The Spatial Distribution of Lipophilic Cations in Gradient Copolymers Regulates Polymer–pDNA Complexation, Polyplex Aggregation, and Intracellular pDNA Delivery

Abstract

Here, we demonstrate that the spatial distribution of lipophilic cations governs the complexation pathways, serum stability, and biological performance of polymer–pDNA complexes (polyplexes). Previous research focused on block/statistical copolymers, whereas gradient copolymers, where the density of lipophilic cations diminishes (gradually or steeply) along polymer backbones, remain underexplored. We engineered gradient copolymers that combine the polyplex colloidal stability of block copolymers with the transfection efficiency of statistical copolymers. We synthesized length- and compositionally equivalent gradient copolymers (G1–G3) along with statistical (S) and block (B) copolymers of 2-(diisopropylamino)ethyl methacrylate and 2-hydroxyethyl methacrylate. We mapped how polymer microstructure governs pDNA loading per polyplex, pDNA conformational changes, and polymer–pDNA binding thermodynamics via static light scattering, circular dichroism spectroscopy, and isothermal titration calorimetry, respectively. While gradient steepness is a powerful design handle to improve polyplex physical properties, augment pDNA delivery capacity, and attenuate polycation-triggered hemolysis, microstructural contrasts did not elicit differences in complement activation.

Graphical Abstract

INTRODUCTION

Therapeutic nucleic acids show promise to improve the lives of millions of people suffering from genetic disorders.¹ However, nucleic acids must overcome intracellular delivery barriers such as cellular uptake, endolysosomal processing, or nuclease-mediated degradation, which curtail their therapeutic functionality.^2,3 Many FDA-approved gene therapeutics rely on viral vectors to overcome these barriers.^4–6 However, engineered viral vectors are difficult to manufacture at scale,^7,8 potentially immunogenic^9–11 and prohibitively expensive.^5,6 Nonviral carriers such as polymers,¹² lipids,¹³ polypeptides,¹⁴ and inorganic nanoparticles¹⁵ are being engineered to deliver nucleic acid payloads to cells. Polymers—which can be economically mass-produced and designed to minimize immune responses—are among the most viable nonviral gene delivery platforms.^16,17 Synthetic advances, particularly in reversibly deactivated radical polymerization, help chemists access arbitrary combinations of compositions, lengths, architectures, and repeat unit spatial distributions, offering unprecedented control over polymer properties.¹⁸

Although cationicity aids polymers in delivering anionic nucleic acids, polycation-induced cytotoxicity,¹⁹ and poor colloidal stability in protein-rich environments²⁰ hinder gene delivery. Polycations aggregate in salt- and serum-containing media due to Debye screening and nonspecific protein adsorption, impeding cellular uptake.^21–23 To improve the cytocompatibility and colloidal stability of polycations while maintaining functional transgene delivery, chemists have investigated the effects of polymer length,^24,25 architecture,^26–28 and cationic charge density.^29,30 Copolymerization proved to be particularly effective; copolymers of cationic monomers with neutral, hydrophilic comonomers overcame trade-offs among serum stability, cytotoxicity, and transfection performance.^31,32

Investigators focused predominantly on statistical copolymers, where cationic and neutral repeat units conform to a stochastic spatial distribution (assuming comparable monomer reactivity ratios), or block copolymers, where cationic and neutral repeat units are segregated within discrete polymer segments. In statistical copolymers, cationic repeat units are well-distributed throughout the length of the polymer backbones. Consequently, statistical copolymers have a higher frequency of contact with anionic cell membranes than equivalent block copolymers.^33–36 Unlike block copolymers,^28,37 statistical copolymers are highly susceptible to aggregation in serum.^28,38,39 Block copolymers that incorporate cationic and neutral hydrophilic comonomers can form polyion complex micelles.^31,40 Here, cationic blocks condense nucleic acids in micellar cores surrounded by densely hydrated hydrophilic shells that prevent aggregation in serum.³⁷ However, this hydrophilic corona may inhibit cellular internalization.^41–44 Several groups have investigated block lengths,^26,45 block order, and monomer identity extensively.^46–49 Although multiblock copolymers form uniformly sized polyplexes in serum, they usually require multistep synthesis and careful characterization to verify chain extension.⁵⁰ On the other hand, gradient copolymers are amenable to one-pot synthesis while still permitting spatial control over the distribution of cationic repeat units. Advances in macromolecular reaction engineering,^51,52 predictive kinetic models,^53,54 semibatch,^51,55–58 and continuous flow polymerization^59,60 allow for exquisite control over monomer spatial distribution in gradient copolymers. However, gradient copolymers remain under the radar in gene delivery.⁶¹

Foreseeing that subtle modifications in gradient steepness may powerfully modulate polymers’ interactions with nucleic acids and cells, we exploited the gradual transition from cationic- to neutral-dominant regions in gradient copolymers to realize cytocompatible, serum-stable, and efficient plasmid (pDNA) carriers. Previously, gradient copolymers were engineered to deliver hydrophobic drugs.^62,63 However, the pDNA delivery capacity of gradient copolymers is entirely unknown.^34,39 Our approach fundamentally advances polymer-mediated pDNA delivery by addressing three gaps in our knowledge base: how does the spatial distribution of cationic groups impact (1) polymer–pDNA complexation, (2) polyplex aggregation in serum-containing media, and (3) pDNA delivery efficiency and cytotoxicity? We modulated the spatial distribution of lipophilic cationic repeat units in gradient copolymers via semibatch reversible addition–fragmentation chain-transfer (RAFT) polymerization. Extensive polymer characterization revealed how the spatial distribution of lipophilic cations impacted physicochemical properties (hydrophobicity, pK_a, and pDNA binding affinity). Static light scattering, circular dichroism spectroscopy, and isothermal titration calorimetry elucidated how polymer microstructures impacted pDNA loading per polyplex, pDNA conformations in polyplexes, and binding thermodynamics, respectively. In serum-containing media, polyplexes formed from block and block-like gradient copolymers resisted aggregation better than polyplexes of statistical copolymers. Furthermore, the distribution of cationic repeat units along the gradient copolymer backbones governed transfection efficiency, cytotoxicity, and hemocompatibility. Statistical copolymers achieved excellent transfection efficiencies, albeit at the cost of low cell viability, whereas block copolymers maintained high cell viability but failed to deliver pDNA. We achieved promising transfection efficiencies (comparable to statistical copolymers) and well-defined polyplex populations in serum-containing media (comparable to block copolymers) merely by tuning the gradient steepness. Our work demonstrates that gradient polymerization is a powerful and under-utilized tool to optimize multiple design objectives in polycation-mediated nucleic acid delivery.

EXPERIMENTAL SECTION

Materials.

N,N-Dimethylformamide (DMF) was purchased from Spectrum Chemicals (New Brunswick, NJ). 4-Cyano-4-(((ethylthio)carbonothioyl)thio)pentanoic acid (CEP; 95%) was used as the chain-transfer agent (CTA) and was purchased from AmBeed (Arlington Heights, IL). The initiator 4,4′-azobis(4-cyanovaleric acid) (V501; >98%), 2-(diisopropylamino)ethyl methacrylate (DIP; 97%, 100 ppm MEHQ), and 2-hydroxyethyl methacrylate (HEMA; ≥99%, ≤50 ppm MEHQ) were purchased from Sigma-Aldrich (St. Louis, MO). DIP and HEMA monomers were purified by passing through basic alumina (Sigma-Aldrich). All other chemicals were used as received.

Polymer Synthesis.

