Poly(β-aminoester) physicochemical properties govern delivery of siRNA from electrostatically assembled coatings

Abstract

Localized short interfering RNA (siRNA) therapy has potential to drive high-specificity molecular-level treatment of a variety of disease states. Unfortunately, effective siRNA therapy suffers from several barriers to its intracellular delivery. Thus, drug delivery systems that package and control the release of therapeutic siRNAs are necessary to overcome these obstacles to clinical translation. Layer-by-layer (LbL) electrostatic assembly of thin film coatings containing siRNA and protonatable, hydrolysable poly(β-aminoester) (PBAE) polymers is one such drug delivery strategy. However, the impact of PBAE physicochemical properties on transfection efficacy of siRNA released from LbL thin film coatings has not been systematically characterized. In this study, we investigate the siRNA transfection efficacy of four structurally similar PBAEs in vitro. We demonstrate that small changes in structure yield large changes in physicochemical properties, such as hydrophobicity, pK_a, and amine chemical structure, driving differences in the interactions between PBAEs and siRNA in polyplexes and in LbL thin film coatings for wound dressings. In our polymer set, Poly3 forms the most stable interactions with siRNA (K_eff,w/w=0.298) to slow release kinetics and enhance transfection of reporter cells in both colloidal and thin film coating approaches. This is due to its unique physiochemical properties: high hydrophobicity (cLogP=7.86), effective pK_a closest to endosomal pH (pK_a=6.21), and high cooperativity in buffering (n_hill=7.2). These properties bestow Poly3 with enhanced endosomal buffering and escape properties. Taken together, this work elucidates the connections between small changes in polymer structure, emergent properties, and polyelectrolyte theory to better understand PBAE transfection efficacy.

Graphical Abstract

1. Introduction

Nucleic acid therapies have enormous potential in that they target and treat disease at the molecular level by directly modulating gene expression.¹ Short interfering RNA (siRNA) is one nucleic acid therapy that has received much interest due to its ability to post-transcriptionally degrade messenger RNA (mRNA) and silence the expression of disease-causing genes. Unfortunately, translation of siRNA to the clinic has been hampered by several delivery challenges, such as its lability, minimal organ and cell-specific targeting, low cellular association and uptake, and endosomal entrapment.² Oligonucleotide chemical modifications and delivery systems that target the liver have facilitated clinically approved siRNA therapies, starting with patisiran in 2018;³ however, extrahepatic and local delivery of siRNA have seen less translational progress, suggesting that further development of drug delivery systems is necessary for successful siRNA therapy. Localized delivery systems hold wide promise for treatment of dermal, ocular, and musculoskeletal diseases, among others.^4–11

Due to delivery challenges, drug carriers for localized siRNA therapy often leverage synthetic transfection polymers to package and traffic nucleic acid therapy.^12,13 These polymers contain protonatable (i.e. conditionally cationic) amines, whose positive charges condense the negatively charged phosphate backbone of the nucleic acid, protecting it from degradation and promoting favorable cellular membrane interactions. Protonatable amines also help nucleic acids to escape endosomal entrapment, potentially through the proton sponge effect, whereby they act as buffers in the acidifying endosome to induce endosomal swelling and eventually siRNA release to the cytoplasm, and through physical disruption of the negatively charged endosomal membrane.¹⁴ Certain protonatable polymers, such as poly-_L-lysine (PLK) or polyethyleneimine (PEI), are commonly used in transfection applications in vitro but are limited by toxicity in vivo.^15–18 Ionizable and degradable poly(β-aminoesters) (PBAEs) are effective for in vitro and in vivo applications due to their low toxicity and high transfection efficiency.^19,20 PBAEs are thus an attractive polymer carrier for local delivery of siRNA therapy.

PBAEs are formed through step polymerization via Michael addition of a diacrylate and an amine.¹⁹ The modularity of PBAE synthesis enables combinations of a wide range of monomers to create transfection-enhancing polymers with divergent properties. The Langer and Anderson groups have screened large libraries of PBAEs to find structures advantageous for DNA delivery^21–26 while the Green group has characterized how changes in polymer structure, such as molecular weight, pK_a, and binding kinetics, mediate plasmid DNA and siRNA delivery.^27–31 Despite some initial studies, less work has been done on siRNA delivery. Additionally, studies have mostly focused on a single physicochemical property and its effect on PBAE mediated transfection efficiency without examination of the ways that many physicochemical properties interact to mediate efficacy. Finally, prior studies largely focus on the solution complexation of polymers and nucleic acids, while translational coatings on scaffolds for localized delivery of nucleic acids have different stoichiometries and structure, thus creating additional design considerations that must be accounted for.

Our lab leverages the layer-by-layer (LbL) assembly approach to electrostatically adsorb PBAE, siRNA, and other polymeric excipients onto the surface of various scaffolds for localized release. As weak polyelectrolytes, PBAEs provide emergent properties when electrostatically assembled into thin film coatings. siRNAs released from PBAE films can interact with and enter cells, escape endosomes, and silence target genes. We have demonstrated that this method of delivery improves diabetic ulcer wound healing in mice,^9,32 reduces excess scarring in a rat burn model,¹⁰ speeds healing of neuroischemic diabetic wounds in rabbits,³³ and transcutaneously vaccinates against HIV.³⁴ However, the impact of PBAE physicochemical properties in LbL siRNA-eluting thin film coatings on siRNA transfection efficacy remains unknown.

In this study, we leverage a set of four polymers with differing hydrophobicity, pK_a, complexation efficiency, binding strength, and disassembly kinetics to elucidate how these properties might collectively mediate transfection efficacy. Ultimately, we show how these properties affect transfection in both colloidal complexes (polyplexes) and from thin film coatings on representative wound dressings. We find that two candidates are poor transfection polymers due to their weak binding equilibrium to siRNA (Poly0) and pK_a above physiological pH (Poly2O). While two polymers, Poly2 and Poly3, transfect siRNA effectively, Poly3 promotes higher silencing than Poly2 in both polyplexes and coatings. Mechanistic elucidation suggests that the lower pK_a of Poly3 may enhance endosomal buffering in early endosomes and that its higher ability to disrupt membranes may enhance endosomal escape, thus promoting higher transfection of siRNA. Poly3’s increased hydrophobicity also slows release, enabling more sustained exposure to siRNA. Ultimately, this work illustrates key criteria in understanding PBAE structure-function relationships for local intracellular siRNA delivery.

2. Methods

2.1. Poly(β-aminoester) (PBAE) synthesis

1,6-hexanediol diacrylate (99% stab. With 90 ppm hydroquinone) and 1,4-butanedioldiacrylate (85+% stab. With 50–105 ppm hydroquinone) were obtained from Alfa Aesar. Triethylene glycol diacrylate was bought from Polysciences, Inc. 1,9-bis(acryloyloxy)nonane (stabilized with MEHQ) and piperazine anhydrous were purchased from TCI. 1,3-bis(4-piperidinyl)propane (97+%, crystalline), anhydrous tetrahydrofuran (THF, 99.8+%, uninhibited), hexanes, and diethyl ether (pesticide grade) were purchased from Thermo Fisher Scientific. Piperazine (anhydrous) was from TCI. New lots of monomers were purchased for synthesis and used without further purification. Poly0 is made through the polymerization of 1,4-butanedioldiacrylate and piperazine. Poly2 is made through the polymerization of 1,6-hexanediol diacrylate and 1,3-bis(4-piperidinyl)propane. Poly2O is synthesized through the polymerization of triethylene glycol diacrylate and 1,3-bis(4-piperidinyl)propane. Poly3 is synthesized by reacting 1,9-bis(acryloyloxy)nonane and 1,3-bis(4-piperidinyl)propane.

PBAE synthesis was performed similarly to previously published work.^19,32,35 Polymers were designed to be amine-terminated, using a 1.02 molar ratio of amine monomer to acrylate monomer. A larger version of Poly3 was made by normalizing the ratio of monomers added with the GC-confirmed purity. 17 mL of anhydrous THF was drawn up into a clean syringe using a Schlenk line to minimize ambient air in the anhydrous solvent. In clean, oven-dried 50 mL round bottom flasks, 10.3 mmol of amine monomer was dissolved by stirring in some of the THF. Once the amine was completely dissolved, 10.1 mmol of the acrylate was added to the flask. The flask was heated in an oil bath at 50°C with stir at 500 rpm. Due to the water-sensitivity of the reaction, the flask was capped with a rubber stopper, and dried nitrogen was bubbled into the mix for at least 10 minutes to purge the flask of ambient air. The reaction apparatus was covered with aluminum foil and left for 48 hours to react.

After 48 hours, the reaction vessel was removed from heat, and polymers were re-precipitated 3x in at least 10-fold volume of ice-cold solvent, redissolving the collected product in 5–10 mL THF between precipitations. Poly2 was re-precipitated in a 50:50 mix of hexanes and diethyl ether. Poly0, Poly2O, and Poly3 were re-precipitated in hexanes. The collected product was a white to off-white powder for Poly0, Poly2, and Poly3. Poly2O was a viscous pale-yellow liquid. The yield for Poly0, Poly2, Poly2O, and Poly3 was 38%, 54.1%, 3.4%, and 31%, respectively. Polymers were stored at room temperature in a vacuum desiccator in nitrogen-flushed vials.

