistency with previously published data²⁴. Therefore, we chose a ligand density of 10^5.9 for all bead formulations in the subsequent studies. It is worth noting that our initial data showed the 2D gel system led to lower T-cell activation rates, as depicted in Supplementary Fig. 15. As a result, we focused on the bead-based system for suspension culture of T cells, a decision that is consistent with previous studies indicating less than optimal T-cell expansion with 2D gel systems⁴⁶.

To further characterize the spatial distribution of antibody conjugation, we performed fluorescence staining and confocal microscopy to examine the conjugated antibodies in individual microbeads. This analysis revealed a distinct ~100 nm fluorescence layer on the bead surface (Supplementary Fig. 16). In addition, we used X-ray photoelectron spectroscopy (XPS) to analyse the bead surface composition. XPS results showed nitrogen presence, confirming antibody surface modification. Quantitatively, an average of 5.3% of alginate on the elastic bead surface and 6.4% on the SynVACs surface were modified, as determined by the amino acid to alginate monomer ratio (Supplementary Fig. 17).

Polyclonal expansion of primary human and mouse T cells

We then used SynVACs, elastic beads and Dynabeads (as a commercial benchmark) for the polyclonal expansion of primary human T cells in peripheral blood mononuclear cells (PBMCs) (Fig. 3a). Four types of SynVAC and elastic bead with distinct mechanical properties were fabricated by conjugating anti-CD3 and anti-CD28 in a 1:1 molar ratio, as detailed in Supplementary Table 1. Dynabeads were used as a benchmark control. Physical interactions between the beads and T cells were examined by using SEM (Fig. 3b). After co-culturing SynVACs and human T cells for 24 h, we observed the formation of the immunologic synapse-like structure (Fig. 3b), which was confirmed by immunofluorescent staining for CD3 epsilon on day 3 (Fig. 3c) and as early as at 1 h (Supplementary Fig. 18). Through quantitative analysis, it was found that SynVACs increased both the number and size of CD3 clusters in human T cells (Fig. 3d,e). The relative size variations of T cells in Fig. 3b,c are primarily due to the inherent growth of T cells upon activation, typically expanding to 12–15 μm by day 3, and potential dehydration-induced cell shrinkage during SEM sample preparation.

Fig. 3 |. Polyclonal expansion of primary human T cells by SynVACs results in a higher TMSC population.

a, Experimental timeline of human primary T cells activated by Dynabeads compared with SynVACs. b, Left: SEM image of interaction between T cells and Dynabeads on day 1. Right: SEM image of interaction between T cells and SynVACs on day 1. Scale bar, 2 μm. c, Immunofluorescence staining of CD3ε, beta-actin and nuclei (4,6-diamidino-2-phenylindole) in primary human T cells after 3 days of culture with Dynabeads, elastic beads and SynVACs. Scale bar, 5 μm. d, Quantification of CD3ε cluster number in primary human T cells after 3 days of culture with Dynabeads, elastic beads and SynVACs. e, Quantification of CD3ε cluster size in primary human T cells (n = 50). f, Fold expansion of human T cells (day 0 = 1) (n = 3). g, PD-1⁺ and TIM3⁺ double-positive T cells of expanded PBMCs being cultured with Dynabeads, elastic beads (E1, E2) and SynVACs (V1, V2), respectively, at day 10 and day 14, as determined by flow cytometry analysis (n = 3). h–k) Flow cytometry analysis of the total number of CD8⁺ T cells (h), CD8⁺ CCR7⁺ CD45RO⁻ CD95^hi TMSCs (i), CD4⁺ T cells (j) and CD4⁺ CD25⁺ FOXP3⁺ T_reg cells (k) in expanded T cells being cultured with Dynabeads, elastic beads and SynVACs, respectively, at day 10 and day 14 (n = 3). The ‘Bare beads’ group was treated with viscoelastic microbeads lacking antibody conjugation, all within an IL-2 enriched medium. By contrast, the other control group comprised T cells solely, which were similarly supplemented with IL-2 but without the inclusion of any beads. In f–k, the data points denote the mean values, and the error bars denote the standard deviations of three biological replicates. Significance was determined by one-way ANOVA and Tukey’s multiple comparison test.

Following these examinations of T cell–beads interactions at early time points, we performed long-term culture for up to 2 weeks, which is in line with standard clinical protocols for CAR T-cell activation and proliferation, and investigated the impacts of SynVAC, elastic beads and Dynabeads on T-cell expansion. It is worth noting that SynVACs resulted in a T-cell expansion rate that was more than fourfold higher than that of the elastic beads at day 14 and 1.5-fold higher than that of Dynabeads (Fig. 3f). The total T-cell counts on days 10 and 14 relative to the initial T-cell seeding number (100,000) are presented in Supplementary Fig. 19. In addition, there was a statistically significant increase in the T-cell expansion rate when using lower-stiffness SynVACs (V1) compared with higher-stiffness SynVACs (V2). We also noted that the proliferation activated by the elastic bead group was less pronounced compared with that by Dynabeads as Dynabeads have been optimized for antibody density and the bead size. When compared with Dynabeads, there are reports on notable or modest improvement by various artificial APC (aAPC) systems^22,47,48. It is possible that, in addition to differences in aAPC, variations in T-cell sources, activation states and culture conditions may account for the differing effects. For example, even when using the same aAPC, the method of interleukin-2 (IL-2) delivery—whether added directly to the solution or released from the aAPC—can substantially impact T-cell proliferation^47,48. In addition, while rapid proliferation is often seen as a positive outcome, it is essential to consider the biological relevance and the quality of the T-cell response. Regarding exhaustion markers PD-1 and LAG-3, no substantial differences were detected among the various conditions after 14 days of culture (Fig. 3g). Remarkably, while Dynabeads prominently promoted CD4-biased skewing (Supplementary Figs. 20 and 21), consistent with a previous report⁴⁷, all SynVACs formulations induced rapid and substantial CD8-biased skewing, with V1 promoting the highest number of CD8⁺ T cells by day 14 (Fig. 3h), which may directly target and eliminate tumour cells, making them more desirable for cancer immunotherapy^49,50. Nevertheless, we recognize the importance of CD4⁺ T cells in supporting this response and sustaining immunological memory, vital for a comprehensive antitumour immune strategy⁵¹. Future studies may explore the synergistic effects of a balanced CD8⁺/CD4⁺ T-cell response to enhance the efficacy and durability of antitumour immunity. Furthermore, the SynVAC formulations (V1) resulted in a sixfold higher proportion of CD8⁺ CCR7⁺ CD45RO⁻ CD95^hi TMSCs compared with Dynabeads and elastic beads (Fig. 3i and Supplementary Fig. 22), based on the analysis by fluorescence-activated cell sorting (FACS) (Supplementary Fig. 23). These TMSCs were also positive for other markers such as CXCR3, CD58 and CD11a (Supplementary Fig. 24). The gating strategy for the related FACS data is depicted in Supplementary Fig. 25. This finding is substantial as TMSCs are known to possess a remarkable capacity for self-renewal, differentiation into effector cells and long-term persistence, where the persistence and functional capacity of the infused T cells are crucial for therapeutic success. Conversely, SynVACs (V1) produced fewer CD4⁺ T cells by day 14 compared with Dynabeads (Fig. 3j and Supplementary Fig. 26a). Consequently, SynVACs (V1) yielded a significantly lower number of CD4⁺ CD25⁺ FOXP3⁺ regulatory T (T_reg) cells compared with Dynabeads (Fig. 3k and Supplementary Fig. 26b). A reduced proportion of T_reg cells in the activated T-cell population may result in enhanced antitumour immune responses, as T_reg cells are known to suppress the activity of other immune cells⁵². In addition to days 10 and 14, early assessments on day 7 revealed phenotypic alterations in T cells and superior initial efficacy of SynVACs over elastic beads (Supplementary Fig. 27), with an increase in TMSCs and CD8⁺ cells, alongside a reduction in T_reg cells, compared with Dynabeads. These trends are consistent with the observations on days 10 and 14, although the effects are less pronounced.

