Quantitative Analysis of Plasmid DNA Distribution and Transgene Expression Following Cell Squeezing

Abstract

Purpose:

Microfluidic cell squeezing is a promising mechanoporation strategy that has been successfully applied to deliver various macromolecules, including proteins and nucleic acids. However, its success is limited when mechanoporation is used alone to facilitate gene delivery. The goal of the current study is to develop a gene delivery strategy that integrates electrotransfection and mechanoporation.

Methods:

We designed and fabricated a microfluidic device for controlled cell squeezing. Mammalian cells were first electrotransfected with pDNA, followed by mechanical squeezing to promote nuclear entry of the delivered DNA. Intracellular pDNA transport and transgene expression were quantitatively evaluated in squeezed and un-squeezed cells.

Results:

Cell squeezing increased cellular uptake and nuclear accumulation of pDNA by more than eightfold compared with un-squeezed controls. Despite this substantial enhancement in DNA delivery, transgene mRNA levels remained unchanged, and protein production was consistently reduced relative to controls. The lack of increased transgene expression might be partially attributable to squeezing-induced chromatin condensation.

Conclusion:

These findings indicate that chromatin condensation limits transgene expression following mechanoporation. Pharmacological inhibition of chromatin condensation may therefore represent a promising strategy to improve the efficiency of mechanoporation-facilitated gene delivery.

Introduction

DNA vector for gene delivery must enter cell nucleus for transgene expression. One strategy to improve the nuclear entry involves synchronizing the cell cycle, followed by DNA transfection shortly before nuclear envelope breakdown (NEBD). Although this strategy is effective, its applications are limited to dividing cells [1]. Another strategy is to incorporate DNA nuclear targeting sequences (DTS) into the vector or attach nuclear localization signal (NLS) peptide to the vector, which can enhance the nuclear entry through the nuclear pore complex (NPC). However, the magnitude of the enhancement varies depending on various factors, such as identity of the DTS or NLS sequence, cell type, and method of transfection [2, 3]. As such there is a need to develop new strategies to accomplish the goal of increased delivery of DNA into the nucleus.

Microfluidic devices have been used to mechanically deform cells to enhance the intracellular delivery of dextran, proteins, and RNAs [4–8], primarily through transient mechanoporation of the plasma membrane. Beyond cytoplasmic delivery, several studies have demonstrated that cell squeezing can mediate plasmid DNA (pDNA) transfection [9, 4, 10, 11], a process that requires successful nuclear entry of pDNA. Although a few reports have achieved transgene expression using mechanoporation alone [12, 4], most studies combine cell squeezing with subsequent electroporation, as mechanoporation by itself often results in limited transfection efficiency [9, 11]. We hypothesized that this limited efficiency is partly attributable to chromatin condensation induced by mechanical deformation. To test this hypothesis, we designed and fabricated a microfluidic device that forces cells through constrictions narrower than the nuclear diameter. Following squeezing, we quantitatively assessed total cellular and intranuclear pDNA levels and measured reporter gene transcription and translation. We also examined chromatin condensation and global transcriptional and translational activities to determine how mechanical deformation influenced gene delivery and transgene expression. Due to the low delivery efficiency by mechanoporation alone, pDNA electrotransfection was performed prior to cell squeezing in all experimental groups to enhance signal detection. This experimental design enabled reliable quantification of intracellular and nuclear pDNA, as well as transgene expression, in un-squeezed control groups for direct comparison. Our results indicated that chromatin condensation represents a limiting factor in cell squeezing-mediated gene delivery.

Materials and Methods

Cell lines and cell culture

COS7, an African green monkey fibroblast-like kidney cell line, EOMA-GFP, a mouse hemangioendothelioma cell line stably expressing pEGFP-N1, and C2C12, a mouse myoblast cell line, were obtained from the Duke University Life Science Facility (LSF). All cells were cultured in DMEM (Gibco 11995–065) supplemented with 10% (v/v) bovine calf serum (Cytvia SH3007204) and 1% penicillin-streptomycin (Gibco 15140–122). For EOMA-GFP cells, DMEM was supplemented with 0.2 mg/ml G418 sulfate (Corning 30–234-CR) to ensure retention of pEGFP-N1. Cells were incubated at 37 °C in 5% CO2 and passaged every 2–3 days to avoid plates of cells reaching confluence.

Plasmids and oligonucleotides

Plasmid pEGFP-N1 referred to as pEGFP was obtained from Nova Lifetech Inc. All other plasmids were obtained from AddGene: pUB-GFP and pEF-GFP were gifts from Connie Cepko (Addgene plasmids # 11154 & #11155), pCAG:GPI-GFP was a gift from Anna-Katerina Hadjantonakis (Addgene plasmid # 32601), and pCXLE-EGFP was a gift from Shinya Yamanaka (Addgene plasmid # 27082). The plasmid pEGFP-C1 EGFP-3XNLS (referred to as pEGFP-NLS) was a gift from Dyche Mullins (Addgene plasmid # 58468). dbDNA was purchased from Touchlight DNA services (CMV-TE-eGFP catalogue # db100–028). Hybridization probes for pEGFP and EGFP-mRNA detection were Probe-EGFP-O4-Sense-C1 (ACD 1040801-C1) for pEGFP, and Probe-EGFP-C2 (ACD 400281-C2) for detection of EGFP-mRNA. Both probes were purchased from Advanced Cell Diagnostics (ACD).

Fabrication of microfluidic device

PDMS devices were cast using Sylgard 184 silicone elastomer and curing agent (VWR 102092–312). Molds for the devices were created on 100 mm silicon wafers (University Wafer 1196) using a custom chrome on quartz photomask (Photo Sciences F10CS5060) and SU-8 10 photoresist provided by the Shared Materials Instrumentation Facility (SMIF) at Duke. To prepare the molds, silicon wafers were washed in piranha solution (equal parts H₂SO₄ and H₂O₂), rinsed with DI water and dried with compressed N₂. SU-8 was then spin-coated onto the cleaned wafer to a thickness of 10 microns according to manufacturer instructions (Kayaku SU-8 datasheet). Briefly, the wafer was vacuum-secured to the chuck of a spin coater and ~1 mL of SU-8 10 was added to the wafer’s center. The wafer was then accelerated to 500 rpm at an acceleration of 100 rpm/second for five seconds, followed by acceleration to 3000 rpm at 300 rpm/second for 30 seconds. The wafer with resist was then placed on a 65°C hotplate for 2 minutes followed by a 95°C hotplate for 5 minutes. The photomask and soft-baked wafer were then mounted on a mask aligner (Karl Suss MA6/BA6) and exposed to constant intensity 365 nm light for 10 seconds. The wafer was then baked at 65°C for 2 minutes, 95°C for 3 minutes, then developed in SU-8 developer (Kayaku Advanced Materials) for 2 minutes. The wafer was then rinsed with isopropyl alcohol and dried with compressed N₂, completing the mold.

To cast devices, Sylgard 184 silicone elastomer and curing agent were mixed at a 10:1 ratio and degassed. The mold was placed in a folded aluminum foil tray and a 15 μL drop of Trichloro(1H,1H,2H,2H-perfluorooctyl)silane (Millipore Sigma 448931) was pipetted into the tray’s corner. The tray was then covered with a plastic lid, and the silane was allowed to evaporate and deposit on the mold for at least 20 minutes. Excess silane was wiped from the corner with a Kim wipe, then 50–60 mL of the PDMS mixture was poured onto the mold. A vacuum chamber was used to remove any bubbles between the PDMS and the mold, then the devices were cured at 60°C for 3 hours. The cured PDMS was then peeled away from the silicon/SU-8 mold and cut to size. Inlet and outlet holes were cut using a 1 mm biopsy punch.

To adhere the PDMS devices to glass slides, glass slides were first cleaned with 10% Isopropyl alcohol and DI water, and PDMS devices were cleaned with Scotch magic tape. PDMS and glass components were then heated on a hot plate to 150°C to evaporate any remaining water. The glass and PDMS faces that were to be bonded were then plasma treated for approximately 1 minute each using a handheld high frequency generator (BD-10A, 115V Electro-Technic Products). The PDMS was then gently and carefully lowered onto the glass slide and gentle compression was applied to the edges when necessary to achieve complete contact between the PDMS and glass without damaging the device features. Devices were then baked at 60°C overnight to complete bonding. Reservoirs were created by cutting off the base of P200 micropipette tips, applying a small amount of mixed PDMS to one end, and placing this over the outlet hole of slide bound PDMS devices. This is followed by baking at 60°C for at least 15 minutes to cure the PDMS attaching the reservoir.

