High-throughput evaluation of polymeric nanoparticles for tissue-targeted gene expression using barcoded plasmid DNA

Abstract

Successful systemic gene delivery requires specific tissue targeting as well as efficient intracellular transfection. Increasingly, research laboratories are fabricating libraries of novel nanoparticles, engineering both new biomaterial structures and composition ratios of multicomponent systems. Yet, methods for screening gene delivery vehicles directly in vivo are often low-throughout, limiting the number of candidate nanoparticles that can be investigated. Here, we report a comprehensive, high-throughput method to evaluate a library of polymeric nanoparticles in vivo for tissue-specific gene delivery. The method involves pairing each nanoparticle formulation with a plasmid DNA (pDNA) that harbors a unique nucleotide sequence serving as the identifying “barcode”. Using real time quantitative PCR (qPCR) for detection of the barcoded pDNA and quantitative reverse transcription PCR (RT-qPCR) for transcribed barcoded mRNA, we can quantify accumulation and transfection in tissues of interest. The barcode pDNA and primers were designed with sufficient sensitivity and specificity to evaluate multiple nanoparticle formulations per mouse, improving screening efficiency. Using this platform, we evaluated the biodistribution and transfection of 8 intravenously administered poly(beta-amino ester) (PBAE) nanoparticle formulations, each with a PBAE polymer of differential structure. Significant levels of nanoparticle accumulation and gene transfection were observed mainly in organs involved in clearance, including spleen, liver, and kidneys. Interestingly, higher levels of transfection of select organs did not necessarily correlate with higher levels of tissue accumulation, highlighting the importance of directly measuring in vivo transfection efficiency as the key barcoded parameter in gene delivery vector optimization. To validate this method, nanoparticle formulations were used individually for luciferase pDNA delivery in vivo. The distribution of luciferase expression in tissues matched the transfection analysis by the barcode qPCR method, confirming that this platform can be used to accurately evaluate systemic gene delivery.

1. Introduction

Gene therapy using exogenous DNA or RNAi is a powerful tool with the potential to address and even cure the underlying genetic origin of disease. However, efficient delivery of the genetic cargo across biological barriers remains a major bottleneck to successful gene therapy [1]. Multiple sequential obstacles exist from the formulation of delivery vehicle for injection to the expression of a target protein in cells [2, 3]. At the systemic level, delivery vehicles must protect the genetic cargo from enzymatic degradation in biological fluid, circumvent clearance by the renal and mononuclear phagocyte system (MPS), and efficiently accumulate in the target tissue. At the cellular level, barriers include cell-specific targeting, cellular uptake, endosomal escape, and cargo release [4].

Non-viral nanoparticle (NP) delivery systems can be engineered to overcome these specific challenges of delivery, while avoiding the design constraints and safety concerns of viral vectors [5–7]. Cellular uptake and endosomal escape can be tuned by modulating the vector’s chemical structure and buffering capacity, among other factors [8–10]. Cell and tissue level active targeting is possible through conjugation targeting ligands, including peptides or small molecules [11]. Passive tissue targeting modification is enhanced by coating NPs with polyethylene glycol (PEG), but PEG may simultaneously prevent cellular uptake [12, 13]. Multifunctional NPs may be designed to sequentially overcome delivery barriers, but they are complex to develop, have interactive effects that are difficult to predict, and are challenging and costly to manufacture. Further, a vector optimized for cellular transfection in vitro may not show efficacy in vivo due to systemic factors, such as non-specific protein adsorption, that change the physical and chemical properties of the carriers [14]. For these reasons, rational design of NPs to overcome all in vivo delivery barriers is a major challenge.

An alternative approach is to screen large libraries of biomaterials for transfection in the target cell or tissue type. This approach has been used broadly, resulting in large libraries of polymer and lipid biomaterials for nucleic acid delivery [15–18]. Poly(beta-amino ester) (PBAE) is a class of biodegradable ionizable polymer that electrostatically complexes with negatively charged nucleic acids to form NPs [19]. Monomer units with distinct structures may be combined during synthesis to create a library of polymers with differential structures and properties, which have been used to successfully transfect multiple types of cells in vitro and in vivo [20–25]. Our group has previously performed high-throughput in vitro screening to demonstrate structure-function relationship of PBAE polymer at the cellular level, including the effect of monomer and end-group structures on cell-type specificity, cellular uptake, and endosomal escape [8, 10, 20, 26, 27]. While ligand- or PEG-modified PBAE NPs have been developed for enhanced tissue targeting [28–30] or penetration [13, 21, 31], respectively, the significance of differential molecular changes to PBAE chemical structure on biodistribution and transfection in vivo is still unclear.

Screening a wide range of biomaterials for tissue-specific in vivo transfection within the context of the entire gene delivery process requires high throughput methods, as in vivo study of multiple vectors is cost- and time-ineffective. Recently, high-throughput in vivo screening methods have been developed to study biodistribution of mRNA-loaded lipid NPs in major organs using “genetic barcode” [32–34]. In this approach, NPs with distinct lipid compositions are used to encapsulate a unique, secondary nucleotide barcode sequence in addition to the siRNA or mRNA payload. This barcode technology enables sensitive quantification of each nanoparticle biodistribution in vivo by deep sequencing methods. However, unlike the previously reported methods that measure the accumulation of the oligonucleotide cargo, a DNA barcode system enables quantification of both plasmid DNA (pDNA) accumulation and its transcribed mRNA expression levels as an elegant way to screen multiple NP formulations, within the same animal, for in vivo transfection to a target tissue. Moreover, by using the actual pDNA cargo as the barcode sequence, the N/P ratio of barcoded NP remains unaltered in comparison to the non-barcoded NP with the pDNA of interest, thereby minimizing potential change in NP’s physicochemical properties.

In this study, we develop a high-throughput method for screening in vivo gene delivery efficiency of a library of PBAE NPs using a single pDNA that harbors a barcode sequence. Both biodistribution and transfection can be analyzed by detecting the primary barcode and the transcribed mRNA of the barcode through qPCR and RT-qPCR analysis, respectively. Also, by inserting the barcode sequence in the pDNA, we eliminate any potential change to the nanoparticle’s physicochemical properties from the addition of a secondary barcode payload. Using the platform, we evaluate specific structures of PBAE polymers and their accumulation and transfection of pDNA in major organs following systemic administration.