Furthermore, our analysis shows that T cells retain CD25 expression when activated by SynVACs, elastic beads and Dynabeads, affirming sustained activation over a 2 week period of cell expansion (Supplementary Fig. 28). In addition, our data show no significant differences in T-cell expansion, CD4/CD8 ratio and CD3⁺ CCR7⁺ CD45RO⁻ CD95^hi TMSCs when using human PBMCs or PBMC-derived T cells (isolated by CD3⁺) as the starting cell source for T-cell expansion (Supplementary Fig. 29). These results justify our choice of PBMCs as a viable starting material for CAR T-cell production, aligning with established practices in both preclinical and clinical research.

Similarly, for the polyclonal activation of primary mouse T cells, we used SynVACs (V1), elastic beads (E1) and Dynabeads. We recorded time-lapse supplementary videos to observe the physical interactions between T cells and SynVACs/Dynabeads (Supplementary Videos 2 and 3). It is worth noting that on day 7, cells in the Dynabeads group formed large aggregates, while SynVACs led to relatively loose, sheet-like structures (Supplementary Fig. 30). This observation suggests that SynVACs may reduce the risk of overstimulation and promote better cell–cell interactions and nutrient exchange, potentially facilitating sustained T-cell activation and expansion. In addition, clusters in the Dynabead group dissipated after day 7, whereas the majority of clusters in the SynVACs groups persisted until day 10. This phenomenon further highlights the potential advantages of SynVACs in T-cell activation and expansion, as they appear to create a more conducive environment for T-cell interactions and persistence. FACS data analysis showed consistent effects of SynVACs on human primary T cells with respect to higher fold expansion, no significant exhaustion, CD8-biased skewing, more CD44⁻CD62L⁺CD95⁺ TMSCs formation, lower T_reg-cell formation and in SynVAC co-culture (Supplementary Fig. 31).

In addressing concerns regarding the role of calcium ions (Ca²⁺) from alginate gels in T-cell activation and differentiation, we conducted a thorough evaluation and show that the calcium released from the alginate beads has a negligible effect on T-cell activation and differentiation. First, SynVACs undergo rigorous washing and dialysis processes and contain minimal free calcium ions. Second, we typically use 500,000 beads in a 24-well plate with 500 ml media to activate an equivalent number of PBMCs, and the calculations suggest a less than 0.5% change of calcium concentration in the culture media. Third, we measured calcium concentration and did not find a significant increase in calcium ion concentration in the culture media (Supplementary Fig. 32). Consequently, we believe that the viscoelastic properties of the alginate, rather than calcium release, are the predominant factors influencing the observed T-cell phenotypes.

In vitro CAR transduction into T cells and tumour-cell-killing activity

Next, we assessed the performance of SynVACs in CAR T-cell generation and in vitro tumour-cell-killing assays. These evaluations are critical for determining the efficiency and effectiveness of SynVACs in generating CAR T cells with high transduction efficiency and potent tumour-killing capabilities. FACS data analysis revealed that SynVAC-activated T cells showed a remarkably higher CAR19 transduction efficiency of up to ~90% at day 6, whereas Dynabeads and elastic beads achieved only 42% and 17%, respectively (Fig. 4a,b), consistent with the average level of CAR transduction in previous reports (20–32%)^10,53. Our further studies with multiple donors confirm the reproducibility of the improved CAR transduction efficiency (Supplementary Fig. 33). Staining with carboxyfluorescein diacetate succinimidyl ester, followed by flow cytometry analysis, showed that the highest proliferation rate was in T cells activated by SynVACs on day 3 post-CAR transduction (Supplementary Fig. 34), which might contribute to the significantly enhanced transduction efficiency.

Fig. 4 |. SynVACs enhance the transduction efficiency and tumour-killing capacity of CAR T cells.

a, CAR expression levels in CAR19 T cells on day 6 post activation by Dynabeads, elastic beads and SynVACs. b, Comparison of CAR transduction efficiency between Dynabeads and SynVACs (n = 5). c, Evaluation of in vitro tumour-killing capability of MCAR T cells against OVCAR3 and OVCAR8 tumour cells, and CAR19 T cells against Nalm6 and Raji tumour cells. CAR T cells were expanded for 13 days using Dynabeads or SynVACs and then co-cultured at various effector-to-target cell ratios (E/T) (n = 3). d, IFNγ secretion was measured via an ELISA assay 24 h post culture with tumour cells (n = 3). e, Phenotypic analysis of tumour-killing efficiency in CAR T cells 24 h post co-culture with tumour cells. f, Quantitative analysis of flow cytometry results (n = 3). The percentage is determined by isotype antibody staining. In b–d and f, the data points denote the mean values, and the error bars denote the standard deviations of five or three biological replicates. In b and f, significance was determined by a two-tailed, unpaired t-test. In d, significance was determined by one-way ANOVA followed by Tukey’s multiple comparison test.

In vitro tumour-cell-killing assays demonstrated that SynVAC-activated CAR T cells showed potent cytotoxic effects against various cancer cell lines, including OVCAR3, OVCAR8, Nalm6 and Raji cells, suggesting that SynVACs can enhance the functionality and specificity of CAR T cells (Fig. 4c). Specifically, the SynVAC-expanded CAR T cells, including CAR19 cells and MSLN-specific CAR-T (MCAR-T) cells, show a more pronounced reduction in live tumour cells across all effector-to-target cell ratios (E/T) compared with the CAR T cells expanded by Dynabeads or elastic beads. This suggests that the SynVAC-expanded CAR T cells possess superior cytotoxic capabilities in vitro. Enzyme-linked immunosorbent assay (ELISA) revealed that SynVAC-activated CAR T cells showed the highest levels of interferon-γ (IFNγ) secretion, while no significant differences were observed between Dynabeads and elastic bead-activated CAR T cells (Fig. 4d). This finding suggests that SynVACs enhance the functional capacity of CAR T cells by promoting the secretion of IFNγ, a critical cytokine involved in antitumour immune responses. Consistent results were observed when using primary mouse T cells where SynVACs V1 promoted the formation of IFNγ⁺ TNF (tissue necrosis factor)⁺ CD8⁺ T cells at day 7 (Supplementary Fig. 35). FACS analysis further demonstrated that post tumour stimulation, SynVAC-activated CAR T cells showed significantly higher expression of proteins such as CD69, perforin and granzyme B, compared with Dynabead-activated CAR T cells (Fig. 4e,f). CD69 is an early activation marker of T cells, while perforin and granzyme B are critical components of the cytotoxic granules responsible for inducing target-cell apoptosis. The enhanced production of these proteins in SynVAC-activated CAR T cells indicates a more potent cytotoxic potential and a stronger activation status compared with their Dynabead-activated counterparts. Taken together, these results reinforce the idea that modulating the mechanical properties, especially the viscoelasticity of synthetic APCs, such as SynVACs, can influence the phenotypic and functional properties of the generated CAR T cells, which could be strategically used for optimizing the CAR T-cell manufacturing process.

Antigen-specific enrichment of MCAR T cells using MSLN/anti-CD28 conjugated SynVACs

We also evaluated the performance of SynVACs in antigen-specific activation of T cells. This assessment is crucial for understanding how efficiently SynVACs can specifically activate T cells in response to the presence of target antigens, which is a key factor in determining the overall effectiveness of immunotherapies. We used MSLN and anti-CD28 conjugated SynVACs to activate MCAR T cells. MSLN is a cell-surface glycoprotein overexpressed in various solid tumours, making it an attractive target for CAR T-cell therapy⁵⁴. The anti-CD28 molecule serves as a co-stimulatory signal, enhancing T-cell activation and expansion. The ligand density of MSLN/anti-CD28 on SynVACs was measured using XPS (Supplementary Fig. 36). On day 5, we found that SynVACs and aAPCs (a cell line overexpressing human CD83/CD86/4–1BBL co-stimulatory receptors and human MSLN) significantly increased the proportion of MCAR T cells from 74.3% to 95.2% and 94.4%, respectively, whereas Dynabeads surprisingly led to a reduced proportion (that is, 74.3% to 63.0%) (Supplementary Fig. 37a). Further investigation revealed that, although aAPCs showed equal capacity as SynVACs in enriching MCAR T cells, they were unable to promote CD8-biased skewing like SynVACs, as evidenced by the CD4/CD8 ratio (Supplementary Fig. 37b). These findings emphasize the advantages of SynVACs in promoting not only the expansion of MCAR T cells but also their preferential CD8-biased skewing, which is crucial for enhancing the cytotoxic potential of CAR T cells in cancer immunotherapy. Moreover, antigen-specific activation using SynVACs further highlights the versatility and adaptability of the SynVAC technology for personalized immunotherapies of various cancer types.