2.2. PBAE characterization

ClogP and molecular weight of the single polymer repeat unit were calculated with ChemDraw 22.0. ¹H-NMR spectra were collected on a Bruker Avance Neo spectrometer at 500.34 MHz and analyzed using MestReNova. Chloroform-d (Cambridge Isotope Labs) was used as the solvent for NMR. MALDI-TOF was measured on a Bruker model MicroFlex MALDI-TOF, with peaks selected using a centroid model of peak selection with signal to noise ratio (SNR) cutoff of 1 in Bruker FlexAnalysis software. M_n, M_w, and pdi were calculated based on these peak masses. Briefly, PBAEs were dissolved at 2 mg/mL in 25 mM pH 5.2 sodium acetate with light orbital shaking. The sample was mixed sinapic acid (the MALDI matrix), spotted, and analyzed. Linear positive acquisition mode and a 1000 to 10000 Da range was employed.

To assess the degradation rate of Poly2 and Poly3, they were dissolved at 1 mg/mL in 100 mM pH 5.2 sodium acetate. They were allowed to degrade in this buffer for 2–3 days, snap-freezing time points with liquid nitrogen. Collections were lyophilized and stored at −80°C until analysis. They were redispersed with ultrapure water, and MALDI-TOF was assessed as before. M_n over time was compared to that at time 0 hours.

2.3. PBAE titration

pK_a of the tertiary amine in the four polymers was measured through acid-base titrations, with adaptation from the literature.^35–37 Briefly, polymer was dissolved at 0.6 mg/mL in 50 mM NaCl and 4.20 mM HCl. Approximately 10 mL of solution was made. The polymer was allowed to dissolve with sonication to break up chunks followed by nutation. The solution was then added to a beaker, set to stir, and a pH probe placed into the solution. 5–30 μL aliquots of 0.1N NaOH were added to the solution and the pH recorded upon stabilization after each aliquot. This was repeated until the titration pH reached 11. Since the starting amount of solution was not completely consistent, the volume of NaOH added was normalized by the initial solution volume (μL 0.1 N NaOH per mL PBAE solution). pK_a was defined as the pH at the half-equivalence point, with the first derivative of the pH versus volume graph used to help find equivalence points (local maxima). Degree of deprotonation (θ) was converted by setting a value of 1 to the equivalence point (by definition) and a value of 0 to the equivalence point minus 2 times the difference between the equivalence point and ½ equivalence point. Other transforms to the form of the Henderson-Hasselbach equation were performed, and the n_hill of deprotonation estimated from the linear range of the log(θ/(1−θ)) vs pH-pK_a graph that occurs near the point where pH equals pK_a.

2.4. PBAE-siRNA binding equilibrium (fluorophore exclusion assay)

The fluorophore exclusion assay was performed with Ribogreen (Invitrogen). siRNA against GFP duplex I from Horizon Discovery (siGFP) was used and was diluted to 1 μg/mL in 50 mM pH 5.2 sodium acetate buffer (hereby referred to as “buffer”), where the concentration of siRNA was verified using the RNA-40 setting on a Nanodrop spectrophotometer (Thermo Fisher Scientific). Ribogreen was added to this mix such that it is in 200-fold dilution. The solution was mixed well and added in 50 μL to a black half-area 96 well plate and allowed to nutate, shielded from light. Meanwhile, PBAE was diluted from 10 μg/mL to 0.041 μg/mL using a 2.5-fold dilution series in the same buffer as siRNA. PBAE was added into the wells at 50 μL and mixed well. After 30 minutes of incubation, fluorescence was read on a plate reader (λ_ex: 485 nm, λ_em: 525 nm). Fluorescence was normalized to wells receiving just buffer without any polymer. Weight to weight (w/w) ratios were converted to N/P ratios using the molecular weight of the polymer repeat unit divided by the number of amines it contains and the molecular weight of the siRNA divided by the number of bases. Thus, to convert from w/w to N/P, the following multiplication factors are used: Poly0 – 2.48, Poly2 – 1.62, Poly2O – 1.52, and Poly3 – 1.47. Curve fitting was performed with a variable slope sigmoidal relationship in Graphpad Prism, which enabled parameterization of K_eff,w/w or K_eff,N/P, as appropriate.

2.5. Polyplex formation

Polyplexes are electrostatically assembled nanoparticles consisting of siRNA and PBAE. They were assembled in 50 mM pH 5.2 sodium acetate buffer. The amount of PBAE relative to siRNA was varied by experiment and within experiments. Briefly, siRNA was diluted in 50 mM buffer with concentration dependent on the experiment. PBAE, diluted to 0.5 mg/mL in 50 mM buffer, was added to the solution and mixed thoroughly. Polyplexes were allowed to form for at least 10 minutes before use.

2.6. Dynamic light scattering (DLS) and zeta potential

To assess how siRNA self-assembled into complexes with siRNA, we performed dynamic light scattering (DLS) and zeta potential measurements. Polyplexes were formed in 50 μL total volume, as above, using a concentration of 25.2 μg/mL siGFP. PBAE was added in at a 0.5, 0.75, 1, and 2 w/w ratio (PBAE:siRNA) and mixed well. Polyplexes were loaded into small volume cuvettes and analyzed with a Nano ZS-90 Zetasizer (Malvern Panalytical) using a refractive index of 1.590 based on the assumptions of polystyrene latex particles and water as the dispersant. Three measurements were made per sample, and the z-average hydrodynamic radius averaged between the measurements. Intensity distributions were also collected. Samples were then diluted in ultrapure water 14-fold to reduce the effects of salt on readings and added to a zeta cell. Three measurements were averaged.

2.7. Polyplex disassembly kinetics

To better assess how polymer and siRNA might dissociate under physiologic conditions, polyplexes were allowed to dissociate in different buffers at 37°C, and timepoints were snap frozen. The binding of siRNA to the PBAEs was then assessed using a gel mobility shift assay, which due to changes in the surface charge of polyplexes causes the siRNA to migrate most when unbound to polymer, to migrate somewhat when bound a small amount of polymer, and to resist migration when fully bound to polymer. Polyplex disassembly was modeled after previous work.³⁸ 60 μL polyplex solution was assembled in 50 mM pH 5.2 sodium acetate buffer with siGFP diluted to 37.3 μg/mL and PBAE diluted to 373 μg/mL (for a 10:1 w/w ratio of PBAE to siRNA). A 10:1 w/w ratio was chosen such that binding should be well-saturated by that point (based on fluorophore exclusion titration assay results). Polyplexes were dispersed at a 5-fold dilution in pH 7.4 phosphate buffered saline (PBS), pH 5.6 PBS, or ionic-strength matched 170 mM pH 5.2 sodium acetate buffer, and incubated at 37°C. After varying amounts of time, polyplexes were snap-frozen in liquid nitrogen and stored at −80°C until measurement. siGFP was prepared at the same concentration and dispersed in PBS or pH 5.2 sodium acetate buffer as a reference. To run the gel mobility shift assay on the unpacked polyplexes, polyplexes were rapidly thawed in warm water and 20 μL was added into the well of pre-cast E-gel EX 4% agarose gels containing SYBR gold (Invitrogen). After running the gels for 12 minutes in an E-Gel electrophoresis system (Invitrogen), they were imaged in the SYBR Gold channel on a ChemiDoc Gel Imaging system (BioRad). In addition to measurements using a gel mobility shift assay, measurements to ascertain polyplex disassembly were also performed using DLS. To do this, polyplexes containing Poly2 or Poly3 that were diluted 5-fold in pH 7.4 PBS and incubated at 37°C over time were assessed with DLS, as described above.

2.8. Cell culture

GFP-expressing HeLa cells (HeLa d2eGFP) were used for all cell experiments, except where noted otherwise. They stably express GFP with a shortened half-life and were a gift from Piyush Jain’s lab at the University of Florida. The plasmid used to express this reporter was CMV-d2eGFP-empty from Phil Sharp’s Lab (Addgene plasmid 26164). To create growth media, Dulbecco’s Modified Eagle Medium (DMEM, Corning) was supplemented with 10% fetal bovine serum (FBS, Gibco) and 1% penicillin-streptomycin (Gibco or Corning). Passaging was performed using 0.25% trypsin-EDTA (Gibco). Routine cell growth was carried out in a humidified incubator at 37°C with 5% CO₂. Periodically, cells were tested for mycoplasma with a MycoAlert kit (Lonza), and all results herein are from mycoplasma negative cells. Cell culture employed best practices.

2.9. Cytotoxicity studies

HeLa cells were incubated with varying concentrations of PBAE diluted in growth media from 250 to 1.95 μg/mL using a 2-fold dilution series. Negative control wells received buffer only. The cells were incubated with polymer for about 20 hours. Toxicity was assessed using PrestoBlue^™ HS Cell Viability Reagent (Invitrogen) according to manufacturer instructions.

2.10. Transfection of PBAE-siRNA polyplexes and endosomal pathway inhibition

HeLa d2eGFP cells were grown to ~70% confluency while avoiding over-confluency between passages. The day before transfection, cells were lifted with Trypsin and washed before plating at 5,000 per well in 96 well plates with 100 μL growth media. Polymer was dissolved at 1 mg/mL in 100 mM pH 5.2 sodium acetate buffer and diluted to 0.5 mg/mL with water, making the buffer 50 mM pH 5.2 sodium acetate buffer. Stocks of siRNA at 50 μM, either siGFP or an AlexaFluor647 (AF647) fluorescently labeled dsDNA siRNA mimic (siD7, IDT, GTCAGAAATAGAAACTGGTCATC and complement, from the literature³⁹), were used. Polyplexes were assembled as described above, with enough for 3 wells plus 13.3% excess in 85 μL total volume. Briefly, siRNA was dissolved at 3.02 μg/mL in buffer, and polymer stock solution was added in various w/w ratios. 0.68 μL RNAiMax lipofectamine transfection reagent was added to siRNA as a positive control (“+L”). Growth media was replaced with 75 μL low-serum OPTI-MEM (Gibco), and 25 μL of polyplexes were added in triplicate to the 96-well plate. Negative control wells received 25 μL of buffer. Plates were returned to the incubator for 5 hours, at which point the media was replaced with growth media.