2. Materials and methods

2.1. Materials

1,4-butanediol diacrylate (B4) (Alfa Aesar, Ward Hill, MA), 1,5-pentanediol diacrylate (B5) (Monomer Polymer & Dajac Labs), 3-amino-1-propanol (S3) (Alfa Aesar), 4-amino-1-pentanol (S4) (Alfa Aesar), 5-amino-1-pentanol (S5) (Alfa Aesar), 2-methylpentane-1,5-diamine (E4) (TCI America, Portland, OR), 2-(3-aminopropylamino)ethanol (E6) (Sigma-Aldrich, St. Louis, MO), and 1-(3-aminopropyl)-4-methylpiperazine (E7) (Alfa Aesar) were purchased and used as received. pEGFP-N1 was purchased from Elim Biopharmaceuticals and amplified by Aldevron (Fargo, ND). pfLuc Luciferase-pcDNA3 was a gift from William Kaelin (Addgene plasmid # 18964) and amplified by Aldevron (Fargo, ND). Purelink Genomic DNA Extraction kit (Thermo Fisher Scientific), PowerUp SYBR Green Master Mix (Applied Biosystems, Foster City, CA), Label IT®-Tracker™ Cy™3 and Cy™5 kit (Mirus Bio LLC, Madison, WI) were obtained from commercial vendors and used per manufacturer’s instructions.

2.2. Polymer synthesis

PBAE polymers are synthesized via a two-step Michael addition reaction as previously described (Figure 1A). First, acrylate-terminated base polymer is created by reacting a diacrylate monomer with a primary amine-containing side chain monomer at a stoichiometric molar ratio of 1.2:1 for 24 h at 90 °C. Then, the base polymer is reacted with 20-fold excess molar amount of primary amine-containing end-capping molecule in THF for 3 h at room temperature. The final polymer is ether-precipitated and stored in DMSO at 100 mg/mL in −20 °C. For this study, a total of 8 PBAE polymers were synthesized using 2 different diacrylate monomers, 1,4-butanediol-diacrylate (B4) and 1,5-pentanediol-diacrylate (B5), 3 side chain monomers, 3-amino-1-propanol (S3), 4-amino-1-butanol (S4), and 5-amino-1-pentanol (S5), and 3 end-capping molecules, 2-methylpentane-1,5-diamine (E4), 2-(3-aminopropylamino)ethanol (E6), and 1-(3-aminopropyl)-4-methylpiperazine (E7) (Figure 1B). Nomenclature for the final polymer follows the label of each monomer used. For example, a polymer synthesized with B5, S3, and E6 is named 536. The molecular weights of the 8 PBAE polymers were determined by gel permeation chromatography (GPC) on a high-performance liquid chromatography (HPLC) system (Agilent) equipped with UV-Vis (Agilent), refractive index (RI) (Wyatt Technology), and Multi-Angle Light Scattering (Wyatt Technology) detectors. Molecular weight information (Mn, Mw, and PDI) was determined for each polymer using a specific refractive index (dn/dc) of 0.06, which was calculated by injecting a known polymer mass, assuming 100% recovery (Table S1).

Figure 1. PBAE polymer synthesis.

(a) PBAE polymer was synthesized using a two-step Michael addition reaction. Acrylate-terminated PBAE base polymer, synthesized by reacting a diacrylate monomer with a primary amine-containing alkanolamine monomer in the first step, is end-capped during the second step with a primary amine-containing molecule. (b) Monomers used in the synthesis of 8 different PBAE polymers.

2.3. Plasmid preparation

A total of five plasmids were used, three encoding for fluorescent proteins and two encoding random nucleotide sequences. In order to ensure that all plasmids were similar in size and had the same backbone, pEGFP-N1 was used as the base plasmid and genes encoding other fluorescent proteins, mOrange and iRFP, or two random sequences were cloned in to replace EGFP gene. The two random plasmids were designed as a proof-of-concept that more plasmids with unique random barcode sequences can be used in a similar high-throughput study. These five plasmids are referred to as pDNA A (GFP), B (mOrange), C (iRFP), D (Noncoding 1), and E (Noncoding 2). All gene sequences can be found in Table S2.

For two plasmids with random noncoding sequences, two random 1500–base pair sequences with 25% fraction each of A, C, G, and T nucleotides were generated using the Aarhus University, Bioinformatics Research Centre, Denmark online tool (URL - http://users-birc.au.dk/~palle/php/fabox/random_sequence_generator.php) Sequences for mOrange and iRFP genes were acquired from Addgene. Restriction cloning sites for NheI and HindIII were encoded on the 5’ and 3’ ends of the DNA fragments, respectively. The double-stranded linear DNAs were custom ordered using gBlock Gene Fragments (IDT, Inc, Skokie, IL). Plasmids were cloned into the pN3 backbone using restriction digestion by NheI-HF and HindIII-HF restriction enzymes (New England Biolabs, Ipswitch, MA), following the manufacturer’s instructions. Purification of DNA fragments was performed by gel electrophoresis and extraction using the QIAquick Gel Extraction Kit (Qiagen, Hilden, Germany), following the manufacturer’s instructions. DNA ligation was performed by combining 100 ng of digested backbone DNA with 700 ng of digested insert, then performing an overnight ligation at 16 °C with T4 DNA ligase (New England Biolabs). Ligated cloned were selected for and amplified using 5-alpha Competent E. coli (New England Biolabs) and Qiagen Maxi Kits. mOrange and iRFP plasmids were also amplified by Aldevron. All plasmids were stored at 1 mg/mL in sterile water at 4 °C.

2.4. Primer optimization

In order to generate forward and reverse primers specific to each pDNA and avoid non-specific amplification of other plasmids or murine genomic DNA, Basic Local Alignment Search Tool (BLAST) from the National Institute of Health’s National Center for Biotechnology Information was used to extensively check for homology. First, Primer-BLAST was used to generate 50 primer candidates for each plasmid that match the conditions listed in Figure S1A. Then, the primer candidates for each plasmid were used as the input query sequence in BLASTn to check for homology against other plasmids as well as the mouse genome (Figure S1B). Primer candidates that showed undesired homology (matching sequence) in the last 6 base pairs or in more than 10 base pairs total were excluded (Figure S1C). Lastly, the top-scoring candidate for each plasmid was checked for hairpin, self-dimerization and hetero-dimerization using IDT OligoAnalyzer 3.1. The final primer sequences were custom ordered (IDT, Inc, Skokie, IL), and stored as 3 μM aliquots in −20 °C. All primer sequences can be found in Table S3.