All copolymers were synthesized via reversible addition–fragmentation chain transfer (RAFT) polymerization (Figure S1A) in a Carousel 12 parallel synthesizer (Heidolph; Wood Dale, IL). All reaction mixtures below were degassed using 3–4 cycles of freeze–pump–thaw. Yields for each reaction ranged between 50 and 70%.

Block Copolymer Synthesis.

To obtain block copolymers, a macroCTA was first synthesized from DIP and chain-extended with HEMA. To synthesize the DIP macroCTA, 0.0032 mol of the DIP monomer was added to 4 mL of DMF, with CEP and V501 (molar ratio of [DIP monomer]:[CTA]:[initiator] = 450:10:1). The reaction mass was degassed, and the reaction was allowed to proceed for 5 h. After 5 h, the reaction mass was quenched in liquid nitrogen. The reaction mixture was acidified with 1 M HCl to dissolve the precipitated polymer. DIP macroCTA was dialyzed against acidified water and lyophilized. Next, 0.00125 mol of HEMA was added to 2 mL of DMF, with macroCTA and V501 (molar ratio of [HEMA]: [macroCTA]:[initiator] = 350:10:1). The reaction mixture was degassed, stirred at 200 rpm, heated to 80°C, and maintained in an inert nitrogen environment for the duration of the reaction (18 h).

Statistical Copolymer Synthesis.

To obtain statistical copolymers, 0.0025 mol of DIP and HEMA were added to 4 mL of DMF with CEP and V501 (molar ratio of [DIP and HEMA monomers]:[CTA]:[initiator] = 700:10:1). The reaction mixture was degassed, stirred at 200 rpm, heated to 80°C, and maintained in an inert nitrogen environment for the duration of the reaction (18 h).

Gradient Copolymer Synthesis.

To obtain gradient copolymers, reaction vials were prepared by adding 0.0025 mol of HEMA to 4 mL of DMF with CEP and V501 (molar ratio of [HEMA]:[CTA]:[initiator] = 350:10:1). Simultaneously, DIP was degassed in a separate vial by purging with nitrogen gas for 30 min ([HEMA]:[DIP] = 1:1). The HEMA reaction mixture was degassed via freeze–pump–thaw, stirred at 200 rpm, and heated to 80°C to start polymerization. After 90 min of polymerizing the HEMA reaction mass, DIP was injected using a NE 1000 syringe pump (New Era Instruments; Farmingdale, NY) over three different time intervals; 30 min for G1, 90 min for G2, and 180 min for G3. After the injections were complete, the syringe was removed, and the reaction was allowed to proceed overnight (18 h).

After 18 h, reactions were quenched with liquid nitrogen, followed by the addition of 1 M HCl to dissolve any precipitated polymers. Polymers were dialyzed (3.5 kDa MWCO, Repligen; Waltham, MA) against acidified water (pH 5.5) over 3–5 days, and the dialysis solution was changed twice daily. Finally, polymers were lyophilized and stored at 4°C.

Polymer Reactivity Ratios.

To determine the reactivity ratios, eight reaction mixtures were prepared at monomer molar feed ratios ranging from 1:9 to 9:1 ([DIP]:[HEMA]) and polymerized to low conversion using the reaction conditions described above. Kinetic studies (Figure S1B) suggested that we could maintain monomer conversions below 10% by quenching the reaction at 20 min. Copolymer compositions were determined via ¹H NMR and plotted against monomer feed compositions. Reactivity ratios were determined by fitting experimental data using the Mayo–Lewis equation, eq 1:

$$F=\frac{R_{\text{DIP}}{f}^2+f(1-f)}{R_{\text{DIP}}{f}^2+2f(1-f)+R_{\text{HEMA}}{(1-f)}^2}$$ (1)

where F estimates DIP incorporation into copolymers (determined from ¹H NMR), f is the mole fraction of the DIP monomer in the monomer feed, and R_DIP and R_HEMA are the reactivity ratios of DIP and HEMA, respectively.

Polymer Characterization.

All NMR spectra were recorded on a JEOL ECA 500 (JEOL; Peabody, MA). Selective one-dimensional ¹H–¹H nuclear Overhauser effect spectroscopy (1D NOESY) samples were prepared by dissolving 50–100 mg of polymer in 0.7 mL of D₂O. To aid polymer dissolution, 10–30 μL of HCl was added. Samples were excited at 1.25 ppm, and 256 scans (500 ms mixing time and 10 s relaxation delay) were collected per sample. ¹H NMR samples were prepared by dissolving ≈10 mg of polymer into 0.7 mL of D₂O, and spectra were recorded with 32 scans and 10 s relaxation delay. Molecular weight distribution was determined via size exclusion chromatography (Agilent; Santa Clara, CA) using refractive index and multiangle light-scattering detectors (SEC-MALS, Wyatt; Santa Barbara, CA) in a 100 mM Na₂SO₄, 1 wt % acetic acid buffer. Three columns were used in series for polymer separation: CATSEC-100, CATSEC-300, and CATSEC-1000 (Eprogen; Downers Grove, IL). The number of DIP and HEMA repeat units for each polymer were calculated following eqs 2 and 3:

$${\text{RU}}_{\text{DIP}}=\frac{n_{\text{DIP}}(M_{\text{n}}-M_{\text{CTA}})}{n_{\text{DIP}}{M}_{\text{DIP}}+(1-n_{\text{DIP}})M_{\text{HEMA}}}$$ (2)

$$\text{R}{\text{U}}_{\text{HEMA}}=R{U}_{\text{DIP}}\frac{1-n_{\text{DIP}}}{n_{\text{DIP}}}$$ (3)

where n_DIP is the molar percent of DIP in the copolymer determined from ¹H NMR, M_n is the copolymer molecular weight determined from SEC-MALS, and M_CTA, M_DIP, and M_HEMA are the molecular weights of the CTA, DIP, and HEMA monomers, respectively. Final copolymer ¹H NMR spectra and SEC traces are displayed in Supporting Information Sections 2–7, respectively.

Polymer ζ-potential was measured using a Brookhaven NanoBrook-90 Plus PALS instrument (Brookhaven Instruments; Holtsville, NY). Polymer pK_a values were measured by using an Orion Star T910 pH titrator (Thermo Fisher Scientific; Waltham, MA). The polymer was dissolved in water (1–5 mg/mL), and 1 M HCl was added to realize an initial pH of ≈2–3. Samples were titrated with 0.05 M NaOH.

The partition coefficient (log P) of the polymers was measured by dissolving the polymer in water at a concentration of 0.5 mg/mL. The UV absorbance of the water solution was determined prior to adding an equal volume of octanol (C_initial). The solutions were stirred overnight and allowed to separate into two layers (water and octanol phases). The final polymer concentrations in the aqueous phase were quantified using a Gensys 150 UV–vis spectrophotometer (Thermo Fisher Scientifc; Waltham, MA) in a 100 mm path length quartz cuvette from 190 to 500 nm. The log P values for each polymer were calculated using the following equation, eq 4:

$$\text{log}P=\text{log}\left(\frac{C_{\text{oct}}}{C_{\text{water}}}\right)=\text{log}\left(\frac{C_{\text{initial}}-C_{\text{water}}}{C_{\text{water}}}\right)$$ (4)

where C_oct, C_water, and C_initial are the concentrations of polymer in the octanol and water phases and the initial polymer concentration in water, respectively, determined from UV–vis.⁶⁴ UV absorbance values were determined from the peak absorbance value at wavelength between 280 and 340 nm, after a baseline correction using the high-wavelength absorbance data (440–500 nm). Raw absorbance spectra are given in Figure S8.

Polyplex Formation.

