Cellular Delivery of Large Functional Proteins and Protein-Nucleic Acid Constructs via Localized Electroporation

Abstract

Delivery of proteins and protein-nucleic-acid constructs into live cells enables a wide range of applications from gene editing to cell-based therapies and intracellular sensing. However, electroporation-based protein delivery remains challenging due to the large sizes of proteins, their low surface charge, and susceptibility to conformational changes that result in loss of function. Here, we use a nano-channel-based localized electroporation platform with multiplexing capabilities to optimize the intracellular delivery of large proteins (β-galactosidase, 472 kDa, 75.38 % efficiency), protein-nucleic acid conjugates (protein spherical nucleic acids (ProSNA), 668 kDa, 80.25 % efficiency), and Cas9-ribonucleoprotein complex (160 kDa, ~60% knock-out and ~24% knock-in) while retaining functionality post-delivery. Importantly, we delivered the largest protein to date using a localized electroporation platform and showed a nearly two-fold improvement in gene editing efficiencies compared to previous reports. Furthermore, using confocal microscopy, we observed enhanced cytosolic delivery of ProSNAs, which may expand opportunities for detection and therapy.

The introduction of exogenous proteins into cells enables the analysis and manipulation of cellular functions such as the perturbation of intracellular signaling pathways¹, genome editing^{2, 3}, regulation of gene expression^{4, 5}, and intracellular sensing.^{6, 7} These capabilities can be leveraged to treat diseases or to elucidate fundamental mechanisms of the cell’s complex machinery. There are several key approaches for the delivery of functional proteins into cells, including physical systems (e.g., electroporation^8–12, cell squeezing^13–15, and hydroporation^{16, 17}), exogenous transfection reagents (e.g., lipid vesicle carriers^{18, 19}), chemical modifications of protein surfaces (e.g., protein spherical nucleic acids^{20, 21} cell penetrating peptides²², hydrophobic modifications²³) and polymer based carriers.^{24, 25} However, due to the large variation in protein size, surface charge, and structure, there are advantages and disadvantages for each platform depending on the characteristics of the protein and application of interest. For instance, recently it has been shown how polymers can be functionalized to carry a range of proteins and peptides into the cytosol. Polymeric carriers can be used for both in vitro and in vivo applications.²⁶ However, forming stable protein polymer complexes is challenging due to the lower charge density of most proteins compared to nucleic acids, cationic polymers have the issue of cell toxicity in some cases and delivery efficiency can be highly dependent on chain length, degree of functionalization and therefore require extensive optimization.²⁶ In the context of in vitro studies, the physical delivery systems like electroporation provide more control over the dosage and the intracellular uniformity of the delivered protein.²⁷ However, delivery of large proteins with traditional electroporation-based systems is challenging because of the relatively small pore sizes generated across cell membranes as compared to other physical membrane disruption systems like microinjection and mechanoporation.^{5, 28, 29} An additional challenge with electroporation systems is the diffusion-dominant molecular transport of weakly charged proteins, which contrasts the electrophoretic-dominant delivery of highly charged nucleic acids.³⁰ Therefore, to successfully deliver a particular protein with electroporation, several conditions must be met: 1) the size of the pores generated by the electric field must be greater than a certain threshold for the protein of interest to directly enter the cytosol, 2) the pores must remain open above this threshold size for enough time for delivery of proteins in sufficient quantity, 3) the structure of the protein must be preserved to retain its functional integrity, and 4) the pores must reseal following delivery to prevent cell death. Extensive optimization of the electric pulse parameters (voltage amplitude, pulse shape, frequency, pulse duration, and number of pulses), in conjunction with other experimental factors such as electroporation buffer composition and protein concentration, is required to satisfy these conditions and achieve efficient cargo delivery.^{8, 9, 30} Recently, we developed a well-plate based localized electroporation device (LEPD) designed for the multiplexing of pulse parameters and experimental conditions to expedite the optimization process needed to achieve efficient delivery of plasmids, oligonucleotides, and siRNA in a variety of cell types, including hard-to-transfect induced pluripotent stem cells, while retaining high cell viability.⁸