Primer sequences were experimentally checked for their specificity toward the corresponding plasmids by quantitative real time polymerase chain reaction (qPCR) and gel electrophoresis. Briefly, 100 ng of each plasmid was amplified against each of the five primer pairs to determine C_T values and generate the melt curve. 2 μL of pDNA, 2 μL of 3 μM forward primer, 2 μL of 3 μM reverse primer, and 14 μL of PowerUp SYBR Green Master Mix solutions were mixed for qPCR amplification. PCR reaction consisted of initial polymerase activation stage at 95 °C for 2 min, followed by 40 cycles of denaturation at 95 °C for 15 sec, annealing at 55 °C for 15 sec, and elongation at 72 °C for 1 min. After PCR amplification, the 25 qPCR products along with DNA ladder were also run through gel electrophoresis in 1% agarose gel.

2.5. Validation of qPCR protocol using fresh tissue lysates spiked with pDNA

A 5–7 week old female BALB/c mouse was euthanized and major organs – liver, kidneys, spleen, lungs, and heart – were harvested. Organs were washed with 1X PBS three times, cut into small pieces with a razor blade, and minced between the frosted ends of two microscope slides. Then, 10 mg of liver sample and 5 mg of samples from all other organs were separately placed into 1.5mL tubes. Tissues were digested using digestion solution provided in the Purelink Genomic DNA Extraction kit. Following the digestion, 10 ng to 10 pg of each pDNA were spiked into digested tissue. Subsequent steps of DNA extraction were followed as instructed and the final purified DNA was then diluted with water ten-fold for liver samples and two-fold for all other organ samples. The same qPCR reaction protocol was followed as described above.

2.6. Nanoparticle formulation and characterization

NPs were formed by bulk mixing of PBAE polymer and pDNA in aqueous conditions to allow electrostatic interaction and particle self-assembly. PBAE polymer in DMSO at 100 mg/mL and pDNA in water at 1 mg/mL were diluted to 12.5 mg/mL and 0.5 mg/mL, respectively, using 25 mM sodium acetate (NaAc) buffer at pH 5.0. Equal volume of polymer and DNA solutions were mixed and incubated for 10 min to form NPs. This ensures the mass ratio of polymer to DNA to be consistent at 25 w/w across the different NPs evaluated.

Hydrodynamic diameter was measured by dynamic light scattering, and zeta potential was measured by electrophoretic mobility using the Zetasizer Pro (Malvern Instruments, Malvern, UK). Immediately before these measurements, NPs were diluted 1:10 in a 1:1 ratio of 25 mM sodium acetate and 1X PBS to a final volume of 1 mL. Encapsulation efficiency of pDNA in NPs was assessed by gel electrophoresis. NP prepared at 25 w/w with 0.03 mg/mL pDNA was combined with 6X Gel Loading Dye with no SDS. Samples were loaded into a 0.8% agarose gel containing 1 μg/mL ethidium bromide. Free unencapsulated pDNA at the same dose was used as a positive control. The gel was run with 90 V applied for 30 min and visualized by UV exposure.

2.7. Föster resonance energy transfer (FRET) between Cy3- and Cy5-labeled NPs

FRET analysis was conducted to investigate any intermixing between different NPs co-dispersed in a single solution, as would occur during a co-injection. pEGFP DNA was labeled with Cy3 and Cy5 fluorophores following a protocol by Wilson et al [35]. Labeling density was measured using a determined using a NanoDrop spectrophotometer (ThermoFisher Scientific, Waltham, MA). Three different pairs of NP were tested: PBAE 447 with Cy3-labeled pEGFP DNA and PBAE 457 with Cy5-labeled pEGFP DNA, PBAE 457 with Cy3-labeled pEGFP DNA and PBAE 536 with Cy5-labeled pEGFP DNA, and PBAE 536 with Cy3-labeled pEGFP DNA and PBAE 447 with Cy5-labeled pEGFP DNA. NPs were prepared separately and incubated together at 1:1 v/v ratio with gentle pipette mixing, and peak emission intensities of Cy3 (565–570 nm) and Cy5 (665–675 nm) was read following Cy3 excitation at 540 nm using spectrofluorophotometry (Shimadzu RF-5301). A positive control NP batch was formulated by complexing either PBAE polymer 447, 536 or 457 with 1:1 mixture of Cy3-labeled and Cy5-labeled DNAs.

2.8. In vitro transfection with mixture of PBAE NPs

HepG2 cells were purchased from ATCC (Manassas, VA), cultured in Minimum Essential Media (MEM) supplemented with 10% FBS, 100 U/mL Penicillin, 100 μg/mL Streptomycin, 100 μM of MEM non-essential amino acids solution, and 1 mM of sodium pyruvate, and maintained in a humidified incubator at 37°C with 5% CO_2. Cells were seeded in a 96-well plate at 10,000 cells / well and allowed to adhere for 24 hr. For in vitro screening (Fig S6), PBAE NPs were freshly prepared with equal amounts of each DNA barcode plasmid (A-E) at 25 w/w for a final DNA concentration of 0.3 μg/μL. For Figure S8, PBAE NPs were synthesized with pDNA A or a mixture of three NPs with pDNA A, D, or E at the same total pDNA dose each for a final DNA concentration of 0.3 or 0.1 μg/μL. 20 μL NPs were added to cells and allowed to incubate for 2 hours at 37^oC. RNA was isolated and reverse transcribed using a Cells-to-CT kit from ThermoFisher Scientific (Waltham, MA) according to the manufacturer’s instructions. For qPCR analysis, 6 μL of sample DNA, 2 μL of 3 μM forward primer, 2 μL of 3 μM reverse primer, and 10 μL of PowerUp SYBR Green Master Mix (Applied Biosystems, Foster City, CA) solutions were mixed for qPCR amplification. qPCR was performed using a StepOnePlus™ Real-Time PCR System (Applied Biosystems, Foster City, CA) with the cycling parameters described above. Threshold and baseline values were standardized across all samples and all runs to ensure accurate comparison. The comparative C_T method was used to quantify relative expression levels [36]. Barcode amplification was normalized to the housekeeping gene GAPDH to quantify NP accumulation of each PBAE with each barcode relative to the genomic DNA content. Then, this value was normalized to non-specific background amplification in untreated cells, by subtracting the ΔC_T of amplification in untreated animals from animals which received barcoded NPs, thereby obtaining ΔΔC_T.

