Perfusion microbioreactor for CAR-Treg manufacturing

Summary

Manufacturing cell and gene therapies (CGTs) at scale presents challenges in cost, product consistency, and adaptability to personalized treatments. Traditional large-volume bioreactors are designed to support cell growth through controlled nutrient delivery and gas exchange, but are poorly suited to the decentralized, small-batch production required for personalized therapies such as chimeric antigen receptor (CAR) T cells. To address this, we have developed the KCL-Microbioreactor (K-MBR), a closed microbioreactor platform based on microfluidic principles. Engineered in polydimethylsiloxane (PDMS), the K-MBR combines spatial confinement, semi-continuous perfusion, and integrated viral transduction in a compact footprint, enabling efficient gene delivery and robust expansion of therapeutic cells. We demonstrate the platform’s utility by generating functional CAR-Tregs targeting HLA-A2, achieving a 92% increase in yield compared to conventional methods. The K-MBR offers a streamlined solution for CGT manufacturing, with potential to reduce production costs and enhance scalability across a broad range of cell therapies.

Graphical abstract

Highlights

• •Perfusion microbioreactor achieves Treg expansion comparable to gold standard G-Rex device
• •Spatial confinement increases lentiviral transduction efficiency of primary human cells
• •Compact, low-volume platform reduces the physical footprint of cell manufacturing
• •Device supports future automation and advances progress toward point-of-care production

Biological sciences

Introduction

The first engineered cell-based therapy was brought to market in 2017, with the FDA approval of Kymriah, a chimeric antigen receptor (CAR) T cell therapy for the treatment of B-cell acute lymphoblastic leukemia (B-ALL). Since then, there have been additional CAR-T cell therapies approved by the FDA for the treatment of hematological malignancies such as lymphoma, leukemia, and multiple myeloma. By genetically engineering patient T cells to express a CAR for a tumor-associated antigen (TAA), CAR-T cells could recognize and induce apoptosis in tumor cells with high specificity. CAR-T cell therapy has been demonstrated to have remarkable efficacy as a cancer immunotherapy, with remission rates exceeding 90% in certain patient groups.¹

While CAR-T cells are typically derived from a broad, undefined range of T cell subsets, often referred to collectively as pan T cells, there is a growing field of evidence supporting the use of CAR regulatory T cells (Tregs) to modulate immune responses.² Derived from patient Tregs, which play a fundamental role in regulating immune cell functions, CAR-Treg therapy can be used to alleviate immune-mediated diseases by resolving aberrant immune responses. Alleles encoding the MHC class I protein have been a key focal point of preclinical validation studies, with a range of allograft models demonstrating long-term persistence and robust efficacy of HLA-A2-specific CAR-Tregs (A2-CAR Tregs).^3,4,5

Ongoing first-in-human studies are now investigating the clinical applicability of A2-CAR Tregs in the context of solid organ transplantation and autoimmune diseases. STEADFAST (TX200-TR101) is a phase I/II clinical trial, assessing the safety and tolerability of A2-CAR Tregs in HLA-A2 mismatched renal transplantation.⁶ A preclinical assessment of TX200-TR101 demonstrated robust prevention of graft-versus-host disease (GvHD) in a xenogeneic HLA-A2⁺ transplant model, highlighting the capacity of antigen-specific CAR-Tregs to promote immune tolerance.⁷ A similar ongoing phase I/II clinical trial, LIBERATE (QEL-001), is investigating the safety and clinical activity of A2-CAR-Tregs in A2-mismatched liver transplantation, with early data suggesting the therapy to be safe and well tolerated.⁸ These findings support ongoing efforts to translate CAR-Treg therapy into the clinic for targeted, antigen-driven immunosuppression.

One of the key drawbacks that restricts the adoption of CAR-T cell therapies as a first- or even second-line treatment is the high price point that significantly restricts patient accessibility. These therapies are typically autologous, requiring a patient-specific manufacturing process involving skilled labor and expensive clinical-grade reagents that are costly and difficult to scale. This involves isolating and activating T cells, using expensive viral vectors for stable gene expression, and expanding the cells ex vivo over 6 to 14 days.

Closed-system cGMP-compliant platforms such as the CliniMACS Prodigy enable automated, end-to-end production of CAR-T cells⁹ and clinical-grade CAR-Tregs.¹⁰ These systems offer potential for decentralized, point-of-care manufacturing but face challenges due to high upfront costs and lower process efficiencies compared to modular platforms.¹¹ These large-scale bioreactors use large fluid volumes to maintain cell concentrations and nutrient supply, essential for producing CAR-based therapy doses exceeding 10⁹ cells. However, this reduces culture efficiency, raising costs and lowering yields. In these systems, viral transduction, with high-cost clinical-grade lentiviruses, often results in low CAR-positivity rates (∼30%), requiring high virus titers and extended expansion times to achieve therapeutic doses.¹²

Compared to such large-scale bioreactors, microbioreactors are emerging as alternative platforms for cell therapy manufacturing, using significantly smaller working volumes that offer increased process control and improved efficiencies. The integration of microfluidic technology in biomanufacturing has the potential to significantly streamline the production of CAR-T cell therapies by providing increased precision and control over fluid dynamics and cell manipulation. By maintaining cells at a high density, microfluidic systems can enhance the efficiency of key processes, while allowing more accurate monitoring and regulation of culture conditions. Microfluidic technology also offers the possibility of parallelizing multiple steps within the manufacturing process, allowing for a more continuous and automated production line, thus minimizing the need for extensive manual intervention.

A number of microfluidic approaches have been developed which support modalities of a CAR-T cell manufacturing workflow including, but not limited to, cell isolation, high-efficiency gene delivery and rapid expansion.^13,14 A microfluidic device developed by Moore et al., demonstrates the ability to use transmembrane flow to co-localize cells and virus particles, driving high-efficiency viral transduction of primary, human T cells.¹⁵ The use of inertial microfluidics can be used for the sorting of specific cell populations, with Elsemary et al. demonstrating this approach for the purification of viable CAR-T cells.¹⁶ While such devices could be used in the context of CAR-T cell manufacturing, the majority of these systems are confined to individual cell culture operations.