Cells were analyzed using flow cytometry after 72 hours. The cells were prepared by washing and fixing with 2% formaldehyde in PBS for 20 minutes at room temperature. A Fortessa flow cytometer (BD) with high throughput module was used to analyze the cellular fluorescence intensity. Channels were analyzed as such: GFP on a 488 nm laser with a 530/30 filter and AF647 on a 640 nm laser with a 670/30 filter. Flowjo (BD) was used to analyze the fluorescence according to the gating strategy in Figure S11.

Transfection experiments were also carried out using 1 or 2 hours as endpoints to assess early cell association. Transfection with inhibition of ATP-dependent cellular processes was carried out at 4°C for 4 hours to assess cellular association with uptake inhibited. For these experiments, given the shorter timeframe, cells were plated at 15,000 per well, rather than 5,000.

Transfection experiments were also carried out under chemical inhibitors of various endocytic pathways, adapting procedures and inhibitor concentrations from the literature.^40–42 HeLa d2eGFP cells were plated at 5,000/well in 96-well plates and allowed to seed overnight. Dynasore, wortmannin, and genistein were obtained from Selleckchem and dissolved in sterile DMSO at 200x concentration. Dissolved stocks were stored at −80°C until use. The next day, growth media without antibiotics was prepared, diluting inhibitors or DMSO alone 1:200 to get a 1x solution of inhibitor in growth media. 1x solution represents 25.8 μg/mL dynasore, 4.3 μg/mL wortmannin, and 54 μg/mL genistein. Cells were pre-treated with 100 μL of inhibitor-containing media for 1 hour before transfection. Meanwhile, polyplexes were assembled with Poly3 at a 10:1 w/w ratio, as above. RNAiMax commercial transfection reagent was used as a positive control, and naked siRNA (no transfection reagent) served as a negative control. 10 μL of polyplexes were added to the media. After 24 hours, cells transfected with siD7 were fixed and analyzed, and after 48 hours cells transfected with siGFP were fixed and analyzed, as explained above.

2.11. Dressing assembly

Dextran sulfate (DS, ultra-pure grade sodium salt, 500 kDa, VWR) was dissolved at 1 mg/mL in 100 mM pH 5.2 sodium acetate buffer. Linear polyethyleneimine (LPEI, transfection grade, 25 kDa, Polysciences Inc.) was dissolved at 0.25 mg/mL in 100 mM pH 6.0 sodium acetate buffer. Wash buffers were prepared at 10 mM, either pH 5.2 or 6.0, using sodium acetate buffer. Assembly of dressings containing siRNA by the LbL process is well described previously.³² The formulation architecture was [PBAE/DS]₂₀[LPEI/RNA]₅₀, where co-adsorbed layers are in brackets and subscript indicates the number of repeats of those layers. Briefly, Tegaderm^™ non-adherent contact layer wound dressings (3M) were cut to size, prepared by plasma cleaning on high for 10 minutes, and soaked in freshly dissolved PBAE (1 mg/mL in pH 5.2 100 mM sodium acetate buffer) for 1 hour to form an initial coating. Dipping of base layers proceeded as following: 1) PBAE for 10 minutes, 2) two washes in 10 mM pH 5.2 buffer for 30 seconds each, 3) DS for 10 minutes, and 4) two washes in 10 mM pH 5.2 buffer for 30 seconds each. This was repeated 20 times. After drying overnight, RNA layers were assembled as follows: 1) 10 minutes in LPEI, 2) 2 washes in 10 mM pH 6.0 buffer for 30 seconds each, 3) 15 minutes in RNA (68 μL dissolved in 1.7 mL pH 6.0 10 mM sodium acetate buffer), and 4) two washes in 10 mM pH 6.0 buffer for 30 seconds each. This process was repeated 50 times. An HMS-DS50 slide stainer (Carl Zeiss) automated the dip assembly process. After drying in ambient conditions for at least one hour, completed dressing formulations were stored at −80°C prior to use.

2.12. Transfection of siRNA-coated dressings

Transfection of siRNA-coated dressings containing different PBAEs was performed similarly to a protocol as previously described.³² In summary, HeLa d2eGFP cells were plated at 7,000 per well in 48-well plates with 200 μL DMEM growth media. Dressings containing siD7 or siGFP were cut to 0.3–0.5 cm² pieces and placed directly into the well with cells. Untreated cells served as additional controls. “+L” wells received 0.3 μL RNAiMax in addition to the dressing to serve as a positive control. After 72 hours of incubation, dressings were removed, cells were trypsinized and washed, and cells were fixed for flow cytometry, as in section 2.10. Cells were analyzed using flow cytometry, also as described in section 2.10. Flow cytometry was analyzed to determine the median fluorescence intensity (MFI) of siRNA-AF647 and GFP channels.

2.13. Release kinetics and total loading

Release kinetics were performed using siD7-coated dressings in a solution of 1% bovine serum albumin (BSA, lyophilized powder, Cohn Analog^™, BioReagent ≥98%, Sigma-Aldrich) dissolved in 1x PBS. Measurement followed established protocols.³² In short, pieces of Tegaderm were placed in 1% BSA in PBS and allowed to elute over time at 37°C. At various timepoints throughout release, the dressing was moved to a tube with fresh solution, and the solution containing releasate was stored for later analysis at −20°C. Releasate was compared to a standard curve and normalized to release at the final timepoint (about 7 days).

Total loading was also performed according to established protocols.³² Loading was measured by placing pieces of the dressing overnight in 3M sodium chloride solution to elute all material from the thin film coatings. Fluorescence measurements against a standard curve that was also dissolved in 3M sodium chloride enabled quantification. Loadings were normalized to the size of the dressing piece.

2.14. Hemolysis assay

An RBC (red blood cell) ex vivo hemolysis assay was performed to assess the ability of PBAEs to disrupt membranes at varying pH, as described in the literature.^43,44 RBCs were collected from anonymous donors according to an IRB-approved protocol by Research Blood Components, LLC. Briefly, RBCs were washed twice with 150 mM sodium chloride to remove plasma. The washed RBCs were divided up into 4 different pHs of PBS (7.4, 6.8, 6.2, and 5.6), which were chosen to represent progression of an acidifying endosome. Blood was then diluted 50-fold in these 4 PBS buffers. Polymers were diluted at 302.11 μg/mL in pH 5.2 50 mM sodium acetate buffer and filtered prior to use. Polyplexes were also assembled at a 5:1 w/w (PBAE:siRNA) ratio in the same buffer as the polymer. These stock solutions, when diluted 20-fold with blood, represent 4 times the normal concentration used during flow cytometry polyplex transfection experiments. A 2-fold dilution was made down to 0.25 times the concentration normally used. Controls were 50 mM pH 5.2 sodium acetate buffer (negative) and 20% Triton X-100 diluted in water (positive). In a v-bottom 96-well plate, 10 μL of 20x polymer, polyplex solution, or controls was added to 190 μL of 50-fold diluted RBCs at each of the pH values. After incubation for 1 hour at 37°C, cells were spun for 5 minutes at 500xg at room temperature, and 100 μL of supernatant was transferred to a new clear 96-well plate. Absorbance was read at 541 nm and normalized to the Triton positive control (100% hemolysis) and the no polymer control (buffer alone, 0% hemolysis).

2.15. Endosomal escape with Gal8-tagged cells

Endosomal escape was reported using a HeLa reporter cell line containing a GFP-tagged galectin 8 (Gal8).⁴⁵ These cells were a gift from Suzie Pun’s Lab at the University of Washington.⁴⁶ The day before transfection, cells were lifted with Trypsin and washed before plating at 20,000 per well in Lab-Tek^®II 8-well chamber slides (Thermo Fisher Scientific) with 300 μL growth media. Fresh media was added to the cells, and ~0.3 cm² pieces of dressing containing siD7 with Poly2 or Poly3 were added to the wells. Negative controls received just buffer that was changed to normal growth media after 4 hours. Cells were incubated for 22 hours, at which point they were washed with PBS, fixed with 4% formaldehyde in PBS, washed again 2x with PBS, and mounted with Prolong Gold Antifade mounting media with DAPI. Imaging was performed on an FV1200 Laser Scanning Confocal Microscope (Olympus) in the microscopy core at the MIT Koch Institute. Representative images were collected by a researcher blinded to the experimental condition. ImageJ was used for image visualization. Brightness and contrast were adjusted to accentuate the nucleus, Gal8 puncta, and siRNA, while reducing background. No other adjustments were made. Puncta were manually counted and divided by the number of nuclei for quantification.

2.16. Statistical analysis

Data visualization and statistical testing was performed using Graphpad Prism (Dotmatics). The specific statistical procedure used is described for each individual experiment. In general, a difference was considered statistically significant with a family-wide error rate of α=0.05. Parametric statistics were used due to the continuous nature of all data and an assumption of underlying distribution normality.