$$\mathrm{\Delta }\mathrm{\Delta }{\mathrm{C}}_{\mathrm{T}}=({\mathrm{C}}_{\text{Tbarcode}\text{x}}-{\mathrm{C}}_{\mathrm{TGAPDH}}){}_{\mathrm{treated}}-({\mathrm{C}}_{\text{Tbarcode}\text{x}}-{\mathrm{C}}_{\mathrm{TGAPDH}}{)}_{\mathrm{untreated}}$$

2.9. High-throughput biodistribution analysis

All in vivo procedures were approved and overseen by the Johns Hopkins Institutional Animal Care and Use Committee (IACUC). Biodistribution of 8 different PBAE polymers was tested in 5–7 week old female BALB/c mice (The Jackson Laboratory, Bar Harbor, ME). For each mouse, a single 200 μL cocktail solution containing five different PBAE NPs was prepared for tail-vein injection. Each of the five NPs was formulated by pairing one of 8 PBAE polymers with one of the five pDNAs to serve as the identifier barcode (Scheme 1). Each NP formulation contained 10 μg of pDNA for a total of 50 μg pDNA in the cocktail solution. Biodistribution of PBAE 447 NPs from the high-throughput samples was compared against that from additional mice injected only with NPs prepared from PBAE polymer 447 and 50 μg pEGFP DNA to validate accuracy of the high-throughput method. Mice were sacrificed 30 min post injection for DNA extraction. Major organs – liver, kidneys, spleen, lungs, and heart – were harvested, washed with PBS three times, cut into small pieces with a razor blade, and minced between the frosted ends of two microscope slides. Then, approximately 10 mg of liver sample and 5 mg of samples from all other organs were separately placed into 1.5 mL tubes, and pDNA was extracted from minced tissues following the manufacturer’s instruction from the Purelink Genomic DNA Extraction kit (Invitrogen, Waltham, MA). Once DNA was purified through the extraction column, it was diluted with water ten-fold for liver samples and three-fold for all other organ samples. The differences in the amount of minced tissues used to extract pDNA between organs was normalized by the total amount of genomic DNA content during qPCR analysis using the comparative C_T method and GAPDH as a housekeeping gene.

Scheme 1.

High-throughput screening of PBAE NP biodistribution and transfection via qPCR and RT-qPCR.

2.10. High-throughput in vivo transfection analysis

5–7 week old female BALB/c mice (The Jackson Laboratory) were injected via tail-vein with the same cocktail solutions as the biodistribution study (Scheme 1). Mice were sacrificed 6 hr post injection for RNA extraction. Organs were harvested and washed in 1X PBS, then cut into small pieces using a razor and minced between frosted microscope slides. 50 mg of each sample was suspended in Trizol and homogenized. A 20% volume of chloroform was added, and tubes were vortexed briefly to emulsify. Samples were incubated at room temperature for 5 minutes, then centrifuged at 12,000 g for 15 minutes at 4 °C. The aqueous phase was transferred to a new tube, then isoproponal was added at 50 % of the original Trizol volume. Samples were inverted to mix, then incubated for 10 minutes at room temperature to allow for RNA precipitation. Samples were centrifuged at 12,000 g for 30 minutes at 4 °C. Isopropanol was decanted, and pelleted RNA was washed with 75% ethanol. After vortexing, samples were centrifuged at 7,500g for 5 minutes at 4 °C. Ethanol was carefully decanted, and the RNA pellet was air dried for 5–10 minutes. The dried pellet was resuspended in 30 μL water, and RNA concentration was determined using a NanoDrop spectrophotometer (ThermoFisher Scientific, Waltham, MA). RNA was purified of DNA using the TURBO DNA-free kit (Invitrogen, Carlsbad, CA) according to manufacturer’s instructions. Using the High Capacity RNA-to-cDNA™ kit (Invitrogen, Carlsbad, CA), 20 μg of RNA was reverse transcribed according to the manufacturer’s instructions. Reverse transcription was performed at 37 °C for 1 hour, followed by 5 minutes at 95 °C to stop the reaction. PCR reactions were comprised of 10 μL PowerUp SYBR Green Master Mix (Applied Biosystems, Foster City, CA), 2 μL forward primer (3 μM), 2 μL reverse primer (3 μM), and 1 μL cDNA. Baseline and thresholds were standardized across all samples. ΔΔC_T values were calculated as previously described. If there was no amplification for a given sample primer combination, the sample was assigned a C_T of 40, corresponding to the maximum amplification cycle, in order to compute ΔΔC_T value.

2.11. In vivo transfection using pfLuc

To form luciferase NPs for in vivo use, pfLuc DNA was complexed with PBAE 456, 536, and 546 at a 25 w/w ratio in 25 mM sodium acetate (pH = 5.0) for a final DNA concentration of 0.25 μg/μL. 200 μL of NPs (50 μg DNA) was injected via the tail vein in 5–7 week old BALB/c mice. After 30 minutes, 1 hour, 3 hours, or 6 hours, animals were injected intraperitoneally with 150 mg/kg D-luciferin (Gold Biotechnology, St. Louis, MO). After 10 minutes, the animals were anesthetized using isoflurane and bioluminescence was imaged using the IVIS Spectrum (Xenogen, Alameda, CA). Images were analyzed using Living Image® 4.7.3 software (PerkinElmer, Waltham, MA).

2.12. Statistical analysis

All statistical analysis was performed with GraphPad Prism 8 software package. For in vivo experiments, statistically significant barcode amplification was calculated by one-tailed Student’s t-test between the experimental −ΔΔC_T values and zero. Statistical significance was defined as p < 0.05 after correcting for multiple comparisons using the Holm-Sidak method.