Commercial microbioreactors such as the Quantum Cell Expansion System (Terumo BCT), C.NEST (Cytena), and BioLector XT (Beckman Coulter) enable high-throughput cell expansion with in-line monitoring. However, they lack the capabilities required to support a broad spectrum of cell culture operations, limiting their suitability as end-to-end platforms for CAR-T cell manufacturing.^17,18 Only a small number of benchtop microbioreactors have been commercialized with additional features tailored to CAR-T workflows. The Mobius Breez (Merck) is a 2 mL single-use system designed for high-throughput perfusion culture, integrating microfluidics and sensing controls. Although early studies indicate its potential for CAR-T cell production,¹⁹ evidence supporting its compatibility with Tregs is limited. Moreover, the system is not currently GMP-compliant, necessitating further validation and process transfer before use in clinical or regulatory-grade settings.

While existing microbioreactors support many steps of the CAR-T cell manufacturing workflow, they are not optimized for the specific needs of Treg culture and expansion, with limited published evidence demonstrating their suitability for CAR-Treg production. This highlights the need for a microbioreactor designed to address these challenges and improve CAR-Treg manufacturing outcomes.

In order to overcome the aforementioned shortcomings of existing bioreactors, we have developed a closed-system perfusion microbioreactor integrating the activation, transduction and expansion steps required for the manufacturing of functional CAR-Tregs from primary, human Tregs. By confining cells to a small fluidic volume while optimizing culture medium exchange parameters, this system could be used to facilitate increases in both cell transduction efficiency and expansion capacity compared to conventional systems. Using an optimized workflow, the K-MBR supported a CAR-Treg manufacturing workflow with a 50% increase in expansion capacity and a 92% increase in CAR-Treg yield, when compared to standard tissue culture plasticware. Culture in the device had no observable impact on cell phenotype, with generated A2-CAR-Tregs having the ability to potently suppress T cell proliferation in an antigen-dependent manner.

Results

Device design and fabrication

A systematic, iterative approach was used to design and fabricate a polydimethylsiloxane (PDMS)-based microbioreactor (K-MBR) with a 1 mL liquid volume capable of supporting cell culture operations required in the CAR-Treg manufacturing process. The K-MBR was composed of two PDMS components, comprising a lower perfusion chamber and an upper bioprocessing chamber, separated by a porous, polycarbonate membrane (Figure 1). The deposition of a thin silica layer on each side of the polycarbonate membrane allowed the exposure of hydroxyl groups (-OH) following oxygen plasma treatment, facilitating the formation of strong covalent bonds with the PDMS components of the device. The use of alternative chemical bonding methods, such as dipodal silanes or adhesives, resulted in membrane fouling or loss of membrane integrity, especially when working with fragile membranes. By using silica-sputtering, the membrane could be integrated in a leak-free manner without any observable impact on membrane integrity (Figure S1).

Figure 1.

Microbioreactor fabrication

(A) Orthogonal view of the microbioreactor, showing multiple components of the device.

(B) Side-view of microbioreactor demonstrates two sets of ports used for A) perfusing the system with fresh culture medium or B) seeding the device with cells.

(C) Microbioreactor components.

(D) Assembled microbioreactors, filled with culture medium.

The device was designed to support long-term cell culture directly on the membrane surface, with a 5 mm gaseous headspace in the bioprocessing chamber allowing for aeration and efficient agitation. The small membrane pore size of 0.05 μm ensured the confinement of the virus particles in the bioprocessing chamber enhancing their interactions with cells to support high-efficiency lentiviral transduction.

Each chamber was fitted with ports at either end of the device which could be closed during cell culture to maintain sterility outside of a laminar flow hood. During cell culture, spent culture medium in the perfusion chamber could be replaced at regular intervals or continuously, ensuring optimal nutrient replenishment and waste product removal rates were met. The design features of the device, housing the perfusion chamber directly below the porous membrane, resulted in a short diffusion pathway between the two chambers, thus maximizing the exchange of solutes.

The K-MBR supports the culture of primary, human Tregs

Freshly isolated human Tregs were introduced into the device and cultured for a 12-day period with a manual 50% culture medium exchange every 3 days, using establsihed culture conditions (Table 1). Throughout the culture, cells were sampled at regular intervals to establish both the cell counts and viability (Figures 2A and 2B); device performance was compared to a 24-well plate with the same cell seeding conditions and medium exchange regime. The K-MBR could support the long-term culture of human Tregs, with no observed differences in viability and mean fold expansion, compared to standard tissue culture plasticware. Flow cytometric analysis was used to investigate cell phenotype, with a gating strategy established to quantify proteins of interest (Figure S2), using the appropriate antibody panel (Table 2). At day 12 post-seeding, there was no observed effect on cell purity (CD4⁺CD25⁺ double-positive cells) or phenotype (Figure 2C).

Table 1.

Culture media and supplementation used for different cultured cell types

Cell type	Culture medium	Supplementation
Tregs	X-VIVO 15	IL-2 (1,000 IU/mL) Rapamycin (100 nM)
T cells	X-VIVO 15	IL-2 (100 IU/mL)
B-LCLs	RPMI-1640	Heat-inactivated FBS (10% v/v) Penicillin (100 IU/mL) Streptomycin (100 μg/mL) GlutaMAX (1 mM)

Figure 2.

K-MBR supports Treg expansion

(A–C) Compared to a 24-well plate, the K-MBR device had no adverse impact on cell (A) proliferation, (B) viability, or (C) phenotype.

(D) over a 12-day period (n = 3 devices). Cells could be re-stimulated and maintained in culture for an additional 7 days without affecting cell proliferation. Statistical significance was determined by one-way ANOVA for timepoint experiments; for day 12 analysis; for flow cytometric analysis t-tests with Šídák correction for multiple comparisons; error bars representing mean ± SEM.

Table 2.

Antibody panel for T cell/Treg characterization

Marker	Fluorophore	Clone	Supplier
CD4	BV605	SK3	BioLegend
CD25	PE	BC96	BioLegend
CD69	PE-Cy7	FN50	BioLegend
CD127	BV785	A019D5	BioLegend
HLA-A2	PE	BB7.2	Bio-Techne
FoxP3	Alexa Fluor 647	206D	BioLegend
Ki-67	BV510	11F6	BioLegend
CTLA-4	PerCP-Cy5.5	BNI3	BioLegend

Following a 12-day culture period, cells were harvested, re-suspended in fresh culture medium, and re-stimulated in their respective culture vessels. Cells were then cultured for an additional 7 days, to investigate the potential of the device to support a second round of simulation and sustained proliferation within a clinically relevant timeframe of less than 3 weeks from initial isolation. The purpose of this assessment was therefore to evaluate the maintenance of proliferative capacity after re-stimulation, rather than to define the long-term growth kinetics. Robust expansion was demonstrated following re-stimulation, showing that Tregs could be maintained in culture in the K-MBR for multiple rounds of stimulation, without impacting the proliferative capacity of the cells compared to standard tissue culture plasticware (Figure 2D).