In this work, we used the multi-well plate LEPD to ascertain the efficient delivery of large proteins (hundreds of kDa) and identify pulse-shape and voltage amplitude parameters that optimize protein delivery. As opposed to commercial systems like bulk electroporation, which come with preset conditions and are difficult to optimize on the user front, the multi-well LEPD confers full experimental optimization capability on the user front. This can be used to individually tune experimental conditions like pulse characteristics for every delivery application. Here, we demonstrate the efficient delivery of a large protein (β-galactosidase, β-gal, MW = 472 kDa), a protein densely surface-functionalized with DNA (protein spherical nucleic acid conjugate, ProSNA, MW = 668 kDa), and a ribonucleoprotein complex (Cas9-RNP, MW = 160 kDa). Furthermore, we used an automated imaging platform to acquire fluorescence images across the multi-well LEPD immediately following electroporation and employed a deep-learning and feature extraction pipelines (CellProfiler) to segment cells and obtain various morphological measurements from each well with single-cell resolution ^{8, 31–34}.

To deliver protein cargoes into both adherent and suspension cell types, while preserving function and viability, we utilized the multi-well LEPD and optimized the electroporation parameters. The LEPD design consists of a polycarbonate nanochannel membrane attached to a cell culture well (seeding capacity ~ 20,000 – 50,000 cells) that is placed in a multi-well plate with embedded gold electrodes (Fig. 1A–D). Each row in the well-plate LEPD is individually addressable by a pulse generator with tunable parameters that include pulse shape, amplitude, frequency, and duration. Upon application of the electric pulses, the electric field is localized at the interface between the nanochannels (diameter = 400 nm) and the cell membrane, which results in the formation of hydrophilic pores in the cell membrane when the transmembrane potential (TMP) exceeds a threshold level (~ 0.2 – 1 V).^35–37 The dynamic process of pore formation, expansion, and resealing is sensitive to several parameters including the TMP and membrane tension, which can be altered by tuning the electrical inputs and the cell culture conditions.^{28, 35} To investigate the system capability for delivery of very large proteins, we selected Alexa Flour 647 (AF-647) tagged β-gal (Fig. 1E) as a model protein because of its large size and well-established activity assay. We also investigated the effect of charge by delivering β-gal ProSNAs, which combine the functionality of both proteins and nucleic acids.^{20, 38, 39}

Figure 1. Multi-well LEPD architecture, localized electroporation mechanism and intracellular delivery.

(A) Schematics of multi-well LEPD system, (B) single LEPD containing cells, and (C) localized electroporation mechanism for protein delivery. (D) Photograph of physical multi-well LEPD assembly, which contains a top PCB electrode holder with Au stub pins, a bottomless multi-well plate, and a bottom PCB with Au pads. Scale bar = 40 mm. (E) Representative fluorescent micrographs of HeLa cells incubated (~20 s) with AF-647-tagged β-gal (472 kDa) proteins (left, control) and after electroporation-induced delivery of AF-647-tagged β-gal (right) with LEPD (~20 s pulse time). Scale bar = 70 μm.