3. Results and Discussion

3.1. Characterization of a set of structurally similar PBAEs reveals varying physicochemical properties

Poly(β-aminoesters) (PBAEs) are notable for their modularity and enhancement of nucleic acid therapeutic transfection. They complex with nucleic acids, and their protonatable amines enhance endosomal escape.²⁰ Like many polymeric systems, their structure drives their function via modulation of physicochemical properties. Although our lab has primarily used one PBAE, Poly2, in our previous works for nucleic acid delivery,^{9,10,32,33,35} we endeavored to assess how polymer properties influence siRNA complexation for more rational design. The structures of these polymers were designed to sample several important parameters, particularly hydrophobicity, pK_a, and chain rigidities between the protonatable amines. We polymerized Poly0 with piperazine and 1,4-butanedioldiacrylate. Poly2, Poly2O, and Poly3 were synthesized by reacting 1,3-bis(4-piperidinyl)propane with 1,6-hexanediol diacrylate, triethylene glycol diacrylate, and 1,9-bis(acryloyloxy)nonane, respectively (Figure 1A). Polymer structures were confirmed with ¹H-NMR, (Figure S1x–4) and polymer sizes, as measured by MALDI-TOF, were 2–3.8 kDa M_n, 2.6–4.4 kDa M_w, and similar between the four polymers (Figure 1B).

Figure 1. Four PBAEs exhibit diverse physiochemical properties.

A) Four polymers of the PBAE class containing a similar structure, but different physicochemical properties, can be synthesized using Michael addition of a diamine and diacrylate. B) MALDI-TOF mass spectra of the PBAEs with sinapic acid as the matrix enables estimation of polymer average molecular weight (M_n, M_w) and polydispersity (pdi). Poly0: M_n 2.0 kDa, M_w 2.7 kDa, pdi 1.37. Poly2: M_n 3.8 kDa, M_w 4.4 kDa, pdi 1.17. Poly2O M_n 2.3 kDa, M_w 3.0 kDa, pdi 1.29. Poly3: M_n 2.3 kDa, M_w 2.6 kDa, pdi 1.13. C) Molecular weight (MW) of the repeat unit for the four polymers. D) Computed octanol water partition coefficient (ClogP) estimates hydrophobicity of the repeat unit for each polymer.

While Poly0 has a smaller repeat unit molecular weight, Poly2, Poly2O, and Poly3 have similar molecular weights of the repeat unit (Figure 1C). This is due to the much smaller piperazine monomer versus the 1,3-bis(4-piperidinyl)propane monomer. It is worth noting that repeat units for all four polymers contained the same number of tertiary amines, suggesting that the repeat unit molecular weight per amine follows the size of the repeat unit itself.

Polymer repeat unit hydrophobicity, as determined with ClogP predictions, ranged from 0.99 to 7.86, with Poly0 being the least hydrophobic and Poly3 being the most (Figure 1D). The glycol moieties between the ester bonds in Poly2O reduce hydrophobicity and enable hydrogen bond acceptance compared to the alkyl chain nature of Poly2, although the number of hydrocarbon units between the esters remains the same. It is known that some hydrophobicity is important for reducing toxicity of cationic protonated amines and enhancing their transfection efficacy, but too much hydrophobicity can also cause toxicity.^{28,29,47–49} While literature has evaluated the role of hydrophobicity in polyplex transfection of siRNAs with PBAE, the role of hydrophobicity in electrostatically assembled thin films adds an additional important component to examine due to its potential to destabilize the film.⁵⁰

3.2. Polymer structure governs amine pK_a

Numerous previous studies have concluded that the acid dissociation constant (pK_a) is a key modulator of transfection efficacy in polymeric and lipidic siRNA delivery systems by varying the protonation of the amine.^31,51–55 By this same principle, pK_a is also important in the assembly of electrostatically assembled thin films, as it dictates the relative frequency of loops, tails, and trains in the adsorbed polymer and defines the amount of polymer needed to complex all of the negative charge units of siRNA.⁵⁶

Using the first derivative of the titration curve (Figure S5), we found equivalence points and calculated pK_a at the half equivalence point for our polymers. The four PBAEs varied in their pK_a despite all containing tertiary amines: 6.91, 7.08, 8.29, and 6.21 for Poly0, Poly2, Poly2O, and Poly3, respectively (Figure 2A). Poly0, Poly2, and Poly3 would exist as mostly deprotonated species at physiological pH (7.4), with the latter having higher deprotonation than the prior two. In contrast, Poly2O would remain mostly protonated at this pH (Figure 2B). This suggests that the buffering capacity of Poly2O is outside the range of an acidifying endosome. Endosomes have a pH of around 6.8–7.4 immediately after formation and acidify to a range of approximately pH 5–6 as the late endosome fuses with the lysosome. Endosomal escape is thought to occur at both early and late endosomal stages.⁵⁷ Thus, Poly0, Poly2, and Poly3 all have pK_a values consistent with potential buffering at endosome-relevant pH ranges, with Poly3 buffering at the most acidic pH range of the tested polymers. Additionally, at lower pH ranges, closer to those of the late endosome, the cationic nature of these polymers might also enable them to physically disrupt cellular membranes. Differences in pK_a amongst tertiary amines highlight the role that amine context plays in regulating acid dissociation.

Figure 2. PBAEs exhibit diversity in pK_a and cooperativity of deprotonation.

A) PBAE (0.6 mg/mL) was titrated with 0.1 N NaOH. pK_a was based on calculating the half equivalence point, using the first derivative of the titration curve to find the equivalence point (Figure S5). Normalized volume was derived from dividing by the initial volume of PBAE solution. B) Superimposed titration curves for the four polymers with normalization of titration coordinate to a measure of the degree of deprotonation (θ). Approximate pH range of the endocytic process denoted with shading. Unshaded regime: above physiologic pH, dark green shaded regime: extracellular space to early endosomal pH, light green shaded regime: late endosomal to lysosomal pH. C) Conversion of titration coordinate diagram to a log-transformed measurement of deprotonation against pH normalized to pK_a. Equation based on a cooperativity modification of the Henderson-Hasselbach equation. D) The linear region of log(θ/1−θ) versus pH-pK_a was used to determine n_hill, which represents the cooperativity and anti-cooperativity that occurs in the titration of polyamines. 95% confidence interval of the n_hill parameter is shown.

In a polymeric system, protonatable units are physically linked, potentially causing cooperativity or anti-cooperativity in their protonation and modifying the apparent pK_a of the polymer compared to a monobasic species.⁵⁸ Titration curves show that pH is maintained close to pK_a in the buffering region for Poly2 and Poly3; meanwhile, pH varies about the pK_a more for Poly0 and Poly2O (Figure 2B). This suggests a more robust buffering capability for Poly2 and Poly3 over Poly0 and Poly2O. The sharpness of this transition about the pK_a can be quantified by calculating the Hill coefficient (n_hill) for the modified Henderson-Hasselbach equation, using transforms of the titrations (Figure 2C), as explained in the literature.^36,37 On a physical level, a higher Hill coefficient suggests rapid deprotonation of the polymer about the pK_a and that deprotonation of an individual amine is coupled to that of its neighbors. We observe the highest cooperativity in deprotonation for Poly2 (n_hill=5.9) and Poly3 (n_hill=7.2), with no cooperativity or slight anti-cooperativity noted for Poly0 (n_hill=0.79) and minimal cooperativity for Poly2O (n_hill=1.6) (Figure 2D). In another polymeric system studying transfection of ribonucleoproteins, higher n_hill was the best predictor of efficacy, and the researchers posited that rapid deprotonation allows for the unbinding of drug from the polymeric carrier to enable engagement with intracellular targets.⁵¹

In our system, comparing Poly0, Poly2, and Poly3, as we increase the alkyl chain length between amine moieties, increasing hydrophobicity, the cooperativity increases. This is consistent with previous work on block copolymers of poly(ethylene oxide) and poly(2-(dipropylamino) ethyl methacrylate), which demonstrated how polymers with increasing inter-amine chain length have increased cooperativity in deprotonation.³⁷ Alkyl chains afford chain flexibility, permitting minimization of electrostatic repulsion between protonated amine groups. In the case of deprotonation, this means that loss of electrostatic repulsion allows the polymer to rapidly rearrange and minimize its folded energy state. This contrasts with polyethyleneimine, another common transfection-enhancing polymer, which has been reported to demonstrate anti-cooperativity (n_hill=0.3) due to the close physical proximity of amine moieties in a way that prevents full protonation at accessible pH levels due to high levels of electrostatic repulsion.^37,58,59

3.3. Physicochemical properties govern interaction with siRNA

While characterization of the polymers themselves provides important understanding of their behavior as transfection agents, their interactions with siRNA are also a critical factor. Using a fluorophore exclusion assay, in which an RNA-binding fluorescent dye is displaced by the binding of polymer to the siRNA, we estimated effective binding equilibria between each PBAE and siRNA (Figure 3A). Poly3 exhibits the strongest relative binding to siRNA (K_eff,w/w = 0.298) followed by Poly2 (K_eff,w/w = 0.698). Poly2O (K_eff,w/w = 1.20) and Poly0 (K_eff,w/w = 2.46) have much weaker effective binding coefficients. The differences are even more apparent when normalizing the w/w ratio as a nitrogen (from the polyamine) to phosphate (from siRNA), or N/P, ratio (Figure 3A). This normalization accounts for the differences in repeat unit molecular weight. To probe the effects of molecular weight on binding affinity, we used Poly3 as an example. A slightly larger version of Poly3 (M_n 3.2 vs 2.3 kDa) shows no differences in binding affinity to siRNA (F test comparing log(K_eff,w/w), p=0.95), although the effects of further increasing polymer size cannot be ruled out. This also suggests that small differences in the molecular weight of Poly2 and Poly3 are likely unable to explain differences in their binding affinity to siRNA (Figure S6).