Rapid Treg expansion with optimized feeding regime

Metabolite analysis revealed a significant build-up of lactate and ammonia (Figure S3), alongside a depletion of glucose, with a 50% culture medium exchange every 3 days. Based on this information, an optimized regime was developed to maintain the lactate concentration below 1.5 g/L to prevent cell proliferation from being stunted (Figures 3A and 3B). This threshold was not selected on the basis of lactate being intrinsically inhibitory to Treg expansion, but rather as an operational limit to prevent nutrient depletion, waste accumulation, and acidification in a system without active pH control. While recent work has in fact demonstrated that lactic acid supplementation can enhance Treg expansion,²⁰ when pH is tightly regulated, in the K-MBR, rising lactate levels were closely associated with glucose depletion and medium acidification (data not shown). In the first three days of culture, 12.5% of the culture medium was exchanged on a daily basis. From day 3 until day 9, 50% of the culture volume was exchanged daily. In the final 3 days, 25% of the culture medium was exchanged each day. This increasing frequency of culture medium exchange allowed efficient Treg expansion in the K-MBR, achieving a 10-fold expansion over a 12-day period (Figure 3C). Treg expansion in the K-MBR was significantly higher than in a 24-well plate when the same feeding regime was used, suggesting additional factors limiting Treg expansion rate in the plate.

Figure 3.

Optimized Treg expansion in K-MBR

(A) Feeding regime optimization for a 12-day Treg expansion; VVD = vessel volumes per day.

(B) An increased culture medium replacement frequency (regime 2) resulted in a reduction in lactate build-up during Treg culture, in a 24-well plate.

(C) Treg expansion in the K-MBR compared to a 24-well plate; data from independent experiments with cells from 3 healthy donors (n = 9 devices).

(D and E) (D) Treg expansion and (E) day 12 viability in the K-MBR compared to the G-Rex 24-well plate (n = 2 devices). Statistical significance was determined by one-way ANOVA for timepoint comparisons and an unpaired t test for day 12 viability; error bars represent mean ± SEM. ∗∗: p ≤ 0.01.

In a separate experiment, long-term Treg culture in the K-MBR was also compared to the G-Rex 24-well plate, using an optimized cell expansion protocol. (ScaleReady). By integrating a silicone membrane, G-Rex devices enable the rapid expansion of suspension cells and are deemed as being the gold standard in the field for cell expansion. Across a 12-day culture period, the K-MBR achieved robust Treg expansion in-line with the G-Rex, with no impact on cell viability (Figures 3D and 3E). Furthermore, the K-MBR operated with 5.1 mL of culture medium, significantly less than the 20 mL required for the G-Rex expansion protocol (8 mL starting volume with two 6 mL media exchanges).

The K-MBR supports the production of CAR-Tregs

In order to demonstrate the ability of the K-MBR to generate CAR-Tregs, the 12-day workflow was adapted to include a 72-h lentiviral transduction step (between day 3 and day 6 of culture), whereby lentivirus particles encoding a CAR targeting HLA-A2 (A2-CAR) (Figure 4) were introduced into the bioprocessing chamber of the device (Figure 5A). Following the 12-day process, Tregs expanded to a greater extent in the K-MBR compared to a 24-well plate (Figure 5B) with no adverse impacts on cell phenotype as they maintained a high purity (CD4⁺CD25⁺) with robust FoxP3, CTLA-4, and Ki-67 expression (Figures 5C and 5F).

Figure 4.

Schematic of HLA-A2 CAR construct

CAR-chimeric antigen receptor; eGFP- enhanced green fluorescent protein. Adapted from Boardman et al. (2017)³

Figure 5.

CAR-Tregs generation with K-MBR

(A) CAR-Treg 12-day manufacturing workflow, with manual steps and media exchanges shown; LV = lentivirus.

(B) Treg expansion.

(C–E) Day 12 assessment of (C) cell phenotype, (D) transduction efficiency, and (E) total CAR-Treg yield. Two independent experiments with cells from 2 healthy donors (n = 6 devices) (F) representative flow plots for Treg phenotyping. Statistical significance was determined by one-way ANOVA for timepoint experiments; for day 12 analysis; for flow cytometric analysis t-tests with Šídák correction for multiple comparisons; error bars representing mean ± SEM. ∗: p ≤ 0.05, ∗∗: p ≤ 0.01 and ∗∗∗: p ≤ 0.001.

With an MOI of 5, a transduction efficiency of 67% could be achieved in the K-MBR, significantly higher than what was observed in the 24-well plate control (Figure 5D). By confining cells and virus particles within a small fluid volume of 0.5 mL in the bioprocessing chamber, an increase in cell-virus interactions was likely responsible for this improved transduction efficiency. While similarly low volumes can, in principle, be used in standard 24-well plates, maintaining such volumes over multiple days is constrained by evaporative loss, which leads to osmotic and pH shifts that can adversely affect cell health. In our system, the closed-system nature of the device, together with the regular media exchange regime, supported adequate nutrient and gas exchange while keeping evaporation to a minimum, allowing the small working volume to be maintained stably throughout culture. In a separate experiment, introducing virus particles into the perfusion chamber of the device did not yield transduction of cells in the bioprocessing chamber, demonstrating that the membrane acted as a barrier preventing the movement of virus particles between chambers (Figure S4). Using this manufacturing protocol, the K-MBR could generate CAR-Tregs with a 92% increase in yield, compared to the 24-well plate (Figure 5E).

A2-CAR-Tregs suppress T cell proliferation in vitro

Following the 12-day workflow, Tregs were harvested and sorted based on their expression of the reporter gene enhanced GFP (eGFP), allowing the purification of A2-CAR-Tregs. Sorted A2-CAR-Tregs were co-cultured with HLA-A2⁺ or HLA-A2^- B-LCLs for a 48-h period, with flow cytometric analysis used to determine CAR-mediated Treg activation via the expression of activation marker CD69. Significant Treg CD69 upregulation was observed in the presence of HLA-A2 expressing B-LCLs compared to Tregs cultured with HLA-A2 negative B-LCLs (Figure 5A). This upregulation of CD69 was observed to the same extent in CAR-Tregs generated in the K-MBR as those generated in the 24-well plate, suggesting that the CAR was successfully expressed following culture in both vessels.