3. Results

3.1. Primers’ specificity to DNA barcodes

Primers were selected via BLAST with the most conservative conditions to ensure highly selective binding to barcoded plasmid used in the study. Three of the five plasmids used in the study each contained an insert serving as the barcode, which can be detected directly as protein or in its transcribed mRNA form (Scheme 1).

The forward and reverse primer pairs for each barcode gene were designed to produce an approximately 100-basepair long amplicon. Every possible combination of the primer pairs and the plasmids were mixed for qPCR reaction to confirm the primers’ specificity to the corresponding plasmids. As shown in Figure 2A, only conditions with correctly matched primers and plasmid resulted in amplification. Also, each set of primers run either alone or mixed with other primer sets did not yield any false positive PCR results from self- or hetero-dimerization. The melt curves also showed a clean single peak only for correctly matched conditions, which indicates that there is no off-target amplification occurring in the PCR reaction (Figure S2). qPCR of fresh tissue lysates from all major organs of a BALB/c mouse showed no amplification from the primers, indicating no off-target amplification of genomic DNA. PCR-amplified products were run on gel electrophoresis to confirm that the amplicons had the expected length, based on our primer design. The gel image shows bands appearing for primer-plasmid matching conditions only at the height level of approximately 100-basepair mark on the DNA ladder (Figure 2B). These observations validate the specificity of each primer set to its corresponding plasmid.

Figure 2. Primer optimization.

(a) C_T values from qPCR reaction of all possible combinations of 5 plasmids and 5 corresponding primers (n=3, mean ± SEM). (b) Gel electrophoresis of PCR-amplified product, showing specific amplicons’ size of approximately 100 base-pairs. For each plasmid, there are 6 conditions with specific primers for pDNA A, B, C, D, E, or no primers going from left to right, as indicated by the index numbers 1–6, respectively. DNA ladder: 1 kbp left and 100 bp right. (c) C_T values from qPCR reaction of tissue lysates from major organs spiked with known concentrations of 5 barcoded plasmid DNAs (n = 4, mean ± SEM)

We also evaluated whether the C_T value from qPCR varies linearly with pDNA mass. A range of pDNA mass from 10 pg to 10 ng was spiked into fresh tissue lysates and subjected to DNA extraction for qPCR. As shown in Figure 2C, each of the 5 barcode plasmids has a linear correlation with R² > 0.96 between its mass and C_T values in the 5 major organs tested. All C_T values collected in the experiments and used in the analysis were within the linear region of these curves. Amplification of each barcode plasmids was not affected by the type of tissue lysates, as demonstrated by the overlapping standard curves in each panel of Figure 2C. Interestingly, however, amplification of 5 barcode plasmids showed varying degree of sensitivity in each type of tissue lysates, as demonstrated by the non-overlapping standard curves in each panel of Figure S3. This requires each unique barcode plasmid to be tested for linearity between C_T value and a range of plasmid mass. In this study, each PBAE NP formulation is tested with 5 different barcode plasmids, and the biodistribution and in vivo transfection data are averaged to mitigate the small difference in sensitivity across barcode plasmids.

3.2. Nanoparticle characterization

PBAE NPs formulated with 8 different PBAE polymers were characterized based on hydrodynamic diameter and surface charge. As shown in Figure 3A, PBAE NP hydrodynamic diameter measured by dynamic light scattering ranged from 100 – 200 nm for all NP formulations. All NPs showed positive zeta potential between 15 – 22 mV, due to the exposure of positively charged polymer on the surface. While positive charge of NPs could cause toxicity, previous literature on PBAE NPs with similar positive surface charge report minimal toxicity both in vitro and in vivo [20, 37]. Also, all NPs showed 100% encapsulation of pDNA, with no evidence that barcode sequence or polymer structure significantly affects encapsulation efficiency (Figure S4).

Figure 3. PBAE NP properties.

(a) Hydrodynamic diameter and zeta potential of 8 PBAE NPs, measured by dynamic light scattering (n = 3, mean ± SEM). (b) Cy3 and (c) Cy5 emission following Cy3 excitation of a solution consisting of either 447 PBAE NP with Cy3-labeled pDNA, 457 PBAE NP with Cy5-labeled pDNA, 447 PBAE NP with a mixture of Cy3- and Cy5-labeled pDNAs, or a mixture of 447 PBAE NP with Cy3-labeled pDNA and 457 PBAE NP with Cy5-labeled pDNA (left to right).

3.3. FRET analysis showing the absence of nanoparticle intermixing

The high-throughput in vivo screening method is based on different PBAE NPs being injected as a cocktail solution into a single animal and subsequently individually identified in tissue lysates. While inorganic nanoparticles (such as gold NPs) and lipid-based NPs (such as liposomes) form discrete NPs, it is conceivable that polyplex NPs, which are formed from molecular interactions between polyelectrolytes of opposing charge, could intermix components together. Such potential intermixing of PBAE polyplexes has not been previously investigated and could conceivably prevent the barcodes from staying matched to a specific NP. Each PBAE NP formulation is formulated with one barcode pDNA, which serves for NP identification. Thus, the capability of PBAE polyplex NPs to resist exchange of DNA cargo in a cocktail solution ensures that mixed NPs encapsulate only their original defining barcode plasmids. To evaluate this, we formulated two separate batches of NPs encapsulating pDNAs labeled with either Cy3 or Cy5 fluorophores, then mixed the two batches into a single solution. Exchange of Cy3-labeled DNA from a NP with Cy5-labeled DNA from another NP in the mixture brings the fluorophores in close proximity to each other within a NP and causes emission of a FRET signal. Negative control NPs were separately prepared with each of Cy3- or Cy5-labeled DNA alone and positive control NPs were prepared with a 1:1 mixture Cy3- and Cy5-labeled DNA within the same NPs.

