Accepted Manuscript Research paper Engineering of small interfering RNA-loaded lipidoid-poly(DL-Lactic-Co-Glycolic Acid) hybrid nanoparticles for highly efficient and safe gene silencing: A quality by design-based approach Kaushik Thanki, Xianghui Zeng, Sarah Justesen, Sarah Tejlmann, Emily Falkenberg, Elize Van Driessche, Hanne Mørck Nielsen, Henrik Franzyk, Camilla Foged PII: DOI: Reference:
S0939-6411(17)30565-9 http://dx.doi.org/10.1016/j.ejpb.2017.07.014 EJPB 12570
To appear in:
European Journal of Pharmaceutics and Biopharmaceutics
Received Date: Revised Date: Accepted Date:
5 May 2017 19 July 2017 25 July 2017
Please cite this article as: K. Thanki, X. Zeng, S. Justesen, S. Tejlmann, E. Falkenberg, E. Van Driessche, H. Mørck Nielsen, H. Franzyk, C. Foged, Engineering of small interfering RNA-loaded lipidoid-poly(DL-Lactic-CoGlycolic Acid) hybrid nanoparticles for highly efficient and safe gene silencing: A quality by design-based approach, European Journal of Pharmaceutics and Biopharmaceutics (2017), doi: http://dx.doi.org/10.1016/j.ejpb. 2017.07.014
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Engineering of Small Interfering RNA-Loaded LipidoidPoly(DL-Lactic-Co-Glycolic Acid) Hybrid Nanoparticles for Highly Efficient and Safe Gene Silencing: A Quality by Design-Based Approach Kaushik Thankia, Xianghui Zenga, Sarah Justesena,b, Sarah Tejlmanna, Emily Falkenberga, Elize Van Driesschea,c Hanne Mørck Nielsena, Henrik Franzykb and Camilla Fogeda,* a
Department of Pharmacy, Faculty of Health and Medical Sciences, University of Copenhagen
Universitetsparken 2, DK-2100 Copenhagen Ø, Denmark b
Department of Drug Design and Pharmacology, Faculty of Health and Medical Sciences,
University of Copenhagen, Jagtvej 162, DK-2100 Copenhagen Ø, Denmark c
Department of Pharmaceutics, Laboratory of General Biochemistry and Physical Pharmacy,
Ghent University Campus Heymans, Ottergemsesteenweg 460, 9000 Gent, Belgium
*
[email protected] (CF)
1
GRAPHICAL ABSTRACT:
KEYWORDS: siRNA delivery, lipidoids, lipid-polymer hybrid nanoparticles, quality-bydesign, optimization, transfection
2
ABSTRACT: Safety and efficacy of therapeutics based on RNA interference, e.g., small interfering RNA (siRNA), are dependent on the optimal engineering of the delivery technology, which is used for intracellular delivery of siRNA to the cytosol of target cells. We investigated the hypothesis that commonly used and poorly tolerated cationic lipids might be replaced with more efficacious and safe lipidoids as the lipid component of siRNA-loaded lipid-polymer hybrid nanoparticles (LPNs) for achieving more efficient gene silencing at lower and safer doses. However, formulation design of such a complex formulation is highly challenging due to a strong interplay between several contributing factors. Hence, critical formulation variables, i.e. the lipidoid content and siRNA:lipidoid ratio, were initially identified, followed by a systematic quality-by-design approach to define the optimal operating space (OOS), eventually resulting in the identification of a robust, highly efficacious and safe formulation. A 17-run design of experiment with an I-optimal approach was performed to systematically assess the effect of selected variables on critical quality attributes (CQAs), i.e. physicochemical properties (hydrodynamic size, zeta potential, siRNA encapsulation/loading) and the biological performance (in vitro gene silencing and cell viability). Model fitting of the obtained data to construct predictive models revealed non-linear relationships for all CQAs, which can be readily overlooked in one-factor-at-a-time optimization approaches. The response surface methodology further enabled the identification of an OOS that met the desired quality target product profile. The optimized lipidoid-modified LPNs revealed more than 50-fold higher in vitro gene silencing at well-tolerated doses and approx. a two-fold increase in siRNA loading as compared to reference LPNs modified with the commonly used cationic lipid dioleyltrimethylammonium propane (DOTAP). Thus, lipidoid-modified LPNs show highly promising prospects for efficient and safe intracellular delivery of siRNA.