In order to assess the ability of the CAR-Tregs to suppress the proliferation of effector T cells (Teffs), a suppression assay was performed. Prior to use in the assay, freshly isolated Teffs were stimulated with TransAct, in order to drive robust expansion beyond that achievable with Dynabeads.²¹ Following isolation and in vitro expansion, Teffs were frozen prior to being used for suppression assays. Cryopreserved donor-matched Teffs were thawed and rested for 24 h before being used. Tregs were co-cultured with Teffs in the presence of either HLA-A2⁺ or HLA-A2^- B-LCLs. CAR-Tregs were generated from HLA-A2^- Tregs to ensure that CAR-Tregs were not stimulated by one another. As co-cultured Teffs did not express a synthetic receptor targeting HLA-A2. The use of HLA-expressing B-LCLs ensured CAR-mediated activation in the Tregs without Teff stimulation. After a 5-day co-culture, Teff proliferation was assessed by the expression of the cell proliferation marker CellTrace Violet. At day 5, it was observed that >95% Tregs exhibited eGFP expression, demonstrating long-term, stable CAR expression. As expected, the A2-targeted CAR-Tregs suppressed the Teff proliferation more potently in the presence of HLA-A2 antigen (Figure 6B). CAR-Tregs generated in the K-MBR were highly functional, showing the ability to suppress the proliferation of Teffs to the same degree as CAR-Tregs generated in a 24-well plate (Figure 6C), quantified by analyzing the cell trace violet (CTV) signal in proliferating Teffs (Figure 6D). These results demonstrate the ability of the K-MBR to generate CAR-Tregs, with no impact on cell viability, phenotype, or function compared to standard tissue culture plasticware.

Figure 6.

Suppression ability of K-MBR-generated Tregs is preserved

(A) Surface expression of activation marker CD69 assessing the ability of CAR-Tregs to respond to target antigen (n = 2 devices). Suppression assay with cells from 2 healthy donors (n = 6 devices) in independent experiments to assess.

(B) CAR-Tregs suppress T cell proliferation in response to the target antigen in a 24-well plate.

(C) Impact of K-MBR culture on CAR-Treg suppressive capacity in the presence of either HLA-A2⁺ B-LCLs or HLA-A2^- B-LCLs.

(D) Representative flow cytometry plots of proliferating Teffs following a 5-day suppression assay. Statistical significance was determined by mixed-model two-way ANOVA for suppression assays; for CD69 expression analysis, t-tests with Šídák correction for multiple comparisons; error bars representing mean ± SEM. ∗: p ≤ 0.05 and ∗∗∗∗: p ≤ 0.0001.

Discussion

The widespread adoption of CAR-based therapies is hampered by the complexity and cost of the multi-step manufacturing process. The manipulation of a large fluid volume in a typical bioreactor gives rise to batch heterogeneity and poor process efficiency. In addition, such bioreactors typically have high upfront costs and require a need for intensive labor, further driving up running costs. The K-MBR addresses these shortcomings by demonstrating the potential of a device that supports the efficient generation of functional CAR-Tregs using a significantly smaller working volume. It was demonstrated that the K-MBR could be used to achieve the production of a homogenous CAR-Treg population, which suppressed the proliferation of Teffs in an A2-dependent manner.

The 1 mL volume, perfusion microbioreactor was developed to allow the high-density culture and manipulation of primary, human Tregs. One of the key challenges faced in developing a microbioreactor for the culture of suspension cells was integrating a means of cell retention to facilitate culture medium perfusion in the device. The use of a porous membrane allowed Tregs to be maintained in culture while performing the replacement of culture medium from the perfusion chamber below. By designing the device with two separate fluidic chambers, culture medium could be replaced without the risk of exerting shear stress on the cells. While culture medium was replenished manually in a laminar flow hood, the integration of microfluidic ports gives rise to the potential of culture medium perfusion being automated using a fluidic pump system.

Tregs could be stimulated and cultured in the device for a 12-day period, with no impact on viability or phenotype compared to those cultured in standard tissue culture plasticware. Using an optimized feeding regime, we were able to achieve a 150% increase in Treg fold-expansion in the K-MBR compared to a standard feeding protocol expanding cells from 1 x 10⁶ to almost 1 x 10⁷ in 12 days. Interestingly, we did not observe an increase in Treg proliferation in plasticware with this optimized regime, suggesting that there were additional factors limiting Treg growth. These data demonstrate the ability of the K-MBR to support robust Treg expansion, supporting a high cell density exceeding that achievable in standard tissue culture plasticware.

While the rate of nutrient replacement and waste removal was kept constant between the K-MBR and the 24-well plate, differences in materials and geometries between the two culture vessels likely impacted the proliferative capacity of cells in culture. The use of thin PDMS substrates (<2 mm) meant that gas exchange could occur between the headspace and the surrounding environment in the incubator at a sufficient rate to support rapid cell expansion. Unlike in a 24-well plate, where gas exchange only occurs at the gas-liquid interface, within the K-MBR, gas exchange also occurs through the sides of the device. The increased Treg growth rate observed in the K-MBR may be the result of an increased oxygen transfer rate (OTR); further experiments with the use of gas sensors could be implemented to characterize these potential differences in gas exchange rates. The K-MBR supports the rapid expansion of Tregs, in line with the capabilities of the G-Rex rapid expansion platform. The degree of expansion achieved in the K-MBR was done so with 25% of the required 20 mL in the G-Rex, demonstrating the potential of manufacturing steps being performed in the microbioreactor in a low-cost manner.

Tregs could be transduced with high efficiency in the K-MBR, with an almost 2-fold increase in CAR-Treg yield compared to standard tissue culture plasticware. The increase in transduction efficiency observed in the K-MBR was likely due to an increased degree of spatial confinement compared to the 24-well plate. While viruses of the Retroviridae family have been demonstrated to travel in solution by Brownian motion, estimates suggest that the relatively short half-life of these viruses limits their displacement to around 500-600 μm in a single half-life.^22,23 As a result, a decrease in liquid height is typically associated with increased cell-virus interactions, as suspended virus particles are more likely to co-localize with sedimented cells.