Electrotransfection and cell squeezing

Cells were detached from culture plates using 0.25% trypsin (Gibco 1505–057), split into aliquots of 1×10⁶ cells, and washed with DPBS (Gibco 14190–144). For each electrotransfection, 1×10⁶ cells were resuspended in 100 μL of pulsing buffer, consisting of 2 μg pDNA in OptiMEM reduced serum medium + GlutaMAX (Gibco 51985–034), and transferred to a 4 mm gap electroporation cuvette (Genesee 40–102). An electric pulse sequence was applied to cells (2 pulses at 150 V for COS7, 250 V for C2C12 cells, 10 msec duration, and 0.1 Hz frequency) using an ECM 830 square wave electroporation system (BTX).

Cells were then immediately mixed with the full medium (1 mL) that had been prewarmed to 37 °C and incubated at 37 °C for 30 minutes in low-adherence 24-well plates. At the end of 30-minute incubation, the cells were collected, pelleted, and resuspended in pDNA free OptiMEM (1×10⁷ cells/mL). In cell squeezing experiments, 1.1×10⁷ cells were first electrotransfected as described above, yielding approximately 1.1 mL cell suspension. Of this volume, 100 μL was reserved for the un-squeezed control group, and the remaining 1.0 mL was loaded into a syringe, which was then fitted with a 20G hypodermic needle (Thomas Scientific 21A00A220) connected to Tygon tubing (Thomas Scientific 1204G82). A syringe pump (PHD 2000, Harvard Apparatus) was used to deliver the cell suspension to the sharpened end of the tubing before it was inserted into the inlet hole of the device. The cell suspension was pumped continuously through squeezing devices at 110 μL/minute until 100 μL of flow-through could be collected at the outlet, which took 5–10 minutes after initiation of flow. Both squeezed and un-squeezed cell suspensions (100 μL each) were plated into separate wells of 6-well plates or into separate #1.5 glass bottom dishes (Cellvis D35-20-1.5-N). Subsequently, 2 mL of pre-warmed full medium was added to each well or dish, and the cells were then incubated at 37°C, 5% CO₂ for indicated time periods prior to analyses.

In selected COS7 cell groups, the full medium was supplemented with RGFP966 (Millipore Sigma SML1652) at a final concentration of 10 μM. RGFP966 was dissolved in DMSO prior to use; therefore, an equivalent volume of DMSO was added to control groups to account for vehicle effects. The supplemented medium was used for continuous treatment during cell culture following electrotransfection in the designated groups, except during squeezing.

Evaluation of plasma membrane and nuclear envelope permeabilization

To determine if cell squeezing permeabilized membranes in cells, three assays were developed. First, COS7 cells were stained with calcein-AM (Thomas Scientific) prior to squeezing. Calcein-AM is cell permeant but, once inside of cells, it is transformed into calcein which is retained in cells that do not have permeabilized membranes. Immediately post squeezing, the calcein fluorescence in all cells was examined under a widefield fluorescence microscope and the mean fluorescence intensity per cell was quantified through image analysis to determine the permeabilization of the plasma membrane to the small molecule marker, calcein. Second, EOMA cells stably expressing GFP was squeezed through 6×6 μm devices. Post squeezing, GFP fluorescence was quantified with flow cytometry to determine the permeabilization of the plasma membrane to the protein marker, GFP, and changes in cell concentration post squeezing. For each replicate, 30,000 individual cells were used in flow cytometry analysis. Finally, COS7 cells were electrotransfected as described above with pDNA encoding EGFP-NLS (pEGFP-NLS). At 24 hours, the EGFP–NLS signal was predominantly localized to the nucleus. For subsequent experiment, the squeezing device was pre-cooled and maintained on ice throughout the procedure. Cell nuclei were stained with Hoechst, after which cells were incubated on ice for 10 minutes to inhibit active nuclear transport, and then passed through 6×6 μm devices. Un-squeezed cells were kept on ice for the same duration. Following cell squeezing, both groups were plated in coverslip-bottomed dishes and imaged with a widefield fluorescence microscope to determine the permeabilization of the nuclear envelope to EGFP.

In situ hybridization with RNA and DNA probes

Cells were washed with DPBS, fixed with 4% PFA in DPBS, and prepared according to the RNAscope Multiplex Fluorescent Reagent Kit v2 user manual (ACD) with some modifications. All reagents were supplied in the RNAscope Multiplex Fluorescent Reagent Kit v2 (ACD 323100) unless otherwise stated. Fixed cell samples were first incubated with hydrogen peroxide for 10 minutes at room temperature, then washed with ultrapure distilled water (Invitrogen 10977–015) before staining with DNA or RNA probes.

To stain pEGFP, the samples were incubated with RNA digestion buffer (25 μg/mL RNase A in 1x Tris-buffered saline with 0.05% TWEEN 20) (Sigma P2287) for 30 minutes at 37 °C, followed by washing twice with DPBS. Then, the samples were treated with protease III for 10 minutes at room temperature before washing twice with DPBS. Samples were incubated in prewarmed DNA denaturation buffer (75% formamide (Roche 11814320001) in 2x SSC (Invitrogen 15557–044)) for 5 minutes at 75 °C to increase hybridization efficiency, washed once with DPBS, then incubated with Probe-EGFP-O4-Sense-C1 for 4 hours at 40 °C. To stain EGFP mRNA, the samples were incubated with Probe-EGFP-C2 for 2 hours at 40 °C.

After probe hybridization in both cases, the samples were washed twice with RNAscope wash buffer, incubated in 5x SSC overnight at room temperature. All samples were then incubated with two hybridization probe amplifiers successively for 30 minutes, then with a third amplifier for 15 minutes, all at 40 °C, washing with RNAscope wash buffer between incubations. After amplification, probe fluorescence signal was developed by incubating first with HRP for 15 minutes, then with Opal 690 (Akoya FP1497001KT) diluted 1:1500 in TSA buffer for 30 minutes both at 40 °C, washing with the RNAscope wash buffer after each incubation. After staining with DNA or RNA probes, the samples were then stained with 10 μg/μL Hoechst for 10 minutes at room temperature, protected from light prior to imaging with fluorescence microscopy.

Immunofluorescence staining

For antibody staining to detect G3BP1 in stress granules, un-squeezed and squeezed cells were incubated for 1, 4, or 16 hours at 37°C. For antibody staining to chromatin markers or HDAC3, un-squeezed and squeezed cells were incubated for 1.5 hours. After incubation, all cells were fixed with 4% PFA for 10 minutes at room temperature, protected from light. Cells were then permeabilized with 0.2% Triton 100X in DPBS for 5 minutes at room temperature. Before and after permeabilization, cells were washed twice with DPBS. Cells were blocked with 10% normal goat serum in DPNS for 1 hour at room temperature. Following blocking, cells were incubated with primary antibodies overnight at 4°C, protected from light. Antibodies were diluted into 1% BSA in PBS for this incubation. Final dilutions used were1:2000 for Rabbit anti-G3BP1 (Proteintech 13057–2-AP), 1:250 for Mouse anti-Histone H3 Ab (Cell Signaling #14269), 1:2000 for Rabbit anti-H3K27me3 (Cell Signaling #9733), 1:1000 for Rabbit anti-H3K9me3 (Cell Signaling #13969), 1:400 for Rabbit anti-H3K9ac (Cell Signaling #9649), and 1:125 for Mouse anti-HDAC3 (Cell Signaling #3949). Cells were then washed twice with DPBS before incubating them with secondary antibodies diluted in 1% BSA in PBS for 1 hour at room temperature, protected from light. Secondary antibodies used were 1:500 Alexa Fluor 488 conjugated goat anti-mouse IgG (Cell Signaling #4408) and 1:1000 Alexa Fluor 568 conjugated goat anti-rabbit IgG (ThermoFisher Scientific A-11011). Cells were then washed with DPBS twice. Nuclei were stained with 10 μg/μL Hoechst (ThermoFisher Scientific H1399) in DPBS for 10 minutes, followed by two washes with DPBS. In samples where whole cell staining was required, cells were stained with 2 μg/μL ATTO647N (ATTO-TEC AD 647N) for 10 minutes, followed by two washes with DPBS. Vectashield antifade mounting medium (Vector Laboratories H-1000–10) was then applied to samples and they were topped with coverslips. Samples were then either imaged immediately or placed in 4°C and protected from light before imaging within 24 hours.