Polymer stock solutions were prepared by dissolving the polymer into UltraPure water to obtain 15.15 nmol of protonatable amine groups per milliliter (nmol of N/mL). Polymer stock solutions were sterile-filtered using 0.22 μm hydrophilic filters. The plasmid (pDNA), pZsgreen N-1 (4708 bp), was purchased from Aldevron (Fargo, ND). Polyplexes were prepared by adding diluted polymer stock (diluted to obtain the targeted N/P ratio) to equal volumes of diluted pDNA stock (diluted to obtain the final desired pDNA concentration). The solutions were gently pipet-mixed and incubated at room temperature for 45 min before characterization.

Polyplex Characterization.

Dynamic Light Scattering and ζ-Potential.

DLS and ζ-potential analyses were performed using a Brookhaven NanoBook-90 Plus PALS (Brookhaven Instruments; Holtsville, NY). For DLS, ζ-potential and PicoGreen assays, the polyplexes were prepared at a pDNA concentration of 50 ng/μL and then further diluted with water to 10 ng/μL. Three measurements for DLS and ζ-potential were averaged for the final reported values. DLS intensity plots for polyplexes in water are given in Figure S10.

PicoGreen Assay.

PicoGreen assays were performed to assess polymer–pDNA binding. PicoGreen was diluted with UltraPure water according to the manufacturer’s recommendation (200×). Polyplexes were diluted to 10 ng/μL, and then mixed with equal volumes of the diluted PicoGreen. Then, polyplexes were incubated with PicoGreen for 30 min in a black 96-well plate on a shaker at 50 rpm. Three wells were designated for each polyplex sample and three wells of uncomplexed pDNA were prepared as controls. Fluorescence was measured in a microplate reader (Synergy H1, BioTek; Winooski, VT) with monochromator settings of excitation/emission 485/520 nm and 70 gains. The three sample measurements were normalized to the uncomplexed pDNA control and averaged.

Static Light Scattering.

The weight-average number of pDNA copies per polyplex $\left(\overline{N}\right)$ was determined via SLS using a Wyatt DAWN 18-angle, light scattering detector following previous work..⁶⁵ Briefly, the polyplexes were prepared in water at an N/P ratio of 5 and a pDNA concentration of 80 ng/μL. The polyplexes were diluted to 4–5 different pDNA concentrations between 5 and 40 ng/μL. Each sample was injected at a flow rate of 0.1 mL/min for 5–10 min until a stable signal was obtained from the light scattering detectors. The polyplex dn/dc values, which are necessary for accurate SLS measurements, were calculated from a weight average of the polymer dn/dc (displayed in Figures S3–S7) and the pDNA dn/dc (0.2600), obtained by using Wyatt Astra 8.2.0 software. Zimm plots were generated using Wyatt Astra 8.2.0 software and used to calculate $\overline{N}$, the weight-average pDNA copy per polyplex (eqs 5 and 6):

$$m_{\text{bound}}=\gamma \times \frac{C_{\text{polymer}}}{C_{\text{pDNA}}}\times M_{\text{pDNA}}$$ (5)

$$\overline{N}=\frac{{\overline{M}}_{\text{polyplex}}}{m_{\text{bound}}+M_{\text{pDNA}}}$$ (6)

where m_bound is the mass of pDNA-bound polymer, γ is the mole fraction of pDNA-bound polymer (set to 1/5; assuming an effective N/P ratio of 1, which is consistent with the results from ITC, and the polyplexes were formed at an N/P ratio of 5), C_polymer and C_pDNA are the polymer and pDNA concentrations used in the polyplex formation, M_pDNA is the molecular weight of the pDNA, and ${\overline{M}}_{\text{polyplex}}$ is the molar mass of the polyplex determined from SLS.

Circular Dichroism.

CD spectra were obtained at 25°C using a JASCO 815 CD spectrometer (Easton, MD) and a 1 mm path length quartz cuvette as described earlier.⁶⁶ Polyplexes were prepared as previously described at a pDNA concentration of 100 μg/mL and an N/P ratio of 5. Three scans were averaged at wavelengths from 200 to 350 nm with a scan rate of 100 nm/min and a resolution of 1 nm. The CD spectra of the uncomplexed polymer showed no significant absorbance in this wavelength range, indicating alterations in the CD signal were due to changes in the pDNA conformation (Figure S11A). CD spectra for the homopolymers are also displayed in Figure S11B.

Isothermal Titration Calorimetry.

Polymer–pDNA binding was measured at 25°C in UltraPure water using a TA Instruments Nano ITC instrument (New Castle, DE). The sample cell was filled with 300 μL of pDNA at a concentration of 0.3 mM phosphate groups. For the injection, 50 μL of polymer solution was prepared at a concentration of 5.5 mM with respect to nitrogen groups (to obtain a final N/P ratio of 3 once the polymer was fully injected). The polymers were injected in 2.46 μL increments with 200 s between each injection. Data were analyzed with TA Instruments Nano-Analyze software using two independent sites models. Raw heat thermograms are given in Figure S12A.

Transmission Electron Microscopy.

Polyplex morphology was visualized using an FEI Talos F200X TEM (Thermo Fisher Scientific, Waltham, MA), operated at 200 kV and equipped with a field emission gun (FEG). Polyplexes were prepared at an N/P ratio of 5 and a pDNA concentration of 50 ng/μL. The polyplex samples (2 μL) were deposited on copper coated Formvar grids (200 mesh) and incubated for 1 min. The grids were washed in UltraPure water four times and dipped in 2% uranyl acetate for 30 s. To remove the excess stain, the grids were washed four more times with UltraPure water. The grids were then dried in a vacuum desiccator overnight. After imaging, micrographs were processed with ImageJ to determine polyplex sizes. The polyplex radii of 20–40 individual polyplexes were measured using ImageJ and averaged to get R_TEM.

Effect of Human Serum on Polyplex Characteristics.

Human serum was separated from whole human blood without the addition of an anticoagulant. The serum was further diluted to 10% v/v with sterile 1× PBS. The polyplexes were prepared at a pDNA concentration of 50 ng/μL. After 45 min incubation, equal volumes of the 10% serum were added to the polyplexes, mixed, and incubated for an additional 60 min at room temperature. Samples were further diluted with UltraPure water to reach a pDNA concentration of 2.5 ng/μL. DLS and PicoGreen measurements were performed as previously described. DLS intensity plots in 10% serum are given in Figure S13.

Cellular Assays.

Cell Culture.

HEK293T cells (human embryonic kidney cells) were purchased from American Type Culture Collection (ATCC; Manassas, VA) and grown in DMEM (Dulbecco’s modified Eagle medium) supplemented with 10% fetal bovine serum (FBS) and 1% antibiotic-antimycotic. The cells were passaged when they reached 80% confluency; confluency and morphology were visualized using an inverted light microscope, an EVOSTM XL Core Imaging System (Thermo Fisher Scientific; Waltham, MA).

pDNA Delivery.

24 h before transfection, HEK293T cells were seeded in 24-well plates at a density of 50 000 cells per well. The cells were incubated at 37°C, 5% CO₂, and 90–95% humidity. Polyplexes were formulated at N/P ratios of 1, 5, 7.5, and 10 as previously described, and pDNA dosing was fixed at 1 μg/well. One volume of polyplexes was diluted with two volumes of reduced serum media (OptiMEM) and added to the cells. Lipofectamine 2000 (LPF2000) and jetPEI were used according to manufacturers’ protocol. After the addition of polyplexes, the cells were incubated for 4 h (37°C, 5% CO₂, and 90–95% humidity). Then, 1 mL of DMEM was added to each well and incubated for an additional 24 h. The next day, the medium was replaced with 1 mL of fresh DMEM and incubated for 48 h (37°C, 5% CO₂, 90–95% humidity). The cells were trypsinized and centrifuged at 1000 rpm for 5 min to collect the cell pellets. The cell pellets were washed with 1× PBS and resuspended in a 1× PBS solution containing Calcein Red viability dye (400 nM) and FBS (2%). The cells were filtered and transferred to 5 mL FACS tubes. GFP expression was measured using flow cytometry with autofluorescence correction applied (488 nm laser line, Cytek Northern Lights spectral flow cytometer; Fremont, CA). Flow cytometry data were analyzed using FCS Express; gating strategies are given in Figure S16.