Each NP sample was excited at the excitation wavelength for Cy3 (540 nm), and emission was measured at the emission wavelength for Cy3 (565–575 nm) and Cy5 (665–675 nm), shown in Figures 3B/C and S5. Cy3 NPs alone showed emission at 565–575 nm, and Cy5 NPs alone showed no emission, because there is minimal excitation of this fluorophore at 540 nm. The positive control NPs showed FRET activity, with 35% decreased signal at 565–575 nm (Cy3 emission) in comparison to Cy3 NP alone, and an increase in signal of >3 RFU at 665–675 nm (Cy5 emission) in comparison to Cy5 NP alone (~1 RFU). This suggests that part of Cy3 emission was able to excite neighboring Cy5-pDNA encapsulated in the same nanoparticle. In the test case, where NPs were formulated separately then combined, minimal FRET activity was observed as indicated by the similar signal intensity at 565–575 nm compared to Cy3 NPs alone. The signal intensity at 665–675 nm for the test case was similar to the negative control at ~1 RFU. Given the high labeling density of Cy3 and Cy5 DNAs (1 fluorophore per 63.4 base pairs and 54.9 base pairs respectively) and high density of plasmid DNA in PBAE NPs reported in previous work [38], we determined that there was not substantial intermixing of plasmid DNAs between PBAE NPs. While these results do not exclude the possibility of small amounts of exchange, the lack of FRET signal in a cocktail solution indicates that the Cy3 and Cy5 fluorophores mostly maintain separation in distinct NPs. Therefore, we can correlate accumulation of unique barcode DNA with its corresponding unique NP.

3.4. In vitro transfection of PBAE NPs with mixed barcoded pDNAs

HepG2 cells were transfected in vitro with 8 PBAE NPs with 120 ng each of barcoded pDNAs A, B, C, D, and E. After 48 hours, qRT-PCR results show significant differences in transfection between NP formulations, with PBAE 447 as the most effective formulation for transfection and PBAE 546 as the least effective (Figure S6). To evaluate differential expression between the barcodes, the percentage of variation from the mean was calculated for each barcode (Figure S7). Barcode A and B had significantly increased expression over C, D, and E within the same NP formulations. This indicates a difference in expression which is sequence dependent. To control for any bias from these differences, we formulated each PBAE NP with each barcode in subsequent studies. This redundancy ensured that variation in biodistribution, transfection efficiency, PCR amplification efficiency, transcript half-life, or plasmid immunogenicity due to the difference in barcode sequences would not bias the results.

To verify that screening results from mixed barcode formulations are representative of transfection results from a single pDNA, NPs were formulated at either a full dose of pDNA A or a mixture of 1/3 dose each of pDNA A, D, or E. RT-qPCR results showed a decrease in −ΔΔCT of Barcode A mRNA from 9.1 for the single, full dose NP to 7.4 for the mix, 1/3 dose NPs, corresponding to a 1/3 decrease in mRNA concentration (Figure S8). This indicates that RT-qPCR results from mixed barcode NPs are predictive of transfection with a full dose NP. Interestingly, decreasing the total DNA concentration by 1/3 from 0.03 μg/μL to 0.01 μg/μL does not result in a similar decrease in transfection measured by RT-qPCR, suggesting that changing the overall DNA concentration may affect transfection levels.

3.5. High-throughput screening of tissue targeting

Utilizing five barcoded pDNAs and their specific primers, we developed a scheme to test biodistribution of five PBAE NP formulations per animal. In each animal, all PBAEs were paired with a different barcode to distinguish the polymers from one another. Further, each PBAE NP was tested separately in 5 animals, and it was paired with a different barcode in each replicate, as shown in Table 1. NPs were freshly prepared by separately combining barcode DNA and PBAE for each formulation, then mixing the five formulations for each animal immediately prior to injection, to minimize exchange of pDNA between NP formulations. The NPs were administered intravenously via the tail vein in BALB/c mice.

Table 1. Barcode NP administration scheme.

Each animal (1–8) listed in the row title was injected with a mixture of 5 NPs, where each NP is comprised one of 5 barcode pDNAs represented by a letter A, B, C, D, or E and a PBAE polymer listed in the column title.

30 minutes after NP injection, the animals were sacrificed and the heart, lungs, spleen, liver, and kidneys were collected. DNA was isolated and purified from the organs, then qPCR was performed for each sample using primer sets for each barcode. By matching each barcode to its corresponding PBAE NP in a particular animal, we calculated the average relative accumulation of each formulation in the major organs (Figure 4). The plots show accumulation of any given PBAE NP across organs to screen for PBAEs that direct NPs to specific organs over others. Statistically significant barcode accumulation over untreated samples was observed in organs involved in clearance, including the spleen (PBAE 546), liver (PBAE 456, 534, 546), and kidneys (PBAE 534). Although not statistically significant, higher barcode accumulation was also observed in the lung, which presents the first capillary bed that the NPs are exposed to once injected intravenously. Accumulation in the heart was negligible for all formulations. The mean highest levels of barcode DNA were detected in the liver and spleen for 7 out of 8 formulations. This is consistent with rapid MPS clearance of NPs in the range of 100–200 nm.

Figure 4. High-throughput screening of NP biodistribution.

Pooled biodistribution data in major organs (liver, spleen, heart, lungs, kidneys) of each PBAE NP formulation with 5 distinct DNA barcodes from 5 different mice. 30 min post administration of barcoded PBAE NPs, the amount of DNA accumulated was quantified by the amplification of DNA barcodes in qPCR (−ΔΔC_T of barcoded plasmid DNA in each PBAE NP normalized to GAPDH). Data represent mean ± SEM of n = 5 for each PBAE NP from a total of 8 mice (one-tailed Student’s t-test between the experimental −ΔΔC_T values and zero with Sidak’s multiple comparisons, * = p < 0.05).

3.6. High-throughput screening of tissue-specific gene expression

While NP accumulation in the target tissue is necessary for gene delivery, cellular uptake, endosomal escape, and nuclear localization are critical steps for transfection that are not captured by biodistribution studies alone. To this end, we further explored the use of barcoded pDNA to directly quantify PBAE NP transfection in the major organs. Transfection was quantified by performing RT-qPCR on transcribed barcode mRNA isolated from treated animals. Animals were treated with the same NP mixtures as described in Table 1, and organs were harvested after 6 hours, a timepoint that shows a notable increase in in vivo expression signal and therefore captures transfection efficacy (Figure S9). The isolated RNA was purified of any contaminating DNA and reverse transcribed into cDNA. RT-qPCR was performed using each set of primers for each sample, and ΔΔC_T values were calculated as previously described.