Fluorescence microscopy and image analysis

Fluorescence images for mRNA and pDNA, and G3BP1 immunofluorescence analyses were taken using an Andor Dragonfly spinning disk confocal plus microscope equipped with a 63x oil objective and a 40 μm pinhole disk. Fluorescence images for all other immunofluorescence analyses were taken with a Leica Stellaris 8 point scanning confocal microscope equipped with a 40x oil objective. Widefield images were taken with an Axio Observer microscope equipped with a 20x or 40x air objective for widefield images. The super-resolution radial fluctuations (SRRF) algorithm as implemented in Andor’s Fusion software was used to acquire super-resolution images with 120–150 nm spatial resolution.

Automatic analysis of intracellular and intranuclear pDNA and mRNA signals were conducted with CellProfiler as described previously [13]. Additionally, shape-based pDNA puncta detection was accomplished using an adaptive object thresholding algorithm developed by Ranefall et. al. [14] that allows weighted consideration of object shape and intensity. Implementing this algorithm in CellProfiler, rather than relying on classical object detection based solely on size and intensity thresholds, enabled accurate segmentation of visually distinct punctate pDNA signals, including dim puncta, while preventing bright, diffuse pDNA patterns from being misclassified as puncta. For quantification of a fluorescence signal on either a per cell or per nucleus basis, nuclei were first segmented using the Hoechst channel. Cell regions within each field of view were then defined by radially expanding each nuclear object until the expanded regions contacted those of neighboring nuclei. In experiments in which cells were stained with ATTO647N, more accurate cell segmentation was achieved by applying a propagation algorithm to identify ATTO647N-positive regions that encompassed the previously segmented nuclei. Following segmentation, integrated fluorescence intensity was quantified from background-subtracted images for each nucleus or corresponding cell. This value was approximately proportional to the total amount of fluorescent molecules detected within the optical section of the samples.

To account for day-to-day experimental variation in fluorescence intensity arising from multiple sources, including but not limited to variability in sample staining efficiency and fluctuations in laser excitation power, the median intensities of each matched un-squeezed and squeezed pair were normalized to their mean value. This within-pair normalization was intended to reduce multiplicative batch effects while preserving the relative difference between the two groups. Because the ratio (or percentage) of intranuclear relative to intracellular intensity is invariant to multiplicative scaling, the percentage was first calculated with the raw data of integrated intensities for each cell, and the median of these percentages across all cells within a group was analyzed without further normalization.

To account for potential increases in antibody staining resulting from reduced cell density in squeezed cell plates, immunofluorescence measurements of chromatin biomarkers were normalized to histone H3. Specifically, cells were co-stained with an anti–histone H3 antibody, and chromatin biomarker signal intensities were normalized to the corresponding histone H3 intensity in each sample.

Flow cytometry

Cells from all samples were washed with DPBS and collected at 24, 48, or 72 hours post electrotransfection, resuspended in 1 to 1.2 mL DMEM, and immediately subjected to flow cytometry analysis (NovoCyte 1050 flow cytometer and the companion software NovoExpress). The procedure was carried out as previously described [15]. Prior to the analysis, propidium iodide (PI) was added to the cell samples to identify cells with compromised plasma membrane integrity.

For each sample, 30,000 single-cell events were acquired by flow cytometry, and single cell gates were set up using un-transfected control samples based on forward and side scattered light. EGFP signal was detected at 488 nm excitation and the gate for distinguishing EGFP-positive cells was set using un-transfected control samples. Transgene expression was evaluated with two metrics: (a) transfection efficiency, defined as the percentage of cells that were PI-negative and EGFP-positive, and (b) expression level, defined as the geometric mean of EGFP intensity per cell in the EGFP-positive population. The cell density in each sample was quantified by flow cytometry based on the number of events acquired per unit volume. The same methods were used for analysis of EOMA-GFP cells immediately post squeezing.

Total RNA synthesis and total protein level measurement

For quantification of total RNA synthesis samples were prepared with the Click-iT RNA Alexa Fluor 594 Imaging kit (Invitrogen C10330) according to manufacturer’s instructions. Briefly, un-squeezed and squeezed cells were incubated with 0.2 mM 5-Ethynyl Uridine (EU) solution for 16 hours. Cells were then fixed with 10% formalin, permeabilized with 0.5% Triton X-100, and incubated with the reaction cocktail prescribed to attach Alexa Fluor 594 to EU incorporated into RNA during the 16-hour synthesis. Before imaging, the nuclei of cells were stained with 10 μg/μL Hoechst for 10 minutes at room temperature. Cells incubated with no EU were used as a negative control.

To measure the total protein content, cells were cultured for 24 or 48 hours post squeezing then trypsinized, counted, and aliquoted such that each experimental group had an identical number of cells. Cells were pelleted by centrifuging at 280 g for 5 minutes and then re-suspended in 250 μL lysis buffer – 1% (v/v) Triton X-100, 50 mM Tris, pH 7.5, 750 mM NaCl, 5 mM EDTA, with 1:100 diluted protease inhibitor cocktail (Sigma-Aldrich P8340) added just before using. Cells were incubated on ice for five minutes and at room temperature for 10 minutes, vortexing every five minutes. Lysed cells were then centrifuged at 17,000 g for 30 minutes. Lysate supernatant was collected, and the Pierce BCA Protein Assay Kit (ThermoFisher Scientific 23227) was used according to manufacturer’s instructions to quantify total protein per cell. Briefly, 25 μL of each standard (working range = 20–2,000 μg/mL) or experimental sample was pipetted into wells of a 96 well plate, followed by the addition of 200 μL working reagent and incubation at 37°C for 30 minutes. A plate reader was then used to measure the absorbance of all samples at 562 nm. A standard curve was prepared and used to convert blank subtracted absorbance readings to protein concentrations (μg per million cells).

Statistical analysis

Statistical analysis was performed with GraphPad Prism software, the specific test used is noted where results are reported. For imaging experiments quantifying nuclear EGFP-NLS and calcein fluorescence intensity per cell, unpaired t-tests were used to determine the significance of differences in average intensity values of samples with >45 cells imaged per group. Paired t-tests were used to determine the significance of differences in all other two group comparisons. For two factor multiple comparisons the tests employed were 2-way ANOVA with Šídák’s post-hoc test. For all statistical analyses, a difference was considered to be statistically significant if P < 0.05 and marginally significant if P ≤ 0.10.

Results

Device design and evaluation

A microfluidic cell squeezing device was designed with four compartments: inlet, outlet, cell filter, and squeezing channels (Fig. 1). The filter was included to prevent any aggregated cells from entering squeezing channels to clog the device. The filter region contains arrays of pillars that form sequential gaps, first 200 μm and then 50 μm in width. The gaps are followed by 22 squeezing channels in parallel. Each channel contains six constriction regions in series, each 6 μm wide and 40 μm long, separated by 120 μm spacer segments. The spacer regions gradually widen to 50 μm before tapering back to 6 μm prior to the next constriction. The device was fabricated by soft lithography using polydimethylsiloxane (PDMS). All features are 10 microns tall, meaning that the dimensions of the constriction region are 6 × 40 × 10 microns. The width of the region was set based on the size of COS7 cell nuclei, which typically have an average long axis length of 14 microns and short axis length of 10 microns – the width was set to 60% of the average short axis length to induce substantial nuclear deformation without compromising cell viability.

Fig. 1. Schematic of cell squeezing device design.

The image illustrates the top view of the mold used to cast the PDMS device. The areas in black are voids through which cells can travel with a height of 10 μm. The inlet and outlet are 2 mm in diameter. The filter consists of two rows of pillars with 200 μm gaps to catch large cell clusters, followed by 6 rows of pillars with 50 μm gaps to catch smaller clusters. After the gaps, all non-squeezing channels have a width of 50 μm or greater. There are 22 squeezing channels in parallel (see the red box inset with two channels), each of which consists of 6 consecutive constriction regions with a width of 6 μm and length of 40 μm. The squeeze regions are separated by 120 μm long spacer regions, the widths of the first and last 40 μm portions of each spacer gradually expand to 50 μm and reduce back to 6 μm, respectively.