3
1
INTRODUCTION
Therapeutics based on RNA interference (RNAi) are promising candidates for the treatment of a variety of serious diseases that are currently lacking definite clinical management, but for which the pathophysiology and genetic targets are known [1]. Such an approach is highly targetspecific as it is typically mediated by double-stranded small interfering RNA (siRNA) capable of selective posttranscriptional silencing of genes involved in pathogenesis. Importantly, siRNA is highly potent as compared to other antisense approaches previously reported [2]. However, intracellular delivery of exogenous, chemically synthesized siRNA to the RNAi pathway in the cytosol remains a major challenge owing to the high molecular weight and net negative charge of siRNA, which confer a practically negligible ability to permeate across the plasma membrane. Furthermore, oligonucleotides are generally highly susceptible to degradation by endogenous nucleases [3]. Nevertheless, these challenges may efficiently be overcome by designing safe nanocarriers that display unique features of protection against decomposition and facilitation of membrane permeation [4].
Previously, a series of nanocarriers were designed and explored for their potential as delivery systems for siRNA [5]. Several such nanocarriers are based on polymers, and they display the advantages of sustained release, colloidal stability, and structural integrity [4, 6]. However, polymeric nanoparticles suffer from poor loading of hydrophilic bioactive compounds. On the other hand, lipid-based nanocarriers also constitute a major class with added benefits of high siRNA loading efficiency and improved transfection efficiency [7-9]. In contrast to polymeric nanoparticles, these are often associated with drug leakage during storage and rapid in vivo clearance from plasma [10]. Hence, a novel delivery system was designed, possessing the advantages of both polymer- and lipid-based formulations, and referred to as lipid-polymer
4
hybrid nanoparticles (LPNs) [11]. The core of LPNs usually consist of biodegradable polymers, e.g., poly(DL-lactic-co-glycolic acid) (PLGA) [12, 13].
A variety of lipids have been used for the preparation of LPNs, e.g., fatty acids [14, 15], glyceryl tristearate [16], glyceryl tripalmitate [17], lecithin [18, 19], dipalmitoylphosphatidyl-choline [20],
dilauroylphosphocholine
[21],
dioleoylphosphatidylcholine
[22],
and
dioleoyl-
trimethylammoniumpropane (DOTAP) [23]. Among these, the most promising results have been obtained with cationic lipids, which may be attributed to the following properties: (i) strong attractive electrostatic interaction with the negatively charged siRNA, (ii) induction of endosomal escape, and (iii) self-assembly to form a uniform shell layer around a polymeric core [24]. However, commonly used cationic lipids such as DOTAP, 1,2-di-O-octadecenyl-3trimethylammonium propane (DOTMA), and dimethyldioctadecylammonium bromide (DDAB) have only a single quaternary ammonium group available for interaction with siRNA. This often leads to problems caused by excessive positive charge, e.g., non-specific protein binding, colloidal instability, drug leakage, and toxicity issues [25, 26]. Recently, there has been a paradigm shift toward exploration of new synthetic lipid materials (termed lipidoids) for improved efficacy of delivery systems [27]. Lipidoids are synthesized by the addition of amines to either acrylates, acrylamides, or epoxides [28]. In contrast to the commonly used cationic lipids, lipidoids contain several secondary and tertiary amines, rendering them more efficient in interacting with siRNA without significantly increasing the net charge of the delivery system [29].
To date, lipidoids have only been employed in formulations known as stable nucleic acid lipid particles (SNALPs) composed of lipidoid, cholesterol and PEGylated phospholipid, thus
5
constituting a vesicular long-circulating delivery system [30]. Furthermore, among liposomes and more advanced lipid nanoparticles, SNALPs have remained the benchmark formulations for testing the therapeutic effects of siRNA in clinical settings until recently, where a series of clinical trials were terminated based either on lack of convincing efficacy, or development of an influenza-like syndrome [31-33]. Second-generation lipid nanoparticles with improved therapeutic efficacy were further evaluated, however, these suffered from infusion-related problems thereby requiring premedication with corticosteroids [34, 35]. These observations warrant exploration of alternative safe and effective delivery systems containing lipidoids as a major component.
Hence, the hypothesis of the present work was that replacement of previously used cationic lipids associated with toxicity with lipidoids as the lipid component of LPNs might enable more efficient gene silencing. This would allow for use of lower doses, thus improving the overall safety and efficacy of the delivery system.