Cell Viability Assays.

The cytotoxicity of the polyplexes was evaluated 48 h post-transfection. First, the medium was carefully aspirated and replaced with a solution of CCK-8 (Cell Counting Kit-8) in FBS-supplemented FluoroBrite. The cells were incubated for 4 h (37°C, 5% CO₂, 90–95% humidity). Then, 100 μL of the supernatant was transferred to clear-bottom 96-well plates. Absorbance measurements were conducted using a Synergy H1 microplate reader (BioTek, Winooski, VT) set to detect absorbance at 460 and 720 nm. Absorbance values at 720 nm were subtracted from those at 460 nm to eliminate the influence of the air bubbles and background noise.

Cellular Uptake.

Polyplex internalization was quantified using Cy5-tagged pDNA. Briefly, pDNA was labeled using the Label IT Nucleic Acid Labeling Kit Cy5 (Mirus, Madison, WI). The Cy5-tagged pDNA was purified using ethanol precipitation, and the concentration was quantified using spectrophotometry. Cells were seeded in 24-well plates the day before pDNA delivery. The day after pDNA delivery, the cells were trypsinized, centrifuged to form cell pellets, and washed successively with 1× PBS and cell scrub buffer (Genlantis; San Diego, CA) to remove extracellularly bound polyplexes. After washing with 1× PBS a second time, the cells were resuspended in 1× PBS with 2% FBS and filtered into FACS tubes. The geometric mean of the Cy5 fluorescence intensity was measured via flow cytometry. Gating strategies are given in (Figure S17).

Widefield Fluorescence Imaging.

Coverslips were placed in 24-well plates 24 h before cell seeding, coated with gelatin (1 mg/mL), and UV sterilized overnight. HEK293T cells were seeded on the sterilized, precoated coverslips (50 000 cells per well) and incubated for 24 h (37°C, 5% CO₂, 90–95% humidity). Transfection was performed as described in the previous sections. After 48 h, the cells were washed with 1× PBS. Hoechst 3342 was added to the cells and incubated for 15 min. The cells were washed twice with ice cold 1× PBS and fixed with 4% v/v formaldehyde for 30 min at room temperature. Once fixation was completed, the cells were washed thrice with ice cold 1× PBS. The coverslips were carefully removed from the well plates and mounted on glass slides using a ProLong Glass Antifade Mountant (Thermo Fisher Scientific; Waltham, MA). The edges of the coverslips were covered with clear nail polish and dried at room temperature for 2 h. GFP expression was observed using a Leica STELLARIS 5 Confocal Microscope (Deerfield, IL) with an LIA chroic laser supply unit. Images were captured using a 40× water immersion objective lens. 405 and 488 nm solid state lasers with HyD spectral detectors were used.

Confocal Imaging.

Quantitative confocal imaging was conducted using a Nikon A1R confocal laser scanning microscope (Melville, NY). Sample preparation and transfection (using Cy5-tagged pDNA) were performed as described above. Then, the cells were stained with Hoechst 3342 and CytoFix Red Lysosomal Stain for 20 min at 37°C. After incubation, the cells were washed twice with 1× PBS and fixed with 4% v/v formaldehyde. The coverslips were washed three times with 1× PBS and then twice with water and mounted on glass slides as previously described. The images were captured using the following laser settings: GFP (488 nm laser, Chroma ET525/50m emission filter), Hoescht 3342 (405 nm laser, Chroma ET450/50m emission filter), Lysobrite Red (561 nm laser, Chroma 600/50m emission filter), and Cy5 (640 nm laser, Chroma ET685/70m emission filter). Images were captured using an oil objective lens (60×, 1.4 NA Plan Apo Lambda).

The images were analyzed using Imaris software (Bitplane, version 10.1.1). The Cell module in Imaris was used to identify the nucleus and cytoplasm of each cell. The nuclei were identified using the DAPI channel, and the cell bodies were identified using the EGFP channel. The images were smoothed with a 1 μm filter and background subtracted with a 4 μm filter to improve the segmentation results. The individual nuclei were then manually corrected as needed to separate the touching nuclei. Both the nuclei and cell bodies were exported as separated surfaces.

To identify polyplex puncta, we used the Spot module in Imaris on the Cy5 channel with an estimated spot diameter of 1 μm. The spots were then filtered visually by “quality” and maximum intensity. To identify spots within individual cells, the resulting spots were filtered by distance to the cell body surface. Similarly, to identify spots within individual nuclei, the spots within individual cells were furthered filtered by distance to the nuclear surfaces. The Imaris Coloc module was used to compute the colocalization of the TRITC and Cy5 signals. The module generates a colocalization score by first computing an intensity histogram of each pixel in the TRITC and Cy5 channels. We used the module’s automatic thresholding algorithm to identify a region of interest. Briefly, the algorithm works by incrementally reducing the threshold value until the correlation reaches zero to define a region of interest (ROI). The module then computed the Pearson’s correlation coefficient (PCC) within this ROI. We note that for several images, the automatic thresholding algorithm failed, likely due to low signal-to-background. We were unable to obtain a PCC value for these images (Table S1).

Hemolysis Measurements.

Human blood was collected in ethylenediaminetetraacetic acid (EDTA)-coated vacutainers. Red blood cells (RBCs) were separated from whole blood via centrifugation at 1500 rpm for 5 min. The supernatant was discarded, while the sedimented RBC pellets were isolated and washed three times with 150 mM saline with 5 min of centrifugation at 1500 rpm between each wash. Washed RBCs were resuspended in 1× PBS and diluted 10×. For each assay, 100 μL of the polyplex solution was added to 100 μL of the resuspended RBC suspension. The total volume was adjusted to 1 mL with 1× PBS. Both negative and positive controls were prepared using the same RBC dilution. The negative control consisted of RBCs dispersed in 1× PBS and the positive control consisted of RBCs suspended in water. All samples, including controls, were incubated for 1 h at 37°C and then centrifuged at 1500 rpm for 5 min. From the supernatant, 100 μL was collected and placed in a clear-bottomed 96-well plate. Absorbance was measured at 540 nm in a microplate reader (Synergy H1; BioTek, Winooski, VT), and the hemolysis percentage was calculated using eq 7:

$$\%\text{Hemolysis}=\frac{A_{\text{Sample}}-A_{\text{PBS}}}{A_{\text{Water}}-A_{\text{PBS}}}\times 100$$ (7)

Complement Activation Assay.

