Respiring cultureware for high-density, scalable, multipurpose cell-based bioproduction

TL;DR

A new type of cell cultureware improves oxygen delivery, enabling high-density 3D cell growth for various bioproduction applications.

Contribution

The development of a scalable, high-density cell respirator with artificial capillaries for efficient oxygen transport in 3D cell culture.

Findings

The HDCR achieves cellularities over 1E8 cells/cm³, near theoretical 3D confluence limits.

The design supports multiple culture types and scales consistently across different formats.

Applications in CAR-T, viral vectors, and antibody production demonstrate its multipurpose utility.

Abstract

In vitro tissue culture remains inefficient due to inferior oxygen transport in polystyrene versus native capillary beds. Taking a bioinspired approach, we engineered respiring cultureware capable of high density, 3D cell culture. Leveraging the oxygen permeability of silicone and finite element modelling, we designed micromolded membranes that provide high oxygen transport (local kLa equivalent >100/hr). The “high density cell respirator” (HDCR) microarchitecture comprises rows of silicone fins that protrude up from a base membrane. The fins act as artificial capillaries to oxygenate the niche between them, where cells expand. Cellularities of >1E8 cells/cm3 are routinely achieved across common cell lines, approaching theoretical limits of 3D confluence. HDCR cultureware is compatible with adherent, suspension, microcarrier, and spheroid cultures, and inherently linearly scalable due to the conserved geometry across 96-well, 24-well, and dish formats. Applications are explored across general cell culture, CAR-T, viral vector, and antibody, demonstrating utility for multipurpose, intensified bioproduction.

Figures

d](#F1)). We refer to this platform as “high-density cell respirator” (HDCR) cultureware. The membrane microarchitecture comprises a series of fins and grooves, analogous to a heat sink, but designed to conduct oxygen. Oxygen permeation through silicone is ~10x faster than through media [[8](#R8)]. Cells expand within the *replicative niche* of the groove, while the gas permeable fins shuttle oxygen from the atmosphere into the culture and further act to confine 3D cell growth below oxygen diffusion limits. Cells are thus oxygenated directly from the bottom of the device, decoupling the relati

d](#F2)), and 13cm tissue culture dish was designed and fabricated based on this microarchitecture and produced via liquid silicone rubber (LSR) injection molding. HeLa GFP+ cells were cultured in the devices as proof-of-concept and displayed robust growth, eventually occupying the 3D volume of the niche by Day 7 ([Figure 2e](#F2)). The ability to seed, expand, harvest, and passage HeLa-GFP cells within the HDCR platform was demonstrated over 5 passages ([Figure 2f](#F2)). Based on cuboidal cell packing of the expansion niche, a theoretical 3D confluence limit of 1E7 cells/cm^2^ was estimated

c](#F3)). When using gas permeable cultureware, oxygen diffusion through the bottom membrane dominates, allowing large amounts of media to overlay cells and effectively decoupling oxygen and media considerations. While flat gas permeable membranes provide high oxygen transmission and are suitable for suspension cells [[11](#R11)], they allow adherent cells to form aggregates. Small aggregates can maintain high viability; however, as they merge and grow, they can become oxygen-diffusion limited and begin to form necrotic cores ([Figure 3d](#F3)). In contrast, the fin-niche architecture serves t

f](#F4)) versus HDCR cultureware ([Figure 4g](#F4)) and imaged with constant exposure settings on a fluorescent microscope. The cells exogenously expressed green fluorescent protein under the strong major immediate-early promoter and enhancer of the human cytomegalovirus (CMV) and thus acted as an easily visualized and quantitative model for exogenous protein production. Flow cytometry was performed to characterize protein production on a per cell basis, comparing cells cultivated on PS (harvested at 50% 2D confluence) to cells grown to 3D confluence in HDCR devices at day 7. Mean fluorescent

a](#F5)). These minimum recoverable seeding densities compare favorably (5 and 50-fold lower) to the limits reported with flat gas permeable devices (1.25E5 cells/cm^2^ and 110-fold expansion) [[12](#R12)].

a](#F6)). iPSC-derived neural progenitor cell (NPC) production was piloted in HDCR plates as a predictable and scalable alternative to a spheroid-based spinner flask process (Shi Lab, City of Hope National Medical Center, CA). Freshly dissociated NPCs were seeded into 24-well HDCR plates and expanded over 2-weeks with regular feeding and counts (n=2) to generate a growth curve ([Figure 6b](#F6),[c](#F6)). Cells were observed to follow an exponential growth curve with a consistent doubling time of 3.5 days. In subsequent scale up studies in HDCR 13cm dish dishes, the resulting NPCs met critical

Full text

Introduction

The in vitro culture of mammalian cells is a fundamental operation performed across the biological sciences, spanning basic research though biotherapeutics discovery, development, and manufacturing. Yet despite nearly 80 years of development since the advent of the Petri dish [1], cell culture remains a tedious and inefficient process. Polystyrene (PS) cultureware (e.g. tissue culture dishes, well-plates, flasks, and multilayer flasks) typically yield only 1E5 cells/cm^2^ [2] in 2D monolayer culture. In contrast, human tissue has an average cellularity of 4.6E8 cells/cm^3^ [3], enabled by 3D growth. It would take over 36× 15cm PS dishes to generate an equivalent number of cells to just 1 cm^3^ of tissue in vivo– which would involve 1.6lbs of plastic, 2 liters of media, 1 entire cell culture incubator, and 5 hours of technician touch-time.

