Safe and Efficient CRISPR Genome Editing of Primary Human T Cells Using a Droplet‐Based Cell Mechanoporation Platform

ABSTRACT

T cell engineering is a transformative strategy for adoptive cell therapy, holding the key to treating a wide array of human diseases. However, clinical translation is limited by current intracellular delivery methods that compromise viability, induce stress responses, and restrict scalability. This study presents a microfluidic droplet mechanoporation system tailored for primary human T cells, enabling efficient, stable, and clinically scalable gene delivery. Delivery of 2000 kDa fluorescein isothiocyanate (FITC)‐dextran achieves ∼98% efficiency and >90% post‐treatment viability, even at high cell densities, supporting the rapid production of therapeutically relevant cell numbers. The platform efficiently delivers mRNA, achieving transfection efficiencies approaching 99%; further, chimeric antigen receptor (CAR)‐encoding mRNA is successfully delivered to generate CAR‐expressing T cells with tunable surface expression. Clustered regularly interspaced short palindromic repeats (CRISPR)‐Cas9 ribonucleoproteins are effectively delivered for both single and multiplex knockouts (TRAC and PDCD‐1), achieving up to a 2.35‐fold higher efficiency than electroporation. Longitudinal analyses confirm preserved viability, proliferation, genome integrity, and T cell phenotypic stability. Collectively, these results establish microfluidic droplet mechanoporation as a safe, efficient, and scalable platform for the clinical manufacturing of engineered T cell therapies.

This study introduces a microfluidic droplet mechanoporation system for the engineering of primary human T cells. The platform enables efficient, scalable, and robust delivery of mRNA for CAR T cell production and CRISPR‐Cas9 ribonucleoproteins for genome editing, while maintaining high cell viability. These results establish a safe and effective method for the clinical manufacturing of T cell‐based immunotherapies.

1. Introduction

Advances in T cell engineering have fundamentally reshaped the landscape of cancer immunotherapy [1]. Adoptive T cell therapy (ACT) has demonstrated substantial clinical promise by enabling the reinfusion of patient‐ or donor‐derived T cells engineered with tumor‐specific T cell receptors (TCR‐T) or synthetic chimeric antigen receptors (CAR‐T) [2]. TCR‐T modification of the endogenous TCR, which physiologically recognizes peptide‐major histocompatibility complex (MHC) molecules, redirects T cells toward tumor‐associated antigens [3, 4]. Reliable outcomes have been achieved with TCRs targeting the MART‐1 [5, 6], gp100 [7], and NY‐ESO‐1 antigens across multiple malignancies [8, 9, 10]. In contrast, CAR‐T therapy leverages antibody‐derived antigen‐binding domains fused to transmembrane, signaling, and costimulatory elements [11], thereby enabling antigen recognition independent of MHC restriction. To date, seven CAR‐T therapies have been approved by the U.S. FDA, primarily targeting CD19 in B cell lymphomas [12, 13, 14, 15, 16] and BCMA in multiple myeloma [17, 18], demonstrating high clinical efficacy.

Despite these successes, ACT remains constrained by several critical challenges. Severe toxicities, heterogeneous patient responses, and limited persistence frequently undermine its efficacy [19, 20]. In TCR‐T therapies, conventional viral transduction introduces risks of mispairing between endogenous and introduced TCR chains, potentially generating autoreactive receptors and reducing therapeutic potency [21]. CAR‐T and TCR‐T therapies also predominantly rely on autologous T cells, which are costly, time‐intensive to manufacture, and variable in quality [22, 23]. Allogeneic “off‐the‐shelf” T cells represent a potential solution; however, endogenous TCRs can recognize host alloantigens and trigger graft‐vs.‐host disease, necessitating precise genomic disruption [24]. Furthermore, T cell exhaustion, driven in part by sustained expression of inhibitory receptors such as programmed cell death protein 1 (PD‐1), diminishes therapeutic persistence and anti‐tumor activity [25, 26, 27, 28]. These limitations highlight the need for multiplex genome editing strategies that simultaneously eliminate endogenous TCRs and checkpoint regulators, thereby enabling safer and more effective ACT.

For T cell engineering, clustered regularly interspaced short palindromic repeats (CRISPR)‐Cas9 technology has emerged as a powerful tool [24, 29, 30]. Its programmability, high efficiency, and capacity for combinatorial knockouts have enabled clinical studies demonstrating that the simultaneous disruption of the TCR and PD‐1 can enhance T cell persistence and tumor clearance [31, 32]. Intracellular delivery of CRISPR components has traditionally relied on electroporation, favored for its operational simplicity and relatively high transfection efficiency [33, 34]. However, electroporation requires high‐voltage electrical pulses that compromise cell viability, induce aberrant transcriptional programs, and raise concerns regarding DNA damage, collectively creating a major bottleneck for next‐generation manufacturing [35, 36].

To address these limitations, we developed a microfluidics‐based cell mechanoporation platform, the Droplet T Cell Pincher (DTCP), specifically designed for the safe and efficient delivery of both mRNA and CRISPR systems into primary human T cells. The platform transiently permeabilizes cell membranes through controlled mechanical deformation within microfluidic droplets, thereby minimizing cytotoxicity while preserving editing efficiency. DTCP supports clinically relevant throughput, enabling the rapid production of therapeutically relevant cell numbers. The proposed system achieved high‐efficiency mRNA transfection, including the generation of CAR‐expressing T cells with tunable surface density, and robust delivery of CRISPR‐Cas9 ribonucleoproteins (RNPs) for efficient single‐ and multiplex editing of TCR and PD‐1. Longitudinal studies confirm that DTCP‐treated cells maintain high viability, proliferation, genomic integrity, and phenotypic stability. These results establish DTCP as a safe, efficient, and scalable platform for the clinical manufacturing of engineered T cell therapies.