Single-cell RNA sequencing analysis of expanded T cells

To further elucidate the differences in gene expression and subpopulations of the CAR T cells being activated by SynVACs (leading formulation) and Dynabeads (as a gold standard control), respectively, we conducted single-cell RNA sequencing (scRNAseq) on day 14 before the expanded cells were used for in vivo studies (Fig. 5a). These included 9,400 T cells activated by Dynabeads and 7,874 T cells activated by SynVACs. Bioinformatic analysis of gene expression mapped in two dimensions via uniform manifold approximation and projection (UMAP) identified 11 clusters (Fig. 5b). Primarily, 7 clusters comprised CD8⁺ T cells, 3 clusters were CD4⁺ T cells, and 1 cluster consisted of double-negative T cells. Figure 5c visualizes the changes in the percentages of the 11 distinct clusters in pie charts. Each cluster’s subtype was then identified based on protein markers and gene signatures (Fig. 5d).

Fig. 5 |. scRNAseq analysis reveals distinct activation patterns and gene expression profiles in CAR T cells activated by SynVACs and Dynabeads.

a, Experimental workflow showing how and when CAR T cells were collected for scRNAseq. b, UMAP plot showing the distribution of CAR T cells activated by Dynabeads or SynVACs with 11 distinct clusters identified at day 14. c, Pie chart of the 11 distinct clusters in SynVACs group showing more CD8⁺ TMSCs (1), less CD4⁺ Th 17 cells (9), less CD8⁺ terminally differentiated effector memory T cells (4) and less T_reg cells (10) compared with the Dynabeads group. d, Gene expression signatures and specific markers used for identifying 11 distinct clusters. e, Individual cells in the UMAP embedding coloured by expression of CD8A, CD4, CCR7 and IFNG. f, Violin plots of overall expressed genes reveal distinct expressions of CCR7 and IFNG genes showing enhanced stemness and tumour-killing efficiency of SynVAC-activated CAR T cells. In f, significance was determined by a z-test.

To explore the CCR7⁺ memory T-cell subpopulations, we focused on clusters 1 and 8, which showed a gene-expression profile of TMSCs, featuring high levels of LEF1, TCF7, CCR7, SELL and IL7R. Consistent with our findings in Figs. 3 and 4, SynVACs increased the percentage of CD8⁺ TMSCs from 6.8% to 14.4% (cluster 1) and CD8⁺ effector memory T cells from 26.2% to 45.1% (clusters 3 and 5) but decreased the percentage of CD8⁺ terminally differentiated effector memory T cells from 14.9% to 4.2% (Supplementary Table 2). It was noted that the expression levels of CD95 and CD28 were relatively low (Supplementary Fig. 38) in cluster 1 (CD8⁺ TMSCs); if we only consider CD95⁺ cells in this cluster, SynVAC also significantly increased this population (4.9%) in comparison to Dynabeads (2.2%). However, SynVACs decreased CD4⁺ T helper 17 cells from 18.0% to 4.3% (cluster 9) and CD4⁺ regulatory T cells from 9.5% to 1.0% (cluster 10), although SynVACs increased CD4⁺ TMSCs from 2.0% to 3.5% (Supplementary Table 2). Cluster 4 is characterized as terminally differentiated due to its higher expression of exhaustion markers such as TIGIT and effector genes such as NKG7, GNLY and GZMB, coupled with lower levels of LEF1 and stemness-associated markers, aligning with the terminally differentiated effector memory T cells profile. Cluster 7 is labelled as exhausted because it presents high CXCR6 expression, indicative of tissue residency, and elevated TIGIT levels, alongside a reduced expression of stemness and effector genes, consistent with an exhausted phenotype. Taken together, scRNAseq analysis provides more comprehensive information on the subpopulations of expanded T cells and confirms the beneficial effects of SynVACs on the increase of T cell stemness and CD8⁺ subpopulations.

Another noteworthy observation was that the SynVAC-activated CAR T cells showed higher expression levels of CD8A, CCR7 and IFNG, compared with Dynabead-activated cells (Fig. 5e,f and Supplementary Fig. 39). CD8A gene is associated with cytotoxic T cells, which are instrumental in eradicating cancer cells. The CCR7 gene is an important marker for stem-like T cells, and its high expression is associated with improved therapeutic potential in CAR T-cell therapy. The elevated gene expression of IFNG in SynVAC-activated CAR T cells indicates a potentially more potent antitumour immune response, as IFNG enhances the ability of the immune system to detect and eradicate cancer cells. The enhanced gene expression of CD8A, CCR7 and IFNG in the SynVACs group suggests amplified antitumour responses and sustained immune protection.

Exploring the heterogeneity between SynVACs and Dynabeads populations, we conducted gene enrichment analysis using the EnrichR 3.1 package, which further revealed that SynVAC-activated CAR T cells showed an upregulation of several critical biological processes and pathways, such as DNA metabolic processes, DNA replication and transcription regulation (Supplementary Fig. 40). These findings suggest that SynVACs-mediated activation may promote CAR T-cell proliferation and expansion, which is essential for robust and lasting antitumour responses. The upregulation of DNA metabolic processes and DNA replication implies that the SynVAC-activated CAR T cells undergo more active cell division and growth compared with those activated by Dynabeads. This increased proliferation capacity may enhance CAR transduction rates and tumour-killing efficacy observed in the SynVACs group. The RNA sequencing data provided is indeed insightful, yet it is understood that the true value lies in perturbing the identified signals to ascertain their functional roles. Moving forward, we plan to conduct targeted perturbation studies to unravel the functional implications of these genetic variations, with the goal of deepening our grasp of T-cell programming and advancing the efficacy of CAR T-cell therapies.

In vivo antitumour efficacy and long-term persistence of CAR T cells in a xenograft B-cell lymphoma model

Subsequently, we assessed the in vivo antitumour efficacy of SynVAC-activated CAR19 T cells in a human lymphoma Raji xenograft mouse model (Fig. 6a). This evaluation is essential to determine the therapeutic potential of SynVAC-activated CAR T cells in a physiologically relevant setting and to provide valuable insights into their ability to target and eliminate cancer cells in a living organism. We observed that in the untreated control group, all mice succumbed to the disease by day 28 (Fig. 6b). By contrast, SynVAC-activated CAR19 T cells effectively eliminated cancer cells in all of the treated mice by 28 days post injection. The Dynabeads group, then again, showed a significant cancer recurrence (Fig. 6c). In addition, survival curves were included to demonstrate the long-term outcomes following treatment (Supplementary Fig. 41). Statistical analysis revealed a significant difference in tumour burden between the Dynabeads and SynVACs treatment groups (Supplementary Fig. 42). Furthermore, on day 40, we examined the persistence of CAR19 T cells in various tissues of the mice and found a marked increase in the SynVACs group across multiple organs and tissues, such as blood, spleen and liver (Fig. 6d). Upon further investigation of the blood, we observed that the phenotype of T cells in the SynVACs group showed significantly elevated levels of CD8-biased skewing, CAR expression and CD62L expression (Fig. 6e,f). The elevated levels of CD62L observed in CAR19 T cells activated by SynVACs indicate a greater presence of T-cell subpopulations such as naive T cells, TMSCs and central memory T cells. This observation has significant implications for the potential effectiveness and durability of CAR T-cell therapy. Additional experiments were conducted to evaluate the performance of CAR T cells activated by elastic beads, as this group is pivotal when asserting the effects of altered viscoelasticity. The CAR T cells activated by elastic beads showed a pattern of cancer recurrence comparable to that of the Dynabeads group (Supplementary Fig. 43). These findings underscore the superior efficacy of SynVAC-activated CAR19 T cells in not only eliminating cancer cells but also maintaining long-term persistence in vivo. The enhanced persistence and phenotype characteristics of CAR19 T cells in the SynVACs group suggest robust T-cell function, which may contribute to better cancer control and reduced chances of recurrence.