The low yield of traditional cultureware arises from oxygen transport limitations due to its poor solubility in aqueous media compared to other nutrients like glucose and amino acids [4]. The media overlaying the cell monolayer imposes a barrier to oxygen diffusing from the atmosphere to the cells, limiting cell density. Shaker flasks or spinner flasks can provide mixing and agitation to better aerate the media; however, suspension culture is aphysiological for most cells. For cells that can be adapted to grow in suspension (e.g. CHO-S, HEK293-S), higher densities can be achieved through stirring, but ultimately shear forces limit maximum cell density by damaging cells [5].

A useful way to characterize cultureware and bioreactors cellular capacity is to consider their oxygen transfer rates – that is the amount of oxygen per volume per time they can provide. In stirred tank reactors, it is often decomposed as OTR = k_L_a*(C_O2,sat_ - C_O2,media_), where the volumetric mass transfer rate, k_L_a, characterizes the number of turnovers of a component per time, usually oxygen [6]. For static cultureware where cells grow in a particular region of the vessel, a local k_L_a equivalent or oxygen turnovers can be calculated for comparison (Appendix A). Taking HEK293 cells for example, which require 1.72E-13 mol O_2_ / hour [7], and allowing oxygen to reach as low as 1% in media from an initial 20.9% in atmosphere, a k_L_a of ~0.5 / hr would be required to support 1M cells/mL, which is typical for glass and PS cultureware (Figure 1a). A k_L_a of 5 / hr, which is typical of shaker and spinner flasks, would support 1E7 cells/mL (Figure 1b).

In nature, an intricately distributed capillary network brings oxygen and nutrient-rich blood into close proximity to cells for efficient diffusional exchange and insulates them from perfusion-induced shear forces (Figure 1c). This allows the human body to achieve an estimated k_L_a of 118/hr at rest (Appendix A). Given the remarkable cellularity achieved by the body compared to existing cultureware or bioreactors, we took a bio-inspired approach. We hypothesized that by recapitulating the capillary bed using highly gas permeable membranes to provide robust oxygenation and a 3D, shear-free microenvironment, cells could be cultivated to tissue-like densities. We aimed to support >1E8 viable cells/mL, representing a significant step towards theoretical and tissue-like 3D confluence limits, and consequently targeted a k_L_a of >50/hr.

The respiring cultureware we report on uses a micromolded gas-permeable silicone membrane optimized for solid-state oxygenation to support high density, 3D cell culture (Figure 1d). We refer to this platform as “high-density cell respirator” (HDCR) cultureware. The membrane microarchitecture comprises a series of fins and grooves, analogous to a heat sink, but designed to conduct oxygen. Oxygen permeation through silicone is ~10x faster than through media [8]. Cells expand within the replicative niche of the groove, while the gas permeable fins shuttle oxygen from the atmosphere into the culture and further act to confine 3D cell growth below oxygen diffusion limits. Cells are thus oxygenated directly from the bottom of the device, decoupling the relationship between media height and oxygen supply. This enables large amounts of media to overlay the cells and provide a reservoir for nutrient/waste exchange via diffusion, which are typically the next growth-limiting components after oxygen.

Results

HDCR Membrane Design, Simulation, and Proof-of-Concept

Capillary spacing in the body is driven by hypoxic signaling (e.g. HIF, VEGF) from cells to ensure uniform oxygenation. This varies by tissue type and metabolic demand but is typically less than 100μm [9]. In this work, we leveraged in silico finite element modelling (FEM) of oxygen transport to identify appropriate spacing and dimensions of the oxygenating fins (i.e. fin width, niche width, and niche/fin height). The goal of the simulation was to optimize the HDCR membrane microarchitecture to support maximal cell density per unit area while maintaining a uniform oxygen microenvironment. A multiphysics model (COMSOL Multiphysics, Sweden) of the “unit cell” architecture (Figure 2a) was developed, with the intention of patterning the design in X and Y to create a family of linearly scalable devices (Figure 2b). A Michaelis-Menten relationship for oxygen uptake rate (OUR) was utilized based on V_O2_ and K_M_, with a threshold for cell viability/hypoxia of 1% [10] along with additional parameters for mass transport in Appendix B - Table 1. A parametric sweep of geometry (500μm fin width × 200–800μm niche width × 400μm-1200μm fin height) identified several promising geometries (Figure 2c). Ultimately, a 400μm wide and 800μm high niche flanked by 500μm thick fins was selected as the microarchitecture to advance to manufacturing based on its high cell yield (predicted 3.6E6 cells/cm^2^) and uniform oxygen microenvironment.