First, we tested the effect of voltage amplitude and pulse-shape on the delivery of AF-647-tagged β-gal proteins into adhered HeLa cells. We applied pulse profiles of different shapes consisting of short single-level pulses (P1), bi-level pulses (P2), or long single-level pulses (P3) (Fig. 2A, Fig. S1). The applied voltage must be sufficiently high to yield a threshold TMP at the cell membrane to generate hydrophilic pores.²⁸ In our previous work where we delivered highly charged molecules like plasmids and nucleic acids, we found that an applied voltage amplitude of 20 V was enough to enable intracellular delivery while retaining viability.⁸ Using a voltage amplitude of 20 V, we observed that the short single-level pulse (P1) resulted in the lowest percentage of cells with detectable signal (21.78 %) and the lowest mean fluorescence intensity (MFI) of AF-647 (Fig. 2B–C). Conversely, the bi-level pulse (P2) and the long single-level pulse (P3) profiles resulted in the successful delivery of β-gal in 75.38% and 75.86% of the cells respectively, with significantly higher MFI of AF-647 compared to the short single-level pulses (P1). The rationale for applying a bi-level pulse was to supply a high enough voltage amplitude for a short duration (V1 = 20 V, t1 = 0.5 ms) to induce hydrophilic pore formation and follow it with a lower voltage (V2 = 10 V, t2 = 1.5 ms) to maintain the pores open for a longer duration to provide sufficient time for the weakly-charged proteins to diffuse into the cytosol. This was done without subjecting the cells to strong electric fields that may result in joule heating or formation of reactive oxygen species, which can be deleterious and lead to cell death. Evidently, P3 (V1 = 20 V, t1 = 2.0 ms) resulted in a significant reduction in cell viability (40.88%) as opposed to 75.03% viability in the case of P2 (Fig. 2D). Taken together, our results show that a bi-level pulse shape enables efficient delivery of β-gal protein while preserving the viability and functional integrity of the cells. The effect of the pulse shape was also tested for the delivery of β-gal ProSNA, and we observed optimal delivery with P2 (Fig. S2). To investigate protein delivery dosage into live cells, we adjusted the number of applied P2 pulses (Fig. 2E) and observed a linear increase with pulse number. Additionally, we utilized an automated imaging platform equipped with deep-learning models for cell segmentation to analyze the morphological features of the cells following electroporation.⁸ Briefly, we trained deep learning models consisting of a U-Net architecture to segment individual cells in the images and employed an image profiling software (CellProfiler) to measure a variety of shape and intensity features (refer to Table S1 for descriptions of the shape features). To compare the morphological outcomes across experiments, we standardized each feature using the robust Z score (R.Z. score) and plotted a heatmap to visualize the differences (Fig. 2F). The analysis revealed that a long pulse shape leads to significant shrinking and rounding of the cells, which is known to be detrimental to their normal function.⁴⁰ We note that the multiplexing capabilities of the multi-well LEPD platform made possible the expedited optimization of the pulse parameters necessary to deliver a large protein like β-gal (464 kDa) using the localized electroporation mechanism.

Figure 2. Delivery of a large protein and pulse shape optimization.

A) Composite fluorescent micrographs of HeLa cells stained with calcein-AM (green) viability marker and Hoechst (blue) nuclear marker following the delivery of β-gal-AF-647 (red) using short single-level pulses (P1: V = 20 V, t = 0.5 ms), bi-level pulses (P2: V₁ = 20 V, V₂ = 10 V, t₁ = 0.5 ms, and t₂ = 1.5 ms) and long single-level pulses (P3: V = 20 V, t = 2.0 ms). Scale-bars = 100 μm. B) Delivery efficiencies of β-gal corresponding to the various applied pulse shapes. C) Mean fluorescence intensity (MFI) fold change of AF-647 normalized with respect to P1. D) Viability percentages of the different pulse treatments. E) Variation in protein dosage with pulse number as shown by the different levels of mean fluorescence intensity (MFI) fold change of intracellular β-gal normalized with respect to a zero-pulse control. F) Heat-map of standardized scores (R.Z. score) of various cell-shape features used to compare morphological outcomes following β-gal delivery with the different pulse inputs. Error bars indicate standard error of the mean. N = 2 or 3 for all bar plots. * p-value < 0.05, ** p-value < 0.01, n.s.= not significant.

One robust method of enhancing cellular delivery of proteins is functionalizing protein cores with a dense shell of radially oriented nucleic acids to form protein spherical nucleic acids (ProSNAs).^{20, 38, 39} SNAs readily enter cells through scavenger receptor A mediated endocytosis, without need of any transfection agent.⁴¹ Additionally, ProSNAs combine the functionality of both proteins and nucleic acids in an independent modular format and have been shown to have the potential to be used in applications like functional enzyme delivery and live cell chemical analysis. Since proteins are inherently weakly charged compared to nucleic acids, we hypothesized that higher charge density of the ProSNA complex would amplify the electrophoretic drive and improve the rapid delivery of the protein cargo via electroporation.