Previous studies have shown that increasing the number of constriction regions enhances intracellular delivery of fluorescent macromolecules [5, 8]. To determine whether this trend also applies to nuclear delivery of pDNA, squeezing devices with either one (1×6 μm) or six (6×6 μm) constriction regions were evaluated in pilot experiments to determine which one was optimal for cell squeezing. Six constrictions corresponded to the upper range reported in prior studies [5, 16]. When cell squeezing was applied at 4 hours post-electrotransfection, 1×6 μm and 6×6 μm devices resulted in ~1.5x and ~7.6x increases, respectively, in the median intranuclear pDNA signal, with comparable effects on cell viability. In another pilot experiment, the delay between electrotransfection and cell squeezing varied between 30 minutes and 4 hours. We observed that the nuclear pDNA level increased with decreasing the time delay; it was nearly four folds higher in the 30-minute group than in the 4-hour group. Based on these results, the 6×6 μm configuration was selected for cell squeezing at 30 minutes post-electrotransfection in subsequent experiments.

Cell squeezing increased cellular uptake and nuclear entry of pDNA

COS7 cells were electrotransfected with unlabeled pEGFP, subsequently squeezed, and cultured in coverslip-bottom dishes. Un-squeezed controls were processed in parallel under the same conditions, except that the squeezing step was omitted. At 16 hours post electrotransfection, cells in both groups were fixed, stained with Hoechst and a fluorescent amplified DNA hybridization probe targeting pEGFP, and imaged with super-resolution confocal microscopy. The images were analyzed for quantification of total intracellular and intranuclear pDNA fluorescence. Intranuclear pDNA distribution was further classified as punctate or diffuse as previously described [13]. Representative images of cells in both groups are shown in Fig. 2a. Unless otherwise specified, fluorescence intensity data are presented as the normalized median pDNA intensities. For each paired experiment, the median values for the two groups were first calculated separately and then normalized to the average value of the two medians. The mean ± SD of the replicate medians for the un-squeezed and squeezed groups were, respectively, 0.16 ± 0.16 and 1.85 ± 0.16 for total intracellular pDNA (Fig. 2b); and 0.21 ± 0.21 and 1.79 ± 0.21 for intranuclear pDNA (Fig. 2c), indicating that cell squeezing increased intracellular pDNA signal by 11.6-fold and intranuclear pDNA signal by 8.5-fold, compared to electrotransfection alone (i.e., un-squeezed control). Using the same dataset, the percentage of intranuclear relative to intracellular intensity was calculated for each cell. The median of the percentage values was determined for each replicate without normalization. The mean ± SD of the percentage across replicates were 3.44% ± 0.82% and 3.48% ± 1.69%, respectively (Fig. 2d), with no detectable difference between the two groups.

Fig. 2. Cell squeezing increased intracellular and intranuclear pDNA levels without altering their relative ratio.

COS7 cells were electrotransfected with pEGFP and subsequently subjected to squeezing in the 6×6 μm microfluidic device. In the matched controls, the cells were un-squeezed. All cells were incubated in the full medium for 16 hours post electrotransfection. Then, they were fixed and stained with Hoechst dye (blue) for nuclei and the DNA probe (red) for EGFP. (a-g) Images near the center of >50 nuclei were acquired with confocal microscopy, with a total of 1,144 nuclei for the un-squeezed cells and 548 nuclei for the squeezed cells across 6 independent repeats. (a) Representative super-resolution images of (i) un-squeezed and (ii) squeezed COS7 cells. Scale bars: 10 μm. Normalized medians of integrated fluorescence intensities were calculated for (b) intracellular pDNA and (c) intranuclear pDNA. (d) Median percentage of intranuclear relative to total intracellular pDNA signal intensity. Normalized median intensities were also calculated separately for (e) intranuclear punctate pDNA and (f) intranuclear diffuse pDNA. (g) Average number of intranuclear pDNA punctates per cell in each group. ns: no significant difference, P > 0.05, ** P < 0.005, ***P < 0.0005, ****P< 0.0001; n=6, paired t-test.

The intranuclear pDNA signals were further divided into punctate and diffuse groups; and their median intensities were 0.42 ± 0.25 and 1.59 ± 0.25 (Fig. 2e) and 0.15 ± 0.17 and 1.85 ± 0.17 (Fig. 2f), respectively, indicating that cell squeezing disproportionately increased the diffuse pDNA signal compared with the punctate fraction. Furthermore, the average numbers of intranuclear pDNA puncta were observed to be 1.26 ± 0.26 and 1.74 ± 0.79 per nucleus in un-squeezed and squeezed cells, respectively (Fig. 2g). Together, the data shown in Fig. 2e and 2g suggested that the squeezing increased puncta size but not the number per nucleus.

To quantitatively evaluate effects of constriction width on pDNA uptake, cells were squeezed through devices containing six 18-μm constriction regions (6×18 μm). In two independent repeats, squeezing in this device increased intracellular pDNA intensity by 5.60- and 1.37-fold and intranuclear pDNA intensity by 1.65- and 3.02-fold (Fig. S1). These increases were substantially smaller than those achieved with the 6-μm constriction regions, indicating that narrower constrictions led to greater cellular and nuclear entry of pDNA.

Cell squeezing permeabilized cell membrane and nuclear envelope

To understand the mechanisms by which cell squeezing enhanced cellular and nuclear entry of pDNA, we next explored the impact of cell squeezing on cell membrane permeability. Previous studies suggest that permeabilizing the plasma membrane is essential for delivery of macromolecules into squeezed cells [7, 8]. To test if cell squeezing permeabilized the cell membrane in the current study, cells were stained with calcein-AM prior to squeezing. Calcein-AM is membrane-permeant, once inside the cells, it is converted to calcein, which is retained intracellularly if the plasma membrane is intact. Immediately after squeezing, a subset of cells exhibited little to no detectable calcein fluorescence (Fig. 3a, yellow arrows). Quantitatively, the calcein fluorescence intensities (mean ± SD) of cells in un-squeezed controls and squeezed in the 6×6 μm devices were 549.7 ± 194.5 au and 191.0 ± 123.1 au respectively (Fig. 3b). When the cells were squeezed in the 6×18 μm devices, the intensity (mean ± SD) was 443.2 ± 188.6 au (Fig. 3b). These results suggested that the plasma membrane was permeabilized after squeezing cells, allowing intracellular calcein to leak out. The permeability increase was substantially larger for cells squeezed in the 6×6 μm device compared to the 6×18 μm device, consistent with the observation of pDNA uptake by squeezed cells.

Fig. 3. Cell squeezing increases the plasma membrane permeability in cells.

(a-b) Increase in membrane permeability to calcein. COS7 cells loaded with calcein (green) was squeezed or un-squeezed in the 6×6 μm or 6×18 μm device. Immediately after the squeezing, typical brightfield and fluorescence images of cells in all groups are shown in (a). Some cells had no detectable calcein signal post squeezing in the 6×6 μm device (yellow arrows). (b) Mean fluorescence intensity per cell in un-squeezed (n = 105), 6×6 μm squeezed (n = 109), and 6×18 μm squeezed (195) cells. ****P < 0.0001; ANOVA with Tukey post-hoc test. Scale bars: 10 μm. Bars are mean ± SEM. (c-e) Increase in membrane permeability to GFP. EOMA cells stably expressing GFP were squeezed or un-squeezed with the 6×6 μm device and analyzed with flow cytometry immediately after squeezing. (c) percentage of GFP+ cells, (d) geometric mean of GFP intensity among GFP+ cells, and (e) concentration of cells in the suspension. **P < 0.005 ***P< 0.0005; n = 6, paired t-test. (f-h) Altered nuclear retention of EGFP-NLS. COS7 cells expressing EGFP-NLS (green) was squeezed or un-squeezed in the 6×6 μm device. (f) High-speed video snapshots of a COS7 cell passing through the constriction region. Adjacent images are separated by 90 msec. Scale bar: 15 μm. (g) Typical images of cells after squeezing. The nuclei were stained with Hoechst (blue). (h) Average fluorescence intensity per nucleus in un-squeezed (n = 174) and squeezed (n = 628) groups. ****P < 0.0001; unpaired t-test. Scale bars: 10 μm. Bars and error bars are mean and SEM respectively.