A family of 96-well, 24-well (Figure 2d), and 13cm tissue culture dish was designed and fabricated based on this microarchitecture and produced via liquid silicone rubber (LSR) injection molding. HeLa GFP+ cells were cultured in the devices as proof-of-concept and displayed robust growth, eventually occupying the 3D volume of the niche by Day 7 (Figure 2e). The ability to seed, expand, harvest, and passage HeLa-GFP cells within the HDCR platform was demonstrated over 5 passages (Figure 2f). Based on cuboidal cell packing of the expansion niche, a theoretical 3D confluence limit of 1E7 cells/cm^2^ was estimated for a 15μm diameter cell (Figure 2g). Experimentally, the HeLa cells approached this value during the culture and even exceeded this value in P5 when fed with twice the usual media volume. Over each passage the cells were expanded >100-fold versus a typical 10-fold expansion in PS cultureware, highlighting the wider dynamic range afforded by the HDCR cultureware.

Oxygen Transmission Rates and Diffusion Limits Across Cultureware

The oxygen transmission rate of the HDCR silicone membrane was measured manometrically along with a flat silicone membrane (0.3mm thin) and PS dish (1mm thick) (Figure 3a). Both silicone membranes provided orders-of-magnitude higher oxygen transmission rates versus PS and the modelled oxygen transmission rate through 2mm of cell culture media. Whereas in manometric testing (air:membrane:air) the silicone fins pose a slight barrier to oxygen diffusion versus flat membranes, in practice (air:membrane:cells) they act as shunts bringing oxygen deep into the 3D cell niche due to the low permeability of cells and media. Simulated values of oxygen flux through each material with a 3.6E6 cells/cm^2^ cell load were recorded and converted to an equivalent local k_L_a over the volume occupied by the cells (Figure 3b). The HDCR architecture was estimated to afford a local, static k_L_a of 114/hr, similar to native capillary beds.

Oxygen diffusion through the overlaying media plays an important role in PS cultureware, accounting for >95% of oxygen transmission (with the balance permeating through the PS). This is sufficient to support monolayer culture but beyond this, cells begin to die (Figure 3c). When using gas permeable cultureware, oxygen diffusion through the bottom membrane dominates, allowing large amounts of media to overlay cells and effectively decoupling oxygen and media considerations. While flat gas permeable membranes provide high oxygen transmission and are suitable for suspension cells [11], they allow adherent cells to form aggregates. Small aggregates can maintain high viability; however, as they merge and grow, they can become oxygen-diffusion limited and begin to form necrotic cores (Figure 3d). In contrast, the fin-niche architecture serves to mechanically confine cell aggregates below an oxygen-diffusion-limited width of 400μm (Figure 3e) as predicted by simulations. This phenomenon is analogous to the hypoxic cores that develop in pancreatic islets during transplantation, which is known to occur around 200–250μm [12]. The confinement and close proximity to oxygenating fins ensures cells remain viable up to 3D confluence and makes the HDCR platform compatible with adherent and spheroid-like cell culture, as well as suspension cells.

Linear Scalability of HDCR Platform

Predictable linear scale up of cell culture remains a major impediment to translating cell and gene therapies and cell-based processes. The linear scalability of devices based on the HDCR microarchitecture was assessed through parallel HEK293T growth curves spanning 96-well, 24-well plates and 13cm dish formats. HEK293T cells were seeded into a reference 15cm PS dish (145cm^2^) or 13cm HDCR dish (136cm^2^) at a typical seeding density of 2.8E4 cells/cm^2^. Cells in the PS dish had hit 2D confluence by day 2 (1.8 ± 1.0E5 cells/cm^2^) and would normally be passaged, whereas the cells in the HDCR dish continued to expand exponentially out to day 7 (3.4 ± 0.5E6 cells/cm^2^), approaching near 3D confluence (Figure 4a). Similar HEK293T expansions were carried out in HDCR 24-well plates (1.9cm^2^/well) and HDCR 96-well plate (0.46cm^2^/well). Growth curves were substantially parallel across the device family and regression analysis showed a good exponential fit for 96-well (R^2^ = 97.6%), 24-well (R^2^ = 97.1%), and 13cm HDCR dish (R^2^ = 99.5%). The inferred cell doubling times of 22.6 hours, 24.4 hours, and 23.8 hours, respectively, are within 8% of each other, resulting in day 7 yields of 4.3E6, 2.9E6, and 3.4E6 cells/cm^2^. Thus, the platform affords predictable linear scalability.

The 13cm HDCR dish yielded 461M cells (equivalent to 1cm^3^ of tissue) versus the 20M cells grown in the reference 15cm PS dish (Figure 4b). The niches were observed to be full and uniform across the HDCR dish (Figure 4c). PS vs HDCR dish growth curves were repeated 3 separate times with technical triplicates to understand intra- and inter-run variability (Table 1). An average cell yield of 3.4 ± 0.5E6 cells per cm^2^ was achieved. Modelling had predicted 3.6E6 cells/cm^2^. This represents a 31x greater yield per area than PS dishes. Viability of the cells in the HDCR 13cm dish at 3D confluence (97±2%) was comparable to that of the viability of cells in the reference PS dish at 2D confluence (97±0.8%).