To study and quantify this effect, we functionalized the AF-647-tagged β-gal protein with single-stranded DNA tagged with fluorescein (FITC) at two different surface densities (18 DNA strands/protein (SNA1) and 25 DNA strands/protein (SNA2)). Both sets of ProSNAs were delivered into adhered HeLa cells using the LEPD with the optimized bi-level pulse conditions (Fig. 3A). The fluorescent micrographs show that the ProSNAs were efficiently introduced into most cells (efficiency = 80.25%) in the field of view. In the control case, where the cells in LEPD were incubated with the ProSNAs without any pulsing for the duration of bilevel pulse conditions (~20 s), we did not observe any significant uptake of ProSNAs into the cells (Fig. 3A, Fig. S3). However, for a much longer incubation period (1 hour), we did observe internalization of β-gal-SNA into the cells (Fig. S3). We next compared the mean fluorescence intensity (MFI) of the AF-647-tagged β-gal bare protein to the two ProSNA constructs delivered via LEPD mediated electroporation. We observed an increase in the average MFI per cell for the ProSNAs compared to the bare β-gal (Fig. 3B). Moreover, the uptake of the ProSNA with the higher DNA surface density (SNA2) was significantly higher (p < 0.05) than the ProSNA with lower DNA density. These results indicate that the negatively charged DNA strands enhance the electrophoretic transport of large proteins into the cytosol.

Figure 3. Delivery of ProSNA and protein activity assay.

A) Representative fluorescent micrographs of HeLa cells following 20 s incubation with AF-647-tagged β-gal-SNA (control), electroporation of AF-647-tagged β-gal, or electroporation of β-gal-SNA complex. The DNA on the β-gal-SNA are tagged with FITC. B) Mean fluorescence intensity (MFI) fold change of AF-647 for different cargo under the same delivery conditions. β-gal (no DNA attached), SNA1 (β-gal-SNA with 18 DNA strands/ β-gal core), SNA2 (β-gal-SNA with 25 DNA strands/ β-gal core). C) Representative images HeLa cells after X-gal assay post electroporation of β-gal, β-gal-SNA, ad blank samples (control). D) Percentage of cells exhibiting β-galactosidase activity 2 hr. following delivery of β-gal, β-gal-SNA and blank samples (control). Error bars indicate standard error of the mean, N = 2 or 3 and * p-value < 0.05, ** p-value < 0.01, n.s.= not significant. Scale-bars = 100 μm.

To determine if the functionality of proteins delivered with the LEPD system is preserved, we used a β-gal activity measurement assay (X-gal) to test for activity of β-gal and β-gal-SNA in the HeLa cells 4 hours after delivery. Active β-gal hydrolyzes the X-gal chromogenic substrate, which results in the formation of a blue precipitate that is detectable with optical microscopy. The micrographs we obtained following incubation with X-gal show the presence of the blue precipitate in cells for both bare β-gal and β-gal-SNA, which indicates that the protein remains active in the cells after electroporation (Fig. 3C). Furthermore, we obtained images of cells incubated with bare β-gal without application of electric pulses as a negative control and did not observe any blue precipitate following the X-gal assay; therefore, the blue precipitate present in the electroporated cells was a result of the delivered exogenous protein (Fig. 3C). We also quantified both the percentage of cells with protein activity and the intensity of the blue precipitate for all the cases: control, bare β-gal and β-gal ProSNA (Fig. 3D, Fig. S4). It is known that perturbations imparted by the delivery process like changes of pH, protein aggregation, denaturing, or shielding of active sites with secondary molecules, may result in conformational changes in proteins and consequent loss of function.⁴² Therefore, for proteins to retain their function, the environmental factors imposed on the proteins in the delivery experiment must be carefully considered. Here, the proteins retained their function in the presence of electric fields and after functionalizing their exterior with nucleic acids.