Barcode transfection of each PBAE across organs is shown in Figure 5. Transfection was predominantly observed in the kidney, spleen, and liver. For the heart and lungs, there was no detectable transfection as PCR amplification was equivalent to amplification in untreated samples for all NP formulations. Transfection was primarily localized to the liver and spleen, with statistically significant spleen transfection detected for 447, 457, and 536 PBAEs, and liver transfection detected for 546 PBAE. Transfection in the kidneys was relatively low / not significant, despite similar levels of barcode DNA accumulation to the liver and spleen. This may be because small fragmented particles or free pDNA ineffective for transfection may preferentially be cleared by the renal system. Interestingly, PBAEs showing similar level of accumulation in an organ did not necessarily resulted in similar level of transfection. For example, comparable amounts of pDNA were delivered to the liver by both 456 and 546 PBAEs, however 546 PBAE NP showed a significant level of transfection in the liver while 456 did not. This difference in transfection efficiency could be explained by the differential effect of each PBAE polymer structure following tissue accumulation in the downstream intracellular delivery steps, including interaction with cell surface, internalization, endosomal escape, and pDNA release. These differences are particularly striking as the chemical structures of 456 and 546 are so similar. In the repeating unit of the polymer, 456 contains 4 carbons between acrylate groups and 5 carbons between the amine group and the alcohol group in the side chain, whereas 546 contains 5 carbons between acrylate groups and 4 carbons between the amine group and the alcohol group in the side chain. Yet, this small molecular difference generates a dramatic biological difference in the ability of the polymer to facilitate gene delivery in a tissue-specific way, in this case transfection of liver. This result highlights the utility of a high-throughput in vivo assay to evaluate differential activity of gene delivery effectiveness between closely related materials from a nanoparticle library. While it is difficult to identify a single parameter from the properties of a given PBAE polymer that directs specific accumulation or transfection in an organ (Table S1), a statistical correlation analysis can elucidate the structure-function relationship of gene delivery polymers [26].

Figure 5. High-throughput screening of in vivo transfection.

Pooled transfection data in organs with detectable transfection signal (liver, spleen, and kidneys) of each PBAE NP formulation with 5 distinct DNA barcodes from 5 different mice. 6 hr post administration of PBAE NPs, mRNA was extracted from organs and relative expression was quantified by RT-qPCR (-ΔΔC_T of mRNA transcription of barcoded plasmid DNA in each PBAE NP normalized to GAPDH). Data represent mean ± SEM of n = 5 for each PBAE NP from a total of 8 mice (one-tailed Student’s t-test between the experimental −ΔΔC_T values and zero with Sidak’s multiple comparisons, * = p < 0.05).

It is also important to note that PBAE NP that showed significant transfection in the liver in vivo was not the optimal candidate from in vitro transfection screening (Figure S6). While there were significant differences in in vitro transfection efficacy with all 8 PBAE NP formulations in HepG2 liver cancer cells, these differences were not predictive of in vivo performance. For example, PBAE 546 showed the lowest in vitro expression in these cells but showed a high degree of liver targeting in vivo as shown by barcode screening and luciferase imaging. This further highlights the significance of the in vivo high-throughput screening method.

3.7. Validation of liver- and spleen-specific gene delivery by 456, 536, and 546 PBAE NPs

To verify that this system accurately predicted transfection patterns for a particular NP formulation, we delivered a reporter plasmid, firefly luciferase (fLuc) intravenously using 456, 536, and 546 PBAE and imaged bioluminescence after 6 hours. Luciferase expression was localized primarily to the spleen for 456, liver and spleen for 536, and liver for 546 PBAE NP (Figure 6). Bioluminescence in heart, lung, and kidney were negligible in all three formulations. These gene expression results from singular injection agree with high-throughput transfection screening results for each of the respective PBAE NPs. The agreement between high-throughput screening and singular injection results suggests that the barcode method is predictive of NPs functional outcome based on specific polymer composition regardless of the nucleic acid sequence of pDNA. In addition, 456 and 536 PBAE NPs exhibited conflicting patterns between the biodistribution and gene expression results, where 456 PBAE NP showed highest accumulation in the liver but highest expression in spleen, and 536 PBAE NP showed similar levels of accumulation across organs but significant expression only in the spleen and liver. This again highlights the importance of directly quantifying transfection at the level of mRNA or protein, rather than biodistribution of NPs and DNA, to evaluate gene delivery in vivo.

Figure 6. Luciferase expression after intravenous administration of PBAE NPs harboring fLuc plasmid DNA.

Quantification and representative images of organ bioluminescence 6 hours after intravenous administration of PBAE 456 (a,d), PBAE 536 (b,e), and PBAE 546 (c,f) NPs. Data represent mean ± SEM of n = 3. Statistical differences between organ luminescence were calculated by one-way ANOVA with Holm Sidak’s multiple comparison test. *: p < 0.05, **: p < 0.01, ****: p < 0.0001

4. Discussion

Many different polymers and lipids libraries have been synthesized to formulate NPs by combinatorial methods. These NPs have extensively been characterized to elucidate structure-function relationships [15, 19, 39]. This often requires high-throughput methods for standardized and efficient screening for optimization. Multi-well plates and automated pipetting robots have enabled fast and efficient in vitro analysis of cytotoxicity, cellular uptake, and transfection [40]. Because these in vitro cell culture studies are rapid and reproducible, they are often used to identify optimal formulations from the library for further in vivo testing. However, many studies have found that in vitro transfection efficacy is a poor predictor of in vivo efficacy because cell culture conditions do not recapitulate the many barriers to systemic gene delivery [41, 42]. Cost, time, and loss of animal life are bottlenecks preventing in vivo screening of large numbers of NP formulations. A high-throughput strategy for in vivo gene therapy experiments with novel methods to investigate multiple NPs in a single animal reduces the labor, cost, and use of animals. The study showed how only 8 mice could be used to perform a biodistribution experiment of 8 NP formulations with n=5, which would otherwise have required 40 mice without high-throughput methods. The barcode strategy also mitigates variability between individual animals unduly influencing observed differences in NP behavior. While each PBAE NP formulation is still evaluated in five separate mice, variability between animals across the five different PBAE NP formulations is reduced, as a complete set of five NP formulations is evaluated in a single mouse and thus different NPs are being evaluated in the same animals instead of in unique animals. Furthermore, just as automation adds efficiency to high throughput, this in vivo screening approach could be adapted for additional NP types per animal and further efficiency on the analysis side obtained by using automated pipettors and high-throughput 384-well qPCR.