Characterization of Cell Cycle Beyond Conventional Confluence Limits

The cell cycle state of cells grown in PS and HDCR cultureware were compared by EdU staining and flow cytometry to evaluate what happens to cells grown beyond conventional 2D confluence limits (Figure 4e). At day 3, when cells grown in the PS dish had just surpassed 2D confluence, the cells were observed to remain proliferative with an S-phase population showing EdU staining indicative of robust DNA replication. Cells in the HDCR dishes showed similar staining. However, when cells in PS dishes were allowed to become significantly overconfluent (day 6, 7), their rate of DNA synthesis dropped markedly. By comparison, cells in the HDCR dishes maintained a higher rate of DNA synthesis.

Protein and Monoclonal Antibody Production During High Density Cell Culture

As an additional confirmation that cells remain productive during high density culture, HEK293-CMV-GFP cells were expanded over 1 week within PS (Figure 4f) versus HDCR cultureware (Figure 4g) and imaged with constant exposure settings on a fluorescent microscope. The cells exogenously expressed green fluorescent protein under the strong major immediate-early promoter and enhancer of the human cytomegalovirus (CMV) and thus acted as an easily visualized and quantitative model for exogenous protein production. Flow cytometry was performed to characterize protein production on a per cell basis, comparing cells cultivated on PS (harvested at 50% 2D confluence) to cells grown to 3D confluence in HDCR devices at day 7. Mean fluorescent intensity (MFI) was comparable on a per cell basis across PS and HDCR dishes, with slightly more protein production in HDCR-cultivated cells (Figure 4h). Consequently, surface productivity was dramatically higher in HDCR.

To further assess protein production in HDCR membranes, the stable humanized IgG1κ monoclonal antibody producing cell line (NISTCHO) was cultivated statically in HDCR 24-well plates versus shaken in PS shake flask (per manufacturer’s recommendation) over 8 days. Total mAb (mg) / (mL) media and cell specific mAb productivity (mg/cell/day) were observed to remain comparable across the two platforms (Figure 4i), despite cell density reaching 8.9E6 cells/cm^2^ in the HDCR (Figure 4j).

Cell Seeding Limits and Dynamic Expansion Range

To understand the minimum seeding density and dynamic expansion range supported by the HDCR membranes, K562 cells were expanded within the dishes from low (2.5E4 cells/cm^2^) and very low (2.5E3 cells/cm^2^) seeding densities, as a proxy for NK/T-cells. In both cases, the cells were able to recover and expand to greater than 3E6 cells/cm^2^, representing 164-fold and 1220-fold expansions, respectively (Figure 5a). These minimum recoverable seeding densities compare favorably (5 and 50-fold lower) to the limits reported with flat gas permeable devices (1.25E5 cells/cm^2^ and 110-fold expansion) [12].

Since gas exchange occurs through the bottom of the HDCR devices, it was hypothesized that there would be an elimination of edge effects observed in PS [13]. K562 cells were seeded (5E4 cells/well) into 24-well PS (n=3) and HDCR plates and expanded over 3 days and 7 days, respectively – due to the lower dynamic range of PS. HDCR plates exhibited no edge effects, substantially less variability (CV=3.9%) versus PS plates (CV=25%), and approximately 15x higher cell yields (Figure 5b).

Enhanced T-Cell Clustering and Expansion

T-cell expansion in HDCR devices was also assessed by researchers at City of Hope (Duarte, CA) using consented donor PBMCs. T cells were plated in PS or HDCR 24-well plates at 1M cells/well, mock or CAR transduced via lentivirus, and expanded over 1 week in complete X-VIVO media containing 100 U/mL recombinant human IL-2 (rhIL-2, Novartis Oncology) and 0.5 ng/mL recombinant human IL-15 (rhIL-15, CellGenix). Enhanced T-cell clustering was observed in CAR-T cells expanded in HDCR versus PS plates (Figure 5c) and this corresponded to 2.2–3.6x greater cell yields by day 7.

Prospective Applications Across R&D and Complex Biologics

HDCR cultureware has been applied to a variety of other cell types, modalities, and applications both internally and with academic, biotech, and pharma collaborators. The niche-fin microarchitecture makes it viable for adherent, suspension, and microcarrier culture (Figure 6a). iPSC-derived neural progenitor cell (NPC) production was piloted in HDCR plates as a predictable and scalable alternative to a spheroid-based spinner flask process (Shi Lab, City of Hope National Medical Center, CA). Freshly dissociated NPCs were seeded into 24-well HDCR plates and expanded over 2-weeks with regular feeding and counts (n=2) to generate a growth curve (Figure 6b,c). Cells were observed to follow an exponential growth curve with a consistent doubling time of 3.5 days. In subsequent scale up studies in HDCR 13cm dish dishes, the resulting NPCs met critical quality attributes for the program around viability, transgene expression, and karyotype. The stable mAb production cell, NISTCHO, was similarly shown to expand well within the dishes to near 3D confluence, despite typically being cultivated as a suspension cell under stirred/shaken conditions (Figure 6d). Human primary dermal fibroblasts, which are contact inhibited, were expandable within the device on microcarriers to provide additional surface area (Figure 6e). For oncolytic virus production, A549 cells were cultivated to high density on microcarriers within HDCR devices and infected with an oncolytic vaccinia virus (CF33 [14]) for 48 hours (MOI 0.1) (Figure 6f). Production in HDCR 13cm dish yielded 3.1E7 pfu/cm^2^ (37.5 pfu/cell) versus T225 PS flasks, which yielded 3.3E6 pfu/cm^2^ (22.9pfu/cell), representing an order-of-magnitude intensification. Finally, for gene therapy vector production HeLa-based AAV2 producer cells were expanded within HDCR 13cm dishes and induced to package AAV with an E-selectin therapeutic transgene, purified, titered, imaged under TEM (Figure 6g), and validated by flow cytometry transduction assay to be bioactive. Productivity of 1.2E11 vg/cm^2^ was achieved with an average per cell productivity of 5.1E4 vg/cell. To date, the the list of cell types that have been successfully cultivated within the HDCR platform includes, HEK293, HeLa, CHO, hybridoma, NS0, Vero, CV-1, A549, K562, L929, iPSC-derived neural stem cells, primary fibroblasts, macrophage, and Sf9.