In addition to the small molecule calcein, we analyzed the cell membrane permeabilization to cytosolic proteins. Specifically, we used GFP as a biomarker for cytosolic proteins. EOMA cells stably expressing GFP were squeezed through the 6×6 μm devices. Immediately after squeezing, the percentage of GFP-positive cells and the geometric mean of the GFP fluorescence intensity per cell decreased to 84.3% ± 3.4% and 96.6% ± 0.85% of those in the un-squeezed controls, respectively (Fig. 3c–d). Although these differences were found to be statistically significant (P < 0.05, paired t-test), the amount of decrease was significantly lower than that of calcein. We also quantified the changes of cell concentration in suspension immediately after squeezing in the 6×6 μm device, and found that the concentration was reduced to approximately 50%, compared to the un-squeezed controls (Fig. 3e), indicating that half of the cells could not pass through the device.

The data shown in Figs. 2b through 2d demonstrated that cell squeezing proportionally increased intracellular and intranuclear pDNA signals, but did not alter their ratio. To determine whether this observation was molecular weight-dependent, we examined effects of cell squeezing on the ratio for a smaller molecule, EGFP. The experiment was conducted using COS7 cells transiently expressing EGFP fused with a nuclear localization signal (EGFP-NLS), which preferentially localized to the nucleus. The experiment was performed on ice to inhibit active nuclear transport. Un-squeezed cells were kept on ice for the same duration. Fig. 3f shows a single COS7 cell passing through a 6×6 μm constriction region. After squeezing, both groups were plated in coverslip-bottomed dishes and imaged using widefield fluorescence microscopy (Fig. 3g). Integrated fluorescence intensity per nucleus (mean ± SEM) in un-squeezed and squeezed cells were 74.20 ± 8.31 au and 23.41 ± 2.24 au, respectively (Fig. 3h), reflecting substantial depletion of EGFP-NLS from the nuclei and a decreased intranuclear to intracellular ratio in squeezed cells. Taken together, the data shown in Figs. 3g and 3h demonstrated that cell squeezing caused transient disruption of nuclear envelope selectivity that allowed passive transport of EGFP-NLS across the nuclear envelope.

We showed earlier that squeezing cells in pDNA-free suspension increased intracellular pDNA signal by approximately one order of magnitude (Fig. 2b). What is the source for the additional intracellular pDNA? We hypothesize that the increased pDNA levels in squeezed cells are attributable to enhanced endocytic uptake of pDNA remaining associated with the plasma membrane following electrotransfection. To test the hypothesis, cells were divided into three groups. At time zero, pDNA was electrotransfected into cells in all groups, followed by cell culture for up to 16 hours. Cells in Groups 1 and 2 were stained with the DNA probe at 1 and 16 hours, respectively, and immediately imaged by fluorescence microscopy without squeezing. Cells in Group 3 were squeezed at 30 min and processed for imaging at 16 hours. The images in Group 1 showed an abundance of pDNA associated with the plasma membrane, with some pDNA internalized via endocytosis (Fig. S2a, left). By 16 hours, substantially less pDNA signal was visible in un-squeezed cells (Group 2) (Fig. S2a, middle), whereas cells squeezed using the 6×6 μm device in Group 3 retained strong pDNA signal (Fig. S2a, right). Quantitatively, the integrated intensity of pDNA signal per cell in sum-intensity projection images (mean ± SEM) was 8.25×10⁷ ± 0.83 ×10⁷ (au) and 1.43×10⁷ ± 0.30 ×10⁷ (au) for un-squeezed cells at 1 and 16 hours, respectively, and 6.04×10⁷ ± 0.83 ×10⁷ (au) for squeezed cells at 16 hours (Fig. S2b), indicating that only 17% of pDNA in un-squeezed cells remained detectable at 16 hours post-electrotransfection relative to 1 hour, presumably due to gradual release of membrane-associated pDNA into the medium during the 16-hour incubation period. In contrast, 73% was retained in squeezed cells. These findings support the hypothesis that cell squeezing post electrotransfection promote internalization of membrane-bound pDNA via endocytic pathways, thereby reducing extracellular loss.

Viability of squeezed cells

Cell squeezing may reduce the cell viability. To quantify the effect, COS7 cells were electrotransfected and then divided into un-squeezed and 6×6 μm squeezed groups. Thereafter, cells in both groups were plated at equal density and cultured for 24 hours. At the end of incubation, detached and adherent cells in each culture dish were collected and counted separately by flow cytometry. Cell density was defined as the total number of detached and adherent cells per unit volume. Detached cells were considered nonviable; therefore, viability analysis by flow cytometry was limited to the adherent cells. Prior to analysis, propidium iodide (PI) was added to these cells to assess plasma membrane integrity. Cell viability was defined as the percentage of PI-negative adherent cells relative to the total number of cells (adherent plus detached) in each culture dish. We observed that PI-positive cells were negligible among adherent cells; therefore, the viability value was approximately equal to the percentage of adherent cells per dish.

We found that the cell density and viability in the squeezed group were 33% ± 11% and 97% ± 1%, respectively, relative to those in the un-squeezed controls (Fig. S3). Combined with the data shown in Fig. 3e, these results indicated that squeezing significantly reduced cell density, primarily due to cell retention within the device and cell killing during the squeezing process. However, among the cells that successfully passed through the squeezing channels and reached the outlet, the majority remained viable during the 24-hour culture period.

Effects of cell squeezing on transgene expression

Cell squeezing substantially increased cellular uptake and nuclear entry of pDNA (Fig. 2b and 2c). Will the increase enhance transgene expression? To answer the question, COS7 cells were prepared under the same conditions as those in the pDNA uptake experiment. Instead of using DNA probe, these cells were stained with Hoechst and an amplified RNA hybridization probe targeting EGFP mRNA, followed by imaging with super-resolution confocal microscopy and subsequent image analysis.

Transfection efficiency at the mRNA level in all samples was quantified with two metrics: (i) %EGFP-mRNA+ cells: defined as the percentage of cells in each image with detectable EGFP-mRNA signal, and (ii) transcription level: defined as the geometric mean of mRNA probe signal intensity per cell among EGFP-mRNA+ cells. Representative images of un-squeezed and squeezed cells are shown in Fig. 4a. %EGFP-mRNA+ cells (mean ± SD) was 85.87 ± 9.74 in un-squeezed group and 87.22 ± 7.58 in squeezed group (Fig. 4b). Transcription level (mean ± SD) was 1.06 ± 0.18 in un-squeezed group and 0.94 ± 0.18 in squeezed group (Fig. 4c). No significant difference was observed for either metric. The data indicated that 6×6 μm squeezing had indistinguishable effects on the transfection efficiency at the mRNA level although squeezing increased the intranuclear level of pDNA by 8.4-fold (Fig. 2c) at 16 hours post electrotransfection. One possible explanation for the lack of the mRNA increase was saturation of the transcriptional machinery. To address this possibility, we repeated the experiment using a reduced pEGFP dose of 0.25 μg per million cells instead of 1 μg during electrotransfection. In the low dose experiment, %EGFP-mRNA+ cells (mean ± SD) decreased to 62.20 ± 8.57 in un-squeezed group and 67.95 ± 22.12 in squeezed group (Fig. S4a). The transcription level (mean ± SD) was 1.07 ± 0.38 in un-squeezed group and 0.93 ± 0.38 in squeezed group (Fig. S4b). Again, no significant differences were observed between the two groups for either metric, suggesting that the lack of an observed increase in transcription despite substantial increase in intranuclear pEGFP in squeezed cells was not due to saturation of transcriptional machinery.

Fig. 4. Effects of cell squeezing on transgene transcription and protein production.