Next, using confocal microscopy we studied the intracellular localization of β-gal-SNAs following direct delivery with LEPD. For comparison, we delivered linear DNA tagged with FITC in separate LEPDs. The cells were transferred to a confocal dish immediately following electroporation and incubated with the nuclear stain Hoechst 4 hours after transfer. Confocal micrographs (Fig. 4A) were acquired with different laser lines to excite the three fluorophores covalently coupled to specific locations: β-gal core tagged with AF-647 (red), DNA tagged with FITC (green), and nuclei stained with Hoechst (blue). The images (Fig. 4A–B) show a uniform distribution of the ProSNAs within the cytosol of the cells which contrasts the punctate fluorescence that is sometimes characteristic of endocytosis-mediated delivery (Fig. S3).³⁸ Images corresponding to the delivery of the linear DNA strands show that the DNA localizes to the nuclear region (Fig. 4A–B, lower panel). We measured the fluorescence intensity along line-segments of select cells for both cases (Fig. 4B) to quantify the observations. Indeed, the intensity of the ProSNAs (AF-647 and FITC, Fig. 4C (upper panel)) is highest outside of the nucleus whereas the free DNA (FITC, Fig. 4C (lower panel)) strongly colocalizes to the nucleus. Furthermore, in the case of ProSNAs, there is a strong colocalization between the β-gal protein core and the attached DNA, which indicates that the DNA shell remains attached to the protein core following electroporation. These results indicate that there is a synergistic enhancement of cytosolic delivery efficiency when combining the ProSNA constructs with the multiplexed LEPD platform. These capabilities can allow for the design and execution of in vitro studies where ProSNAs can be utilized for sensing cytosolic molecules using DNA (e.g., mRNA detection) or for therapeutic applications (e.g., gene knockdown using antisense oligonucleotides).^{6, 43}

Figure 4. Intracellular localization analysis of ProSNAs and linear DNA.

A) Representative fluorescent micrographs obtained by confocal imaging of HeLa cells after delivery of AF647 conjugated β-gal-SNA complex (upper panel) and FITC tagged linear DNA (lower panel) respectively. Scale bar = 50 μm. B) Zoomed in view of a representative cell delivered with ProSNA (top) and linear DNA (bottom) respectively. Scale bar = 25 μm. C) Line Scans of fluorescence intensities across representative β-gal-SNA delivered and linear DNA delivered cells. The DNA on the β-gal-SNA are tagged with FITC.

Following the successful delivery of large protein constructs like β-gal and ProSNA, we evaluated the capability of the LEPD to enable genetic modifications in cells via exogenous protein-RNA complex delivery. The CRISPR-Cas9 system has been extensively used in biomedical research and clinical applications due to its simplicity, ease of use, and precise gene editing capability.⁴⁴ The CRISPR-Cas9 gene editing system is typically used in three formats: plasmid DNA, mRNA and RNP (ribonucleoprotein) complexes. The RNP format uses a preformed complex of the Cas9 protein and the guide RNA which obviates the need for transcription or translation. Therefore, in the RNP format the onset of editing is much faster, avoids the possibility of permanent recombination into the host genome and potential damage to endogenous genes, and is known to have the lowest off targets effects among all the three formats of delivery.^45–47 Previously, there have been studies where localized electroporation systems were used to successfully deliver Cas9 RNPs into cells bringing about specific genetic modifications. ^{9, 30} However, the genetic modifications attempted were confined to non-homogenous end-joining (NHEJ) pathways as the data reported contained only the knockout efficiencies. Here we report the capability of the LEPD to successfully deliver a CRISPR-Cas9 based RNP complex along with a single-stranded oligodeoxynucleotide (ssODN) repair template into EGFP expressing K562 cells and bring about a specific homology directed repair (HDR) based genetic modification. Specifically, we used the bi-level pulse shape (P2) and co-delivered a Cas9 RNP complex and a ssODN to knock out the EGFP gene and subsequently introduce a blue fluorescent protein (BFP) encoding gene at the same locus. The knockout efficiency was evaluated for cases where we applied 400, 800 and 1600 P2 pulses to deliver the gene editing cargo. The highest knockout efficiency was in the case of 1600 pulses (Fig. S5). We also used lipofectamine 3000 to deliver the gene editing cargo, however the knockout efficiencies were lower than LEPD (Fig. S6). Representative phase contrast and fluorescent micrographs in Fig. 5A show the successful knock-out of EGFP and knock-in of the BFP expressing gene in many of the K562 cells. We used flow cytometry to quantify the gene editing efficiency (Fig. S7). The data is presented in Fig. 5B, where the stacked bar plots represent the percentage of cells that had both BFP knock-in and EGFP knock-out (blue bar), just EGFP knock-out (grey bar) and no gene editing (green bar). There was successful knock-in of the BFP gene in 23.6% of the cells, while about 34.7 % of cells just had EGFP knockout with no knock-in. Therefore, the total editing efficiency was 58.3%, which is substantially higher than previously reported editing efficiencies (25–30%) for localized electroporation systems ^{9, 30}.