2. Results and Discussion

2.1. Platform Design and Operation Principle

The DTCP platform leverages droplet microfluidics to co‐encapsulate cells and functional macromolecules and to drive them through a microscale constriction. Passage through this constriction induces transient disruption of both the plasma and nuclear membranes, thereby enabling rapid cytosolic and nuclear access for exogenous cargos. In our earlier work, this principle enabled the successful delivery of diverse CRISPR modalities into different immortalized cell lines [37]. However, those initial designs were ineffective for primary human T cell engineering and lacked the throughput and operational robustness required for translational or clinical‐scale applications. Accordingly, rather than pursuing incremental optimization, we undertook a comprehensive re‐engineering of the platform to specifically accommodate the biological and mechanical constraints of primary human T cells and to rigorously assess their suitability for CRISPR genome engineering and CAR‐T manufacturing (Figure 1A,B).

FIGURE 1.

Overview of the droplet T cell pincher (DTCP) platform for engineering primary human T cells. (A) Schematic illustration of the DTCP‐based manufacturing process for T cell immunotherapies. (B) Illustration of T cells with endogenous receptors, the mechanism of droplet cell mechanoporation, and the resulting engineered T cells lacking endogenous receptors while expressing CARs. (C) High‐speed microscopy images depicting stable droplet generation with T cells during the DTCP process. (D) Bright‐field (BF) and fluorescence (FL) images showing the delivery of 2000 kDa FITC‐dextran into primary T cells via DTCP (Scale bars: 50 µm).

The resulting DTCP architecture incorporates several structural innovations while retaining the droplet‐based operational framework (Figure 1C; Movie S1). Notably, the channel cross‐sectional area was reduced to match the substantially smaller dimensions of T cells with distinct mechanical properties [38, 39, 40], ensuring sufficient shear‐mediated membrane permeabilization. This geometric downsizing, however, dramatically increased hydraulic pressure, which initially led to elevated cell death, accumulation of non‐viable cells at the inlet filter, frequent clogging and, often delamination of the bonded chip under excessive pressure (Figure S1).

To overcome these limitations, we implemented a series of additional architectural adjustments aimed at reducing flow resistance and improving operational robustness. These included the elimination of the acceleration sheath‐fluid channels, the streamlining of channel curvatures, the enlargement of the cell inlet reservoir, redesign of the cell inlet filter geometry, and the shortening of the overall channel length. Collectively, these modifications substantially lowered hydraulic resistance, mitigated clogging under high cell density operation, and preserved cell integrity (Figure S2).

We next assessed the operational stability of the redesigned platform under continuous use. The device could be operated continuously for approximately 40 min without user intervention or performance degradation. Representative time‐lapse images demonstrate stable and uninterrupted droplet generation and cell processing throughout this period (Figure S3), and this level of chip stability was consistently reproduced across independent runs. Together, these results indicate that the revised DTCP architecture substantially improves clogging resistance, mechanical robustness, and the effective device lifetime during continuous operation.

Having established the stability of the platform, we evaluated its capacity for intracellular delivery. Using optimized operating parameters (see Section 2.2), fluorescein isothiocyanate (FITC)–conjugated dextrans were employed as model macromolecular cargo. Robust intracellular fluorescence was observed only in DTCP‐processed T cells, whereas the endocytosis control group exhibited negligible uptake (Figure 1D). These findings confirm that the DTCP can reliably permeabilize primary human T cells and mediate the efficient internalization of macromolecules, establishing a foundation for subsequent genome editing and cell engineering applications.

2.2. Delivery Characterization: Channel Geometry and Flow Rate Dependence

Primary human T cells differ from immortalized cell lines in terms of their mechanical and structural properties, including their smaller size, higher nucleus‐to‐cytoplasm ratios, and distinct viscoelastic properties [38, 39, 40]. These features imply that their behavior during droplet cell mechanoporation strongly depends on the geometry of the microchannel and the applied flow conditions. To examine delivery performance, we systematically evaluated the intracellular delivery of 2000 kDa FITC‐dextran into T cells as a model macromolecule.

Delivery outcomes were quantified across a series of channel configurations, namely constriction width, length, and flow rate (i.e., the Reynolds number; Re), by quantifying delivery efficiency, mean fluorescence intensity (MFI), viability, and yield (defined as the product of the MFI fold change and relative viability) (Figure 2A). Owing to donor‐to‐donor variability in baseline viability, all post‐treatment viabilities were normalized relative to each donor's endocytosis control. For translational relevance, experimental conditions that reduced relative viability below 80% were excluded from further consideration.

FIGURE 2.

Characterization of 2000 kDa FITC‐dextran delivery into primary human T cells under varying DTCP operational and geometric conditions. (A) Schematic illustration of the DTCP parameters that affect delivery performance. (B–E) Delivery efficiency, mean fluorescence intensity (MFI) fold change, relative viability, and overall yield at different inlet flow rates. (F–I) Delivery efficiency, MFI fold change, relative viability, and yield across varying constriction gap widths. (J–M) Delivery efficiency, MFI fold change, relative viability, and yield across varying constriction gap lengths. Statistical significance: n.s., not significant; ^*, ^**, ^***, and ^**** indicate p <0.05, 0.01, 0.001, and 0.0001, respectively, determined using one‐way ANOVA. Error bars represent the mean ± standard deviation (SD), and each dot corresponds to an individual donor.