Fig. 6 |. In vivo efficacy of SynVAC-activated CAR19 T cells in a human lymphoma Raji xenograft mouse model.

a, Experimental workflow showing SynVAC-activated CAR19 T-cell generation and therapy on a human lymphoma Raji xenograft mouse model, where 1 × 10⁶ Raji cells were intravenously injected into NSG mice on day 0 and 3 × 10⁶ CAR19 T cells were intravenously injected on day 4. BLI was used to monitor tumour growth. b, BLI of NSG mice inoculated with luciferase Raji cells and then either mock treated (Vehicle) or treated with CAR19 T cells activated by SynVACs or Dynabeads at various time points (n = 5). c, Quantification of bioluminescence signal (n = 5). TBL, total bioluminescent signal; p/s, photons per second. d, CD3⁺, CD45⁺ double positive cells in different organs at day 40 post injection of CAR T cells activated by Dynabeads/SynVACs, as determined by FACs analysis. e, Characterization of CD4/CD8 ratio, CAR expression and CD62L expression in blood. f, Quantification of the flow cytometry results in e (n = 5). In f, the data points denote the mean values, and the error bars denote the standard deviations of five mice. Significance was determined by a two-tailed, unpaired t-test.

In vivo antitumour efficacy and long-term persistence of MCAR T cells in a xenograft ovarian cancer model

Finally, we assessed the in vivo antitumour efficacy of SynVAC-activated MCAR T cells in a human ovarian cancer xenograft mouse model (Fig. 7a). While we have used lymphoma animal models, their limitations in fully representing solid tumours necessitate a further investigation of SynVACs’ effectiveness in solid tumour contexts for a more comprehensive understanding of their therapeutic potential. By day 33, untreated control mice had all developed ovarian solid tumours, while CAR T cells activated by Dynabeads only slightly decreased the tumour growth trend. By contrast, we observed that although SynVACs were unable to completely eradicate cancer cells, they significantly reduced tumour cell recurrence and even had a probability of completely preventing tumour cell recurrence in one of the eight mice (Fig. 7b,c). A significant difference in tumour burden between the Dynabeads and SynVACs treatment groups was evident upon statistical analysis (Supplementary Fig. 44). In addition, on day 40, we examined the metastasis of ovarian tumour cells in different tissues and organs of mice. Compared with the vehicle control group and the Dynabead-activated CAR T-cell treatment group, the SynVAC group effectively prevented tumour cell proliferation and metastasis, particularly in the lungs, bone tissue and brain tissue (Fig. 7d,e). Specifically, we have focused on the infiltration and phenotypic characterization of CAR T cells in the peritoneal ascites of experimental mice, which also reflects their presence at the ovarian cancer tumour sites. Our findings reveal a notable increase in T-cell retention in the SynVACs group, rising from 1.75% to 7.06%, in comparison to the Dynabeads group. Moreover, the proportion of CAR T cells within the tumour notably increased from 27.3% to 61.6%. In addition, we observed a significant rise in the percentage of CD62L⁺ CAR T cells within the tumour, escalating from 23.4% to 43.3% (Fig. 7f,g). These findings reveal that CAR T cells activated by SynVACs not only inhibit tumour growth and metastasis but also promote the in vivo longevity of CAR T cells, increase the proportion of CAR-expressing cells and elevate the population of CD62L⁺ T cells such as naive T cells, TMSCs and central memory T cells, especially in solid tumour environments such as ovarian cancer. In summary, SynVACs show great potential for enhancing the effectiveness of CAR T-cell therapy against solid tumours, leading to better cancer control and reduced chances of recurrence.

Fig. 7 |. In vivo efficacy of SynVAC-activated MCAR T cells in a human ovarian solid tumour xenograft mouse model.

a, Experimental workflow illustrating SynVAC-activated MCAR T-cell generation and therapy in a human ovarian solid tumour xenograft mouse model, where 1 × 10⁶ OVCAR8 cells were intraperitoneally injected into NSG mice on day 0 and 3 × 10⁶ MCAR T cells were intravenously injected on day 4. BLI is used to monitor tumour growth. b, BLI of NSG mice inoculated with luciferized OVCAR8 cells and then either mock treated (Vehicle) or treated with MCAR T cells activated by SynVACs or Dynabeads at various time points after treatment. c, Quantification of bioluminescence signal. d, BLI of ovarian cancer metastases in various organs on day 40. GI, gastrointestinal. e, Quantification of bioluminescence signal in different organs (n = 8). f, Phenotype and CAR expression level of MCAR T cells collected from peritoneal ascites of the experimental mice, as determined by flow cytometry. g, Quantification of the flow cytometry data (n = 8). In e and g, the data points denote the mean values, and the error bars denote the standard deviations of eight mice. In e, significance was determined by one-way ANOVA and Tukey’s multiple comparison test. In g, significance was determined by a two-tailed, unpaired t-test.

Discussion

We developed a microfluidic method for generating SynVACs that mimic the mechanical properties of APCs and presents activation signals for T-cell engineering. Our synthetic APCs integrate both chemical and mechanical programmability to mimic the properties of natural APCs. This endeavour not only exemplifies the forefront of biomimetic engineering but also offers potential advancements in immunotherapy by closely mirroring the complex functionalities of natural APCs. First, microfluidic technology enables precise control over the size and shape of the microspheres, resulting in highly uniform particle populations. This uniformity is crucial for ensuring consistent and reproducible interactions with T cells. Second, microfluidic techniques facilitate the rapid and efficient production of microspheres, potentially enabling large-scale manufacturing of SynVACs for widespread application. This scalability is essential for translating our research findings into practical, real-world solutions. Third, thanks to their mechanical properties, microbead-based activation systems more closely replicate the physiological interactions between T cells and APCs, thereby facilitating a more natural engagement of T cells. Most importantly, SynVACs that mimic the mechanical properties of APCs can facilitate biomimetic T cell–SynVAC interactions, leading to more effective T-cell activation, CD8⁺ T cell-biased skewing, higher CAR transduction efficiency, enhanced CAR T cell stemness and more robust and durable immune responses against cancer cells, potentially improving remission durations and reducing the risk of cancer recurrence. Although SynVAC and Dynabeads result in similar T-cell expansion rates overall, differences in the expanded T-cell subpopulations may account for the different therapeutic effects in vivo. The underlying molecular mechanisms of SynVAC’s effects on T-cell activation and expansion ex vivo remain to be investigated. It is likely that the viscoelastic properties of SynVAC alter the kinetic activation of TCRs and thus affect T-cell expansion.

We found that SynVACs significantly enhance CAR transduction efficiency compared with Dynabeads, which has been further supported by in vivo experiments. The phenotypic analysis and CAR expression levels of CAR19 T cells collected from the blood reveal a notably high proportion of CAR-expressing cells. A higher transduction efficiency ensures that a larger proportion of T cells express the desired CAR or transgene, which can lead to more potent and effective therapeutic responses against target cells, such as tumour cells⁵⁵. In addition, with higher transduction efficiencies, a smaller number of starting T cells may be required to generate a therapeutic dose of CAR T cells, potentially preserving more of the patient’s healthy T-cell population and minimizing unwanted cytokine storms.

Moreover, we have discovered that T cells activated by SynVACs show a significantly higher tumour-killing efficiency compared with those activated by Dynabeads, evidenced by elevated expression levels of CD69, perforin, granzyme B and IFNγ. Elevated CD69 expression suggests that SynVACs may promote a more efficient and rapid activation of T cells compared with conventional methods. Increased expression of perforin and granzyme B molecules in SynVAC-activated T cells highlights their heightened cytotoxic potential and their ability to effectively eliminate target cells, such as cancer cells. Elevated IFNγ expression in SynVAC-activated T cells underscores their enhanced functionality and capacity to mount robust antitumour immune responses. Taken together, these findings show that SynVACs not only significantly enhance T-cell activation and expansion but also promote the generation of highly functional and cytotoxic T cells endowed with potent antitumour capabilities. This outcome is consistent with the observed CD8⁺ cell-biased skewing induced by SynVACs.