Figure 5. CRISPR/Cas9 delivery and gene editing.

A) Representative phase contrast and fluorescent micrographs of K562 cells, 7 days after CRISPR Cas9-gRNA RNP complex and ssODN delivery. The gRNA targeted the endogenous EGFP and the ssODN has a BFP template to facilitate homology directed repair (HDR) (left: Phase contrast, middle: EGFP, right: BFP). Scale bar = 100 μm B) Stacked bar plots of flow cytometry data (Fig. S7) to show the average efficiency (N=3) of knockout (% of EGFP- cells) and knock-in (% of BFP+ cells).

In summary we have investigated the cellular delivery of large functional proteins using the multiplexed LEPD system and found that application of a bilevel pulse shape results in superior performance both in terms of delivery efficiency and viability compared to other pulse parameters. Moreover, we observed that the efficiency and dosage is improved by using ProSNAs compared to bare protein, even though the ProSNA conjugate is substantially larger. We hypothesize that the charged DNA strands of the ProSNA enhance electrophoretic transport, leading to increased molecular delivery into the cell interior. Moreover, we used a protein activity assay to show that both the bare protein and the ProSNA retain their activity after delivery, which indicates that the employed pulse parameters and experimental conditions do not compromise the protein structure. We also characterized intracellular localization of the delivered molecules and observed that DNA delivered using the Pro-SNA architecture is mostly homogeneously distributed in the cytoplasm as opposed to linear DNA, which is localized to the nucleus after delivery. Furthermore, using the optimized conditions for delivery of large proteins, we delivered a CRISPR Cas9-gRNA RNP complex along with a ssODN to knock out EGFP and subsequently knock in BFP in K562 cells with higher editing efficiencies than previously reported for localized electroporation systems. This work expands the capabilities of nanochannel membrane based electroporation devices previously demonstrated by Espinosa and others,⁹ by increasing the multiplexing capacity and throughput using the well-plate architecture. A high-throughput platform with multiplexing is essential for optimizing electroporation experiments due to the vast number of experimental variables (e.g., pulse parameters, various cell types, buffers, etc). Here, the multiplexed design of the LEPD along with the AI assisted image analysis was used to quickly optimize the pulse conditions for delivering large proteins and protein-based constructs.

Moreover, this work shows that in addition to previous demonstrations of the LEPD which were confined to the delivery of highly charged species like plasmids and small molecules like siRNA⁸, the LEPD can also be used for (1) direct delivery of large proteins into the cytosol while retaining protein functionality, without being dependent on secondary mechanisms like endosomal escape which is an issue with most chemical carriers (2) taking advantage of engineered nanostructures like ProSNAs to enhance molecular delivery, and can potentially be used in cell engineering workflows that entail delivery of large protein complexes like CRISPR-RNP to carry out CRISPR-mediated gene editing.