Flow rate characterization showed that stable droplet generation stability was maintained up to 1.0 mL/h. At this flow rate, relative viability modestly decreased to 89.0%, while this was offset by substantial gains in MFI fold change, yield, and throughput (Figure 2B–E). These results indicate that increased shear and pressure at higher flow rates enhance molecular uptake, provided that viability remains within acceptable limits.

Constriction width exerted an even stronger influence on delivery outcomes. Narrower constrictions enhanced delivery efficiency and MFI fold change but decreased relative viability consistently across donors (Figure 2F–H). A width of 3 µm yielded the highest levels of cargo uptake, whereas average viability dropped to 74.1% and the geometry resulted in practical challenges, including frequent clogging and an increased risk of chip rupture during prolonged operation. In contrast, a 4 µm constriction width provided a more balanced performance, combining robust delivery with preserved viability, and was therefore selected as the optimal width (Figure 2I).

Constriction length further modulated intracellular delivery. Lengths of 20 and 35 µm significantly improved the MFI fold change and yield compared with a length of 5 µm, while maintaining relative viability above 80% (Figure 2J–L). Statistical analysis confirmed that the yields at 20 µm (p = 0.01) and 35 µm (p = 0.02) were significantly higher than those at 5 µm, with no significant difference between 20 and 35 µm (p = 0.82) (Figure 2M). Considering throughput requirements and the elevated hydraulic pressures associated with longer constrictions, a length of 20 µm was selected as the most suitable length.

These systematic evaluations establish the characteristic response of T cells to DTCP‐mediated mechanoporation. The combined optimal operating parameters, a 4 µm width, a 20 µm length, and a flow rate of 1.0 mL/h, achieve a favorable balance of high delivery efficiency, yield, and cell viability. These operational conditions were subsequently used for the intracellular delivery of functional cargos, including mRNA and CRISPR‐Cas9 RNP complexes.

2.3. Delivery Characterization: Donor, Cargo Size, and Cell Density Insensitivity

After identifying the optimal operating parameters, the robustness of DTCP performance relative to biological and operational variability was evaluated. Specifically, sensitivity to donor source, molecular cargo size, and input cell concentration, three factors that frequently confound intracellular delivery platforms and limit translational utility, were tested.

To assess whether donor‐to‐donor differences influence performance, 2000 kDa FITC‐dextran was delivered into primary T cells from 18 independent healthy donors. The delivery efficiency and viability were measured before and after DTCP treatment. Paired line plots confirmed consistent trends across all donors, with statistically significant improvements in delivery efficiency (Figure 3A,B). Importantly, inter‐donor variability was minimal, with coefficients of variation (CV; defined as CV = standard deviation (SD) / mean × 100) of 0.9% and 4.1% for delivery efficiency and viability, respectively. These results demonstrate that DCTP‐mediated mechanoporation is largely insensitive to intrinsic donor heterogeneity, a critical requirement for clinical manufacturing where biological variability is unavoidable.

FIGURE 3.

Assessment of DTCP performance across donors, cargo sizes, and cell concentrations. (A) Delivery efficiency of primary T cells from 18 donors following DTCP treatment. (B) Relative viability of T cells from the same donors. (C) Delivery efficiency across different FITC‐dextran molecular weights. (D) Relative viability for each FITC‐dextran size. (E) Representative fluorescence intensity histograms for 2000 kDa FITC‐dextran at varying cell concentrations. (F) Delivery efficiency, (G) MFI fold change, and (H) relative viability as a function of cell concentration. Each data point is based on flow cytometry analysis performed independently on different donors. All error bars indicate the mean ± standard deviation (SD). Each dot represents an individual donor. Statistical analyses were performed using one‐way ANOVA with Bonferroni post‐hoc tests.

We subsequently investigated whether intracellular delivery is affected by the size of the delivered cargo. To this end, FITC‐dextran molecules spanning a broad molecular weight range (3–5, 40, 150, and 2000 kDa) were delivered at equal concentrations (0.3 mg/mL). Across all tested sizes, the delivery efficiency and relative viability remained stable, with delivery efficiencies consistently near 98% and relative viability close to 90%. No significant differences were observed among the groups (Figure 3C,D), demonstrating that the DTCP platform can internalize biomolecules across a wide size range without compromising delivery performance, supporting its versatility for diverse therapeutic cargos.

Finally, the impact of input T cell concentration was examined across a range of 0.2, 0.5, 1, 2, and 4 × 10⁸ cells/mL. The delivery efficiency and relative viability remained unchanged across this entire range (Figure 3E–H). Notably, the MFI fold change did not decrease even at 4 × 10⁸ cells/mL.; instead, it exceeded the level observed at 2 × 10⁷ cells/mL, which possibly reflects enhanced FITC‐dextran uptake arising from increased cell–cell interactions within droplets after constriction (Figure S4; Movie S2) [41, 42]. The preservation of high efficiency and viability at elevated cell densities further indicates that DTCP is amenable to high‐throughput operation. At its maximum processing rate, the platform can exceed 6.7 × 10⁶ cells/min, enabling the generation of approximately 1 × 10⁸ engineered T cells, which is comparable to clinically relevant doses [43], within only 15 min of operation.