A recent study demonstrated that culturing CD8⁺ T cells in slow-relaxing (1,000–10,000 s at 60% stress relaxation) viscoelastic collagen gel for 3 days before or after co-culture with Dynabeads enhances the tumour-killing activity of T cells, while fast-relaxing gel increases long-term memory genes⁵⁶. It is worth noting that this study explored the effects of a viscoelastic matrix with a different experimental system from ours, leading to some different findings. First, their primary objective was to investigate the mechanical properties of a 3D collagen matrix on T cells, which is more relevant to in vivo conditions when T cells reside in collagen-rich tissues. T cells were in the collagen matrix for a 3 day period, and then Dynabeads were used for T-cell expansion. By contrast, our approach is to develop viscoelastic artificial cells to replace Dynabeads, enabling the activation and expansion of T cells for CAR T-cell production ex vivo. Second, we use non-adherent alginate microbeads for suspension culture, exclusively presenting T-cell activation signals. By contrast, the 3D collagen matrix involves potential cell–matrix interactions. Third, the stiffness of the collagen gel was below 1 kPa, in contrast to our 10 kPa microspheres. Fourth, the previous study indicates that a slow-relaxing gel is more effective in inducing tumour-cell killing, while a fast-relaxing gel increases the expression of long-term memory genes; by contrast, our approach demonstrates that a fast-relaxing gel is more effective in activating T cells for both tumour-killing activity and the formation of TMSCs. Finally, we focus on investigating the activation and differentiation of naive T cells, rather than CD8⁺ cells as in the previous study. Another recent article presents the development of the multifunctional scaffold, an alginate-based technology that accelerates in vivo CAR T-cell manufacturing to a single day, enhancing their persistence and efficacy against distal tumours in mice²⁴. Although alginate is used, the study does not explore the impact of viscoelasticity on T-cell activation. Therefore, our objectives, experimental systems and findings are distinct from those presented in the recent publications, offering novel insights into the mechanochemical effects of viscoelastic artificial cells.

Stem-like T cells are a unique subset of T cells that possess properties of both stem cells and immune cells, including TMSCs, naive T cells and central memory T cells. These cells can self-renew and differentiate into various effector and memory T-cell subsets^12,57. In this study, we demonstrated that CAR T cells activated by SynVACs showed a higher proportion of cells expressing high levels of CCR7 across the entire RNA transcriptome. This observation is particularly substantial because cells expressing high levels of CCR7 are associated with central memory T cells and TMSCs, which are known to show increased self-renewal capacity, long-term persistence and improved antitumour efficacy. The expansion of these CCR7-expressing subpopulations suggests that SynVACs might promote the generation of more potent CAR T cells with a greater ability to control tumour growth and achieve durable responses in cancer immunotherapy.

The observation that SynVACs greatly enhance the long-term persistence of CAR T cells in vivo in a B lymphoma xenograft model has substantial implications for adoptive cell therapies. Persistence is an essential factor influencing the therapeutic efficacy of CAR T cells, as it determines the duration of their antitumour activity within the host. The long-term persistence of CAR T cells allows for the continuous recognition and elimination of tumour cells, leading to a more effective and durable treatment response. This finding underscores the potential advantages of using SynVACs for CAR T-cell activation and expansion, as it may ultimately lead to substantially enhanced clinical outcomes in patients receiving CAR T-cell therapy.

In the ovarian solid tumour xenograft models, we observed a remarkable reduction in tumour size when treated with SynVAC-activated CAR T cells compared with both Dynabead-activated CAR T cells and the vehicle control group. Furthermore, we found a significant decrease in metastatic cancer cells within vital organs such as the lung, pancreas and uterus in the SynVAC-treated group. These findings emphasize the superior antitumour efficacy of SynVAC-activated CAR T cells in combating not only primary tumours but also metastatic cancer cells that have spread to distant organs. This enhanced therapeutic performance may be attributed to the significantly enhanced expansion, persistence and function of the SynVAC-activated T cells. If these cells can effectively infiltrate and target tumour cells, this may ultimately lead to better tumour control and potentially prolong patient survival.

In summary, we have shown the remarkable potential of SynVACs as a tool for T-cell activation and expansion. We hypothesize that the viscoelastic properties of SynVACs may more closely mimic the natural mechanical cues that T cells encounter within lymphoid tissues, potentially leading to more physiologically relevant activation and proliferation responses. This hypothesis warrants further investigation. However, existing evidence indicates that T cells can sense and respond to the mechanical properties of their microenvironment^58,59. Our findings highlight the advantages of SynVACs over traditional Dynabeads, paving the way for more effective and durable CAR T-cell therapies, and have substantial implications for the development of next-generation adoptive cell therapies.

Methods

High-throughput microfluidic method for fabrication of alginate viscoelastic microbeads

The microfluidic device was designed using AutoCAD 2022, followed by fabrication via conventional photolithography technique. Initially, photosensitive epoxy (SU-8 2015, MicroChem) was spun onto a 4-inch silicon wafer to a thickness of 13 μm. This wafer underwent a soft bake at 95 °C for 6 min and ultraviolet exposure through a chrome mask bearing the desired channel patterns. Following a 10 min development phase in SU-8 developer (MicroChem) and a rinse with isopropyl alcohol, the master mould took shape on the wafer. Polydimethylsiloxane pre-polymer and curing agent were mixed, poured onto the silicon substrate and cured at 65 °C for 3 h to produce a polydimethylsiloxane slab. This slab was then subjected to punch-outs to create inlet and outlet points, followed by a rinse with 50% ethanol to clean the channel and an oxygen plasma treatment using Plasma Prep II (SPI Supplies) to prepare for bonding. The assembled devices were placed in a 65 °C oven for roughly 30 min to strengthen bonding and left at room temperature for 24 h to ensure optimal hydrophobicity before experiments. A custom ‘torch-like’ component was designed and 3D-printed using an Elegoo 4K printer (ELEGOO) and sterilized with 70% ethanol before being fitted to the microfluidic device to act as a reservoir. This reservoir, possessing a 5 ml capacity, served as a storage area for the SynVACs produced in the oil phase, facilitating manual collection later on. During the production process, a microscope was used for real-time monitoring of the size and production rate of the alginate beads to promptly address any unexpected issues within the chip.

Fabrication of the SynVACs and elastic beads with well-defined mechanical properties

To minimize batch-to-batch variation, we used Novamatrix’s PRONOVA sodium alginate, manufactured under Good Manufacturing Practice guidelines and adhering to ISO standards. This particular grade has been regulated by the FDA as Generally Regarded as Safe for clinical use, ensuring its safety and reliability. Alginate viscoelastic microbeads were fabricated within the microfluidic device via a pH-induced internal gelation method^60,61. The core stream was loaded with a fluid comprising 3% alginate (VLVG, very low viscosity grade alginate, molecular weight 70 kDa, NovaMatrix), 1% CMC (used to stabilize laminar flow) and Ca–EDTA in deionized water. The concentration of Ca–EDTA was adjusted to modulate the stiffness of the resultant viscoelastic beads. The sheath stream contained 1% surfactant (157 FSH, Krytox) and 0.5% acetic acid, all dissolved in fluorocarbon oil (Novec 7500, 3M). Both fluids, sterilized with a 0.22 μm filter before use, were fed into the microfluidic device via separate inlets, facilitated by syringe pumps (Harvard Apparatus). After the formation and temporary storage of the microbeads in the reservoir, they were collected into a new tube containing 20% perfluoro-1-octanol (Sigma) and 0.2% acetic acid in fluorocarbon oil, which promoted crosslinking. Subsequently, HEPES-C buffer (20 mM HEPES, 140 mM sodium chloride, 5 mM potassium chloride, 2 mM calcium chloride, pH = 7.2) was added to the collection tube. The tube was centrifuged at 1,000 g for 1 min to transition the microbeads from the oil phase to the HEPES buffer. Finally, the buffer containing the microbeads was collected, rinsed twice (6,000 g, 5 min) for a thorough cleaning and stored at 4 °C for further experimentation.

Elastic beads were fabricated by converting ionically crosslinked viscoelastic microbeads into permanently crosslinked covalent microbeads. To begin with, we first fabricated viscoelastic microbeads using fluid containing 3% alginate (MVG (medium viscosity grade alginate)/VLVG = 1:2, average molecular weight = 120 kDa) and 25 mM Ca–EDTA. The resulting microbeads were resuspended in MES–Ca buffer (100 mM 2-(N-morpholino) ethane sulfonic acid, 300 mM sodium chloride, 2 mM calcium chloride, pH = 6) for 1 h to adjust the pH to 6. Subsequently, the microbeads were gathered by centrifugation (6,000 g, 5 min) and resuspended in MES–Ca buffer containing 320 mM 1-ethyl-3-(3-dimethyl aminopropyl) carbodiimide, hydroxy benzotriazole (HoBt) and AAD for overnight incubation. The concentration of AAD and HoBt was adjusted to modulate the stiffness of the resultant elastic beads. Finally, a sodium citrate buffer (77 mM NaCl, 55 mM sodium citrate) was introduced to the elastic beads for 2 h to chelate calcium ions and completely remove the ionic crosslink bonds. The covalent beads were then rinsed with HEPES-T buffer (HEPES buffer with 0.5% Tween20) and stored at 4 °C for future use.