Transcription analysis. COS7 cells were prepared using the same protocol as described in Fig. 2 except a hybridization probe recognizing EGFP-mRNA was used. Cell nuclei are stained with Hoechst (blue), and EGFP-mRNA probe was stained with Opal 690 (red). Z-stack images spanning the height of cells with a step size of 1 μm were taken of >70 cells per group with a total of 619 nuclei for the un-squeezed (US) cells and 571 nuclei for the 6×6 μm squeezed cells across 6 independent repeats. Images were sum-intensity projected ahead of automated image analysis to quantify total EGFP-mRNA signal within cells. (a) Representative sum-intensity projected confocal images of COS7 cells that were (i) un-squeezed and (ii) squeezed. Scale bars: 10 μm. (b) Percent of cells with EGFP-mRNA signal higher than a non-transfected control. (c) Transcription level defined as the geometric mean of EGFP-mRNA signal in EGFP-mRNA+ cells. The data for each matched pair of un-squeezed and squeezed groups were normalized by the mean of the two values to account for day-to-day variation in intensity values. ns: no significant difference, P > 0.05; n=6, paired t-test. (d) Translation analysis. COS7 cells were prepared using the same protocol as described in Fig. 2. Flow cytometry analysis was performed at 24, 48, or 72 hours post electrotransfection. n=6 independent repeats. The results are shown as percentage of GFP positive cells (left) and expression level defined as geometric mean of EGFP signal (right). Un-squeezed: black circles; 6×6 μm squeezed: blue triangles. * P < 0.05, ** P < 0.005, *** P < 0.001; n=6, 2-way ANOVA with Šídák’s post-hoc.

Transfection efficiency at the protein level was quantified with two metrics as well: (i) transfection efficiency: defined as the percentage of viable cells positive for EGFP fluorescence, and (ii) expression level: defined as the geometric mean of fluorescence intensity among EGFP-positive cells. Our data showed that relative to un-squeezed cells, squeezing through 6×6 μm constrictions resulted in 0.70 ± 0.06, 0.64 ± 0.06, and 0.60 ± 0.06-fold changes in transfection efficiency at 24, 48, and 72 hours post electrotransfection, respectively (Fig. 4d, left), and led to 1.12 ± 0.16, 1.31 ± 0.16, and 1.08 ± 0.11-fold changes in expression level at the same time points (Fig. 4d, right). In a replicated experiment using a lower pEGFP dose (0.25 ug per million cells), cell squeezing resulted in 0.75 ± 0.11, 0.71 ± 0.12, and 0.63 ± 0.10-fold changes in transfection efficiency (Fig. S4c, left), and 1.40 ± 0.21, 1.48 ± 0.16, and 1.36 ± 0.07-fold changes in expression level (Fig. S4c, right). To validate the flow cytometry findings, we performed Western blot analysis of EGFP. The EGFP band intensity is approximately proportional to the product of transfection efficiency and expression level; thus, it provides an independent measure of overall protein production. The relative EGFP band intensity in squeezed cells was 0.896-fold of that in un-squeezed controls (Fig. S5), consistent with the trend observed in the flow cytometric analyses (Figs. 4d and S4c).

The transgene expression results indicated that cell squeezing consistently reduced transfection efficiency at EGFP protein level, despite minimal effects on its mRNA level. One possible explanation is that cell squeezing impairs nuclear export of EGFP mRNA for protein production. To test this possibility, we analyzed EGFP mRNA images shown in Fig. 4 and Fig. S4, along with all other images from the same experimental groups, and quantified the percentage of intranuclear EGFP mRNA relative to total intracellular signal. No difference was observed between un-squeezed and squeezed cells (Fig. S6), indicating that nuclear export of EGFP mRNA was not inhibited by cell squeezing. Taken together, the results above suggest that cell squeezing may impair pDNA transcription and/or subsequent translation during transgene expression.

Effects of cell squeezing on bulk transcription and translation

In addition to EGFP expression, we determined how cell squeezing impacted bulk transcription and translation. To quantify bulk transcription, 5-ethynyl uridine (EU) was fluorescently labeled; and the total RNA synthesis was measured by quantifying the fluorescence intensity of EU incorporated into nascent RNA over a 16-hour incubation period following squeezing, using widefield microscopy and image analysis. Similar to the pDNA measurements shown in Fig. 2, total RNA fluorescence intensity is presented as the normalized median RNA intensities (Fig. 5a). The mean ± SD of the replicate medians for squeezed cells was 0.72 ± 0.22 relative to un-squeezed controls. Total protein levels were quantified using the BCA assay. At 24 hours post squeezing, the mean ± SD protein concentrations (μg per million cells) were 2,304 ± 1,004 for un-squeezed cells and 2,118 ± 988 for squeezed cells. At 48 hours, the corresponding values were 2,602 ± 594 and 2,730 ± 717, respectively (Fig. 5b). Together, the RNA synthesis and total protein data indicate that cell squeezing with the 6×6 μm device reduced bulk transcription, while having insignificant effects on total protein levels.

Fig. 5. Effects of cell squeezing on bulk transcription and translation.

(a) Bulk transcription analysis. Un-squeezed and 6×6 μm squeezed COS7 cells were prepared in parallel and incubated for 16 hours in medium containing 5-Ethynyl Uridine (EU). At the end of incubation, cells were stained using the Click-iT RNA kit to detect RNA synthesized during the incubation period. Widefield fluorescence images of >130 cells were taken per group in each of 6 independent repeats for a total of 1,189 un-squeezed cells and 1,113 squeezed cells. (left) Representative images of (i) un-squeezed and (ii) 6×6 μm squeezed cells. Nuclei are stained with Hoechst (blue) and newly synthesized RNA is stained with Alexa Fluor 594 (red). Scale bars: 10 μm. (right) median of RNA intensity per cell – matched pairs of squeezed and un-squeezed values were normalized by dividing each value by the average of both to account for day-to-day variation in intensity values. *P < 0.05; n=6, paired t-test. (b) Bulk translation quantification. Un-squeezed and 6×6 μm squeezed samples of equal cell number were prepared at 24- or 48-hours post squeeze, and the BCA assay was used to quantify total protein per million cells. ns: no significant difference, P > 0.05; n = 6 (24 hours), n = 5 (48 hours), 2-way ANOVA with Šídák’s post-hoc.

Effects of cell squeezing on stress granule formation

Gene transcription is regulated by complex interactions among transcription factors, regulatory elements, and signaling pathways, in addition to nuclear concentration of pDNA. To further elucidate how cell squeezing affected EGFP expression, we investigated potential mechanisms by which squeezing could reduce transcription. We first examined the formation of stress granules in squeezed cells after immunofluorescence staining of the biomarker G3BP1 (Fig. S7a); cells treated with 0.5 mM sodium arsenite served as a positive control (Fig. S7b). Given that cell squeezing caused membrane permeabilization (Fig. 3) and that stress granules can form in response to membrane disruption [17], squeezing might induce stress granule formation. We observed that G3BP1 signal intensity appeared elevated after squeezing (Fig. S7a and S7c). However, this increase was attributable to reduced cell density in squeezed samples, which altered antibody staining level rather than reflecting genuine stress granule induction (Fig. S7d). After correction for the cell number, the data indicated that squeezing through the 6×6 μm device did not lead to more stress granules compared with un-squeezed controls.

Effects of cell squeezing on chromatin condensation

A second potential mechanism underlying the observed transcriptional changes in squeezed cells relates chromatin condensation, which can lead to transcriptional repression. Previous studies have shown that nuclear compression-induced chromatin condensation reduces levels of active RNA polymerase II [18, 19], which is required for pDNA transcription. In addition, transfected pDNA can associate with histone proteins to form mini-chromosomes [20, 21], a heterochromatin-like structure that may limit RNA polymerase II access and thereby suppress transcription. To determine whether chromatin condensation occurred under our experimental conditions, cells were electrically pulsed and squeezed using 6×6 μm devices as described above. To avoid interference from EGFP fluorescence during immunostaining, pDNA encoding EGFP was not added to the pulsing buffer. At 1.5 hours post-squeezing, cells were immunostained to assess levels of the heterochromatin markers H3K27me3 and H3K9me3, as well as the euchromatin marker H3K9ac, using confocal microscopy. The integrated intensity of chromatin marker signal per cell was normalized to the H3 signal to control for variability in antibody staining between cells or between experimental groups.

Cell squeezing increased the normalized H3K27me3 intensity to 1.72 ± 0.39 relative to un-squeezed controls, while decreasing the normalized intensities of H3K9me3 and H3K9ac to 0.746 ± 0.096 and 0.699 ± 0.21, respectively (Fig. 6). H3K27me3 is a marker of facultative heterochromatin that forms in response to specific stimuli and is associated with transcriptional repression [22]. In contrast, H3K9me3 is typically found in permanently condensed genomic regions and plays a structural role with less direct involvement in dynamic transcriptional regulation [22]. The significant increase in facultative heterochromatin, together with decreases in constitutive heterochromatin and euchromatin markers, suggests that cell squeezing promotes region-specific chromatin condensation, which may contribute to the observed reduction in transcription.