Figure 3. Poly2 and Poly3 exhibit greater association with siRNA in self-assembled polyplexes than Poly0 or Poly2O.

A) In the fluorophore exclusion titration assay, binding of polymer to siRNA displaces a pre-bound fluorophore, decreasing the fluorescent signal. With this assay, the effective equilibrium binding coefficient (K_eff) can be assessed between the polymers, where a lower value indicates stronger binding of polymer to RNA. Relative binding is shown as both a w/w ratio (PBAE:siRNA) and a nitrogen (from the polymer) to phosphate (from the siRNA) ratio. The data was fit using a variable slope sigmoidal relationship in Graphpad Prism. K_eff,w/w was estimated from the curve fit. Representative of 2 biological replicates. n=3 technical replicates. Mean±S.D. B) Relationship between hydrophobicity (as ClogP of the polymer repeat unit) and the binding equilibrium between polymer and siRNA (as K_eff,w/w). A least squares linear regression showing relationship was fit. R²=0.99. Color indicates the corresponding polymer. C) z-average particle hydrodynamic size and polydispersity index (PDI) measured using dynamic light scattering (DLS) for PBAE and siRNA complexed into polyplexes at four w/w ratios in 50 mM sodium acetate buffer. Representative of 2 biological replicates. n=3 technical replicates. Mean±S.D. D) Zeta potential surface charge of assembled polyplexes containing PBAE and siRNA at four w/w ratios after dilution 14-fold in water to reduce charge shielding. Representative of 2 biological replicates. n=3 technical replicates. Mean±S.D.

Interestingly, we observe a negative association between the effective PBAE-siRNA binding equilibrium, K_eff, and the calculated polymer hydrophobicity (ClogP_repeat) (Figure 3B). This suggests that increasing hydrophobicity in the range of the polymers tested enhances binding of PBAE to siRNA. While siRNA is canonically thought to bind to polycations through electrostatic interactions of its phosphate backbone, some evidence suggests that hydrophobicity may play a role in mediating this effect under certain circumstances.⁶⁰ It seems that this trend is not broadly generalizable, as some systems reported in literature have shown decreased binding to nucleic acid with increased hydrophobicity while others have shown the opposite, signifying that the effect is likely highly dependent on the specific polymer system and solution conditions.^39,61,62 In molecular simulations of cationic amphiphiles binding to DNA, it was shown that for sufficiently hydrophobic systems, increased chain flexibility enables cooperativity in charge neutralization whereby multiple cationic groups within a single polymer chain can associate with each phosphate group.⁶³ While PBAEs are not traditional amphiphiles with a single charge unit and single hydrophobic tail, polymer physics suggests that they may form a 3D conformation in solution that effectively enables this cooperativity, especially in the presence of salt that can shield some of the present charges. Thus, we propose that the effect of hydrophobicity on binding is due to a modulation of electrostatic interactions with extended chain length, not specific hydrophobic interactions between the nucleic acid and polymer. This effect is alike to that governing cooperativity in amine titration.

When PBAE and siRNA interact in solution, they self-assemble into electrostatically bound complexes called polyplexes, with higher packaging efficiency usually associated with lower polyplex size. Dynamic light scattering (DLS) was used to probe polyplex formation with the four polymers at w/w ratios around their charge inversion point. At a low w/w ratio of PBAE to siRNA (0.5), polyplexes existed mostly as 300–350 nm particles bearing negative zeta potential in the range of −23 to −30 mV, with Poly2 and Poly3 exhibiting more negative surface potential (Figure 3C–D). As the w/w ratio increases up to 2, Poly0 never fully charge reverses, existing as large aggregates over 1000 nm in size that are not colloidally stable. In contrast, Poly2 and Poly3 transition from aggregates with low colloidal stability around a w/w ratio of 0.75–1 into polyplexes with full charge reversal to zeta potentials of about +29 mV at a w/w ratio of 2. Interestingly, we observe that even at a w/w ratio of 1, where Poly3 exhibits near neutral surface charge, the formed particles are not large aggregates as is observed for other polymers around their charge transition point. This may be due to chain flexibility in the more hydrophobic Poly3 enabling regional areas of positive and negative charge that increase colloidal stability despite an overall neutral charge. Poly2O seems to similarly transition in charge reversal across the range of w/w ratios tested, but not to the same extent as Poly2 and Poly3. The progression of charge reversal by w/w ratio also follows trends observed in measurements of PBAE-siRNA binding, with Poly2 and Poly3 undergoing charge reversal at a lower w/w ratio than Poly2O or Poly0. We do note that while the observed polyplexes are larger than what is typically seen, we assessed polyplexes only around the charge inversion point, where incomplete charge overcompensation likely led to some degree of aggregation.

3.4. Polymer physicochemical properties govern dissociation of PBAE from siRNA

While it is important for the PBAE to package siRNA, it must remain stable at physiologic pH while also disassembling the siRNA when needed for intracellular silencing bioactivity. Since disassembly occurs after endosomal escape and at cytoplasmic pH, we carried out a polyplex disassembly assay at pH 7.4. Polyplexes were assembled at a 10:1 w/w ratio at pH 5.2 in 50 mM sodium acetate buffer. They were then released in PBS at 37°C, with samples collected at various time points and flash frozen in liquid nitrogen (Figure 4A). A 10:1 w/w ratio was used given that the fluorophore exclusion assay suggested that all four polymers should be fully bound above that w/w ratio. A gel mobility shift assay was used to determine extent and kinetics of decomplexation, where neutralized complexed siRNA does not migrate towards the positive terminus of the agarose gel, but naked siRNA does migrate. We would expect that stable polyplexes might unpack only at later time points. We observe that even from time 0, Poly0 polyplexes destabilize at pH 7.4 and release siRNA. Poly2O polyplexes also release naked siRNA but more slowly than Poly0. From hour 0–8, partial migration of siRNA in the gel suggests a small amount of Poly2O is still bound to it. By 24–48 hours, the siRNA appears to have completely unpacked from Poly2O. In contrast, Poly2 polyplexes release siRNA at 48 hours, suggesting their stability. Poly3 polyplexes release a small amount of siRNA at time 0, but the naked siRNA is rapidly re-assembled into polyplexes, probably due to reorganization in the new pH environment. After that initial time point, the polyplexes remain stable through 48 hours (Figure 4B). To better understand whether the change in polyplex stability was due to a pH change or an ionic strength change, we released the same polyplexes in 170 mM ionic strength pH 5.2 sodium acetate buffer to match that of PBS or PBS itself titrated to pH 5.6. Although the high ionic strength and acidity of 170 mM sodium acetate buffer caused siRNA to smear on the gel, released siRNA is only seen with Poly0, while other polyplexes remain stable. While this release is also seen for Poly0 at time 0 with PBS at pH 5.6, Poly2 does not release siRNA suggesting a high charge density and strong association with siRNA. Poly2O does release over time, like in PBS at pH 7.4, Poly3 releases a small amount of siRNA early on that later reassociates (Figure S7). This inability to form stable polyplexes with Poly0 aligns with the poor Poly0-siRNA binding that we demonstrated in other assays.

Figure 4. Poly2 and Poly3 exhibit slower polyplex disassembly of siRNA in pH 7.4 biologically-relevant conditions.

A) Polyplexes containing siRNA with Poly0, Poly2, Poly2O, or Poly3 were assembled in pH 5.2 sodium acetate buffer at a 10:1 w/w ratio (PBAE:siRNA) to ensure complete charge reversal. Disassembly of these polyplexes was carried out in PBS at pH 7.4 and 37°C, and samples were rapidly frozen in liquid nitrogen after various time points to assess the remaining degree of binding between polymer and siRNA. B) Gel mobility shift assay in 4% agarose gels was used to assess the degree of disassembly between PBAE and siRNA. Comparison to siRNA (si) as reference. Representative of two biological replicates. Images brightened and contrast-enhanced to better visualize bands. Polyplexes were also released in PBS at pH 7.4 and 37°C over 8 hours for DLS measurements. DLS was used to assess z-average size (C) and derived count rate (D), which is the counts of particles concentration normalized to linear attenuator setting. Horizontal dashed lines indicate z-avg sizes for the polyplexes in pH 5.2 sodium acetate buffer before disassembly in PBS. One biological replicate. n=3 technical replicates. Mean±S.D.

To better understand what might be occurring in the early time points for more stable polymers, we incubated polyplexes made of Poly2 or Poly3 with siRNA in PBS at pH 7.4 but probed the polyplexes during the first 8 hours with DLS. At a 10:1 w/w ratio, Poly2 polyplexes (276 nm) were larger than those of Poly3 (105 nm). Both remained relatively stable when initially placed into pH 7.4 PBS; however, over time, the Poly2 polyplexes began to aggregate, increasing in size and decreasing derived count rate, likely as reduced charge units on the polymer reduced surface charge. In contrast, Poly3 remained stable in size over the whole 8 hours. Interestingly, the derived count rate also increased over this time, suggesting some type of polyplex rearrangement process, but the significance of this is not yet understood (Figure 4C–D). It could be that even in a pH above pK_a, Poly3 continues to form polyplexes in solution, as we observe evidence of a slight shift in the size intensity distribution of polyplexes from time 0 to time 1 hour that then remains relatively constant for the remainder of the experiment (Figure S8). Furthermore, we observe degradation half-lives of Poly2 and Poly3 of 71 and 123 hours, respectively (Figure S9). While the specific conditions we used led to longer degradation half-lives than reported in the literature,¹⁹ it suggests that disassembly of polyplexes over the observed time scales is minorly impacted by polymer degradation.