To evaluate complement activation, polyplexes were prepared at N/P ratios of 5 and 10 as previously described. Human serum samples (from different donors) were collected and stored at −20°C. Before experimentation, the human serum samples were thawed to room temperature and gently pipet-mixed. For the polyplex samples, 5 μL of the polyplex solution was mixed with 45 μL of human serum. For the negative control, 5 μL of a diluted pDNA solution was mixed with 45 μL of the human serum. For the positive control, 2.5 μL of zymosan (10 mg/mL) was mixed with 47.5 μL of human serum. All samples were pipet-mixed and incubated for 60 min at 37°C with gentle shaking. After incubation, the samples were further diluted with a specimen diluent (1:40 for SC5b-9 and 1:70 for C 4d). Diluted samples were analyzed for SC5b-9 and C 4d protein content using the SC5b-9 Plus ELISA kit and C 4d Plus ELISA kit (Quidel; San Diego, CA), following the manufacturer’s protocol. ELISA absorbance values were converted to protein content (ng/mL) using a calibration curve.

Statistical Analysis.

All experiments were performed in triplicate. Data are presented as the mean ± standard deviation. All statistical analyses were performed using Python. Tukey’s honestly significant difference (HSD) test was used to perform pairwise comparison of sample means. Statistical significance was defined as a p-value less than 0.05. Statistical analyses were performed to determine the significance of the observed differences in the transfection efficiency, cell viability, hemolysis, complement activation, and cellular uptake across the different polyplex formulations.

RESULTS AND DISCUSSION

Statistical copolymers of lipophilic, cationic monomer 2-(diisopropylamino)ethyl methacrylate (DIP) and hydrophilic, neutral monomer 2-hydroxyethyl methacrylate (HEMA) exhibited promising pDNA delivery characteristics but formed polyplexes that aggregated severely in salt- and serum-containing media, hindering progression to in vivo studies.⁶⁷ We hypothesized that modulating the spatial arrangement of lipophilic DIP cationic repeat units will impart colloidal stability in salt- and serum-containing media, while conserving intracellular pDNA delivery capacity. To tune gradient steepness, we programmed comonomer addition rates during semibatch reversible addition–fragmentation chain-transfer (RAFT) polymerization. Afterward, we compared polymer–pDNA binding interactions, polyplex aggregation propensity in serum, pDNA delivery performance, and hemocompatibility across statistical, block, and gradient copolymers.

The Spatial Distribution of Lipophilic Cations Dictates Polymer pK_a and the Hydrophobic–Hydrophilic Phase Balance.

We synthesized block (B), statistical (S), and gradient copolymers (G1–G3) via RAFT (Figure 1A), reasoning that RAFT will facilitate precise control over the molecular weight distribution, dispersity, and spatial distribution of repeat units. Reactivity ratios (R_DIP = 1.449 and R_HEMA = 1.132, Figure S1C) predicted that, despite a slight preference for DIP, propagating copolymers add DIP and HEMA monomers with similar probability.⁶⁸ Since DIP and HEMA had comparable reactivity ratios (≈1), we reasoned that the addition rate of comonomers during semibatch polymerization would tune gradient steepness.⁵⁶ Thus, to synthesize gradient copolymers, the same amount of DIP monomer was gradually introduced to a polymerizing HEMA reaction mass over three different injection times (30, 90, and 180 min for G1, G2, and G3, respectively, Figure S1D). Decreasing the addition rate of DIP realized copolymers with decreasing gradient steepness (from G1 to G3).

Figure 1.

We synthesized length- and compositionally equivalent copolymers with varied spatial distributions of lipophilic cations. We measured the resulting differences in polymer pK_a and hydrophobic–hydrophilic phase balance (log P). (A) Five polymers, statistical–S, block–B, and gradient—G1, G2, and G3, were polymerized from cationic (DIP) and neutral (HEMA) monomers. (B) The ν parameter from 1D NOESY quantifies the likelihood that DIP repeat units lie adjacent to HEMA. B and block-like gradient copolymers, G1, exhibited the lowest ν values. Cross peaks were the most intense in S and G3. (C) Hydrophobicity (log P) and the ν parameter were sensitive to the spatial distribution of DIP, with B exhibiting the highest octanol–water coefficienct (log P, which parametrizes hydrophobicity). B also had the lowest pK_a due to the close proximity of the DIP repeat units. ^aM_n and dispersity (Đ) from SEC-MALS. ^bMole percent DIP from ¹H NMR. ^cpK_a from titration. ^dPolymer ζ-potential from electrophoresis. ^elog P from UV–vis. ^fν from 1D NOESY. Created with Biorender.com.

Furthermore, we exploited the comparable reactivity ratios of DIP and HEMA to obtain statistical copolymers, S, via one-pot batch polymerization. To realize block copolymers, B, we first synthesized DIP macromolecular chain-transfer agents (macroCTAs) and then chain-extended macroCTAs with HEMA. All five copolymers (S, B, G1, G2, and G3) were characterized via size exclusion chromatography with multiangle light scattering (SEC-MALS) and ¹H nuclear magnetic resonance spectroscopy (¹H NMR) to determine the molecular weight distribution and composition (see Figure S21 for ¹H NMR spectra), respectively (Figure 1C). All five copolymers contained between 40–46 DIP repeat units and 41–48 HEMA repeat units; we obtained length- and composition-matched copolymers albeit with systematically varied spatial distribution of cationic repeat units (DIP). While B had the highest dispersity (1.25), likely due to the two-step synthesis required, the dispersity values in these five copolymers indicated fidelity to RAFT polymerization kinetics. Despite altering comonomer addition schemes to modulate the spatial distribution of DIP repeat units, we obtained low dispersities, consistent polymer lengths, and comparable incorporation of DIP throughout (Figure 1C).

pK_a measurements revealed that the spatial distribution of DIP repeat units impacts polycation protonation equilibria, which in turn impacts pDNA binding propensity.⁶⁹ DIP repeat units are clustered together in B, suppressing polycation protonation^70,71 whereas DIP repeat units are interspersed with neutral HEMA repeat units in S and G1–G3, which have higher pK_a values (titration curves are given in Figures S3–S7).

To capture and quantify variations in the spatial distribution of DIP repeat units across S, B, and G1–G3, we performed one-dimensional ¹H–¹H nuclear Overhauser effect spectroscopy (1D NOESY). By exciting protons from DIP (at 3.4 ppm), we collected nuclear Overhauser effect signals from protons within a 4–5 Å proximity to the excited protons.^62,72 The ν parameter is the ratio of peak intensities arising from HEMA protons (4.0 ppm) neighboring excited DIP protons (3.4 ppm) to the total peak intensities from DIP and HEMA (eq 8):

$$\nu =\frac{I_{\text{HEMA}}}{I_{\text{HEMA}}+I_{\text{DIP}}}\times 100\text{\%}$$ (8)

where I_DIP and I_HEMA are the integration values for the HEMA and DIP peaks highlighted in Figure 1B. We normalized ν values to S (ν_S = 100%) (Figure 1A). This ratio estimates the abundance of polymer segments where DIP and HEMA appear successively, capturing the local compositional heterogeneity within the copolymer segments (Figure 1B). Given the comparable reactivity ratios of DIP and HEMA, DIP repeat units in statistical copolymers are equally likely to be found adjacent to HEMA or DIP repeat units. In this scenario, ν values will be higher. However, if the spatial distribution of DIP resembles block copolymers, we should observe lower ν values since DIP and HEMA repeat units will be spatially segregated. Based on ν estimates, we concluded that DIP repeat units form blocky segments in G1 but distribute more statistically among HEMA repeat units in G3.