With the 12-day CAR-Treg manufacturing workflow, a 6-fold expansion rate could be achieved. While significantly higher than the expansion capacity achieved in standard plasticware, a drop in expansion was observed compared to the 10-fold expansion reached with wild-type (WT) Tregs. The incorporation of viral transduction into the workflow required a washing step to be performed at day 6, whereby cells were removed from the device and washed to remove residual virus. This step, in combination with the potential adverse effects of the lentiviral vector on cell health, likely explains the drop in the expansion capacity of the engineered Tregs compared to WT Tregs. Further device iterations could incorporate a means of flushing out virus particles from the device while ensuring cell retention in the bioprocessing chamber, therefore removing the need for a manual washing step.

CALR-Tregs generated in the K-MBR potently suppressed the proliferation of Teffs in a 24-well plate, with no observable impact on cell function compared to those generated in standard plasticware. While it has been demonstrated that CAR-T cell phenotype/function can differ dramatically based on the device used to carry out the manufacturing process,¹¹ these data suggest that the effector functions of the CAR-Tregs were not affected. The suppression observed was highly dependent on the presence of HLA-A2 on the surface of the target cells, demonstrating the role of the A2-CAR in driving CAR-Treg activation and function. As the most polymorphic MHC allele group, with over 540 reported alleles, HLA-A2 mismatches are common in the etiology of GvHD and transplant rejection.²⁴ A2-CAR-Tregs could be used to directly target and suppress the proliferation of A2+ T cells in the donor tissues or exert bystander suppression in the local milieu surrounding the tissue, to prevent the onset of GvHD and promote tissue tolerance.

Existing bioreactors used for autologous cell therapy manufacturing are limited in their ability to support the generation of CAR-T cells in a highly efficient, closed-loop manner. While such closed-loop bioreactors are commercially available, they are plagued by batch heterogeneity, low process efficiency, and still require a degree of manual intervention. We demonstrate that the K-MBR perfusion microbioreactor can generate functional CAR-Tregs targeting HLA-A2 in a highly efficient manner. The perfusion microbioreactor supports high transduction efficiency and robust cell expansion, comparable with the G-Rex platform. With a culture surface area of 1.5 cm², the K-MBR was able to support a cell density up to 6.5 x 10⁶ cells/cm², compared to 5.0 x 10⁶ cells/cm² in the G-Rex platform, across a 12-day culture period. By maintaining cells at a high density, our approach generates CAR-Tregs with a reduced physical footprint compared to that of large-scale bioreactors. By further reducing space and infrastructure requirements, it brings point-of-care cell therapy production closer to feasibility, streamlining workflows and enabling a more accessible, patient-specific approach. Moreover, our system has the potential to be fully automated, aligning with the growing demand for more controlled and reproducible operations in the CAR-Treg manufacturing process.

While further steps are necessary to demonstrate that such microbioreactors can be scaled up further to reach the high final cell numbers achievable in large-scale bioreactors, this research paves the way for the development of a closed-loop microbioreactor suitable for clinical point-of-care CAR-Treg production.

Limitations of the study

One key limitation of this study is that all experiments were performed using cells from healthy donors, and the performance of the K-MBR with patient-derived cells, including those from immunocompromised or clinically relevant disease settings, remains to be determined. While the platform was able to support high-density cell culture and efficient CAR-Treg manufacturing at the laboratory scale, further work is required to demonstrate robust scale-out, integration of automation, and compatibility with fully cGMP-compliant workflows. In addition, although comparable outcomes were observed across devices, a more comprehensive assessment of inter-device variability and long-term process reproducibility is necessary to support clinical manufacturing. Functional characterization was limited to in vitro assays, while in vivo persistence, stability, and safety of CAR-Tregs generated in the K-MBR were not evaluated.

Resource availability

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Ciro Chiappini (ciro.chiappini@kcl.ac.uk).

Materials availability

Materials from this study are available through Ciro Chiappini upon request.

Data and code availability

All data reported in this article will be shared by Ciro Chiappini upon request; this article does not report original code. All datasets generated and analyzed in this study, including raw flow cytometry files and source data, are available from the lead contact upon reasonable request.

Acknowledgments

C.C. acknowledges funding from the European Union under the ERC Starting Grant ENBION 759577 and the medical research council under the Confidence in Concept award (MC_PC_18052); MFX (MicrofluidX) Ltd. and W.E. acknowledge funding from the MRC DTP iCASE Studentship Scheme.

Author contributions

W.E.- contributed methodology, investigation, formal analysis, project administration, and writing of original draft; N.S.- contributed methodology, investigation, and formal analysis; Y.W.- contributed methodology, investigation, and formal analysis; Y.L.- contributed methodology, investigation, and formal analysis C.W.- contributed methodology and investigation; D.M.- contributed methodology and investigation; C.S.- contributed methodology and investigation; M.R.- contributed methodology and investigation; C.M.C and A.E. - contributed conceptualization and consultation as well as provided additional funding and in-kind contributions, e.g., use of company facilities and resources; G.L.- contributed supervision, methodology, formal analysis and project administration; C.C.- conceptualization, funding acquisition, methodology, project administration, supervision, visualization, and writing of original draft.