3.5. Assessing transfection efficacy of PBAE polyplexes

Ultimately, we wanted to observe how the emergent properties of PBAE-siRNA systems derived from polymer physicochemical characteristics might govern cellular interactions. While we wanted to eventually understand transfection with these polymers from thin film coatings, polyplexes were first used to serve as a proxy to understand the polymer’s transfection efficacy in a simplified system. We also believe that some thin film architectures release as particles that may behave like polyplexes.⁶⁴ First, polymer toxicity was assessed in HeLa cells; a PrestoBlue metabolic cytotoxicity assay revealed viability above 70% up to at least 15.6 μg/mL concentrations for all the polymers, with Poly2 displaying the greatest toxicity. For reference, polyplexes at a 1:1 w/w ratio with siRNA use a polymer concentration of 0.76 μg/mL in transfection experiments. Poly0 and Poly2O remained biocompatible even at 250 μg/mL (Figure S10).

We then tested transfection efficacy. HeLa cells expressing a fluorescent reporter (GFP) were incubated for 3 days with polyplexes containing a fluorescently labeled siRNA mimic composed of dsDNA (siD7) or an siRNA against GFP (siGFP). The mimic, although composed of DNA bases, has a similar size and structure to siRNA, so it is believed that it serves as a representative model.³⁹ The siD7 enabled assessment of cellular association while siGFP allowed assessment of bioactivity (Figure 5A, gating in Figure S11). At the lowest w/w ratio tested, 5:1, Poly2 and Poly3 exhibited 4.3–8.8-fold higher siRNA association with the HeLa cells than buffer alone, but the difference between these two polymers was not significant (Ordinary One-way ANOVA with Tukey multiple hypothesis correction, p=0.7947), nor was the difference between any of the polymers used. In contrast, Poly0 and Poly2O exhibited no higher association with the HeLa cells than buffer alone, suggesting that polyplexes synthesized from Poly0 and Poly2O do not interact well with HeLa cells and have very low uptake (Figure 5B).

Figure 5. Poly2 and Poly3 with siGFP promote higher silencing efficiency in GFP-expressing cells.

A) PBAEs were mixed with siRNA - either an AlexaFluor647 (AF647) fluorescently labeled dsDNA siRNA mimic (siD7) or a GFP-targeting siRNA (siGFP) - at pH 5.2 to assemble polyplexes. Polyplexes were incubated with GFP-expressing HeLa cells for 3 days, after which the signal from siD7 (association) and GFP (reporter silencing) was assessed with flow cytometry. At a 5:1 w/w ratio of PBAE to siRNA, levels of B) cell association with siD7 and C) GFP-expression with siGFP in GFP-expressing cells were measured. Median fluorescence intensity (MFI) with the four PBAEs was compared to buffer alone (negative control and reference point for normalization) and siRNA with Lipofectamine RNAiMax commercial transfection reagent (“+L”). Representative of 2 biological replicates. n=3–6. Mean±S.D. One-way ANOVA with Tukey-corrected pairwise comparisons. Only comparisons between the polymers are shown. *p≤0.05, **p≤0.01, ***p≤0.001.

While polyplexes must associate with cells to transfect them, association does not always portend bioactivity. Additional barriers beyond association must be overcome: uptake, endosomal escape, release of siRNA, association with the silencing machinery, transcript silencing, and turnover of resulting protein according to its half-life. Thus, reporter cells expressing GFP serve as a good model to test polymer-mediated surmounting of these barriers. After delivering siGFP with polyplexes containing each of the four PBAEs, we noted that Poly3 provided significantly enhanced GFP reporter silencing at 3 days (48%) than Poly0 (3%), Poly2 (16%), or Poly2O (8%) compared to buffer alone (Ordinary One-way ANOVA with Tukey’s multiple hypothesis correction, p=0.0001, 0.0032, and 0.003, respectively) (Figure 5C), suggesting that Poly3 is a better transfection polymer than the others. While Poly2 did show high levels of siRNA association at a 5:1 w/w ratio, it demonstrated only low levels of silencing; however, when included at a higher 10:1 w/w ratio, Poly2 demonstrated significantly higher silencing over Poly0 and Poly2O (Ordinary two-way ANOVA with Tukey’s multiple hypothesis correction, comparing the four polymers at different w/w ratios, p<0.0001 for both comparisons). At this higher w/w ratio, the difference in transfection between Poly2 and Poly3 disappeared, and silencing levels approached those with the RNAiMax positive control. Similar trends were observed at 15:1 and 20:1 w/w ratios (Figure S12). Overall, Poly3 serves as the best transfection polymer, although Poly2 also serves as a viable transfection polymer when used at higher w/w ratios. The importance of w/w ratio is particularly important in thin film coatings, since the w/w ratio in these self-assembled systems is often lower than that which can be achieved in colloidal systems.^35,64 Higher w/w ratios may enhance transfection, as excess polycation existing in a free, unbound form has been shown to aid endosomal escape.⁶⁵ To also assess the role of PBAE molecular weight on transfection, our 3.2 kDa Poly3 led to slightly higher association with HeLa cells than our normal 2.3 kDa Poly3 at higher w/w ratios, but no differences were observed in the resulting GFP reporter silencing (Figure S13). This suggests that a small increase in Poly3 size does not impact PBAE-mediated transfection, but further experiments are needed.

It is also important to understand the endosomal mechanisms that may be driving uptake of PBAE polyplexes. Although chemical inhibitors can have off-target effects,⁶⁶ they remain one approach to elucidating uptake mechanisms. Poly3 as a representative PBAE demonstrates inhibition of uptake after 24 hours with the dynamin inhibitor dynasore (25.8 μg/mL) but inhibition of caveolin-mediated endocytosis with genistein (54 μg/mL) does not, suggesting a role for clathrin-mediated or fast endophilin-mediated endocytosis in the uptake of PBAEs. After 48 hours, inhibition of caveolin-mediated endocytosis decreases the level of silencing effect with PBAE, as does dynamin inhibition with dynasore, suggesting roles for both endosomal mechanisms in transfection. Macropinocytosis inhibition with wortmannin (4.3 μg/mL) played a minimal role in uptake or transfection efficacy (Figure S14). Previous studies of PBAE uptake mechanisms in breast cancer cells, suggested a role for caveolin-mediated endocytosis in uptake of PBAEs and a combination of clathrin- and caveolin-mediated endocytosis for transfection efficacy.³⁰ Meanwhile, uptake of polyplexes containing plasmid DNA and PBAEs similarly highlighted dynamin-dependent uptake independent of caveolin, aligning with our findings.⁶⁷ In contrast, lipid nanoparticles have been shown to primarily depend on macropinocytosis and dynamin-dependent processes for uptake.⁶⁸ Altogether, the trends observed here are likely dependent on the specific PBAE, the time scale used and the concordance of observation time scale with that of different uptake mechanisms, and the dosing, potentially explaining some differences between our observations and the literature. In particular, others have found that the ratio of different uptake mechanisms varies slightly based on the specific PBAE used, its molecular weight, and end group functionalization,^30,41 but more work will be needed to connect physicochemical properties to endocytosis mechanism.

3.6. Assessing transfection efficacy of PBAE-containing oligo-coated dressings

Since the electrostatic assembly of LbL thin film coatings does differ in many ways from polyplexes as we have already alluded to,⁶⁹ our ultimate goal was to assess how transfection efficacy of siRNA from the surface of wound dressings may vary with differing PBAE layers. Specifically, differences in how siRNA and polymer interact in assembly of polyplex versus thin film systems could lead to differing transfection efficiencies. Since we have previously demonstrated a formulation for delivery of oligonucleotide therapy from the surface of wound dressings in vitro and in vivo,³² we sought to leverage this formulation while varying only the PBAE component used. We focused on Poly2 and Poly3, given that they showed the most promise in experiments using polyplexes. The architecture comprised 20 alternating layers of PBAE and dextran sulfate (DS, ~500 kDa) followed by 50 bilayers of linear polyethyleneimine (LPEI, 25 kDa) and siRNA (~14–16 kDa) (Figure 6A). After interlayer diffusion processes, we envisioned that the siRNA in this formulation is associated with both the LPEI and PBAE.^70–72

Figure 6. siRNA dressings assembled with Poly3 promote higher cell association and reporter gene silencing bioactivity than those assembled with Poly2.

A) The layer-by-layer (LbL) approach was used to coat wound dressings with siRNA and transfection-enhancing polymers. PBAE (Poly2 or Poly3) is alternatively adsorbed with dextran sulfate (DS) 20 times to form hydrolyzable base layers. RNA layers contain 50 alternating layers of linear polyethyleneimine (LPEI) and siRNA. B) Total loading of fluorescent siRNA on the dressings was assessed with total release in 3M NaCl and compared to a standard curve. n=7–13. Mean±S.D. Student’s t-test. C) Schematic demonstrating transfection of oligo-coated dressings into GFP-expressing cells. Loading of fluorescently labeled non-targeting oligo (siD7) can be used to demonstrate cell association upon release and loading of siRNA targeting GFP (siGFP) can be used to assess silencing bioactivity of released siRNA. The opposite oligo can be used as a control to account for any effects that the polymers themselves may impart. D) Median fluorescent intensity (MFI) association of AF647 fluorescent non-targeting siRNA (siD7) with HeLa cells and E) MFI silencing of GFP in GFP-expressing HeLa cells when siGFP is released from dressings with Poly2 or Poly3. “+L” represents samples that received lipofectamine RNAiMax commercial transfection reagent in addition to the dressing (positive control). Observations normalized to average MFI for Poly2 and Poly3. n=5–10 for association and 2–10 for silencing (with n=2 from the +L positive controls). Mean±S.D. One-way ANOVA with Tukey multiple comparison correction – showing comparisons between Poly2 and Poly3 only. ns – non-significant, *p≤0.05, **p≤0.01.