Fig. 6. Effects of cell squeezing on chromatin condensation.

Un-squeezed and squeezed COS7 cells were fixed 1.5 hours after squeezing and were stained with Hoechst for nucleus segmentation, a mouse anti-histone H3 antibody, and rabbit antibodies for (a) H3k27me3, (b) H3k9me3, and (c) H3k9ac. Mouse and rabbit secondary Abs with Alexa Fluor 488 (not shown) and Alexa Fluor 568 (red) were used to tag primary antibodies for H3 and the various chromatin state markers. Z-stack confocal images were taken across the full depth of nuclei in each FOV with a step size of 1 μm. Sum intensity projected z-stacks were analyzed for total H3 and chromatin state marker signal per nucleus. Representative slices of each chromatin marker within cell nuclei are shown on the left; and quantitative analysis is shown to the right of representative images. Total chromatin state marker intensity was normalized by total histone H3 intensity to account for cell-to-cell or group-to-group variation in antibody staining. Two sets of pairs in panels (b) and (c) have values that are too similar to be visually distinct, resulting in the appearance of three sets of pairs instead of four. Median H3-normalized intensity values for matched pairs of un-squeezed and squeezed replicates were further normalized to account for experiment-to-experiment variation in intensity values by dividing each value by the average of two values. 4 independent repeats with >35 nuclei were imaged in each group for a total of 220 un-squeezed cells and 148 squeezed cells for H3k27me3, 206 un-squeezed cells and 165 squeezed cells for H3k9me3, and 204 un-squeezed cells and 160 squeezed cells for H3k9ac. ns: no significant difference (P = 0.10), *P < 0.05, n = 4, paired t-test. Scale bars: 5 μm.

Previous studies have shown that HDAC3, a histone deacetylase that catalyzes the removal of acetyl groups from histone proteins, plays a critical role in chromatin condensation under conditions of geometric confinement or nuclear compression [23, 18, 19, 24]. Most of these studies report increased nuclear localization of HDAC3 in response to such mechanical treatments [23, 18, 24]. To determine whether this response also occurred in cells squeezed using 6×6 μm device, HDAC3 was immunostained in both un-squeezed and squeezed samples. Cells were co-stained with Hoechst to enable nuclear segmentation and with ATTO647N, an NHS-ester dye that binds cellular proteins, to facilitate whole-cell segmentation (Fig. S8a). Confocal microscopy was performed at 1.5 hours post squeezing, followed by quantitative image analysis to measure total HDAC3 intensity per cell and the percentage of HDAC3 localized within the nucleus. The median percentage of nuclear HDAC3 was 77.95 ± 2.92% in un-squeezed cells and 85.75 ± 1.19% in squeezed cells (Fig. S8b).

The significant increases in facultative heterochromatin (Fig. 6a) and intranuclear HDAC3 (Fig. S8) in squeezed cells suggested that HDAC3-mediated chromatin condensation may contribute to the reduced transgene expression in squeezed cells. To test this possibility, we examined whether inhibition of HDAC3 with RGFP966 could restore transgene expression. Four experimental groups were included: (i) un-squeezed (US) and untreated, (ii) squeezed (6×6 μm) and untreated, (iii) un-squeezed and treated with RGFP966, and (iv) squeezed (6×6 μm) and treated with RGFP966. The percentages of EGFP+ cells (mean ± SD) were 27.31% ± 7.75%, 19.20% ± 5.87%, 47.86% ± 10.45%, 33.51% ± 9.02% in Groups i through iv, respectively (Fig. 7a). In the same groups, the expression levels per EGFP+ cell (mean ± SD) were $\left(2.42\pm 0.65\right)\times {10}^5$, $\left(2.58\pm 0.47\right)\times {10}^5$, $\left(4.54\pm 1.15\right)\times {10}^5$, and $\left(3.80\pm 0.86\right)\times {10}^5$ au, respectively (Fig. 7b). The ratio of transfection efficiency between squeezed and un-squeezed cells was approximately 0.7, regardless of RGFP966 treatment. For expression level, the squeezed to un-squeezed ratio was ~1.1 in untreated samples but ~0.8 in RGFP966-treated samples. On the other hand, HDAC3 inhibition increased transfection efficiency to a similar extent in both squeezed and un-squeezed cells but did not alter their relative relationship, compared with untreated groups. In contrast, HDAC3 inhibition enhanced expression level more strongly in un-squeezed cells than in squeezed cells. Notably, in the absence of HDAC3 inhibition, squeezed cells exhibited slightly higher expression level than un-squeezed cells, whereas under HDAC3 inhibition, squeezed cells showed lower expression level. These results suggest that cell squeezing and HDAC3 inhibition may exert opposing effects on certain chromatin condensation pathways that regulate transgene expression.

Fig. 7. HDAC3 inhibitor RGFP966 improves transgene expression in squeezed cells.

COS7 cells were electrotransfected with pEGFP then incubated for 30 minutes with or without 10 μM RGFP966. Treated and untreated cells were then further split into un-squeezed and 6×6 μm squeezed cells as previously described and incubated for 24 hours before flow cytometry was used to measure transgene expression. (a) Percentage of GFP positive cells and (b) expression level presented as geometric mean of EGFP signal intensity, both normalized to un-squeezed, untreated group. All bars are mean ± SD. *P < 0.05, **P < 0.005; n = 4, 2-way ANOVA with Šídák’s post-hoc.

Discussion

We investigated the mechanisms underlying gene delivery by microfluidic cell squeezing. Squeezing markedly increased cellular uptake and nuclear accumulation of pDNA compared with un-squeezed controls. However, this increase did not enhance transgene expression, likely due to nuclear compression-induced chromatin condensation that suppressed transcription. This effect may partly explain the limited adoption of mechanoporation for gene delivery. Importantly, treatment with an HDAC3 inhibitor RGFP966 partially restored transgene expression, implicating chromatin remodeling in the observed transcriptional repression.

Data in our study indicated that the increase in intracellular pDNA after squeezing (Fig. 2b) was due to endocytosis of plasma membrane-associated pDNA because this was the only pDNA source available for cellular entry. Previous studies have shown that membrane damage induces compensatory vesicle trafficking responses, including endocytosis and exocytosis [25–28]. Although endocytosis may be transiently suppressed immediately after membrane injury, it is subsequently upregulated to remove damaged membrane components and excess lipids delivered during repair [29, 30]. A dynamic balance between transient suppression and subsequent upregulation of endocytic activity may therefore contribute to the increased intracellular pDNA observed after squeezing. Further studies are needed to directly investigate this mechanism [30].

Squeezed cells exhibited >90% viability at 24 hours post-squeezing (Fig. S3b), indicating that disruptions of the plasma membrane and nuclear envelope are largely transient and reversible. These results are consistent with previous reports [9, 31, 5, 8]. However, total cell density decreased by 50–70% immediately after squeezing (Figs. 3e and S3a). This reduction was primarily due to cell retention within the microfluidic device and cell killing during squeezing. Consistent with this interpretation, flow-through samples contained increased levels of cellular debris compared with un-squeezed controls. Collectively, these findings indicated that although surviving cells maintained high viability, a substantial fraction of cells either failed to pass through the device or did not withstand the process. Future efforts should focus on optimizing device design and operating conditions to reduce cell retention and damage, thereby improving overall throughput.