Declaration of interests

G.L. is a Founder and consultant of Quell Therapeutics. C.M.C. and A.E. are founders and shareholders of MFX Ltd. The other authors declare no interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Brilliant Violet 605™ anti-human CD4 Antibody (SK3)	Biolegend	Cat#344645; RRID: AB_2734347
PE anti-human CD25 Antibody (BC96)	Biolegend	Cat#302605; RRID: AB_314275
PE/Cyanine7 anti-human CD69 Antibody (FN50)	Biolegend	Cat#310911; RRID: AB_314846
Brilliant Violet 785™ anti-human CD127 (IL-7Rα) Antibody (A019D5)	Biolegend	Cat#351329; RRID: AB_11219610
PE anti-human HLA-A2 Antibody (BB7.2)	Biolegend	Cat#343305; RRID: AB_1877228
Alexa Fluor® 647 anti-human FOXP3 Antibody (206D)	Biolegend	Cat#320113; RRID: AB_439753
Brilliant Violet 510™ anti-human Ki-67 Antibody (11F6)	Biolegend	Cat#350517; RRID: AB_2941433
PerCP/Cyanine5.5 anti-human CD152 (CTLA-4) Antibody (BNI3)	Biolegend	Cat#369607; RRID: AB_262967
Bacterial and virus strains
HLA-A2 CAR second-generation pLNT/SSFV	Boardman et al.3	N/A
Biological samples
Peripheral blood leukocyte cones	NHSBT Tooting	Cat#NC24
Chemicals, peptides, and recombinant proteins
CellTrace™ Violet Cell Proliferation Kit	ThermoFisher	Cat#C34557
Dynabeads™ Human T-Activator CD3/CD28	Gibco	Cat#11131D
T cell TransAct™, human	Miltenyi	Cat#130-111-160
RosetteSep™ Human CD4+ T cell Enrichment Cocktail	Stemcell	Cat#15062
Lymphoprep™	Stemcell	Cat#18060
CD25 MicroBeads II, human	Miltenyi	Cat#130-092-983
LS columns	Miltenyi	Cat#130-042-401
Recombinant Human IL-2 Protein	R&D Systems	Cat#202-IL
Rapamycin	LC Laboratories	Cat#R5000200MG
PEG-it™	System Biosciences	Cat#LV810A-1
LIVE/DEAD™ Fixable Near-IR Dead Cell Stain Kit	ThermoFisher	Cat#L34975
Experimental models: Cell lines
SPO (HLA-A2+DR11+) B-LCLs	Boardman et al.3	N/A
BM21 (HLA-A2−DR11+) B-LCLs	Boardman et al.3	N/A
Software and algorithms
FlowJo	FlowJo	https://www.flowjo.com/flowjo/download
Prism	GraphPad	https://www.graphpad.com/features
Leica LasX	Leica Microsystems	https://www.leica-microsystems.com/products/microscope-software/p/leica-las-x-ls/downloads/

Experimental model and study participant details

Epstein-Barr virus (EBV) transformed cell lines

Cryopreserved B-lymphoblastic cell lines (B-LCLs) were thawed and re-suspended at 0.2 x 10⁶ cells/mL in supplemented RPMI-1640 with the addition of 1 mM glutaMAX (Gibco), 10% heat-inactivated foetal bovine serum (FBS), 100 IU/mL penicillin and 100 μg/mL streptomycin (ThermoFisher, UK). Cells were maintained in culture (5% CO₂, 37°C) at an appropriate density and expanded to reach suitable cell numbers prior to use. The Epstein-Barr Virus (EBV) transformed cell lines SPO B-LCLs (HLA-A2⁺) and BM21 B-LCLs (HLA-A2^-) (Merck) are part of the Human Leukocyte Antigen (HLA) Typed Collection maintained by the European Collection of Cell Cultures (ECACC) and were obtained from Boardman et al.³ These cells were originally derived from a peripheral B cell line established from a male donor. Before use, cryopreserved cell lines were thawed and re-suspended in pre-warmed culture medium before being rested overnight. The next day, cells were collected and re-suspended in fresh culture medium to remove residual DMSO, prior to use for cell culture experiments. Cell line authentication testing was not performed. All cell lines tested negative for mycoplasma contamination prior to experimental use.

Primary T-cells

For cell culture experiments, Tregs were re-suspended at 1 x 10⁶ cells/mL in X-VIVO 15 culture media (Lonza) supplemented with 1,000 IU/mL IL-2 and 100 nM rapamycin. Human T-Activator CD3/CD28 Dynabeads™ (Thermo Fisher Scientific) were added at a 1:1 cell to bead ratio and cells were maintained in culture for up to 12 days. Dynabeads were selected as a source of CD3/CD28 antigens for Tregs, with data demonstrating their ability to drive robust expansion compared to other stimulation methods.²⁵ At day 12, Dynabeads™ were magnetically removed and cells were washed and re-suspended in culture medium at 1× 10⁶ cells/mL with the addition of fresh Dynabeads™ when re-stimulation was required. For culture in the G-Rex 24-well plate, an optimised cell expansion protocol was used (ScaleReady), whereby 1 x 10⁶ Tregs were seeded in 8 mL, with the replacement of 6 mL culture medium at day 4 and day 8.

Teffs were re-suspended at 1 x 10⁶ cells/mL unless otherwise stated, in X-VIVO 15 culture medium supplemented with 100 IU/mL IL-2. T cell TransAct™ (Miltenyi Biotec) was added at according to the manufacturer’s protocol (1:100 titre) and cells were maintained in culture for up to 12 days. Re-stimulation was performed by washing cells and re-suspending in culture medium with fresh T cell TransAct™. The various supplemented culture media used for different cell types is summarised (Table 1).

Method details

Replica moulding and K-MBR assembly

A CO₂ laser-ablated (Trotec) polymethylmethacrylate (PMMA) negative mould was designed and fabricated for each of the two PDMS components of the K-MBR microbioreactor. Following replica moulding, 2 mm port holes were punched in the top PDMS substrate, which contained the bioprocessing chamber. These ports made it possible to access both the bioprocessing and perfusion chambers of the device, allowing fluid manipulation.

Membrane silica deposition

To facilitate the bonding between the polycarbonate membrane and PDMS substrates, a silica (SiO₂) intermediate layer was deposited on both sides of the membranes using a previously described approach.²⁶ A magnetron SF sputterer (Korvus Technology) was used (100 W RF power, 35 min on each side, with an argon flow rate of 32 sccm and oxygen flow rate of 8 sccm) to deposit SiO₂ on both sides of the polycarbonate membrane, resulting in a SiO₂ intermediate layer with a thickness of around 100 nm on each side of the membrane. The sputtered membrane was laser cut to size before being exposed to oxygen plasma on both sides (100W, 50sccm, 1 min) along with the two PDMS components. Device components were brought into conformal contact with one other and the assembled device was incubated at 80°C for 30 min to ensure strong bonding between the device components.

K-MBR device preparation and use

Assembled devices were fitted with polyetheretherketone (PEEK) tubing (Cole-Parmer), secured using silicon adhesive, and attached to silicon tubing (VWR) allowing fluid manipulation in a leak-free manner. For cell culture experiments, ports were closed using luer plugs (microfluidic Chip Shop) to maintain sterility in the devices. Prior to use, the K-MBR was either flushed with 70% ethanol or steam autoclaved at 121°C for 15 min, before being rinsed thoroughly with sterile PBS and equilibrated with culture medium for a minimum of 2 h.