The loading of siD7 between the Poly2 and Poly3 formulation was not significantly different, although we note some heterogeneity in loading across the surface of the wound dressing (Figure 6B). Poly3 dressings were shown to have slower release of siD7 than Poly2 formulations, which is believed to be due to increased hydrophobicity of Poly3 excluding more water, slowing surface erosion and polymer hydrolysis (Figure S15). We demonstrated slower hydrolysis of Poly3 compared to Poly2 (Figure S9). In our system, we did not observe film instability with Poly3 as previously reported.⁵⁰

Like polyplexes, dressings were tested for transfection efficacy by direct incubation with GFP-expressing HeLa cells for 3 days. Fluorescently labeled siRNA (siD7)-containing dressings were used to assess cellular association of released siRNA and siGFP was used to assess silencing bioactivity of released siRNA. Dressings containing the other siRNA were used as a control to account for any effects that the polymeric thin film coatings themselves might induce (Figure 6C). Dressings were also co-incubated with Lipofectamine RNAiMax as a positive control to help any released siRNA enter cells, thus representing the maximum possible transfection. Poly3 dressings demonstrated approximately 230 times the median fluorescence intensity (MFI) of a non-fluorescent siRNA formulation containing siGFP, while Poly2 dressings demonstrated only 45 times MFI of the same. This suggests that Poly3 dressings promoted about 5 times more cellular association than Poly2 dressings at 3 days, a difference that was statistically significant (Ordinary one-way ANOVA with Tukey’s multiple hypothesis correction, p=0.0007) (Figure 6D). This contrasts with polymer performance in polyplexes, where no significant difference in association was observed. The high amount of siRNA loading, sustained release, differences in the w/w ratio of polymer to siRNA, and addition of other transfection-enhancing polymers in the formulations may have enabled differences in association to be amplified in dressing transfection over polyplex transfection. As with the polyplexes, cellular association is only one step of effective transfection, so we also measured silencing bioactivity of the reporter GFP. Poly2 dressings containing siGFP significantly reduced MFI by about 20% over non-targeting siD7 dressings (Ordinary One-way ANOVA with Tukey’s multiple hypothesis correction, p=0.0236), while Poly3 dressings containing siGFP significantly reduced the same by 39% (p=0.0002). Thus, Poly3 dressings had nearly double the silencing efficacy of Poly2 dressings (p=0.0325) (Figure 6E), consistent with trends observed in polyplexes. Lower w/w ratios of polycation to siRNA in coated dressings compared to those achievable in polyplexes, along with slowed release kinetics and potential release of naked siRNA likely drove incomplete silencing.^35,64

While there are similarities in the transfection of siRNA from polyplexes and electrostatically assembled coatings on dressings, we were intrigued by the comparability of the observed trends. This suggests the key role of the PBAE in mediating the degree of transfection efficacy despite the presence of other excipients in the dressing coatings. Additionally, as true surface erosion would have siRNA eluting prior to PBAE, the observed results suggest a few key observations: 1) there is likely a high level of interlayer diffusion in the coatings, allowing the weakly charged polymeric layers to rearrange and reassociate with different electrostatic interactions, especially given the PBAE hydrophobicity and smaller molecular weight of many of the employed materials,^70–72 2) dressings may not release by pure surface erosion and might elute larger chunks that can be taken up or slowly erode over time, as observed in other systems,⁶⁴ 3) siRNA release kinetics may play a role in modulating transfection efficacy, and 4) released siRNA from thin films is likely associated with both PBAE and LPEI; however, any or all of these mechanisms governing differences between Poly2 and Poly3 are complemented by fundamental differences in polymer transfection ability, as demonstrated by polyplex transfection.

3.7. Poly2 and Poly3 polyplexes demonstrate differences in mediating uptake and endosomal membrane disruption

Given the differences in transfection observed between Poly2 and Poly3 and their different physicochemical properties, we wanted to better connect physicochemical properties to steps in the transfection process. Thus, having already established differences in polymer pK_a and cooperativity in titration that might mediate differences in transfection efficacy, we investigated if differences in transfection efficacy were additionally driven by differences in 1) polymer-mediated association with the cellular membrane, 2) membrane disruption potential, 3) enhanced endosomal escape capability, or a combination of any or all mechanisms. As in the LbL experiments, we focused on only the two working PBAEs, Poly2 and Poly3. Polyplexes were once again employed over LbL dressings to isolate the role of PBAE from that of other excipients used in LbL or from release kinetic differences.

In association experiments, it is hard to tell whether detected siRNA is on the surface or internalized. Thus, to isolate differences in binding to the cell membrane, we employed methods to limit uptake and internalization processes (Figure 7A). First, we tried blocking any ATP-driven processes by performing polyplex transfection at 4°C for 4 hours and found that in general there was very low association of siRNA with the cell membrane in Poly2, Poly3, or positive control RNAiMax (“+L”) groups compared to those cells receiving no siRNA (“buffer”). Furthermore, there was no significant difference in membrane association between Poly2 and Poly3 polyplexes (Ordinary one-way ANOVA with Tukey’s multiple comparison, p=0.9099) (Figure 7B). Because cold temperatures can also influence the diffusion of polyplexes and fluidity of cellular membranes, we assessed immediate membrane interaction of polyplexes at body temperature (37°C). We chose to investigate this on short timescales of 1 or 2 hours after transfection, since studies in lipid nanoparticle systems have shown uptake and endosomal escape to lag by about 1–2 hours,^68,73 although transfection with PEI has shown a lag of only about 30 minutes.⁷⁴ Similar to transfection at 4°C, transfection at 37°C for 1 hour showed little cellular association and no significant difference between Poly2 and Poly3 (p=0.9918) (Figure 7C). By 2 hours of transfection, association had increased from that observed at 1 hour, but no significant differences between Poly2 and Poly3 were detected (p=0.9915) (Figure 7D). In all experiments, the RNAiMax groups had significantly increased association over Poly2 and Poly3, showing the effectiveness of commercial systems at rapid cellular association in vitro. Together, this data suggests that differences in binding of polyplexes made from Poly2 or Poly3 to the cell membrane do not significantly contribute to observed transfection differences.

Figure 7. Poly2 and Poly3 exhibit similar membrane association at early time points and with ATP-dependent uptake processes inhibited.

A) PBAEs were mixed with siD7 at a 5:1 w/w ratio to assemble polyplexes, which are then incubated with HeLa cells at varying temperatures and for varying amounts of time to assess cellular membrane association. Cells were incubated with polyplexes at B) 4°C for 4 hours to slow any ATP-dependent metabolic uptake process, C) 37°C for 1 hour to capture early cellular binding interactions of siRNA and D) 37°C for 2 hours to assess slightly later interactions between cell membrane and siRNA. Comparison was to buffer alone (negative control and reference point for normalization) and siRNA with RNAiMax commercial transfection reagent (“+L”). n=3–4. Mean±S.D. One-way ANOVA with Tukey multiple comparison correction – only showing comparisons between Poly2 and Poly3. ns – non-significant.

Next, we sought to evaluate if Poly2 and Poly3 may differ in their ability to physically disrupt the endosomal membrane, since it is another crucial step to transfection. We recapitulated this process ex vivo by measuring the polymers’ ability to lyse red blood cell (RBC) membranes at varying pH values,¹⁴ although this assay has been shown to be an imperfect predictor of transfection efficacy.^43,45 In this assay, hemolysis of RBCs is measured after treatment with different transfection systems within a physiological pH range (7.4, 6.8, 6.2, and 5.6) that recapitulates internalization and endosomal acidification (Figure 8A). We find that across a range of pH values, Poly3 polyplexes display higher levels of hemolysis than Poly2 polyplexes. In fact, Poly2, even at concentrations up to 4 times (4x) those used for in vitro transfection, exhibits very low membrane disruption. In contrast, Poly3 demonstrates up to 24% hemolysis at 1x concentration, and is pH-dependent. Interestingly, we find that as the pH drops from 6.8 to 5.6, so does hemolysis. This suggests that Poly3 could disrupt early endosome membranes, but further acidification causes further protonation and rearrangement of polymer chains with siRNA in a way that reduces its ability to disrupt membranes (Figure 8B–C). Similar trends exist for polymers alone (Figures S16). There is a low amount of hemolysis with Poly3 polyplexes at pH 7.4 (8–18% depending on concentration), when they would be expected to have practically no protonation. This suggests that hydrophobicity itself may also lead to some degree of hemolysis, although we note that this hemolysis contrasts with the non-toxicity of Poly2 and Poly3 up to about 5 times the concentration used in the 1x group (Figure S10). This also points to the delicate balance between cytotoxicity and transfection efficacy, which is a well-known issue with cationic polymeric oligonucleotide therapy transfection delivery systems.¹⁷ In a local delivery approach, the toxicity is likely minimal and could be further reduced by binding to proteins in the tissue microenvironment. Overall, results of this assay suggest that part of the difference in transfection between Poly2 and Poly3 may be related to the higher potential for Poly3 to physically disrupt cell membranes over Poly2.