We initially hypothesized that cell squeezing might enhance nuclear entry of pDNA by transiently disrupting nuclear envelope selectivity. This rationale was based on previous studies reporting rapid transgene expression following treatments that compromise nuclear envelope barrier function [9, 32]. In the present study, squeezing similarly caused transient loss of nuclear envelope selectivity, permitting passive redistribution of EGFP-NLS between nucleus and cytoplasm (Fig. 3g and 3h). However, this loss of selectivity is likely insufficient to permit passive transport of pDNA across the nuclear envelope, which is substantially larger than EGFP-NLS. As a result, squeezing enhanced pDNA accumulation in both cytoplasmic and nuclear compartments, but no difference was observed in the percentage of total intracellular pDNA localized to the nucleus between squeezed and un-squeezed groups. This observation is consistent with previous experimental results and with statistical model predictions for cells with an intact nuclear envelope, which demonstrate that nuclear and cytoplasmic pDNA copy numbers increase proportionally with increasing pDNA dose during electrotransfection, while their ratio remains constant [33, 34]. Collectively, the results discussed above suggest that squeezing does not sufficiently disrupt nuclear envelope selectivity to permit enhanced nuclear entry of large molecules such as pDNA, leaving the mechanisms governing pDNA nuclear import and intracellular partitioning largely unaffected. One caveat in the discussion is that a substantial fraction of transfected cells may have undergone division by the time intracellular pDNA distribution is analyzed [13]. Because mitosis involves transient nuclear envelope breakdown, cell division can facilitate nuclear entry of pDNA. Accordingly, future studies should examine whether cell squeezing enhances pDNA nuclear entry in non-dividing cells. If confirmed, this approach could represent a valuable strategy for improving gene delivery in primary cells [35].

Despite the increased nuclear accumulation of pDNA in squeezed cells, pDNA-derived mRNA levels were unchanged compared with un-squeezed controls (Figs. 4a, 4b, 4c, S4a and S4b). Moreover, cell squeezing reduced total RNA synthesis (Fig. 5a). These findings prompted us to investigate potential mechanisms by which squeezing might suppress transcription. We first considered whether stress granule formation could account for the apparent reduction in transcription. Stress granules are membrane-less organelles that generate in response to various cellular stresses. They can sequester mRNA and, in some contexts, promote its degradation [36–38]. In addition, stress granules have been implicated in cellular responses to membrane damage [17], which may occur during cell squeezing. However, we detected no increase in stress granule formation following squeezing (Fig. S7). Although the stress granule formation does not appear to contribute to reduced transgene expression in squeezed cells, these data provide useful insight, considering that the potential for mechanical stress to induce stress granules remains largely unexplored in the current literature.

Previous studies have shown that nuclear compression, whether during cell migration through narrow channels [19] or by externally applying compressive force [18], increases chromatin condensation and reduces global transcription. Chromatin condensation could similarly suppress transcription of transfected pDNA. Accordingly, we hypothesized that mechanical compression of the nucleus during cell squeezing promoted chromatin condensation, thereby reducing transcription. To test this hypothesis, we measured H3K27me3, H3K9me3, and H3K9ac, markers of facultative heterochromatin, constitutive heterochromatin, and euchromatin, respectively, in squeezed and un-squeezed cells. Squeezing increased H3K27me3 but decreased H3K9ac, consistent with enhanced chromatin condensation and transcriptional repression, as well as with previous reports showing that nuclear compression elevates facultative heterochromatin and reduces euchromatin [18, 19]. However, those studies also reported increased constitutive heterochromatin (H3K9me3), in contrast to our findings.

One possible explanation is the duration of compression. In our system, nuclear deformation lasted only seconds or less, whereas previous studies applied compression for extended periods: hours during migration [19] or approximately one hour under external force [18]. It is plausible that both transient and prolonged compression upregulate facultative heterochromatin and reduce euchromatin, but the prolonged compression may be required to increase constitutive heterochromatin that plays a critical role in maintaining nuclear mechanical stability [39].

Song et al. reported decreased H3K9me3 but no change in H3K27me3 or H3K9ac after forcing cells through 7-μm channels [40]. Although their study and ours both used microfluidic constrictions for cell squeezing, the mechanical stress differed substantially. Song et al. used a single 7-μm constriction at 20 μL/min, whereas our device contained six 6-μm constrictions and operated at a 5.5-fold higher flow rate. Smaller channel size, increased constriction regions, and higher flow rate are all known to increase mechanical stress during squeezing [5, 6]. Consistent with the stress difference, Song et al. observed minimal disruption of nuclear envelope selectivity, while we detected transient loss of nuclear envelope selectivity for EGFP-NLS (Fig. 3g and 3h), indicating that our protocol imposed greater nuclear deformation. These findings suggest that the extent and nature of chromatin remodeling depend on the magnitude and duration of mechanical stress. While our results demonstrate a consistent increase in facultative heterochromatin in squeezed COS7 cells, further studies are needed to determine whether this response generalizes to other cell types and to define how parameters such as compression duration, constriction geometry, and flow rate influence chromatin condensation.

To further elucidate the mechanisms of chromatin condensation, we examined the role of HDAC3. Consistent with previous reports [23, 18, 19, 24], we observed a significant increase in nuclear HDAC3 localization in squeezed cells (Fig. S8). This redistribution suggests that enhanced HDAC3 activity within the nucleus contributes to the increase in facultative heterochromatin following squeezing. We next evaluated the effect of treatment with an HDAC3 inhibitor, RGFP966, which has been shown to reduce chromatin condensation [41], on transgene expression in both squeezed and un-squeezed cells. Although the treatment increased transgene expression in both groups (Fig. 7), squeezing reduced transfection efficiency by approximately 30% regardless of inhibitor treatment (Fig. 7a). These results indicated that cell squeezing and HDAC3 inhibition influenced transfection efficiency through distinct mechanisms. For instance, both HDAC3 and HDAC6 have been implicated in tubulin deacetylation, a process that decreases microtubule stability [42, 43]. Inhibition of HDAC activity can therefore stabilize microtubules and enhance intracellular trafficking of transfected pDNA [44, 45]. Accordingly, RGFP966 may improve transfection efficiency in part by promoting microtubule-dependent transport of pDNA, whereas the reduction in transfection efficiency observed after cell squeezing likely results from a distinct, currently unknown mechanism. Another notable observation is that cell squeezing can exert opposing effects on expression level per cell, depending RGFP966 treatment. In untreated cells, squeezing consistently caused no change or increased expression levels at both low and high pEGFP doses (Fig. S4c, right; Fig. 4d, right; Fig. 7b). In contrast, in RGFP966-treated cells, squeezing reduced expression levels by approximately 20% (Fig. 7b). These findings suggest that HDAC3 inhibition may partially counteract the transcriptional effects associated with cell squeezing. With further optimization of inhibitor dose and treatment duration, HDAC3 inhibition could potentially counteract squeezing-induced chromatin condensation and thereby enable the increased nuclear pDNA to yield higher transgene expression.

Although EGFP mRNA levels were similar between un-squeezed and squeezed cells after electrotransfection, EGFP protein production was reduced in squeezed cells. This discrepancy could result from (i) impaired nuclear export of mRNA, (ii) misfolding of EGFP that reduces detectable fluorescence, and/or (iii) decreased translation efficiency. Analysis of EGFP mRNA localization revealed no difference in the fraction of transcripts retained in the nucleus between groups (Fig. S6), indicating that nuclear export was not affected. To evaluate whether impaired EGFP folding contributed to the decreased fluorescence signal in flow cytometry analysis, we quantified EGFP by Western blot, which detects both properly folded and misfolded protein [46]. Western blot measurements can be directly compared with the product of transfection efficiency and per-cell expression level obtained by flow cytometry [29], as both reflect total EGFP protein abundance per sample. Consistently, a 10–22% reduction in EGFP protein level was observed in squeezed cells (Figs. 4 and S5), arguing against the loss of fluorescence due to protein misfolding. These results suggest that cell squeezing reduces translation of pDNA-derived mRNA, although the underlying mechanisms remain unclear. Elucidating these mechanisms will be essential for optimizing cell squeezing as a strategy to enhance mechanoporation-mediated gene delivery.

In summary, we demonstrated that microfluidic cell squeezing substantially increased cellular uptake and nuclear accumulation of pDNA without markedly compromising cell viability. However, this enhanced delivery did not translate into proportional increases in transfection efficiency or transgene expression relative to un-squeezed controls. Our data indicated that this limitation was partly attributable to squeezing-induced chromatin condensation, particularly related facultative heterochromatin, which likely suppresses transcriptional activity. Inhibition of HDAC3 increased transgene expression in both squeezed and un-squeezed cells, presumably by reducing chromatin compaction. Further studies are required to define the molecular mechanisms underlying translational suppression and chromatin remodeling following squeezing, and to identify optimal HDAC inhibitors or treatment regimens that can be combined with cell squeezing to fully leverage the increased nuclear pDNA for enhanced transgene expression.

Data Availability:

All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplemental Information.