At substantially higher input cell concentrations, additional physical and analytical complexities emerge that complicate precise mechanical control and data interpretation. At elevated densities, droplets more frequently encapsulate multiple cells with heterogeneous contact configurations, resulting in nonlinear deformation histories and less predictable per‐cell delivery doses. Accordingly, a concentration of 5 × 10⁷ cells/mL was selected for subsequent DTCP experiments to ensure controlled and reproducible mechanoporation. Notably, even under this conservative operating condition, the prior validation of uninterrupted continuous operation for approximately 40 min on a single chip (Figure S3) indicates that a single run can reliably process at least ∼3 × 10⁷ cells without user intervention, a yield that falls within the lower range of clinically relevant T cell doses reported for adoptive cell therapies [43, 44].

Collectively, these results show that DTCP maintains high delivery efficiency and viability across diverse donor sources, molecular cargo sizes, and input cell densities. This robustness highlights its suitability for reproducible and large‐scale genome engineering in clinical and translational settings.

2.4. EGFP‐mRNA and CAR‐mRNA Transfection of T Cells via DTCP Platform

Building on the robustness established above, the ability of the DTCP platform to deliver functional biomolecules into primary human T cells was evaluated. As an initial test, EGFP‐encoding mRNA was delivered to confirm efficient cytoplasmic delivery and subsequent protein expression. At a concentration of 20 µg/mL, strong EGFP fluorescence was observed in the majority of treated cells (Figures 4A; Figure S5). To examine dose dependence, EGFP‐mRNA was delivered at 10, 20, and 40 µg/mL, followed by flow cytometry analysis. The transfection efficiency and MFI increased with dose, reaching >96% transfection efficiency and ∼90% viability at 20 µg/mL, along with an approximately 19‐fold increase in MFI relative to endocytosis controls (Figure 4B–E). These results confirm that DTCP enables efficient and cytocompatible mRNA delivery into primary T cells.

FIGURE 4.

EGFP‐mRNA and CAR‐encoding mRNA transfection using the DTCP platform. (A) Bright‐field and fluorescence images showing EGFP expression (20 µg/mL mRNA) via endocytosis and DTCP after 24 h. (B) Fluorescence intensity histograms of a representative donor. (C) MFI fold change, (D) delivery efficiency, and (E) relative viability of EGFP mRNA expression as a function of mRNA concentration. (F) Bright‐field and fluorescence images showing CAR expression (100 µg/mL mRNA) via endocytosis and DTCP after 24 h. (G) Fluorescence intensity histograms of a representative donor. (H) Bar plots showing transfection efficiency (left) and cell viability (right) after treatment with increasing CAR mRNA doses (100 and 200 µg/mL) compared with the endocytosis control (Ctrl). (I) Transfection efficiency and (J) cell viability for CAR mRNA delivery at 100 and 200 µg/mL relative to the control. All data points are based on flow cytometry analysis from n = 3 independent donors. Error bars represent the mean ± standard deviation (SD), and each dot indicates an individual donor (Scale bars: 50 µm).

Encouraged by these findings, we next applied the platform to transfect T cells with CAR‐encoding mRNA. GFP‐anti‐CD19‐CAR‐mRNA was internalized at 100 µg/mL to validate the delivery of large transcripts encoding full‐length CARs. Robust GFP fluorescence confirmed the successful expression of the CAR construct (Figure 4F; Figure S5C,D). The delivery performance was then evaluated at two doses, 100 and 200 µg/mL. The transfection efficiencies at these concentrations reached approximately 70% and 90%, respectively, whereas cell viability remained high (∼92%) at both concentrations (Figure 4H).

To directly quantify CAR surface expression, anti‐CD19‐CAR‐mRNA was delivered and cells were analyzed by immunofluorescence staining using an APC‐labeled anti‐FMC63 antibody. At 50 µg/mL, 36.3% of the cells expressed CAR on the cell surface, which increased to 53.5% at 100 µg/mL, indicating dose‐dependent modulation of surface expression (Figure 4I). Importantly, relative viability remained high under both conditions (94.1% and 91.9%, respectively; Figure 4J).

These results demonstrate that the DTCP platform enables efficient, high‐viability engineering of primary human T cells with both reporter and therapeutically relevant mRNAs. Moreover, the ability to modulate CAR expression levels by adjusting the delivered mRNA dose highlights the platform's flexibility and utility for rapid, tunable, and scalable CAR‐T cell engineering.

2.5. Superior CRISPR‐Cas9 Genome Editing in T Cells via DTCP

CRISPR‐Cas9 introduces double‐strand breaks (DSBs) that are repaired via endogenous DNA repair machinery, generating insertions or deletions (indels) that disrupt gene function [45]. In T cells, such genome edits can eliminate receptor expression; for instance, TRAC knockout eliminates the endogenous TCR, whereas PDCD‐1 knockout suppresses PD‐1 signaling (Figure 5A). We hypothesized that the DTCP platform would enable efficient delivery of CRISPR‐Cas9 RNP complexes into T cells, thereby supporting the development of allogeneic CAR‐T products (via TRAC disruption) and enhancing therapeutic potency (through PD‐1 disruption).

FIGURE 5.