Discussion

We set out to develop a respiring cultureware platform to provide researchers with a more efficient means of cultivating cells. Targeting cell densities of >1E8 cells/mL within the bed of the device, we estimated a local k_L_a equivalent of >50/hr would be necessary to meet the oxygen needs of the cells. The bioinspired “capillary bed” microarchitecture we converged on leverages the gas permeable property of silicone to achieve a k_L_a of >100/hr under static conditions, exceeding our specification.

HDCR devices were shown to routinely enable culture of HEK293 and several other cell types to densities of >1E8 cells/mL within the expansion niches. On a surface area basis (3.4E6 HEK293T cells/cm^2^), this represents a 31-fold intensification versus PS devices. The equivalent of 1cm^3^ of tissue can now be cultivated in a single 13cm dish, saving considerable incubator space and reducing plastic waste and technician touch-time by around 90%. Process intensification enabled by HDCR cultureware has afforded collaborating scientists with an abundant supply of challenging-to-produce viral vectors (e.g. murine CF33-mCD19t) for novel combination therapies [15]. It has also enabled multiple vector candidates to be advanced in parallel and research associates to be freed up from cell culture to higher value tasks. From a logistical perspective, the greater dynamic range of HDCR devices extends the possible duration between passaging, providing flexibility to scientists balancing multiple projects. Rather than having to passage plates on 2–3 day intervals (10-fold expansion), one can expand over a week (100-fold expansion) and yield functional cells.

Across protein production, mAbs, cell therapies, oncolytic viruses, and AAV vectors we have shown that per cell productivity can be largely maintained despite the increase in cell density, leading to meaningful intensification. These pilot applications highlight the versatile nature of the platform and its utility across adherent, suspension, microcarrier, and spheroid culture. The HDCR architecture provides more physiologically relevant, high density, 3D culture conditions than PS, yet retains its simplicity for predictable, linear scale up. Cultivating cells within a 3D format may even prove advantageous, such as with the increased clustering and expansion rate of T cells. While this data has provided hints at how the HDCR cultureware platform may be useful in the biological sciences and for advanced therapeutics production, further rigorous process development, optimization, and benchmarking will be required by modality-specific subject matter experts. Devices are being made widely available to the academic and biopharma community to find applications where better cultureware can accelerate discoveries and cures.

In subsequent work, we will show how the scalability of the HDCR can be extended further by stacking membranes in Z to create high-capacity perfusable cartridges with 1–100B+ cell capacity. This may help to overcome production bottlenecks for cell and gene therapies and reduce scale up risk. The past 100 years of 2D cell culture has proven remarkably fruitful despite poorly recapitulating cell growth conditions in the body. We are hopeful that the biomimetic, high-density, 3D microenvironment of HDCR cultureware will generate new insights.

Materials and Methods

Finite Element Modelling of HDCR Architecture

A multiphysics finite element model (COMSOL Multiphysics, Sweden) of a representative unit-cell of a PS dish (1mm thick), flat gas permeable silicone membrane (300μm thick), and HDCR membrane (300μm thick, variable fin-niche dimensions) was developed using the parameters in Appendix B. A Michaelis-Menten relationship for oxygen consumption was utilized based on V_O2_ and K_M_, with a threshold for cell viability/hypoxia of 1%. Maximum cellularity was taken as 1E8 cells/mL. Parametric sweeps were conducted to identify optimal fin and niche dimensions (500μm fin width × 200–800μm niche width × 400μm-1200μm fin height) that maximized cell yield per area while ensuring no hypoxia even at 100% 3D confluence. A 400μm wide and 800μm high niche flanked by 500μm thick fins was selected as the HDCR microarchitecture to advance to manufacturing based on its high cell density and homogenous oxygen microenvironment. A maximal cell density of 3.6E6 cells/cm^2^ was predicted.

Liquid Silicone Rubber Injection Molding of HDCR Plates

Singe cavity aluminum molds were designed, CNC machined, polished, and cleaned for manual injection molding. Sylgard 184 was prepared (10:1 ratio) per manufacturer’s recommendation, degassed by centrifugation and vacuum, and injected into the mold at 75psi. The molds were baked at 130°C for 60 min, parts were demolded and washed with isopropanol and fully cured at 150°C for 60 min. HDCR devices were placed in autoclave pouches and steam sterilized.