2 2.1
MATERIALS AND METHODS Materials
2′-O-methyl-modified dicer substrate asymmetric siRNA duplexes directed against enhanced green fluorescent protein (EGFP-siRNA) and scrambled negative control were generously provided by GlaxoSmithKline (Stevenage, UK) as dried, purified and desalted duplexes (Supplementary data, Table S1). The siRNA duplexes were re-annealed according to a standard protocol recommended by IDT (Coralville, IA, USA). PLGA was from Wako Pure Chemical Industries (Osaka, JP) and PolySciTech (Akina, West Lafayette, IN, USA) (Table S2).
6
Polyvinylalcohol (PVA) 403 with an 80.0% degree of hydrolysis was provided by Kuraray (Osaka, JP). The DOTAP hydrochloride, cholesterol and N-palmitoyl-sphingosine-1{succinyl[methoxy(polyethylene glycol)2000]} (C16 PEG2000 ceramide) were from Avanti Polar Lipids (Alabaster, AL, USA). Heparin and octyl-β-D-glucopyranoside (OG) were obtained from Sigma-Aldrich (St. Louis, MO, USA). Quant-iT™ RiboGreen® RNA Reagent and Tris–EDTA buffer (10 mM Tris, 1 mM EDTA, pH 7.5) (TE buffer) were acquired from Molecular Probes, Invitrogen (Paisley, UK). Primers were obtained from TAG Copenhagen (Copenhagen, DK). The Sybr Green® Master mix was supplied by Roche (Basel, CH). RNase-free diethyl pyrocarbonate (DEPC)-treated Milli-Q water was used for all buffers and dilutions. Additional chemicals were of analytical grade and purchased from Sigma-Aldrich.
2.2 The
Synthesis, purification and characterization of lipidoids lipidoid
mixture
was
synthesized
as
previously
reported
[27].
In
brief,
N‐dodecylprop‐2‐enamide (Mw 239.40; 3 mmol), synthesized and purified according to the previously reported procedure [27], was added to triethylenetetramine (Mw 145.23;15 mmol) in a thick-walled 25-ml glass vial. The vial was then placed in an oil bath (90°C), and the mixture was melted. A small magnet was added to the vial, which was screw-capped and stirred for 6.5 days at 90°C, resulting in the formation of a L mix, consisting primarily of L4, L5 and L6 (Figure 1). The obtained Lmix was analyzed by thin layer chromatography (TLC, Silica gel 60; Merck Millipore, Merck KGaA, Darmstadt, DE) eluting with CH2Cl2–MeOH–NH3 (100:10:1, v/v/v) (Supplementary data, Figure S1). Fractions enriched in L4, L5 and L6 were obtained by vacuum liquid chromatography (on a column packed with Silica gel 60H from Merck) using a gradient elution with mixtures of CH2Cl2, MeOH and NH3 starting from 350:10:1 (v/v/v) and
7
subsequently increasing polarity to 150:10:1 (v/v/v). The purity of the fractions was assessed by analytical TLC, and pure fractions were concentrated by evaporating the solvents in vacuum. The isolated lipidoid fractions were subsequently characterized by 1H NMR and high resolution mass spectroscopy (Figures S2-S5). Further, HPLC in conjunction with an evaporative light scattering detector (ELSD) was also employed to estimate of the identity and purity of the isolated lipidoid fractions (Figure S6)[36].
Figure 1. Schematic representation of the synthesis of lipidoid. Lipidoids were synthesized via an aza-Michael addition reaction of triethylenetetramine with N-dodecylacrylamide to give a cationic core displaying lipophilic tails. The resulting lipidoid mixture (Lmix) was mainly composed of lipidoid 4 (L4), lipidoid 5 (L5) and lipidoid 6 (L6), representing an increasing degree of alkylation (e.g, L4 designates triethylenetetramine carrying four N-dodecylacrylamide-derived tails).