CRISPR‐Cas9 RNP‐mediated genome editing in primary T cells for TRAC and PDCD‐1 knockout. (A) Schematic illustration of gene knockout via CRISPR‐Cas9 RNP delivery. (B) Representative T7E1 assay gels and TRAC knockout efficiency for three independent donors treated with electroporation (EP) or DTCP (NC: negative control). (C) Representative T7E1 assay gels and PDCD‐1 knockout efficiency for three independent donors treated with EP or DTCP. (D) Relative PD‐1 expression in untreated control (NC) and DTCP‐treated cells without (left) and with (right) restimulation. (E) Normalized PD‐1 knockout efficiency in three independent donors without (left) and with (right) restimulation, based on PD‐1 immunofluorescence staining. (F) Relative PD‐1 expression in untreated control (NC) and EP‐treated cells (EP) without (left) and with (right) restimulation. (G) Schematic illustration of simultaneous multiplexed gene knockout using two CRISPR‐Cas9 RNPs. (H) Representative T7E1 assay gels and TRAC knockout efficiency in independent donors treated with EP or DTCP. (I) Representative T7E1 assay gels and PDCD‐1 knockout efficiency in independent donors treated with EP or DTCP. n.s. indicates no statistical difference; ^*, ^**, ^***, and ^**** denote p values <0.05, 0.01, 0.001, and 0.0001, respectively. Statistical analyses were performed using one‐way ANOVA. Error bars represent the mean ± standard deviation (SD), and each dot indicates an individual donor.

Using optimized mechanoporation parameters, CRISPR‐Cas9 RNPs (100 pmol per 10⁷ cells), targeting either TRAC or PDCD‐1, were delivered into T cells at a density of 5 × 10⁷ cells/mL. For comparison, electroporation (EP) was performed at the manufacturer's recommended density of 2 × 10⁷ cells/mL for T cells, with an identical molar RNP dose per cell. The donor source and activation duration were matched between platforms to ensure direct comparability. Gene disruption was quantified three days post‐delivery using a T7E1 assay on extracted genomic DNA [37]. The representative gels exhibited substantially higher editing efficiencies with DTCP: a TRAC knockout efficiency of 55.8% compared with only 17.3% with EP (Figure 5B). Similarly, PDCD‐1 disruption reached 38.6% with DTCP vs. 16.7% with EP (Figure 5C), corresponding to a greater than a twofold improvement at both loci.

We then examined whether genome disruption resulted in corresponding suppression at the protein level, focusing on PD‐1 expression. Because PD‐1 expression is activation‐dependent [46], cells were either left untreated or restimulated with CD3/CD28 following RNP delivery. In untreated controls, restimulation increased PD‐1 expression from 12.3% to 37.7% (p = 0.01). In contrast, the DTCP‐treated cells exhibited only a marginal increase from 4.9% to 7.6% (p = 0.14), while maintaining significantly lower expression than controls under both conditions (resulting in knockout efficiencies of 60.7% and 78.7%, Figure 5D,E). By comparison, the electroporated cells showed a pronounced PD‐1 upregulation, with expression increasing to 3.3% without restimulation and 23.0% with restimulation (p = 0.03, Figure 5F). Mock electroporation controls confirmed this effect, showing elevated PD‐1 levels even without RNP delivery, consistent with electroporation‐induced stress and post‐transcriptional upregulation (Figure S6A).

Cell viability analysis further highlighted the cytocompatibility advantage of DTCP. Without restimulation, DTCP‐treated cells maintained 95.6% viability, compared with 79.1% for EP‐treated cells. Following CD3/CD28 restimulation, the difference widened further (94.9% for DTCP vs. 41.7% for EP), highlighting the cytotoxic effects of electroporation (Figure S6B). These findings demonstrate that electroporation not only compromises baseline viability but also sensitizes cells to activation‐induced cytotoxicity, whereas DTCP preserves both survival and functional resilience.

Finally, to evaluate multiplex genome editing capacity, we co‐delivered sgRNAs targeting TRAC and PDCD‐1 at a 1:1 ratio, while maintaining a constant total RNP concentration. DTCP achieved dual knockout efficiencies of 51.0% for TRAC and 45.1% for PDCD‐1, compared with 22.9% and 18.1%, respectively, using electroporation (Figure 5G–I). This substantial gain underscores the ability of DTCP to support combinatorial genome editing without compromising viability or scalability.

To further resolve multiplex genome editing outcomes at the single‐cell level, we analyzed 2D flow cytometry distributions of simultaneously measured TCR and PD‐1 expression, enabling direct visualization of combinatorial knockout states within individual cells following multiplex editing (Figure S7A). Representative 2D dot plots (Figure S7B–D) illustrate the distribution of editing outcomes across three independent donors. Quantitative analysis revealed that the double‐positive (TCR⁺ PD‐1⁺) population was reduced by approximately 61.5%, decreasing from 11.87% to 4.57%. When normalized to the initial condition, TCR expression decreased by 41.2% (from 80.20% to 47.14%), while PD‐1 expression was reduced by 55.6% (from 17.60% to 7.81%). In contrast, the double‐negative (TCR⁻ PD‐1⁻) population increased substantially, from 14.08% to 49.67%, consistent with efficient and balanced dual gene disruption (Figure S7C).

These results establish DTCP as a robust and clinically relevant alternative to electroporation for CRISPR‐Cas9 genome editing in primary human T cells. By enabling efficient single and multiplexed knockouts while preserving high viability and minimizing stress‐induced transcriptional artifacts, DTCP provides a powerful platform for safe and scalable genome engineering relevant to allogeneic CAR‐T manufacturing.

2.6. Assessment of T Cell Phenotypic Stability and Genomic Integrity Post‐Treatment

To ensure translational and clinical applicability, DTCP treatment must not compromise T cell viability, genomic integrity, or phenotype. We therefore conducted a comprehensive safety assessment, comparing DTCP to electroporation by considering key parameters including viability, proliferation, DNA damage, DNA repair responses, and subset composition.