Preparation of SynVACs and elastic beads for T-cell stimulation

SynVACs with well-defined spatial organization of ligands may enhance CAR T-cell activation, resulting in improved expansion, persistence and therapeutic efficacy⁶². To prepare SynVACs and elastic beads across different scenarios, tetrazine–TCO ligation was used to covalently conjugate antibodies/proteins onto the microbeads. Briefly, 4 million microbeads were resuspended in 800 μl MES buffer containing 200 mM 1-ethyl-3-(3-dimethyl aminopropyl) carbodiimide hydrochloride and 200 mM sulfo-NHS (N-hydroxysuccinimide) for an overnight reaction to activate carboxylic groups. Microbeads were then washed with HEPES-CT buffer (HEPES buffer containing 2 mM calcium chloride and 0.5% Tween20) to adjust the pH to 7. TCO⁻PEG6-amine (2 μmol per million beads, Click Chemistry Tools) was added to the solution and allowed to conjugate onto the microbeads overnight at room temperature. The reaction mixture was placed on a roller during this process to ensure thorough mixing. The microbeads were dialysed against HEPES-CT buffer using a 1000KD dialysis bag (Spectrum Laboratories) to remove excess TCO (HEPES-CT buffer was replaced 2–3 times per day). Antibody–tetrazine conjugation was performed according to the manufacturer’s protocol. Anti-CD3 (Biolegend 317302) and anti-CD28 (Biolegend,302902) antibodies were mixed at a 1:1 ratio for polyclonal T-cell activation. MSLN protein (Acro Biosystems, MSN-H526x) and anti-CD28 antibodies were mixed at a 1:1 ratio for antigen-specific CAR T enrichment. The mixture was concentrated in 100 μl phosphate-buffered saline (PBS) buffer using an Amicon Ultra-2 spin column (Sigma Aldrich). The concentrated mixture was mixed with tetrazine-PEG5-NHS (Sigma Aldrich) at a molar ratio of 1:5 and reacted for 30 min. The antibody/protein–tetrazine complex was then desalted using a spin column and rinsed with PBS 5 times to remove unreacted tetrazine-PEG5-NHS. The purified antibody/protein–tetrazine complex was mixed with glycerol at a 1:1 ratio and preserved at −20 °C. To refine the ligand density of SynVACs and elastic beads, we adjusted the dose of the antibody/protein–tetrazine for the tetrazine–TCO ligation with TCO-labelled microbeads.

Mechanical property and ligand density characterization of SynVACs

Rheological measurements were performed using an Anton Parr Rheometer. A 3% alginate solution (70 kDa, VLVG) was crosslinked with various concentrations of calcium crosslinker to generate an 8 mm disk of gel, supplemented with 1% CMC. The disk had an average thickness of about 2 mm. For the formation of the elastic hydrogel, a previously established protocol for elastic beads was followed. Specifically, a 3% alginate solution (120 kDa, MVG/VLVG = 1:2) was crosslinked with a certain concentration of calcium crosslinker to create an 8 mm gel disk. Gel samples were equilibrated with MES buffer (pH = 6) for 1 h, followed by immersion in an MES solution containing different concentrations of AAD and corresponding HoBt under constant stirring. This ensured an even distribution of AAD within the alginate matrix. The mixture was left to react at room temperature overnight, allowing the formation of an alginate gel with both ionic and covalent bonds. Subsequently, the gel samples were immersed in a sodium citrate buffer for 2 h to chelate and remove the calcium ions from the gel. Last, the resulting covalent alginate gel was thoroughly rinsed with deionized water and equilibrated in HEPES buffer for 24 h to ensure the complete removal of unreacted components and stabilization of the gel’s mechanical properties.

For the rheological test, an 8 mm PP025 measuring plate was used. The viscoelastic gel sample was carefully placed onto the centre of the rheometer plate with a spatula. The cantilever was lowered to the preferred gap height (1 mm was used in this study). Oscillatory strain sweeps (0.1–500%, 1 Hz), oscillatory frequency sweeps (0.1–100 Hz, 1% strain) and time sweeps (0.5% strain, 1 Hz, 2 min) were all conducted at room temperature to measure the storage (G′, Pa) and loss (G″, Pa) moduli. A shear ramp (0.01–100 s⁻¹) was used to examine the relationship between viscosity and shear rate. All experiments were repeated at least thrice. Rheological measurements of elastic hydrogels were performed using the same procedure.

The compressive moduli and stress relaxation properties of the viscoelastic gel were evaluated via compression tests on gel disks (8 mm in diameter, 2 mm thick, equilibrated in RPMI for 24 h) using a method from a previously published study⁶³. The gel disks were compressed with a deformation rate of 1/120 mm s⁻¹ for 30 s using a Chatillon TCD225 series force measurement system. The slope of the stress–strain curves (first 5–10% of strain) was used as the initial compressive modulus. Thereafter, the strain was held constant while the load was recorded over time. Stress relaxation was calculated by measuring the time it took for the stress to decrease to half from the maximum stress. No prestress was applied to the gels for these measurements. Compression and stress relaxation measurements of elastic hydrogels were performed using the same procedure.

The average antibody density per SynVAC was determined by quantitative flow cytometry. Briefly, 1 × 10⁵ SynVACs from each experimental group were conjugated with the appropriate antibody: Anti-CD3 FITC (Biolegend 300305). SynVACs were then washed and resuspended in HEPES-CT buffer for analysis in a Flow Cytometer (BD LSRFortessa Cell Analyzer). A standard curve was constructed using Quantum Simply Cellular anti-Mouse IgG beads (number 815, Bang Laboratories) stained with anti-CD3 FITC. For elastic beads, the number of antibodies per bead was determined using the same procedure. The average MSLN protein density per SynVAC was determined based on a standard curve constructed using Quantum Simply Cellular anti-Mouse IgG beads (number 815, Bang Laboratories) stained with anti-MSLN FITC.

AFM for SynVACs mechanical property and cell mechanics methods

APCs (primary monocytes) are extracted from human PBMCs using the CD14 MicroBeads kit (Miltenyi Biotec) on the same day as the test. SynVACs, elastic beads and APCs were placed on a JPK NanoWizard 4a BioScience AFM and indented by a Bruker SAA-SPH-1UM probe with a spring constant k = ~0.25 N m⁻¹ (the exact k value of each probe was determined by laser Doppler velocimetry calibration and used for the specific test). After the force spectroscopies were obtained, Young’s modulus was determined by fitting the data to a Hertz/Sneddon model using JPK Data Processing (v.3.4)⁶⁴. To measure viscoelasticity, the height after reaching 10 nN was maintained constant on the surface of SynVACs, elastic beads and APCs, and the stress relaxation profile was obtained by recording the vertical deflection force over the relaxation time.

XPS analysis for SynVACs ligand density

The elastic beads and SynVACs were carefully prepared and mounted for XPS measurements using an Axis Ultra DLD spectrometer (Kratos Analytical). The analysis was conducted under ultrahigh vacuum conditions to prevent contamination and interference from atmospheric gases. We used a monochromatic Al Kα X-ray source (λ = 1,486.6 eV) for excitation and scanned a wide range of binding energies to capture the complete elemental profile of the bead surfaces. The core level spectra were acquired for carbon (C 1 s) and nitrogen (N 1 s), and the peak areas were integrated to calculate the atomic ratios. The presence of nitrogen on the surfaces was particularly indicative of antibody conjugation, given that antibodies contain nitrogen-rich amino acids. The data were analysed and quantified to determine the percentage of antibody conjugation on the alginate monomer at bead surfaces.