Several methods for high-throughput in vivo screening using nucleic acid barcodes have been recently reported. Dahlman et al. used deep sequencing methods to simultaneously characterize biodistribution of dozens of lipid NPs harboring barcoded oligonucleotides [32, 33]. This innovative approach highlights the power of barcoding methods to impact the field of drug and gene delivery. While oligonucleotide barcodes offer flexibility and versatility due to their small size, they are used as a passive tag to measure biodistribution rather than functional delivery. We chose to incorporate DNA barcodes in plasmid vectors to assess biodistribution (extracellular barriers) and transfection (intracellular barriers) in target tissues using the same barcodes. Because properties of NPs may be affected by the properties of their cargo depending on the NP system, including potentially by adding a noncoding nucleic acid barcode tag or conjugating labeling molecules to the nucleic acid cargo, the presented strategy also allows us to evaluate NPs specifically formulated for delivery of their standard pDNA cargo without an additional labeled component. We show that directly measuring transfection is critical, as biodistribution results did not accurately predict in vivo transfection. Particles that accumulate in tissues may become entrapped in mucosal or extracellular matrix barriers [43], sequestered in vesicles [44], or exocytosed into the interstitial space [45]. Successful gene therapy is dependent on overcoming all extracellular and intracellular barriers, ultimately resulting in the transcription and translation of a therapeutic protein.

The type of genetic cargo, including DNA, mRNA, siRNA, and short oligonucleotide, can interact with the vectors differently and affect the NP’s physicochemical properties as well as biodistribution profile. For example, Guimaraes et al. showed that NPs with the same lipid but either mRNA or DNA oligonucleotide exhibited different levels of accumulation in liver and spleen [34]. We simplified NP’s nucleic acid cargo to a single plasmid vector harboring both the barcode and a functional gene, eliminating the need for a secondary barcode oligonucleotide and the possibility of its differential molecular interaction with various vectors. This strategy also allows broad utilization of the technology with other NP formulations. Moreover, our study shows that randomly generated nucleotide sequences can function well as barcodes, which indicates that many NPs with similar random barcodes can be evaluated simultaneously as a larger NP cocktail within a single animal. Increased number of barcodes would increase the utility of the high-throughput screening strategy. In contrast to prior work, our system uses qPCR for barcode quantification, which is routinely used with well-established analysis methods. Therefore, this strategy is easily adaptable to different NP systems and laboratories to characterize gene delivery.

Quantifying transfection at the cellular level would add further value to this approach for gene therapy development. To achieve this degree of granularity, cell types of interest could be sorted using immunocytochemistry and flow cytometry or other single cell methodologies, then barcode accumulation and transfection could be quantified in the population of interest. This would enable identification of NPs that target individual cancer cells within a heterogeneous population, specific immune cells, and other phenotypes affected by genetic disease.

Our reported biodistribution results from high-throughput barcode NP screening agree with the current understanding of NP pharmacokinetics, which has been extensively reported in the literature [46, 47]. Particles with diameter < ~10 nm are rapidly cleared by the renal system and accumulate in the kidneys, bladder, and urine [48]. NPs larger than ~20 nm are cleared by the MPS in the liver and spleen [49]. Our characterization of PBAE NPs by DLS sizing confirms that all PBAE NPs tested are between 100 and 200 nm, suggesting that MPS clearance would dominate their pharmacokinetics. In agreement with these principles, we found by high throughput barcode screening that PBAE NPs accumulate in the liver and spleen at 30 min post administration. Biodistribution was evaluated at 30 min post administration due to short half-life (10 min) of PBAE NPs in the blood [50] and potential DNA degradation, while in vivo transfection was quantified at 6 hr based on similar methods reported by previous literature on luciferase gene expression using PBAE NPs [24, 51]. Interestingly, we also observed barcode accumulation in the kidneys, despite the average NP size being far above the maximum for renal clearance. Because transfection in the kidney was negligible, we hypothesize that barcode DNA accumulating in the kidney was unencapsulated DNA or very small degraded NP fragments. We also observed consistent NP accumulation in the lung. NP entrapment in the lung capillaries has been observed and well-characterized in the literature [1]. Aside from NP size, aggregation with serum proteins may also play a role in lung accumulation, so the effects of PBAE NP physicochemical properties on protein adsorption should be further studied in future work [52]. Across the PBAE NPs evaluated in this study, while we found lung accumulation, we found that these entrapped NPs were not successful for gene delivery, potentially due to cell-specific transfection efficacy often demonstrated with PBAE polymers [20, 23, 24]. Success for gene delivery depended on polymer structure, with small, seemingly insignificant changes to structure of one or two carbons, making a significant difference to gene therapy performance, from tissue-specific accumulation to cell-specific transfection. This proof-of-concept study demonstrates that there is much to be learned by extending higher throughput nanobiotechnology studies from in vitro to in vivo, to better understand differential nanomaterial function in biological systems.

5. Conclusion

With gene therapy emerging as a viable and versatile approach to treat or potentially cure various diseases, optimizing non-viral delivery vectors has become an active area of research. In this study, we explored the role of a polymer’s chemical structure to direct tissue-specific nanoparticle targeting and gene transfection. To this end, we also successfully developed an innovative high-throughput method using pDNA itself as a barcode that consequently reduces the number of animals used and mitigates variability between animals. Using the method, we demonstrated certain PBAE polymeric nanoparticles are capable of delivering and transfecting pDNA in the liver and/or spleen. We also showed that tissue accumulation of PBAE NPs does not necessarily correlate with in vivo gene expression, emphasizing the importance of in vivo transfection screening to predict the therapeutic efficacy of gene therapy. Both polymer structure and tissue type were important to determine transfection efficiency. Finally, we validated the high-throughput screening method by showing correlation between its mRNA result and in vivo protein expression of the firefly luciferase reporter gene using the same PBAE NPs.