Expansion of HeLa-GFP+ Cells in HDCR Plates

HDCR 24-well plates received 5.3E4 HeLa-GFP cells (Cell Biolabs Inc, San Diego, CA) + 3.5 mL of complete growth medium (DMEM High Glucose + 10% FBS) (Thermo Fisher Scientific, Waltham, MA) to initiate growth and maintained in a 37C, 5% CO_2_, humidified incubator. Fluorescent images were captured daily and full 3.5 mL media exchanges occurred on days 4,5, and 6 of this 7-day culture. Fluorescent images were captured on the ECHO Revolve inverted fluorescent microscope (56% LED power + 125 ms exposure) on days 0–5 of culture. Fluorescent imaging settings were changed on D6 (56% LED power + 25 ms exposure time) and D7 (56% LED power + 5 ms exposure time).

HeLa-GFP cells were inoculated into a HDCR 24-well plate following the same initial experimental set up. Cells were expanded over 8 days and with full 3.5 mL media exchanges of complete media occurred on days 4,5,6, and 7. Passage 2 was allowed to grow until day 9 which necessitated an extra 3.5 mL media exchange on day 8. Passage 5 was allowed to expand until day 10, with additional feedings on day 8 and 9, and was fed with double the amount of media/exchange (8 mL/exchange). Duplicate wells within the HDCR 24-well plate were harvested and averaged for near daily cell counting. To harvest, media was removed via the lowest SureShelves^™^ and 3 mL of PBS + 5 mM EDTA was introduced via the lowest SureShelves^™^ to wash cells. PBS was then removed via the lowest SureShelves^™^ and 0.75 mL of TrypLE^™^ Express Enzyme (Thermo Fisher Scientific, Waltham, MA) was introduced to the 0.25 mL of PBS + cells remaining in the well via a wide bore P1000 pipette. Cell mass was gently fluidized 3–4 × via pipetting and returned to the incubator for 7 minutes. Mixing and incubation step was repeated for cell densities >3E6/ well to allow for full cell dissociation. After complete dissociation, 1 mL of complete media was introduced to 1 mL of cell suspension to halt enzyme, full well volume was mixed to ensure homogenous suspension and 200 ul of suspension was taken for cell counting on the Countess 3 cell counter (Thermo Fisher Scientific, Waltham, MA).

Manometric Measurement of Cultureware Oxygen Transmissivity

Test specimens were harvested from a PS cell culture dish (Nunc, Thermo Fisher Scientific, Waltham, MA) and HDCR 13cm dish (XDemics, CA) and hermetically mounted to a backing ring if needed to accommodate the manometer. Flat silicone membrane test specimens (Sylgard 184, Dow Corning, USA) were also prepared by mixing per manufacturers’ recommended ratio, centrifuging and vacuum degassing, and compression molding into 300μm thick, 90mm wide circular molds, which were cured at 80°C for 90 minutes. All parts were cleaned with isopropanol and passed visual inspection for defects. Manometric measurements of oxygen transmission rate per ASTM D1434 were carried out in triplicate (Labthink International, Inc). Transmissibility data was divided by sample thickness to determine gas permeability values.

Calculation of local kLa equivalent

$[eqn]$ was scaled to identify the maximum cellular oxygen uptake rate at which the minimum oxygen tension within the HDCR membrane reached 7.6 Torr, the threshold assumed for cessation of cellular activity. This scaling factor was determined to be $[eqn]$ . At this condition, oxygen flux, viable cell density, oxygen flux per cell, and oxygen turnover were calculated.

Simulations were performed for the HDCR membrane, a flat silicone membrane, and conventional polystyrene cultureware. All configurations used the same initial cellular density and cellular volume to enable direct comparison.

Viable cell density was defined as the number of cells per unit surface area experiencing an oxygen partial pressure greater than 7.6 Torr. Oxygen flux per cell was calculated as the total oxygen flux into the cellular region divided by the viable cell density. Oxygen turnover, expressed as the volumetric mass transfer coefficient (kLa), was calculated as:

$$[eqn]$$

J is the oxygen flux and $[eqn]$ is the oxygen solubility constant at atmospheric conditions.

Viability Staining of Low and High Density HEK293 Across Cultureware

HEK293T cells (ATCC CRL-3216; American Type Culture Collection, Manassas, VA) were cultured on polystyrene 24-well plates, HDCR flat-bottom silicone 24-well plates, or HDCR capillary 24-well plates. Cultures were maintained in DMEM high- glucose supplemented with 10% fetal bovine serum (FBS) to support exponential growth. Twenty-four hours prior to imaging, cultures underwent a final medium exchange with FluoroBrite^™^ DMEM (Gibco) supplemented with 4 mM L-glutamine and 10% FBS to minimize background fluorescence. On the day of imaging, adherent cells cultured on polystyrene plates were washed once with 2 mL phosphate-buffered saline (PBS), after which 0.25 mL PBS and 0.25 mL of reconstituted LIVE/DEAD^™^ Cell Imaging Kit (488/570; Invitrogen) solution were added. For cells cultured on HDCR flat-bottom and capillary membranes, culture medium was removed at the lowest SureShelves^™^, and membranes were gently rinsed with 3 mL PBS. PBS was subsequently removed at the lowest SureShelves^™^, leaving 0.25 mL PBS in each well. An additional 0.25 mL of reconstituted LIVE/DEAD^™^ Cell Imaging Kit solution was added and gently mixed, minimizing disruption of the cell layer.