Longitudinal monitoring of cell viability and proliferation revealed distinct differences in recovery profiles between platforms (Figure 6B,C). DTCP‐treated cells exhibited only a transient decrease in viability to 74.8% at day 2 post‐treatment, followed by a recovery to 78.0% by day 4, with stable viability thereafter. In contrast, EP‐treated cells showed a continuous decline, decreasing from 83.9% to 68.5% over 16 days. Measurement of viable cell density every two days confirmed a generally slow expansion across groups owing to the absence of additional stimulation after initial CD3/CD28 activation. By day 16, the fold expansion was 6.54 in untreated controls, 5.93 for DTCP‐treated cells, and 5.69 for EP‐treated cells, indicating no significant differences under baseline conditions. Upon CD3/CD28 restimulation at day 8, DTCP‐treated cells maintained >80% viability through day 16, whereas EP‐treated cells declined sharply to 64.2% (Figure S8A). Correspondingly, relative viable cell density showed >9.2‐fold expansion in DTCP‐treated cells compared with <7.8‐fold for EP‐treated cells, suggesting that electroporation impairs long‐term proliferative resilience (Figure S8B).

FIGURE 6.

Post‐treatment analysis of T cells following DTCP and electroporation to evaluate platform stability. (A) Schematic illustration of post‐transfection analyses. (B) Viability and (C) viable cell density tracked over 8 days after treatment with EP or DTCP. Statistical analyses were performed using one‐way ANOVA. (D) Representative confocal images and (E) flow cytometry quantification of PerCP‐conjugated γH2A.X immunofluorescence staining to assess DNA damage (NC: negative control; EP: electroporation; DTCP: droplet cell mechanoporation; PC: positive control). (F) RT‐qPCR analysis of DNA repair‐associated gene expression (XPA) following EP, DTCP, and positive control treatments. Statistical analyses were performed using Levene's test. (G) Representative flow cytometry histograms of surface CD4 and CD8 expression from four independent donors following EP or DTCP treatment. (H) Quantification of CD4⁺/CD8⁺ subset ratios across donors. n.s. indicates no statistical difference; ^*, ^**, ^***, and ^**** denote p values <0.05, 0.01, 0.001, and 0.0001, respectively. Error bars represent the mean ± standard deviation (SD), and each dot indicates an individual donor.

As genome instability or unintended DNA damage can bias editing outcomes [47], we assessed whether DTCP or electroporation introduced detectable nuclear damage. Positive controls were generated by exposing T cells to ultraviolet C (UVC) for 1 h, a condition known to induce DNA lesions [48, 49]. Samples collected 3 h post‐treatment were stained for phosphorylated histone H2A variant X on serine 139 (γH2A.X) and counterstained with DAPI to visualize the DNA damage foci (Figure 6D). Confocal imaging showed minimal γH2A.X signal in untreated, DTCP‐, and EP‐treated cells, whereas UVC‐exposed controls exhibited strong, nuclear‐localized foci. Flow cytometry across four independent donors confirmed no increase in DNA damage in DTCP‐treated samples, whereas 32.8% of EP‐treated donors showed a modest but detectable increase (Figure 6E).

To further probe the subtle perturbations in DNA repair pathways, we quantified the expression of a representative repair gene, XPA, which mediates the recognition of UV‐induced DNA lesions [50]. RT‐qPCR analysis (normalized to GAPDH) showed minimal perturbation following DTCP treatment. XPA expression increased by only ∼6.4% in DTCP‐treated cells, whereas EP‐treated cells showed ∼40.4% upregulation (Figure 6F). Given this variability, mean‐based comparisons alone were insufficient; we therefore applied Levene's test to evaluate the variance across conditions. When comparing XPA expression levels, DCP‐treated cells were statistically indistinguishable from untreated T cells, whereas EP‐treated cells and UVC‐positive controls showed significant variance. Gel electrophoresis confirmed successful amplification of GAPDH and XPA across all donors, ruling out gene loss or degradation as a possible explanation for the reduced expression in selected EP‐treated samples (Figure S9). These findings indicate that DTCP preserves DNA repair homeostasis, further supporting its safety profile.

Finally, we evaluated T cell subset composition, as balanced CD4⁺/CD8⁺ populations are critical for ACT efficacy [51]. Flow cytometry analysis of 5,000 cells per donor (Figure 6G) showed that fluorescence intensity, corresponding to antibody‐derived signals for surface CD4 and CD8 expression, was largely preserved in DTCP‐treated cells relative to untreated controls, with only minor shifts in intensity. These histograms therefore reflect the population‐level distribution of CD4⁺ and CD8⁺ expression levels across individual T cells. In contrast, electroporated cells exhibited more pronounced alterations, including the emergence of a distinct subpopulation beyond 10² fluorescence units. Across the four donors, the CD4⁺/CD8⁺ ratios remained broadly comparable between treatments and controls, although EP‐treated cells showed a modest reduction in CD8⁺ proportions (Figure 6H).

Overall, these results demonstrate that DTCP preserves T cell viability, proliferation, genomic integrity, DNA repair stability, and subset composition. Together, this comprehensive safety profile supports the use of DTCP as a robust and clinically compatible platform for therapeutic T cell manufacturing.