Figure 8. Poly3 polyplexes exhibit more pH-dependent membrane disruption than Poly2 polyplexes.

A) PBAEs were mixed with siGFP at a 5:1 w/w ratio to assemble polyplexes, which were then incubated at varying concentrations with RBCs dissolved in PBS at pH 7.4, 6.8, 6.2, or 5.6. After 1 hour of incubation, absorbance of the supernatant at 541 nm was read, quantifying free hemoglobin. B) Poly2 and C) Poly3 polyplexes were assessed for normalized hemolysis. Hemolysis normalized to 1% Triton X-100 (1.0) and pH 5.2 50 mM sodium acetate buffer alone (0.0). Polyplexes and controls dissolved 1 to 20 in RBCs. 1x concentration represents Poly2 at 3.77 μg/mL and siRNA at 0.755 μg/mL. Bar color representative of pH. Representative trial of 2 biological replicates from two blood donors. n=3 technical replicates. Mean±S.D.

3.8. Poly2 and Poly3 promote endosomal escape to enable delivery of siRNA to the cytoplasm

For siRNA to be functional, it must escape endosomes and enter the cytoplasm, where it can interact with target transcripts and induce silencing. While we characterized how endosomal buffering might differ between Poly2 and Poly3 along with how physical membrane disruption may differ between the two polymers, we sought to visualize endosomal escape. Upon endosomal escape, cytosolic Gal8 binds newly exposed endosomal glycans on the ruptured membrane to provide stabilization, creating distinct clusters of Gal8.^73,75 Gal8 was coupled to a GFP reporter and expressed in HeLa cells (HeLa Gal8-GFP).⁴⁶ This allows fluorescence microscopic visualization of the endosomal escape process. We compared endosomal escape events in cells treated with Poly2 versus Poly3 dressings, with buffer treatment as a negative control (Figure 9A). Qualitatively, we visualize endosomal escape events as distinct nanometer sized puncta. While buffer conditions induce minimal endosomal escape events, both Poly2 and Poly3 dressings yield abundant endosomal escape events. We observe a higher amount of fluorescently labeled siRNA associated with cells transfected with Poly3 dressings compared to that with Poly2 dressings, consistent with flow data. Gal8 signal provides a snapshot of endosomal escape events in a specific time frame, while analysis of siRNA signal itself may have accumulation over time. Areas of diffuse siRNA signal are observed with Poly3 dressings, but this is less present with Poly2 dressings. In both cases, puncta of siRNA signal is likely indicative of siRNA trapped in endosomes. This finding suggests enhanced endosomal escape of siRNA over time with Poly3 versus Poly2 (Figure 9B). Ultimately, quantification of Gal8 puncta per cell corroborates our qualitative observations, with both Poly2 and Poly3 dressings demonstrating enhanced endosomal escape events over buffer alone (One-way ANOVA with Tukey multiple comparison correction, p<0.0001 for both). No significant difference is observed between Poly2 and Poly3 dressings (p=0.48, Figure 9C). Overall, these results concur with the findings from other assays while specifically linking enhanced endosomal escape to enhanced efficacy of Poly3 dressings.

Figure 9. Poly2 and Poly3 dressings promote endosomal escape, although more in the latter.

A) Dressings formulated with fluorescently labeled siRNA (siRNA-AF647) and Poly2 or Poly3 were incubated with HeLa Gal8-GFP cells. Compared to buffer alone (negative control). 22 hours later, endosomal escape was assessed, where Gal8-GFP puncta are indicative of escaped endosomes. B) Representative max projection of confocal images. 30x. Scale bar represents 50 μm. Representative images. Yellow arrowheads depict examples of Gal8 puncta. Green dashed line inset shows magnified view of Gal8 puncta. Magenta dashed line inset shows magnified view of siRNA-AF647 signal. Pink arrowhead shows siRNA that is likely in endosomes. Lilac arrowheads show diffuse cytoplasmic siRNA signal, associated with endosomal escape events. Brightness and contrast set to same thresholds across images. C) Manual quantification of Gal8 puncta normalized to the number of nuclei. n=3 images per condition. One-way ANOVA with Tukey multiple comparison correction. Mean±S.D. ****p≤0.0001.

3.9. Connecting physicochemical properties and transfection efficacy

While the small nature of this four polymer PBAE set precludes understanding of macro-level trends, mechanistic studies enable us to elucidate why certain polymers in our set worked well for transfecting siRNA from thin film coatings on scaffolds and others did not. We demonstrate that for given polymer molecular weight, increasing PBAE hydrophobicity lowers the pK_a of the protonatable amines and enhances the polymer buffering capacity. These trends are observed in Poly2 and Poly3, although the slightly lower pK_a of Poly3 (6.21) over Poly2 (7.08) is more in line with what others have reported as serving as an effective buffer in the acidifying endosome. Poly0 had a fast pH transition about its pK_a and thus low buffering capacity, while the hydrophilicity and longer chain length between amines of Poly2O over Poly2 raised the prior’s pK_a to a level whereby it remains somewhat charged at all physiologically relevant pHs. This prevents it from buffering within a range that would be experienced in the acidifying endosome. The confluence of hydrophobicity, chain flexibility, and charge density led to Poly3 demonstrating the strongest binding to siRNA at pH 5.2, followed by Poly2; these polymers formed colloidally stable polyplexes at appropriate w/w ratios. Binding strength and hydrophobicity also gave rise to enhanced stability of Poly2 and Poly3 polyplexes, although the prior demonstrated aggregation at pH 7.4 while the latter did not. When assembling PBAE and siRNA into electrostatically assembled LbL thin film coatings, the higher hydrophobicity of Poly3 does not seem to demonstrate electrostatic instability as previously reported,⁵⁰ but rather slows the release kinetics of siRNA from the dressings, potentially through the local exclusion of water and enhanced binding strength to siRNA.

The interactions between PBAE and siRNA control the degree of siRNA-mediated silencing of a reporter in HeLa cells in both polyplex form and LbL wound dressing coatings. We further demonstrate the key role of the PBAE component on siRNA transfection from LbL thin films, in which Poly3 mediates higher transfection efficiency than Poly2. Ultimately, we demonstrate that Poly3 demonstrates enhanced buffering capacity in a more relevant endosomal pH regime than Poly2, to enhance the proton sponge effect. Additionally, Poly3 can induce greater membrane disruption than Poly2 in a pH-dependent manner to help mediate endosomal escape. Ultimately, we show enhanced endosomal escape with Poly3. In summary, the physicochemical differences in a small group of carefully selected PBAEs with subtle structural changes and given molecular weight mediate large differences in transfection of siRNA from thin film coatings on wound dressings.

3.10. Limitations

This study is limited in that it studied only one well-known polymer class and the library was relatively small. Only one polymer molecular weight was investigated, and it is known that polymer molecular weight can influence transfection of nucleic acids, although the trends by which it does so have not been well defined and are both cell-type and polymer-dependent.^21,28,76,77 The molecular weights of PBAEs used in this study are within the range used by other PBAE studies.^23,31,78 Although the amine monomer was consistently kept in excess during synthesis, allowing for a predominance of amine-terminated polymers, the unknown role of end groups remains an important consideration in the effects of our PBAEs. Future studies can better assess the structure-function relationships that may exist based on polymer molecular weight and end group effects.

Additionally, while we attempted to explain from fundamentals how polymer physicochemical properties might drive changes in chemical interactions with siRNA, this study provided only limited proof of the actual 3D conformation of the polymers as a function of buffer pH, ionic strength, and polymer size. This is particularly relevant since it appears that our most hydrophobic polymer, Poly3, interacts with siRNA and itself in ways that are independent of just the protonatable amine units. Future studies will explore a larger library of PBAEs for thin film-mediated delivery of siRNA and seek to better characterize the polymer physics underlying the observed trends.

4. Conclusions

The localized delivery of siRNA therapy to treat various diseases is challenged by entry of the therapeutic into cells and out of endosomes. While our lab has developed electrostatic coatings on wound dressings containing siRNA and transfection-enhancing polymers for various local delivery applications, minimal work has characterized the role of PBAE physicochemical properties on transfection efficiency. In this study, we investigate the role of polymer hydrophobicity and pK_a at given polymer molecular weights on emergent properties such as binding equilibrium to siRNA, complexation characteristics, siRNA disassembly kinetics, colloidal stability, and siRNA release kinetics from coated wound dressings. We then tie these properties to important biomaterials properties including polymer toxicity, transfection efficiency as both polyplexes and electrostatically assembled dressings, association with the cell membrane, membrane disruption ability, and endosomal escape. Ultimately, we find that Poly3, which has the greatest hydrophobicity, a pK_a in the range of an early endosome, high buffering cooperativity, the highest binding equilibrium to siRNA, and high colloidal stability, leads to the highest transfection efficacy in a wound dressing coating formulation. The conclusions of this study may serve useful in the rational design of PBAEs for siRNA and oligonucleotide therapy transfection systems to local areas of disease. Importantly, while our previous works have used mostly Poly2,^9,10,32,35 this work suggests that incorporating Poly3 into our oligonucleotide-eluting electrostatically assembled dressings may help to enhance intracellular delivery. In addition to in vivo investigation of Poly3-containing coated wound dressings, future work will seek to elucidate how physicochemical differences in PBAEs may mediate the intracellular delivery of other therapies for local delivery, including proteins, small molecules, and larger nucleic acid therapies.

Data availability

Data will be made available on request.