Cell handling

All aseptic cell culture was performed in class II biological safety cabinets, with cultured cells being kept in all cases in incubators at 37°C, supplied with 5% CO₂. Cells were typically washed and pelleted using 300g centrifugation at 4°C, unless stated otherwise. Cell counts were obtained either manually using trypan blue dye (Sigma-Aldrich) and a haemocytometer, or with a NucleoCounter NC-202 (ChemoMetec). For cell counts and viability measurements prior to experimental endpoints, cell suspensions were mixed thoroughly by pipetting, and a 20 μL aliquot was taken for analysis.

Primary T cell isolation

Cells were isolated from leukocyte cones obtained from the NHSBT (National Health Service Blood and Transplantation, Tooting, London, UK), obtained from anonymised healthy volunteers with informed consent. Peripheral blood from leukocyte cones was diluted 1:1 in PBS and incubated at room temperature for 20 min with RosetteSep Human CD4⁺ T cell Enrichment Cocktail (StemCell Technologies). The reagent contains antibody complexes that crosslink CD4⁻ cells with glycophorin A on erythrocytes, allowing the isolation of CD4⁺ cells by negative selection. Following incubation, blood samples were diluted further with PBS and layered onto Lymphoprep density gradient medium (StemCell Technologies). Samples were then centrifuged at 600g for 20 min with low acceleration and deceleration, facilitating the separation of lymphocytes by density gradient centrifugation. CD4⁺ peripheral blood lymphocyte cells (PBMCs) were isolated from the interphase layer, washed twice with PBS and re-suspended in magnetic-activated cell sorting (MACS) buffer- PBS supplemented with 0.5% bovine serum albumin (BSA) and 5 mM Ethylenediaminetetraacetic acid (EDTA), at 1 x 10⁷ cells/mL.

The cell suspension was mixed with CD25 Microbeads II (Miltenyi Biotec) and incubated at 4°C for 15 min. Cells were washed and re-suspended at 2 x 10⁸ cells/mL in MACS buffer before being passed through a MACS LS column placed on a magnetic stand (Miltenyi Biotec). The column was rinsed three times with MACS buffer, washing out CD25⁻ cells while capturing bead-bound CD25⁺ cells in the column. The column was then removed from the magnetic stand and a plunger was used to force 5 mL MACS buffer through the column, eluting the CD25⁺ fraction. This CD4⁺CD25⁺ population has been demonstrated to exhibit high expression of the canonical Treg marker FoxP3. CD4⁺CD25⁺ Tregs were used fresh for all cell culture experiments while CD4⁺CD25⁻ Teffs were either used fresh for cell culture experiments or cryopreserved for use in suppression assays. For cryopreservation, cells were re-suspended between 1 x 10⁷ and 5 x 10⁷ cells/mL in freezing medium composed of 90% heat-inactivated fetal bovine serum (FBS) (Gibco) and 10% dimethyl sulfoxide (DMSO) (Sigma-Aldrich). Cells were stored at −80°C overnight before being transferred to liquid nitrogen for long-term storage.

Flow cytometry

For phenotypic analysis using flow cytometry, between 0.1 - 0.5 x 10⁶ cells were re-suspended in 100 μL PBS per sample with the addition of antibodies for antigens of interest, according to the manufacturers’ protocols. For viability testing, LIVE/DEAD™ Fixable Near-IR Dead Cell Stain (Invitrogen) was included in the panel at a dilution of 1:1,000. Samples were incubated at 4°C for 20 min before being washed with 1 mL PBS. For intracellular staining, cells were fixed and permeabilised using Foxp3/Transcription Factor Fixation/Permeabilization kit (Invitrogen) for 30 min before being washed and re-suspended in 100 μL permeabilisation buffer containing intracellular antibodies. Samples were incubated at 4°C for 30 min before being washed with PBS and re-suspended in 300 μL PBS before acquisition on the BD LSRFortessa analyser (BD Biosciences). A panel of anti-human antibodies was used for the characterisation of human T-cells and Tregs, to assess cell purity and function (Table 2).

Anti-FoxP3, Ki-67 and CTLA-4 antibodies were used to quantify the intracellular expression of these proteins; all others listed were used to quantify surface marker expression.

Fluorescence-activated cell sorting (FACS)

Following the 12-day full CAR-Treg manufacturing process, harvested Tregs had their anti-CD3/CD28 Dynabeads™ removed and were re-suspended at 1 x 10⁷ cells/mL in PBS and sorted into sterile tubes containing X-VIVO 15 culture medium based on eGFP expression using a BD FACS Aria II (BD Biosciences). Sorted GFP⁺ CAR-Tregs were re-suspended in X-VIVO 15 at the required concentration and used for downstream operations.

Purity testing and HLA typing

Following cell isolation and prior to cell culture experiments, cells were stained and flow cytometric analysis was performed to assess cell viability and purity. Following the selection of single, live lymphocytes, using the aforementioned gating strategy, Treg purity was assessed by quantifying the surface expression of CD4 alongside the expression of additional markers- CD25, CD127 and FoxP3.

For cell culture experiments that required HLA haplotyping, a 10 μL blood sample was taken during the cell isolation process and incubated with 200 μL Ack Lysing buffer (Gibco) for 5 min to lyse red blood cells (RBCs). Samples were washed with PBS and labeled with anti-human HLA-A2 (Bio-Techne) for flow cytometric analysis.

Antigen specific activation

Both HLA-A2⁺ and HLA-A2^- B-LCLs were firstly incubated at 1 x 10⁷ cells/mL in 50 μg/μL mitomycin C (Sigma-Aldrich) solution for 50 min at 37°C before being washed with PBS and re-suspended in X-VIVO 15. By inhibiting DNA synthesis, mitomycin C was used to inhibit B-LCL cell proliferation without impacting their ability to stimulate Tregs. To determine CAR-specific Treg activation, sorted A2 CAR-Tregs were co-cultured with either the HLA-A2⁺ or HLA-A2^- B-LCLs at a ratio of 2:1 Tregs to B-LCLs, in a 96-well round bottomed plate.

Following 48 h of culture, cells were harvested and washed with PBS before being stained for LIVE/DEAD, CD4 and CD69, as previously described. Cells were washed and re-suspended in PBS before acquisition, to assess the degree of antigen specific Treg activation. CAR-Tregs were identified as live CD4⁺ cells expressing the lentiviral-CAR (LV-CAR) eGFP reporter gene, with the quantification of CD69 surface expression demonstrating the degree of Treg activation. Tregs were also cultured in the absence of B-LCLs, to identify the baseline CD69 expression of unstimulated Tregs.