To strengthen conclusions from 1D NOESY, we measured the octanol–water partition coefficient (log P). Log P measures the tendency of polymers to leave the aqueous phase and partition into an organic phase (usually octanol) and parametrizes polymer lipophilicity. We expect the spatial distribution of the lipophilic cation DIP (which bears bulky hydrophobic isopropyl substituents) to modulate log P. Furthermore, protonated DIP repeat units will be less lipophilic than deprotonated ones. Polymers with evenly scattered DIP repeat units (such as S), will be protonated to a greater extent and therefore, less lipophilic. Whereas largely contiguous DIP segments (such as B) exhibit lower protonation degrees (pK_a of 6.2 for Bversus 6.6 for), intensifying lipophilicity (and the log P). Hence, log P variations provide further evidence of differences in the spatial distribution of DIP repeat units across our five copolymers. Among gradient copolymers, G1 had the highest log P while G3 did not partition into octanol (exhibiting the highest hydrophilicity). Overall, log P measurements confirmed findings from 1D NOESY: G3 has the most statistical arrangement of DIP repeat units with the highest ν and lowest log P among gradient copolymers, whereas G1 has the most block-like distribution, with the lowest ν and highest log P (Figure 1C).

Having synthesized and characterized length- and composition-matched copolymers with varied spatial distributions of DIP, we then asked how the microstructure impacts pDNA loading per polyplex, polyplex size, and morphology.

Statistical, Block, and Gradient Copolymers Form Polyplexes of Comparable Size But Differentially Load pDNA.

Polyplexes with high pDNA loading may enjoy superior transfection efficiencies,^21,73 thanks to higher nuclear import of pDNA.^74,75 However, high pDNA loading per polyplex tends to balloon polyplex size,⁶⁵ which in turn may exacerbate intracellular delivery challenges such as polyplex internalization. Additionally, transfection efficiencies can be highly dependent on the timing of polyplex delivery, cell cycle, and transfection reagents, so higher pDNA loadings do not explicitly correlate with higher transfection efficiencies.^76,77 Our goal was to tune the spatial distribution of DIP such that we could decouple pDNA loading per polyplex from polyplex size (hydrodynamic radii, R_h). To study how microstructural contrasts impacted polyplex size and composition, we performed dynamic light scattering (DLS) to measure polyplex size, transmission electron microscopy (TEM) to visualize polyplex morphology, and static light scattering (SLS) to compute the number of pDNA molecules per polyplex.

SLS enabled Zimm analyses (Figure 2) of polyplexes formed from B, S, and G1–G3. First, we calculated the polyplex weight-averaged molecular weights from the y-intercept of the Zimm plot. From this molecular weight, we calculated the number of pDNA copies per polyplex (Figure 2).⁶⁵ B polyplexes loaded the most pDNA molecules and G3 the lowest. Among gradient copolymers, the block-like G1 loaded the most pDNA. This result mirrors observations from 1D NOESY and log P which suggested that DIP spatial distribution in G1 was most similar to B. The second virial coefficient, A₂ parametrizes polyplex–water interactions. Only B exhibited negative A₂ values, indicating pronounced hydrophobicity (consistent with log P data in Figure 1C).

Figure 2.

Polyplexes formed from S, B, and G1–G3 exhibited near-identical hydrodynamic radii but varied pDNA loading levels per polyplex. (A) Weight-averaged molecular weights (Zimm analyses) estimated the number of pDNA molecules per polyplex. (B) All five copolymers formed spherical assemblies with pDNA, with polyplex cores sized around 40 nm. Scale bars are 500 and 100 nm for the images and insets, respectively. (C) B loaded the most pDNA per polyplex and also exhibited a negative second virial coefficient A₂, indicating unfavorable polyplex–solvent interactions. While G1 polyplexes mirrored the pDNA binding behavior of B, G3 emulated S polyplexes by exhibiting the lowest pDNA loading per polyplex. ^aPolyplex radius from TEM. ^bHydrodynamic radius from DLS. ^cPolyplex ζ-potential from electrophoresis. ^dPolyplex molecular weight, radius of gyration, pDNA loading, and A₂ from SLS. All displayed values are for an N/P ratio of 5. R_h values for other N/P ratios showed similar results and are displayed in Figure S9.

Irrespective of the copolymer microstructure, polyplexes were sized around 50 nm. This result was consistent across DLS, SLS, and TEM (Figure 2). TEM micrographs (Figure 2) revealed spherical and relatively monodisperse polyplexes for all five polymers. All five groups of polyplexes exhibited comparable radii of gyration (R_g) and hydrodynamic radii (R_h). Collectively, our results demonstrate that we can modulate pDNA loading per polyplex while maintaining a consistent polyplex size merely by controlling the spatial distribution of cationic repeat units. We attribute this finding to two factors: (1) the spatial organization of cationic repeat units dictates pDNA conformational changes during polyplex formation, and (2) polymer–pDNA complexation thermodynamics are microstructure-dependent. To test this, we performed circular dichroism (CD) spectroscopy to map pDNA structural changes and isothermal titration calorimetry (ITC) to compare polymer–pDNA complexation thermodynamic parameters across microstructures.

pDNA Conformation in Polyplexes and Polymer–pDNA Binding Thermodynamics Are Both Microstructure-Dependent.

CD and ITC elucidated how the spatial arrangement of DIP shaped polymer–pDNA binding interactions. CD (Figure 3B), which detects helicity changes during polycation-mediated pDNA condensation, revealed conformational differences among pDNA in the five groups of polyplexes. Free pDNA (not polymer-bound, negative control) had positive and negative bands at 280 and 250 nm, respectively, consistent with the native B-type helical pDNA conformations. Interestingly, the CD spectra of S and G1–G3 were indistinguishable from each other but markedly different from free pDNA. Compared to free pDNA, the spectra for these polyplexes shifted right, and their molar ellipticity increased. These spectral changes suggest that pDNA transforms from native helical B-type conformations into C-type structures in S and G1–G3 polyplexes.⁷⁸ Unlike with other copolymers, the molar ellipticity did not increase as dramatically when pDNA formed polyplexes with B. This suggests that B copolymers rapidly neutralize anionic charges in polymer–pDNA binding pairs without forcing pDNA to adopt non-native conformations. SLS indicated that B loads the most pDNA whereas CD spectra reveal that B does not perturb native pDNA conformations as dramatically as S and G1–G3. These attributes of B are likely related; the high density of cationic repeat units augments the pDNA loading without distorting pDNA conformations.

Figure 3.

Spatial distribution of the cationic repeat unit governs pDNA conformation and polymer–pDNA complexation thermodynamics. (A) The dissociation constant (K_d) describes polycation–pDNA binding affinity. (B) B distorted the native pDNA helical conformations to a lesser extent than did S and G1–G3. (C) Wiseman plots were fitted to ITC data (points) using two independent site models (lines). Thermodynamic binding parameters are tabulated in (D). B has the highest pDNA binding affinity (lowest values of K_d1), whereas S and G1–G3 exhibited weaker pDNA binding. Created with Biorender.com.

Next, in Figure 3D, we tabulated results for the first binding event, while thermodynamic parameters describing the second binding event are given in Figure S12B.