3. Conclusion

This study establishes the DTCP platform as a robust and scalable technology for engineering primary human T cells. Through re‐engineering the platform for T cell‐specific processing, delivery efficiencies exceeding 98% and relative cell viabilities above 90% were achieved upon delivering 2000 kDa FITC‐dextran, confirming efficient intracellular delivery without compromising cell viability. Notably, the platform maintained high performance under high cell density processing (4 × 10⁸ cells/mL), supporting its suitability for the rapid generation of clinically relevant T cell numbers. Beyond the current device configuration, a recently reported parallelized microfluidic design [52] is expected to further increase processing throughput and support the handling of substantially larger cell populations.

We further demonstrated the platform's versatility by delivering multiple biomolecular modalities. The delivery of GFP‐CD19‐CAR mRNA resulted in approximately 90% transfection efficiency, validating DTCP as an effective, non‐integrating approach for transient receptor expression. For CRISPR‐Cas9 RNP delivery, DTCP consistently outperformed electroporation, yielding up to a 2.4‐fold higher knockout efficiency at both TRAC and PDCD‐1 loci, including simultaneous multiplexed disruption. These results highlight DTCP's capacity to support efficient genome engineering in primary T cells.

Longitudinal assessment further confirmed that DTCP‐treated cells maintained high viability and stable proliferation for more than one week following treatment. Genomic integrity was preserved, as indicated by minimal γH2A.X induction and stable expression of a representative DNA repair‐associated gene. Phenotypic analysis showed maintenance of CD8⁺ cytotoxic T cell proportions, indicating retention of key functional characteristics following delivery.

Beyond demonstrating delivery efficiency, this work advances mechanoporation from platform‐level characterization toward biologically and translationally relevant applications. By validating DTCP in primary human T cells, demonstrating CAR expression, clinically relevant genome editing (TRAC and PDCD1), multiplexed knockout, and immune‐cell‐specific and phenotypic readouts, this study positions DTCP as a practical enabling technology for next‐generation T cell engineering.

Given its high editing performance, DTCP is expected to be compatible with emerging genome editing modalities, such as base [53] and prime editors [54]. While this study focuses on delivery efficiency and editing outcomes, future work should extend toward functional validation, including cytotoxicity assays, alloreactivity assessments, and in vivo efficacy studies. Prior reports indicate that partial disruption of key loci can attenuate alloreactive responses and enhance antitumor function, underscoring the potential translational relevance of the editing efficiencies achieved here [55, 56]. Systematic definition of functional thresholds relevant to clinical performance, therefore represent an important next step. Beyond T cells, the gentle delivery mechanism and scalability of DTCP suggest broad applicability to other hard‐to‐transfect cell types, such as NK cells, γδ T cells, and stem cells, supporting the development of next‐generation cell‐based immunotherapies.

4. Experimental Section/Methods

4.1. Fabrication of Microfluidic Device

The microchannel mold was prepared by etching a silicon wafer using a deep reactive ion etching (DRIE) Bosch process, and Polydimethylsiloxane (PDMS; Sylgard 184, Dow Corning, USA) microfluidic channels were replicated using standard soft lithography. The inlets and outlets were created by punching holes into the PDMS using a pin vise and the device was cleaned using isopropanol and tweezers. PDMS layers were bonded to standard glass slides via oxygen plasma treatment (CUTE, Femto Science, Republic of Korea), followed by baking for at least 24 h in a 75°C or 2 h in a 200°C hot plate to ensure robust adhesion and hydrophobicity.

4.2. Preparation of Primary Human T Lymphocytes

Peripheral blood mononuclear cells (PBMCs) were acquired in accordance with Institutional Review Board approval (SMC 2021‐01‐091). CD3‐positive T cells were isolated via MACS (Miltenyi Biotec, Germany) using LS columns and CD3‐conjugated MicroBeads. Isolated T cells were provided by Prof. D. Cho (Samsung Medical Center, Sungkyunkwan University). For activation, T cells were cultured in X‐VIVO 10 serum‐free medium (Lonza, Switzerland) supplemented with 25 µL/mL ImmunoCult human CD3/CD28 T cell activator (STEMCELL Technologies, Canada) and 13 IU/mL recombinant human IL‐2 (Peprotech, USA) at 1 × 10⁶ cells/mL in 24‐well plates at 37°C with 5% CO₂. Cell proliferation and size distribution were quantitatively monitored using a LUNA‐FX7 automated cell counter (Logos Biosystems, Republic of Korea).

4.3. Intracellular Delivery Procedure

Droplets were generated using Bio‐Rad droplet generation oil filtered through a combination of hydrophilic and hydrophobic polytetrafluoroethylene bead filters (Advantec, Taiwan). Activated T cells were suspended with functional molecules and loaded into Luer‐Lock syringes (BD, USA). Opti‐MEM (Thermo Fisher, USA) was used for the delivery of eGFP mRNA (996 nt, TriLink, USA), anti‐CD19 CAR mRNA (provided by MxT Biotech, Republic of Korea), and CRISPR‐Cas9 RNP complexes. FITC‐dextran (3–2000 kDa; Sigma‐Aldrich, USA) was resuspended in X‐VIVO medium with T cells at a concentration of 2 × 10⁷ to 4 × 10⁸ cells/mL.