Treg suppression assay

Sorted CAR-Tregs were re-suspended at 1 x 10⁶ cells/mL in X-VIVO 15 and serial diluted in wells of a 96-well round bottomed plate. Autologous Teffs were incubated at 1 x 10⁷ cells/mL in 5 μM cell trace violet (CTV) solution at 37°C for 20 min before being topped up with culture medium and incubated for a further 5 min at 37°C. CTV staining can be used to monitor cell proliferation, with each subsequent generation of proliferating cells expressing a lower CTV signal as a result of dye dilution.²⁷ Following staining, Teffs were re-suspended in X-VIVO 15 at 2 x 10⁶ cells/mL and added to the plate, with 50 μL CTV-labelled Teffs per well. By serial diluting the CAR-Tregs, a range of Treg:Teff ratios could be achieved in the plate, spanning from 1:1 to 1:64.

Mitomycin C treated B-LCLs, either HLA-A2⁺ or HLA-A2^- were re-suspended at 1 x 10⁶ cells/mL and 50 μL cell suspension was added per well, to achieve a final volume in each well of 200 μL. Cells were co-cultured for a 5-day period before being analysed with flow cytometry to assess the degree of Teff proliferation in each sample. The degree by which CAR-Tregs suppressed Teff proliferation was determined by taking the inverse of Teff proliferation (CTV^lo) and normalising it to Teff proliferation when stimulated with anti-CD3/CD28 Dynabeads™ at a 40:1 cell to bead ratio.

A gating strategy was used to firstly identify eGFP⁻ Teffs before gating on CTV^lo cells to quantify the degree of Teff proliferation and hence the suppressive capacity of the co-cultured CAR-Tregs.

Metabolite analysis

For metabolite analysis, a minimum volume of 150 μL supernatant was collected from cell culture vessels and analyzed using the BioProfile FLEX2 (Nova Biomedical) with chemistry and osmometry modules. For supernatant sampling in the K-MBR, spent culture medium was collected from the perfusion chamber. When working with sample sizes <150 μL, samples were diluted with Milli-Q water, with calculations performed accordingly to account for the dilution factor.

Chimeric antigen receptor (CAR) construct

A previously described lentiviral HLA-A2 -CAR, with a CD28 hinge domain, CD28⁻CD3ζ signalling domain and an enhanced green fluorescent protein (eGFP) reporter gene under an SFFV promoter³ (Figure 4) was used for the transduction of primary, human Tregs to generate CAR-Tregs targeting HLA-A2. As previously described, HEK293T cells were co-transfected with the transfer plasmid, pΔ8.91 and pCMV-VSV-G plasmids at a mass ratio of 4:3:1 using polyethylenimine (3:1 PEI:DNA wt/wt; Sigma-Aldrich, Gillingham, Dorset, UK). Viral supernatant was collected 48–56 h post-transfection, and lentiviral particles were concentrated using PEG-it™ (System Biosciences, Bar Hill, Cambridgeshire, UK).

Lentiviral transduction

For the transduction of primary human T-cells and Tregs, cells were typically stimulated for 72 h prior to being washed and re-suspended in fresh culture medium. Viral particles were thawed and added directly to the cell suspension at the required concentration to achieve the desired MOI. When performing the full CAR-Treg manufacturing process in the K-MBR microbioreactor and 24-well plate control, cells were counted at day 3 post-seeding and the viral vector was added directly to the culture vessel at the required concentration, following a 50% culture medium exchange.

Following the required transduction period, typically 72 h unless stated otherwise, cells were harvested and washed. For the full CAR-Treg manufacturing process in the K-MBR microbioreactor and 24-well plate, cells were re-suspended in fresh culture medium and returned to the corresponding culture vessel. One of two methods was used for determining transduction efficiency, depending on the LV-CAR used. For cells transduced with the HLA-A2 LV-CAR, transduction efficiency was determined by quantifying eGFP reporter gene expression using flow cytometry.

Virus precipitation

PEG-it virus precipitation solution (System Biosciences) was used for the precipitation of virus particles from cell culture supernatants. Supernatants were re-suspended in PEG-it solution and refrigerated at 4°C overnight. 12 h after re-suspension, the supernatant-containing solution was centrifuged to pellet the virus particles (1,500 x g, 30 min, 4°C) and all residual fluid was carefully aspirated. Lentivirus particles were then re-suspended in X-VIVO 15 culture media at the desired concentration.

Bright-field and immunofluorescence

Both bright-field and immunofluorescence imaging were performed on a Leica DMi8 inverted microscope (Leica Microsystems). The tile scanning feature was used in the Leica LAS X software, for bioprocessing chamber live imaging.

Scanning electron microscopy

Scanning electron microscopy (SEM) images were captured using a Carl Zeiss XB1540 Crossbeam SEM/FIB equipped with an InLens detector, operated at 3 kV. Prior to imaging, the samples were sputter-coated with gold using a sputter coater (Edwards Ltd.).

Quantification and statistical analysis

Flow cytometric analysis

A typical manual gating strategy was used for flow cytometric analysis with FlowJo software, with sequential gating performed on lymphocytes, single cells and live cells prior to the quantification of any other markers. This gating strategy was used in all flow cytometric analyses, unless otherwise stated, prior to the analysis of additional marker expression levels.

Statistical analysis

All data plotted in graphs were presented as mean values with standard error of the mean. Statistical tests were performed using Graphpad Prism, as described in figure legends. For timepoint experiments, assessing cell growth in the K-MBR and standard tissue culture plasticware, a one-way ANOVA was used. For the analysis of flow cytometric data at a single timepoint, for example, day 12 post-seeding, t-tests with Šídák correction for multiple comparisons was performed. For the comparison of Treg suppression following culture in the K-MBR and standard tissue culture plasticware, differences in curves were assessed with a mixed-model two-way ANOVA. Unless otherwise stated, plotted error bars for pooled data represent mean ± SEM. For all statistical tests performed, statistical significance was illustrated according to the following p-value cutoffs:

ns: p > 0.05; ∗: p ≤ 0.05; ∗∗: p ≤ 0.01; ∗∗∗: p ≤ 0.001; ∗∗∗∗: p ≤ 0.0001.

Published: March 5, 2026