The enthalpic components of ITC results are associated with intermolecular drivers of complexation such as electrostatic and hydrophobic interactions but are unhelpful in understanding microstructural differences.⁷⁹ All five polyplexes were formed under endothermic (positive ΔH) conditions, confirming that the polymer–pDNA complexation process is entropically driven. This result is consistent with the literature on polyelectrolyte complexes (PECs), which posits that the release of counterions and water molecules drives PEC formation.^80,81 However, the analysis of TΔS is convoluted by contributions from multiple factors: dehydration of deprotonated hydrophobic DIP repeat units and electrostatic polymer–pDNA neutralization which release water molecules and counterions, increasing entropy.⁸² We attribute the large TΔS of B to the release of water molecules and counterions from protonated DIP segments after binding pDNA. Interestingly, S showed the highest change in the entropy of all of the microstructures despite having a far lower hydrophobicity (log P) than B. However, S has a higher pK_a than B; consequently, DIP repeat units in S are protonated to a greater extent, liberating more counterions upon pDNA complexation. Despite the multiple confounding factors (electrostatic interactions, polycation protonation degree, and hydrophobicity) in polymer–pDNA complexation, we established that electrostatic and hydrophobic contributions to ΔH and TΔS are governed by the spatial distribution of DIP.⁸³

The pDNA binding affinities (K_d, Figure 3A) of S and G1–G3 were comparable. However, the K_d of B was far lower than those of other microstructures, suggesting that polymer–pDNA binding was strongest in B. The clustered arrangement of DIP repeat units increases the density of cationic DIP groups in B, amplifying binding interactions between pDNA and DIP segments, and allowing B to electrostatically neutralize pDNA without distorting native pDNA conformations. In contrast, pDNA molecules must adopt non-native conformations to interact with DIP cations that are scattered widely between neutral HEMA repeat units in S and G1–G3. Overall, B exhibited pDNA binding profiles distinct from those of S and G1–G3. CD and ITC revealed that B bound to pDNA strongly without perturbing native pDNA conformations. Varying the spatial distribution of cationic repeat units elicited diverse pDNA binding profiles; we expected that this will modify the biological performance of the polyplexes.

Block-Like Gradient Copolymers Quell Polyplex Aggregation in Serum-Containing Media.

Although all polyplexes exhibited comparable sizes in water (Figure 2), we expected microstructural differences to dictate serum-induced changes in polyplex size and pDNA encapsulation efficiency. Hence, we performed DLS, TEM, and PicoGreen assays to evaluate polyplex colloidal stability and pDNA encapsulation efficiency in serum-containing media. Incubation in human serum triggered polyplex aggregation for S and G3 across all N/P ratios (except at the lowest N/P ratio of 1), whereas G1 and G2 aggregated only at the highest N/P ratio of 10 (Figure 4A,B). Intensity-weighted DLS histograms are shown in Figure S13. We suspect that block-like microstructures (B, G1, and G2) form polyplexes with core–shell morphologies, with cores consisting of pDNA-bound DIP-rich polymer segments and hydrophilic HEMA-rich coronae.,^31,40,84 HEMA-rich polyplex shells may inhibit nonspecific adsorption of serum proteins through hydration repulsion.^37,85,86 TEM (Figure 4B) confirmed aggregation trends from DLS (Figure 4A).

Figure 4.

G1 emulates B in overcoming polyplex colloidal instability in serum-containing media. (A) G1 and G2 resisted aggregation at lower N/P ratios (1, 5, and 7.5), while S and G3 formed serum-stable polyplexes only at the lowest N/P ratios. (B) TEM confirmed polyplex aggregation in serum (N/P of 5). Scale bars are 2 μm.

We tracked the competitive displacement of pDNA by anionic serum proteins via dye exclusion assays. Here, pDNA release is accompanied by an increase in the fluorescence intensity. All five polyplexes exhibited lower pDNA encapsulation efficiency in serum (Figure 4A) than in water (Figure S9). However, B suffered the highest level of serum-triggered pDNA release. Here, excess polycations are hindered from associating with polyplexes due to the high charge density of B and the resultant electrostatic repulsion between polyplexes and free polycations. Consequently, pDNA is more accessible to serum proteins and PicoGreen molecules within the B polyplexes. Other polyplexes (S and G1–G3) may sequester excess polymers (due to their lower charge density), which retard serum-triggered pDNA displacement and dye intercalation. Additionally, G1–G3 and S alter the pDNA native conformation to a greater extent (Figure 3B) than does B, which may further frustrate dye intercalation between nucleobases. Unlike S, G1 formed serum-stable polyplexes that resisted aggregation. Among gradient copolymers, G1 most closely resembles microstructure B. However, unlike B, G1 avoided serum-triggered pDNA release. These results indicate that the spatial distribution of cations is a powerful, under-utilized design handle to augment polyplex colloidal stability and pDNA encapsulation efficiency in serum-containing media.

G1 Exhibits the Best pDNA Delivery Performance among Gradient Copolymers.

Earlier, Correia³⁹ reported that block and gradient copolymers were less cytotoxic than statistical copolymers, but they did not evaluate intracellular pDNA delivery. For the first time, we report a head-to-head comparison of the pDNA delivery efficiency across block, statistical, and gradient copolymers. Flow cytometry was used to measure green fluorescent protein (GFP) expression in HEK293T cells after polyplexes (formed with GFP-encoding pDNA). Across all four N/P ratios tested, S exhibited the highest transfection efficiency (69 ± 1%, N/P = 10) and the lowest cell viability (26 ± 1%, N/P = 10) (Figure 5A). B elicited near-zero levels of GFP expression but were well-tolerated by cells (85 ± 2% cell viability even at the highest N/P ratio tested). Interestingly, the transfection efficiencies of gradient copolymers (G1–G3) were slightly lower than that of S, spanning a range between 48% and 64% (N/P = 10). The cell viability (27–30%) of gradient copolymers was comparable to that of S. Among all the gradient copolymers studies, G1 achieved the highest transfection efficiency (64 ± 1%, N/P = 10) and cell viability (29 ± 1%, N/P = 10). GFP expression trends from flow cytometry were qualitatively verified via wide field fluorescence imaging (Figure 5C, additional micrographs are in Figure S14).

Figure 5.

We compared pDNA delivery efficiency and cellular viability across the five microstructures via flow cytometry. (A) S achieved the highest transfection efficiency and lowest cell viability whereas B exhibited the opposite trend. G1–G3 exhibited transfection efficiencies and a cell viability intermediate between S and B. (B) Pareto plots visualized trade-offs between transfection efficiency and cytotoxicity. T × V is the product of transfection efficiency (T) and viability (V). G1 achieved the highest cell viability and transfection efficiency among gradient copolymers. (C) Representative fluorescence micrographs of GFP expression mirrored flow cytometric measurements. Scale bars are 40 μm.

S achieved high transfection efficiency but at the cost of cytotoxicity, whereas the opposite trend prevailed in block copolymers. Polymers wherein DIP is distributed along the entire length of the backbone (S) interact with cell membranes more frequently than polymers featuring a segregated DIP distribution (B). Consequently, S disrupts cell membranes to a greater extent than does B. The cytotoxicity and efficiency profiles of G1–G3 were intermediate between S and B, suggesting the existence of a Pareto frontier.⁸⁷ Pareto analysis visualized trade-offs between transfection efficiency and cytotoxicity (Figure 5B). T × V, the product of transfection efficiency (T) and cell viability (V), compares the population of GFP-expressing viable cells across all polyplexes (five polymers, four N/P ratios). The bottom right region of the Pareto plot represents high efficiency low viability polyplexes (predominantly S), whereas the top left corner is occupied by low efficiency high viability polyplexes (mostly from B). Interestingly, G3 (gradient copolymers with the highest DIP–HEMA proximity) polyplexes occupy the center of the plot between these two extremes, whereas G1 (where DIP repeat units tend to neighbor other DIP units rather than HEMA) achieved the highest T × V value among all gradient copolymers. The spatial arrangement of lipophilic cations in gradient copolymers determines Pareto-optimality when considering toxicity–efficiency trade-offs.

The block-like G1, which loads more pDNA per polyplex than S (Figure 2), achieved transfection efficiency comparable to that of S without triggering severe aggregation in serum and emerged as the most promising gradient copolymer in our investigation. To better understand these transfection results, we investigated how the spatial distribution of DIP impacts polyplex internalization and intracellular fate.