Biosafety evaluation of SynVACs for CAR T expansion

A biosafety test was conducted to ensure that alginate-based SynVACs could be effectively separated from T cells through physical centrifugation, thereby ensuring no residual SynVACs in the activated T cells before therapeutic applications. In detail, pre-labelled SynVACs were co-cultured with primary mouse T cells at a 1:1 ratio for 24 h. Post-culture, the cells were stained with Hoechst 33342 (1:1,000, Thermo Fisher) at room temperature for 10 min. The co-culture mixture was then subjected to centrifugation (600 g, 5 min) and resuspended in buffer solution. The bead-to-cell ratio before and after the washing procedure was confirmed via FACS analysis. To ensure the absence of residual alginate monomers, which could potentially trigger innate immune responses if co-injected with CAR T cells, the alginate residue in the supernatant was quantified via high-performance liquid chromatography. For high-performance liquid chromatography analysis, a carbon stationary phase (Kromasil 300–5-E18, 4.6 × 250 mm) was chosen. Elution was performed with a mobile phase consisting of 40% acetonitrile and 60% deionized water, at a flow rate of 1 ml min⁻¹, and ultraviolet detection was set at 254 nm. The system was equilibrated with the mobile phase for 20 min before the first injection. A sodium alginate standard solution was prepared by dissolving 250 mg of accurately weighed VLVG in 5 ml distilled water to create a 5% stock solution. Calibration standards were prepared by diluting varying amounts of the VLVG stock solution to yield a concentration range of 0.05 to 20 μg ml⁻¹ using distilled water.

T-cell culture

The Human Jurkat T-cell line and Jurkat NFAT-zsGreen reporter cell line were gifts from the Christopher Seet Lab at University of California, Los Angeles (UCLA). Primary mouse T cells were isolated from the spleen of C57BL/6 mice using a pan T-cell isolation kit, with CD3⁺ T cells procured for polyclonal activation studies. Healthy donors of human PBMCs were sourced from the UCLA/Center for AIDS Research (CFAR) Virology Core Laboratory, in compliance with federal and state regulations, with no identifying information provided. The human Burkitt’s lymphoma cell line Raji, acute lymphoblastic leukaemia cell line NALM6, ovarian cancer cell lines OVCAR3 and OVCAR8, chronic myelogenous leukaemia cell line K562 and human embryonic kidney 293T cells were obtained from the American Type Culture Collection (ATCC).

Stable tumour cell lines expressing firefly luciferase and enhanced GFP dual-reporters (FG) were generated by transducing parental tumour cell lines with lentiviral vectors encoding the target gene(s). Post lentivector transduction (72 h), cells were flow cytometry sorted to isolate gene-engineered cells, thereby establishing stable cell lines. Four stable tumour cell lines were created for this study, including Raji-FG, NALM6-FG, OVCAR3-FG and OVCAR8-FG cell lines. An aAPC was generated by engineering the K562 human chronic myelogenous leukaemia cell line (ATCC) to overexpress human CD83/CD86/4–1BBL co-stimulatory receptors. The aAPC–MSLN cell lines were further engineered from the parental aAPC line to overexpress human MSLN.

Jurkat cells and primary mouse T cells were cultured in ATCC modified RPMI 1640 medium enriched with 10% FBS, 1% penicillin/streptomycin, 50 μM 2-mercaptoethanol and 100 mg ml⁻¹ normocin. The reporter Jurkat media was supplemented with an additional 1 ng ml⁻¹ puromycin. Human PBMCs from healthy donors were cultured in complete lymphocyte culture medium composed of RPMI 1640 supplemented with 10% FBS, 1% penicillin/streptomycin/glutamine, 1% minimal essential medium non-essential amino acids, 10 mM HEPES, 1 mM sodium pyruvate, 50 mM 2-mercaptoethanol and 100 mg ml⁻¹ normocin. Raji, NALM6, OVCAR3 and OVCAR8 cell lines were maintained in RPMI 1640 medium supplemented with 10% FBS and 1% penicillin/streptomycin/glutamine.

Cell-viability assays

After Jurket T cells were mixed with SynVACs or elastic beads at a 1:1 ratio, the mixture was incubated for 3 h in a 96-well plate with 1 × 10⁴ cells per well. To assess the cell viability, the LIVE/DEAD Cell Imaging Kit (Invitrogen, R37601) was used according to the manufacturer’s protocol. Fluorescence images were collected using a Zeiss Axio Observer Z1 inverted fluorescence microscope and analysed using ImageJ software.

In vitro polyclonal T-cell expansion studies

Isolated mouse primary T cells or human PBMCs were co-cultured with SynVACs, elastic beads or Dynabeads (T Cell-Activator, Gibco), respectively. An initial number of 5 × 10⁵ PBMCs (roughly 20% of the total number were T cells) were seeded in the starting culture media supplied with 30 IU ml⁻¹ recombinant IL-2 (Biolegend) and activated by SynVACs, elastic beads or Dynabeads at a bead-to-cell ratio of 1:1. For the negative controls, one group called ‘Bare beads’ was treated with viscoelastic microbeads without antibody conjugation in IL-2-enriched media, while the other control group consisted of T cells alone, also supplemented with IL-2 but without any beads. Fresh media containing 30 IU ml⁻¹ IL-2 was added to the cells to keep the cell density below 2 × 10⁶ cells per ml through the whole culture process. After CAR transduction, cells were further expanded in media containing IL-2. Cell numbers were counted on day 7, day 10 and day 14. Fold expansion was calculated by dividing the number of cells at the respective time point by the number of cells seeded at the start of the culture.

Lentiviral vectors for CAR introduction

Lentiviral vectors used in this study were all constructed from a parental lentivector pMNDW, which contains an MND promoter and a woodchuck hepatitis virus posttranscriptional regulatory element (WPRE)⁶⁵. The Lenti/CAR19 vector was constructed by inserting into the pMNDW a synthetic gene encoding CD19-targeting CAR. The Lenti/MCAR vector was constructed by inserting into the pMNDW a synthetic gene encoding MSLN-targeting CAR. The synthetic gene fragments were obtained from GenScript and Integrated DNA Technologies. Lentiviruses were produced using human embryonic kidney 293T cells (ATCC), following a standard transfection protocol using the Trans-IT-Lenti Transfection Reagent (Mirus Bio) and a centrifugation concentration protocol using the Amicon Ultra Centrifugal Filter Units, according to the manufacturer’s instructions (MilliporeSigma)⁶⁶.

In vitro CAR T-cell generation

On day 0, human PBMCs from healthy donors were activated by SynVACs, elastic beads or Dynabeads (T Cell-Activator, Gibco), respectively. An initial number of 5 × 10⁵ PBMCs (roughly 20% of the total number were T cells) were seeded in the starting culture media supplied with 30 IU ml⁻¹ recombinant IL-2 and activated by SynVACs, elastic beads or Dynabeads at a bead-to-cell ratio of 1:1. On day 2, cells were transduced with Lenti/CAR19 or Lenti/MCAR virus for another 24 h. The resulting CAR T cells were expanded for another 2–3 weeks in C10 medium; fresh media supplied with 30 IU ml⁻¹ recombinant IL-2 was supplemented if needed, and then the generated CAR T cells were cryopreserved for future use.

In vitro antigen-specific CAR T-cell enrichment studies

Human MCAR T cells were re-stimulated using MSLN-conjugated SynVACs, MSLN-expressing artificial-presenting cells or Dynabeads, respectively. An initial number of 1 × 10⁶ MCAR T cells (roughly 70% of the total number were CAR⁺ cells) were seeded in the starting culture media supplied with 30 IU ml⁻¹ recombinant IL-2 (Biolegend) and activated by SynVACs at a bead-to-cell ratio of 1:1. CAR expression and T-cell phenotype was analysed using flow cytometry on day 5.

Flow cytometry analysis of T-cell phenotypes

Cells were collected at certain time points, and the T-cell phenotypes were evaluated using flow cytometry. Four hours before the flow analysis, a protein transport inhibitor (BD Biosciences) was added to the aliquoted cells to increase the signal of intracellular markers. Fold expansion was calculated at the same time, by dividing the number of cells at a specific time point by the number of seeded cells at the beginning of the culture. To evaluate the CAR transduction rate, we analysed CAR expression using a fluorophore-conjugated antibody specific to the extracellular domain of the CAR construct (such as anti-MSLN, anti-CD19) on the T cell surface via flow cytometry on days 7 and 10.