CRISPR‐Cas9 RNP complexes were assembled using 30 µm Alt‐R S.p. Cas9 nuclease V3 protein and 10 µm sgRNA (IDT, USA). The TRAC sgRNA sequence was CAGGGTTCTGGATATCTGTGGG, and the PDCD‐1 sgRNA sequence was GTCTGGGCGGTGCTACAACT. Protein and sgRNA were mixed at a 1:1 molar ratio and incubated at room temperature for 10 min. Delivery was performed at 100 pmol per 1 × 10⁷ cells. For dual knockout experiments, each sgRNA was used at half the single‐knockout amount. Syringe pumps (Harvard Apparatus, USA) were used to control the oil and cell‐cargo suspension flow. Droplets were coalesced using 1H,1H,2H,2H‐Perfluoro‐1‐octanol (Sigma‐Aldrich, USA), and the cells were washed and resuspended in the X‐VIVO medium supplemented with IL‐2. Cells were incubated for 18, 24, and 72 h after FITC‐dextran delivery, mRNA transfection, and CRISPR‐mediated gene knockout, respectively. Electroporation controls were performed using the Neon transfection system (Thermo Fisher, USA) following the manufacturer‐recommended conditions for primary human T cells (1600 V, 10 ms, 3 pulses) at 2 × 10⁷ cells/mL.

4.4. Flow Cytometry Assay

Cells were harvested and suspended in a flow cytometry buffer (Invitrogen, Thermo Fisher Scientific) with 10 µg/mL of PI (Lonza, Switzerland) to assess the delivery efficiency and cell viability. CAR expression was evaluated using an APC‐conjugated anti‐FMC63 antibody (1:50; Acro Biosystems, USA). Immunophenotyping employed a PE‐conjugated anti‐human CD3 antibody, a FITC‐conjugated anti‐human CD4 antibody, a PE‐conjugated anti‐human CD8a antibody, and a FITC‐conjugated anti‐human PD‐1 antibody (BioLegend, USA). Data were acquired using a Guava EasyCyte (Luminex, USA) and a FACSLyric (BD, USA). Gated cells were analyzed for fluorescence signals surpassing the top 5% of the control cells.

For DNA damage analysis, T cells were irradiated with UVC and treated with 1% H₂O₂ for 1 h as positive controls. Untreated, DTCP‐treated, EP‐treated, and positive control cells were permeabilized and fixed 3 h post‐treatment, stained with a PerCP‐eFluor conjugated γH2A.X antibody (eBioscience, USA), washed, and analyzed via flow cytometry.

4.5. Confocal Imaging of γH2A.X

Cells were prepared as described above, with additional DAPI staining for nuclear visualization. Cells were attached to poly‐L‐lysine‐coated glass slides via centrifugation (Cyto‐Centrifuge TXT3, Nasco, Republic of Korea) and imaged using a Leica DM6 B upright fluorescence microscope (Leica Microsystems, Germany).

4.6. RT‐qPCR for DNA Repair‐Associated Genes

Total RNA was extracted using the Monarch Total RNA miniprep kit (Monarch, NEB, USA), converted into cDNA, and quantified using SYBR Green‐based RT‐qPCR (QuantStudio, Thermo Fisher). Expression levels were normalized to those of GAPDH.

4.7. DNA Substrate Construction and Primer Design

gDNA was extracted using the QuickExtract DNA extraction solution (Biosearch Technologies, UK), and PCR amplification was performed with Quick Taq HS DyeMix DNA polymerase (TOYOBO, Japan) using the following primers: TRAC forward; CAC TGA AAT CAT GGC CTC TTG G, TRAC reverse; GGG CTT AGA ATG AGG CCT AGA A, PDCD‐1 Forward; CCT GCC CAG GAG CAA AGA, and PDCD‐1 reverse; CAC GTG GAT GTG GAG GAA GAG. PCR conditions: 94°C 2 min (pre‐denaturation), followed by 30–40 cycles of 94°C 30 sec, 53°C 30  s, 68°C 1 min. PCR products were purified using silica columns (Monarch, NEB).

4.8. T7E1 Assay

100 ng of PCR amplicons were denatured at 95°C for 5 min, re‐annealed via gradual cooling (95°C–85°C at 2°C/s; 85°C–25°C at 0.1°C/s). The products were digested with T7 endonuclease 1 (NEB, USA) at 37°C for 15 min and resolved on 1% agarose gels stained with ethidium bromide or SYBR gold (Thermo Fisher). Bands were visualized using the iBright imaging system (Thermo Fisher).

4.9. High‐Speed Microscopy and Imaging

High‐speed images were recorded using a Phantom VEO 701L (Vision Research, USA), and fluorescence images were captured with an Axio Observer A1 (Carl Zeiss, Germany). ImageJ software was used for image analysis. The flow cytometry data were analyzed with Guava software, Incyte 3.3 and FlowJo (Becton Dickinson, USA).

4.10. Statistical Analysis

Statistical analyses were conducted using one‐way ANOVA, and post‐hoc comparisons were performed using Tukey's HSD and Bonferroni tests (SPSS v27, IBM, USA). A p‐value <0.05 was considered statistically significant. Data visualization was performed using OriginPro (OriginLab, USA) and FlowJo (Becton Dickinson); the error bars indicate the mean ± standard deviation (SD).

Funding

National Research Foundation of Korea (NRF) grants funded by the Korean government (MSIT; Ministry of Science and ICT) (2021R1A2C2006224, RS‐2023‐00242443, and RS‐2023‐00218543, A.J.C.).

Conflicts of Interest

A.J.C. declares the following competing interest. A financial interest in MxT Biotech, which is commercializing the proposed technology.

Supporting information

Supporting File: smll72716‐sup‐0001‐SuppMat.pdf.

Supporting File: smll72716‐sup‐0002‐MovieS1.mp4.

Supporting File: smll72716‐sup‐0003‐MovieS2.mp4.

Data Availability Statement

Data from this study are available from the corresponding author upon reasonable request (Email: ac467@korea.ac.kr).