Polyplex Uptake Is Highest in Statistical And Lowest in Block But Moderate among Gradient Copolymers; Intracellular Fate Is Unaffected by the Microstructure.

We labeled pDNA payloads with Cy5 fluorophores and compared cellular uptake across polymers via flow cytometric measurements of the Cy5 intensity (Figure 6A). Polyplex uptake was highest for S, followed by those of G1, G2, and G3. B, which did not mediate any transgene expression, exhibited the lowest cellular uptake. The inability of B to shuttle pDNA payloads across cell membranes may explain its subpar transfection efficiency. Interestingly, G1, where the spatial arrangement of DIP is the most segregated among gradient copolymers, formed polyplexes that were internalized more efficiently than G2 and G3. This trend in polyplex internalization (G1 > G2 > G3) mirrors transfection performance trends. Our cellular uptake studies identify internalization as the bottleneck limiting the transfection performance of gradient copolymers. Subtle changes in gradient steepness impact polyplex–membrane associations, culminating in cellular uptake. We speculated that similar interactions between polymers and endosomal membranes steer polyplex intracellular trajectories and nuclear import. Hence, we compared the intracellular localization of polyplexes in various cellular compartments via quantitative confocal microscopy (Figure 6B).

Figure 6.

We compared polyplex cellular uptake and intracellular distribution via flow cytometry and confocal microscopy, respectively. (A) pDNA was labeled with Cy5 tags before polyplex formation for transfection. Subsequently, we measured the geometric median of Cy5 fluorescence intensity to quantify polyplex uptake via flow cytometry. S had the highest uptake followed by G1, G2, G3, and B. (B) Representative confocal micrographs with GFP (green), nuclei (blue), lysosomes (yellow), and Cy5 (red) channels highlighted. Scale bars are 50 μm. (C,D) We enumerated internalized polyplexes and calculated their partitioning between nuclear and cytoplasmic regions. Scale bars are 15 μm.

Consistent with our flow cytometric findings, we could not detect any internalized polyplexes for B. For S and G1–G3, we enumerated the number of internalized polyplexes per cell and classified each polyplex as nuclear and cytoplasmic (Figure 6C). Polyplexes distributed between nuclear and cytoplasmic volumes in an identical fashion for all four polymers, with 20–40% of polyplexes occupying nuclear regions irrespective of the polymer microstructure (Figure 6D). Next, we quantified the likelihood that polyplexes occupy lysosomal compartments by calculating Pearson’s correlation coefficient (PCC), which quantifies the colocalization of fluorescent signals from Cy5-tagged polyplexes and lysosomal markers. We did not observe significant differences in PCC across the microstructures (Table S11). However, labeling pDNA with hydrophobic Cy5 may have interfered with polyplex intracellular distribution. In the future, label-free methods^88–91 will be explored. After studying the impact of polymer microstructure on cellular uptake and intracellular polyplex distribution, we asked whether the microstructure governed hemocompatibility and complement activation.

Polycation-Triggered Hemolysis Is Microstructure-Sensitive; Complement Activation Is Not.

To understand how variations in the spatial distribution of DIP impact hemocompatibility, we measured the absorbance of hemoglobin liberated by polycation-triggered red blood cell (RBC) lysis (Figure 7A). S triggered severe RBC lysis, particularly at higher N/P ratios of 7.5 and 10 (Figure 7B,C). In S, the dispersed arrangement of DIP along the polymer backbone intensifies interactions with the RBC membranes and promotes RBC lysis. In contrast, B and G1–G3 induced ≈40–50% less RBC lysis than did S. Distributing DIP along a gradient rather than uniformly along the polymer backbone frustrates contact between polymers and RBC membranes and attenuates RBC lysis.^38,63 Hemolysis measurements for uncomplexed (free of pDNA) polymers (Figure S15A,B) follow similar trends with S causing the most RBC lysis.

Figure 7.

RBC lysis is microstructure-dependent, whereas complement activation is insensitive to microstructural differences. (A) Schematic overview of RBC lysis and complement activation assays. (B,C) S induced the highest levels of red blood cell lysis and B the least; hemolysis levels were low across G1–G3. The pink coloration of supernatant provides a visual indicator of hemolysis extent. (D) Enzyme-linked immunosorbent assay (ELISA) measurements of SC5b-9 and C4d revealed that complement activation was comparable to unbound pDNA and independent of polymer microstructure. Created with Biorender.com.

To correlate the polymer microstructure with polyplex-mediated complement activation, we performed SC5b-9 and C4d assays.⁹² SC5b-9 measures the formation of membrane attack complexes, while C4d estimates the activation of the classical and lectin-mediated pathways.^93–95 Complement activation was detectable in both SC5b-9 and C4d assays for unbound pDNA controls (Figure 7D). This agrees with prior observations on the inherent immunogenicity of pDNA and the tendency of C4d-binding proteins to associate with pDNA.^96,97 Complement activation was slightly lower in Lipofectamine 2000 than in JetPEI and our copolymers. Lipoplexes encapsulate pDNA in their interior compartments, where pDNA cannot interact with serum proteins, which may explain their lower complement activation. For SC5b-9, we initially anticipated that polyplexes formed at higher N/P ratios (where excess unbound polymers are abundant) will be more activating than those at lower N/P ratios (which contain fewer polymers). Our rationale for this was that free polymers might induce the formation of membrane attack complexes. However, microstructure-based differences in both SC5b-9 and C4d were insignificant, even when we repeated these assays with free polymers without added pDNA rather than polyplexes (Figure S15C). Neither the N/P ratio nor polymer microstructure proved influential in our complement activation studies. We concluded that polymer microstructural variations do not contribute substantially to complement activation.

CONCLUSIONS

In this work, we exploited microstructural differences across length-matched and compositionally identical statistical (S), block (B), and gradient copolymers (G1–G3) to augment the physical properties and transfection performance of polymer–pDNA complexes (polyplexes). Extensive physicochemical characterization revealed microstructure-dependent differences in pDNA complexation pathways, pDNA loading per polyplex, and interactions with serum proteins. We varied the spatial distribution of lipophilic cations (DIP) by tuning the monomer addition rate during semibatch RAFT polymerization to access copolymers with modulated gradient steepness. Through 1D NOESY and octanol–water parition coefficient (log P) measurements, we visualized differences in the spatial distribution of lipophilic cations in the block-like (G1) and the more statistical (G3) gradient copolymers. B and G1 loaded more pDNA per polyplex than S and G3, while maintaining similar radii of gyration, hydrodynamic radii, and morphology. B perturbed native pDNA conformations the least, while S and G1–G3 transformed the native B-type pDNA helical structures to C-type. ITC determined that B had the lowest K_d (0.7 μM), approximately 3–4× lower than the other microstructures, indicating that polymer–pDNA binding is strongest in B. K_d values were similar across the other four microstructures (≈ 2.5 μM), indicating minimal differences in binding strength between S and G1–G3. In serum-containing media, S and G3 exhibited severe aggregation, even at low N/P ratios. However, “blockier” gradient copolymers (G1 and G2) aggregated only at higher N/P ratios of 10. Like B, G1 loaded high amounts of pDNA per polyplex, avoided aggregation in serum-containing media, and was well-tolerated by red blood cells. Tailoring the spatial distribution of lipophilic cations in gradient copolymers is an effective strategy to traverse the Pareto frontier and navigate toxicity–efficiency trade-offs. G1 mediated higher transgene expression than the other two gradient copolymers without triggering a concomitant increase in cytotoxicity. In future work, we will employ continuous flow polymerization to further fine-tune the spatial arrangement of lipophilic cations and augment transfection performance in clinically relevant cellular targets.