Growth Curves Across HDCR Platform

HEK293T cells were seeded in NUNC 15 cm tissue culture treated dishes, Fisher Brand 24- well tissue culture treated 24- well plates, HDCR 13-cm dishes, and HDCR 24- well plates at a seeding density of 2.8E4 cells/cm^2^. Cells were maintained in DMEM High Glucose + 10% FBS. Polystyrene cultureware received 0.2 mL of media/cm^2^ of surface area and HDCR devices received 1.75 mL of media/cm^2^ of plate surface area. Polystyrene cultures received full media exchanges on days 3–7 and HDCR cultures receiving full media exchanges on days 4–7. All devices were harvested on days 3–7 for viability, counting, and EdU staining.

Final samples were pelleted at 500 × g for 5 minutes to assesses representative cell mass.

EdU Staining of Cell Cycle at High Density

To assess proliferation dynamics, cells were pulse-labeled with 10 μM EdU (5-ethynyl-2′-deoxyuridine) for 2 hours at different time points under standard culture conditions. Dishes had their spent media gently removed and transferred to a conical tube to introduce EdU. 15cm PS dishes received 30 mL of 10 μM EdU containing spent media and 13cm HDCR dishes received 60mL of 10 μM EdU containing spent media. After each labeling period, cells were immediately harvested, washed with PBS, and fixed using 4% paraformaldehyde. Permeabilization was performed with 0.5% Triton X-100, and EdU incorporation was detected using the Click-iT^®^ Plus EdU Alexa Fluor^®^ 647 Flow Cytometry Assay Kit (Invitrogen, #C10635) according to the manufacturer’s protocol. DNA content was stained with FxCycle^™^ Violet (Invitrogen, #F19347) for cell cycle analysis. Samples were acquired on a Cytek Aurora flow cytometer, and data were analyzed using FlowJo software (v10). The proliferation rate was assessed based on the mean fluorescence intensity (MFI) of EdU-Alexa Fluor^®^ 647, reflecting DNA synthesis activity. Cell cycle distribution was determined by FxCycle^™^ Violet and EdU-positive populations.

Flow Cytometry Characterization of GFP Protein Production

HEK293-GFP cells were cultured in parallel using standard PS and HDCR dishes (DMEM + 10% FBS, 37°C, 5% CO_2_). Cells were harvested (50% 2D confluence for PS; day 7 for HDCR) and analyzed on a Cytek Aurora flow cytometer (488 nm excitation, 525/40 nm emission). GFP fluorescence intensity was measured using standardized instrument settings, with ≥10,000 gated events per sample. Data was analyzed in FlowJo v10, comparing mean fluorescence intensity (MFI) between culture dishes. Unstained cells served as autofluorescence controls. Fluorescence/cm^2^ was calculated as MFI × cells/cm^2^ at 2D confluence (PS) vs day 7 (HDCR).

Characterization of mAb Production

The NISTCHO monoclonal antibody (mAb)–producing cell line (RM 8675; NISTCHO Lot QL24AA7176–1; National Institute of Standards and Technology, Rockville, MD) was cultured in EX-CELL^®^ CD CHO Fusion medium (Sigma-Aldrich) supplemented with 6 mM L-glutamine (Gibco). Cells were maintained in suspension and routinely passaged according to the manufacturer’s recommendations in vented, plain-bottom PETG Erlenmeyer flasks (Nalgene^™^) at 37 °C in a humidified incubator with 5% CO_2_, using orbital shaking at 150 rpm (19-mm throw). Cells were expanded in a 250-mL shake flask to a density of 2.2E6 cells/mL, then diluted with fresh medium to 3E5 cells/mL. This inoculum was used to initiate cultures in Expansify^™^ 24-well plates (3.5 mL per well) and 125-mL PETG shake flasks (30 mL per flask). Cell density and mAb titer were measured daily from day 0 to day 10. At each time point, 200 μL was withdrawn from each culture, centrifuged at 500 × g for 5 min, and the cell-free supernatant was stored at −80 °C. An equivalent volume of fresh medium was added to maintain constant culture volume. Monoclonal antibody concentrations were quantified using a Human IgG (Total) ELISA kit (Invitrogen), following the manufacturer’s instructions.

Seeding Density Limits with K562

K562 cells (ATCC #: CCL-243) (American Type Culture Collection, Manassas, Virginia) were seeded at 2.5E4 or 2.5E3 cells/cm^2^ per well of HDCR 24- well plate. 2.5E3 cells/cm^2^ received 1 mL of RPMI 1640 + 10% FBS (Thermo Fisher Scientific, Waltham, MA) and 2.5E4 cells/cm^2^ received 3.5 mL of media on D0. Higher seeding density condition received daily 3.5 mL media exchanges on days 5–9 of culture, whereas the lower seeding density condition received 3.5 mL media exchanges on days 6–9 of culture.