All flow cytometry stains were performed in PBS for 30 min on ice. The samples were stained with Mouse Fc Block (anti-mouse CD16/32) or Human Fc Receptor Blocking Solution (TrueStain FcX) before antibody staining. Antibody staining was performed at a certain dilution according to the manufacturer’s instructions. Fluorochrome-conjugated antibodies specific for human CD4 (Clone OKT4), CD8 (Clone SK1), CD45 (Clone H130), TCRαβ (Clone I26), CD3 (Clone HIT3a), CD4 (Clone OKT4), CD8 (Clone SK1), CD45RO (Clone UCHL1), CD58 (Clone TS2/9), CD11a (Clone TS2/4), CXCR3 (Clone G025H7), CD19 (Clone HIB19), granzyme B (Clone QA16A02), perforin (Clone dG9), CD69 (Clone FN50), CD45RA (Clone HI100), CD62L (Clone DREG-56), CD95 (Clone G043H7), CD25 (Clone BC96), PD-1 (Clone A17188A), Tim-3 (F38–2E2), FOXP3 (Clone 206D), IFNγ (Clone B27), TNF (Clone Mab11), mouse CD4 (Clone GK1.5), CD8 (Clone 53–6.7), IFNγ (Clone XMG1.2), TNF (Clone MP6-XT22), PD-1 (Clone 29F.1A12), Tim-3 (Clone B8.2E12), CCR7 (Clone 4B12), CD25 (Clone PC61), CD 44 (Clone IM7), CD95 (Clone SA367H8), CD62L (Clone MEL-14), CD95 (Clone SA367H8), Sca-1 (Clone D7) and streptavidin were purchased from Biolegend. Fluorochrome-conjugated antibodies specific for human MSLN (Clone 420411) were purchased from R&D Systems. Goat anti-mouse IgG F(ab′)2 secondary antibody was purchased from Thermo Fisher. Human Fc Receptor Blocking Solution (TrueStain FcX) was purchased from Biolegend, and Mouse Fc Block (anti-mouse CD16/32) was purchased from BD Biosciences. Intracellular cytokines were stained using a Cell Fixation/Permeabilization Kit (BD Biosciences). Stained cells were analysed using LSRII (BD Biosciences). FlowJo v10 software was used to analyse the data.

SEM imaging of the interactions between microbeads and T cells

Human CD3⁺ T cells were isolated from PBMCs using a human pan T isolation kit (Miltenyi Biotec) and stimulated with SynVACs or Dynabeads for 24 h. The activated human T cells, along with SynVACs or Dynabeads, were then fixed using 4% glutaraldehyde, refrigerated for 2 h and subsequently post-fixed with 1% osmium tetroxide for another 2 h. After fixation, the T cells were rinsed with HEPES-C buffer and then dehydrated through a graduated series of ethanol concentrations (75%, 85% and 95%), with each step lasting 30 min. The samples were then dried using a critical point dryer. The specimens were subsequently mounted on specimen stubs and sputter-coated with a gold-palladium layer to prepare them for SEM imaging.

Immunofluorescent visualization of T-cell activation

Human CD3⁺ T cells were isolated from PBMCs using a human pan T isolation kit (Miltenyi Biotec) and stimulated with SynVACs or Dynabeads for 72 h. On day 3, activated human T cells were collected and made to adhere to a 12 mm cover glass (Citoglas) that had been pre-treated with poly-l-lysine (Sigma). The cells were then fixed with 100% methanol at −20 °C for 5 min, followed by blocking with HEPES-C buffer containing 5% donkey serum. The samples were subsequently incubated with primary antibodies against CD3ε (1:400, Abcam, ab52959), β-actin (1:800, Cell Signaling, number 4970) and NFAT1 (1:50, Cell Signaling, number 4389) at 4 °C overnight. After triple washing with HEPES-CT buffer, the cells were incubated with an appropriate secondary antibody for 1 h. Nuclei were visualized by staining with 4,6-diamidino-2-phenylindole (D3571, Thermo Fisher) at a 1:1,000 dilution for 10 min. Confocal images were acquired using a Leica SP8-STED confocal microscope and further analysed using ImageJ software (v.1.53a).

scRNAseq

CAR T cells activated by SynVACs and Dynabeads were cultured and collected at day 14, followed by sorting with a FACSAria II flow cytometer. The sorted cells were immediately dispatched to the UCLA Technology Center for Genomics and Bioinformatics (TCGB) Core for single-cell TCR sequencing. The sequencing was executed with a 10X Genomics Chromium Controller Single Cell Sequencing System, as per the manufacturer’s guidelines and the TCGB Core’s standard protocol. Library preparation was accomplished using the Illumina TruSeq RNA Sample Prep Kit (catalogue number FC-122–1001), and the sequencing was performed with 150 bp paired-end reads (5,000 reads per cell) on an Illumina NovaSeq system. The processed cell matrix, data tables (such as expression values) and metadata have been made available in the public repository Gene Expression Omnibus database (accession code GSE242531).

Enzyme-linked immunosorbent cytokine assays

ELISAs for detecting human cytokines were performed following a standard protocol from BD Biosciences. Supernatants from cell culture assays were collected and assayed to quantify human IFNγ. The capture and biotinylated pairs for detecting cytokines were purchased from BD Biosciences. The streptavidin–horseradish peroxidase conjugate was purchased from Invitrogen. Human cytokine standards were purchased from eBioscience. The samples were analysed for absorbance at 450 nm using an Infinite M1000 microplate reader (Tecan).

In vitro tumour-killing assay

Tumour cells (1 × 10⁴ cells per well) were co-cultured with effector cells (at ratios indicated in the figure legends) in T-cell culture medium in Corning 96-well clear-bottom black plates for 24 h. At the end of the culture, live tumour cells were quantified by adding d-luciferin (150 μg ml⁻¹; Caliper Life Science) to cell cultures and reading out luciferase activities using an Infinite M1000 microplate reader (Tecan).

Animals

NOD.Cg-PrkdcSCIDIl2rgtm1Wjl/SzJ (NSG) mice were housed in UCLA’s animal facilities. Mice aged 6–10 weeks were used for all experiments, unless stated otherwise. The Institutional Animal Care and Use Committee at UCLA approved all animal procedures. The mice were bred and kept under specific pathogen-free conditions, and all experiments followed the animal care and use guidelines set by UCLA’s Division of Laboratory Animal Medicine.

In vivo bioluminescence live animal imaging

Bioluminescence live animal imaging (BLI) was performed using a Spectral Advanced Molecular Imaging (AMI) HTX imaging system (Spectral Instrument Imaging). Live animal imaging was acquired 5 min after intraperitoneal (i.p.) injection of d-luciferin (1 mg per mouse) for total body bioluminescence, and 15 min after i.p. injection of d-luciferin (3 mg per mouse) for tissue bioluminescence. Imaging results were analysed using AURA imaging software (v.4.0) (Spectral Instrument Imaging).

In vivo antitumour efficacy study of CAR19 T cells in human Raji xenograft NSG mouse model

The experimental design is shown in Fig. 6a. Briefly, on day 0, NSG mice received intravenous (i.v.) inoculation of Raji-FG cells (1 × 10⁶ cells per mouse). On day 4, the experimental mice received i.v. injection of vehicle (100 μl PBS per mouse) or CAR19 T cells (3 × 10⁶ CAR T cells in 100 μl PBS per mouse). During the experiment, mice were monitored for their tumour loads to be measured using BLI. On day 40, the experimental mice were euthanized, and their tissues were collected for further analysis.

In vivo antitumour efficacy study of MCAR T cells in human OVCAR8 xenograft NSG mouse model

The experimental design is shown in Fig. 7a. Briefly, on day 0, NSG mice received i.p. inoculation of OVCAR8-FG cells (1 × 10⁶ cells per mouse). On day 4, the experimental mice received i.v. injection of vehicle (100 μl PBS per mouse) or MCAR T cells (3 × 10⁶ CAR T cells in 100 μl PBS per mouse). During the experiment, mice were monitored for their tumour loads to be measured using BLI. On day 40, the experimental mice were euthanized, and their tissues were collected for further analysis.

Statistical analysis

Data are represented as the mean ± standard deviation. When necessary, a two-tailed Student’s t-test was used to identify statistically significant differences between the two groups, using GraphPad Prism 8 for computations. In instances of comparison among more than two groups, a one-way analysis of variance (ANOVA) was conducted, followed by Tukey’s multiple comparison test. Levels of statistical significance were denoted as follows: not significant, P > 0.05; *P < 0.05; **P < 0.01; ***P < 0.001.

Supplementary Material

Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41551-024-01272-w.

Data availability

The main data supporting the results in this study are available within the paper and its Supplementary Information. All data generated in this study, including source data for the figures, are available via figshare at https://doi.org/10.6084/m9.figshare.25928314 (ref. 67). Source data are provided with this paper.