Evaluation of Edge Effects

K562 cells were seeded at 5 × 10^4^ cells per well in surface-treated tissue-culture polystyrene 24-well plates (Fisherbrand) and HDCR 24-well plates in RPMI 1640 medium (Gibco) supplemented with 10% fetal bovine serum (FBS) (Fisherbrand), in triplicate. Medium volumes were 1 mL per well for polystyrene plates and 3.5 mL per well for HDCR plates. For HDCR plates, complete medium was exchanged on days 4, 5 and 6 post-seeding using the 0.25 mL shelf for medium aspiration and the 1 mL shelf for medium addition to restore a final volume of 3.5 mL per well. Cells were counted on day 3 post-seeding for polystyrene plates and on day 7 post-seeding for HDCR plates using a Countess^™^ 3 automated cell counter (Invitrogen).

CAR-T Cell Production

Leukapheresis products were obtained from consented research participants (healthy donors) under protocols approved by the City of Hope (COH) Internal Review Board (IRB). On the day of leukapheresis, peripheral blood mononuclear cells (PBMC) were isolated by density gradient centrifugation over FicollPaque (GE Healthcare) followed by multiple washes in PBS/EDTA (Miltenyi Biotec). Cells were rested overnight at room temperature (RT) on a rotator, and subsequently washed and resuspended in X-VIVO-15 (Lonza) with 10% FBS (complete X-VIVO). PBMCs were immediately frozen in CryoStor^®^ CS5 cryopreservation media (BioLife Solutions) until further processing. Freshly thawed PBMC were washed once and cultured using either 24-well PS plates or HDCR in complete X-VIVO containing 100 U/mL recombinant human IL-2 (rhIL-2, Novartis Oncology) and 0.5 ng/mL recombinant human IL-15 (rhIL-15, CellGenix). For CAR lentiviral transduction, T cells were cultured with ImmunoCult^™^ Human CD3/CD28/CD2 T cell activator (STEMCELL Technologies), protamine sulfate (APP Pharmaceuticals), cytokine mixture (as stated above) and desired lentivirus at MOI of 1 one day following bead stimulation. Cells were then cultured in and replenished with fresh complete X-VIVO containing cytokines every 2–3 days. Cells were imaged and counted of 7 days. CAR T cells were positively selected for CD19t using the EasySepTM CD19 Positive Enrichment Kit I or II (StemCell Technologies) according to the manufacturer’s protocol. Purity and phenotype of CAR T cells were verified by flow cytometry.

iPSC-Derived NPC Expansion

Freshly dissociated NPCs (p3) (Shi Lab, City of Hope National Medical Center) were seeded in HDCR 24-well plates at 1.5E5 cells/cm^2^ in 2mL of NPMM-A medium (duplicate). Media was replaced (2mL) every other day for first week, daily thereafter. At semi-regular intervals (1–3 days) cells were harvested from wells in duplicate, enzymatically dissociated, and counted using a Cellometer.

NISTCHO Expansion

HDCR 24-well plate wells were seeded with 1E4 NISTCHO cells in 3.5 mL of EX-CELL^®^ CD CHO Fusion medium. A single well was chosen as a representative well for daily imaging. One well was sacrificed for counting/day. All remaining wells received a full media exchange of 3.5 mL on day 7 only and maintained until day 11 (9.05E6 cells).

Primary Human Fibroblast Culture

5E4 Primary Dermal Fibroblast Normal; Human, Neonatal (American Type Culture Collection, Manassas, Virginia) were cultured in complete Fibroblast Growth Kit-Low serum (American Type Culture Collection, Manassas, Virginia) and seeded on 50 μL of prepared Cytodex-3 microcarrier/well of HDCR 24- well plate. Full 3.5 mL media exchange was performed on da 2 of cultures. Wells imaged on day 4.

CF33 Oncolytic Vaccinia Virus Production

A549 cells were seeded into HDCR devices onto Cytodex-3 microcarriers and expanded to 8E5 cells/cm^2^, with regular feeding of DMEM+10% FBS. At time of infection, media was removed and replaced with 0.5mL/cm^2^ DMEM and CF33 vaccinia virus at 0.1 MOI for 1 hour. Thereafter, 0.5mL/cm^2^ of DMEM+20% FBS was added. Cells were harvested 48 hours later for purification of CF33 virus and plaque titering.

AAV2-E-Selection Vector Production

HeLa-based AAV2 producer cell lines carrying a therapeutic transgene (CMV-hSELE) were seeded in an HDCR 13cm dish and allowed to expand to ~3E6 cells/cm^2^. The cells were infected with a helper virus to induce production for 72 hours. Cells were harvested, pelleted, and lysed. The lysate was clarified and AAV was purified by AAVX resin on an AKTA Pure 150. Titers were determined by qPCR with custom probe against the transgene. Potency of the AAV2 was validated by transduction of HeLa cells and flow cytometry validating expression of E-Selectin.