8
2.3
Formulation development of siRNA-loaded, lipidoid-modified PLGA hybrid nanoparticles
Lipidoid-modified LPNs loaded with siRNA were prepared by using the double emulsion solvent evaporation method, essentially as reported previously [24], but with a minor modification of the procedure. Briefly, siRNA was dissolved in 125 µl of HEPES buffer (5 mM, pH 7.4) and added to 250 µl of CH2Cl2 containing lipidoid and PLGA (total solid concentration 60 mg/ml), resulting in the formation of a primary w1/o emulsion. The lipidoid content was varied according to the experimental design and subsequently adjusted with PLGA to keep the total solid concentration constant. The primary w1/o emulsion was probe-sonicated for 90 s in an ice bath at an amplitude of 50 (Misonix, Qsonica, LLC. CT, USA), phase inversed by addition of 1 ml 2 % (w/v) PVA, and vortexed vigorously for 1 min resulting in the formation of a secondary w 1/o/w2 emulsion. The secondary emulsion was subsequently probe-sonicated for 60 s at an amplitude of 50 in an ice bath to reduce the droplet size, and subsequently transferred to a 25 ml beaker containing a magnet and stirred for 45 min. Additional 5 ml of 2 % (w/v) PVA solution was added to stabilize the emulsion. The prepared LPNs were recovered by centrifugation and subjected to a washing step to remove un-encapsulated siRNA and excess PVA. The washing step comprised of centrifugation of the nanoparticles at varying centrifugal forces, i.e. 6,000 g for 5 min, 12,000 g for 5 min, 21,000 g for 5 min, 34,000 g for 5 min and 48,000 g for 10 min at 4°C, and the LPNs were subsequently resuspended in DEPC-treated water. The formulations were lyophilized using trehalose (5% w/v) as protectant. The type of lipidoid (L 4, L5, L6 and Lmix, respectively), the lipidoid content (0-20%, w/w) and the ratio of siRNA:lipidoid (1:10-1:50, w/w) were tested during optimization.
9
2.4
Preparation of control formulations
A stock solution (30 mg/ml) of L5 was prepared by dissolving it in dimethyl sulfoxide (DMSO) containing 1% (v/v) trifluoroacetic acid. Lipoplexes were prepared by addition of the stock solution (approx. 9 µl) to an siRNA solution (1 µM, 1 ml) prepared in TE-buffer. The lipoplexes were vigorously vortexed before use.
Lipidoid-based SNALPs were prepared by using the previously reported EtOH destabilization method [27], but with slight modifications. The SNALPs were composed of L 5, cholesterol, and C16 PEG2000 ceramide at molar ratios of 42:48:10. Considering the low solubility of L 5 in EtOH, an in situ buffer formation technique was adopted for the preparation of SNALPs. Briefly, L5 (5.64 mg) was dispersed in glacial acetic acid (37 µl) and subsequently dissolved in absolute EtOH (1.5 ml). To this, cholesterol (1 ml, 1.86 mg/ml in EtOH) and C16 PEG2000 ceramide (1 ml, 2.63 mg/ml in EtOH) were added. The EtOH solution of lipids was rapidly mixed with an aqueous solution of NaOAc (6.5 ml, 88 mg/ml in DEPC-treated water), which results in spontaneous formation of un-loaded SNALPs at a lipid concentration o 5.2). An siRNA solution was prepared by mixing siRNA stock (75 µl, 1 mM), absolute EtOH (350 µl), and NaOAc buffer (575 µl, 50 mM, pH 5.2). The preformed SNALPs (10 ml) were subsequently co-incubated with the siRNA solution (1 ml) at 37°C for 30 min. The loading ratio of siRNA to SNALPs was kept as 1:7.5 (w/w) [27]. Excess EtOH and unentrapped siRNA were removed by dialyzing the SNALPs against 1000x volume of phosphate-buffered saline (PBS), pH 7.4, for 2 h using a 100 kD molecular weight cut off dialysis membrane cassette (Float-A-Lyzer®, G2, Spectrum Laboratories, Rancho Dominguez, CA, USA).
10
The double emulsion solvent evaporation procedure was also used to prepare DOTAP-modified LPNs as previously reported [24] and as explained above. The formulations were prepared at a theoretical siRNA loading of 1% (w/w of total solid content), and the DOTAP content was 10% (w/w of total solid content).
2.5
Physicochemical characterization
The intensity-weighted mean hydrodynamic diameter (z-average) and polydispersity index (PDI) were determined by dynamic light scattering using the photon correlation spectroscopy technique. Samples (diluted to approx. 3 mg/ml solid content) were analyzed at 25°C employing a Zetasizer Nano ZS (Malvern Instruments, Worcestershire, UK) equipped with a 633 nm laser and 173° detection optics. For viscosity and refractive index, the values of water were used. The particle size distribution was reflected in the PDI, which ranges from 0 for a monodisperse to 1.0 for an entirely heterodisperse dispersion. The zeta potential of the formulations (diluted to approx. 3 mg/ml solid content in MilliQ-water) was measured using laser-Doppler microelectrophoresis. The measurements were performed on three independent batches. The Zetasizer Software version 7.11 (Malvern Instruments Ltd) was used for data acquisition and analysis.
The encapsulation efficiency of the LPNs was measured using a previously reported procedure [24], with minor modifications. Briefly, a volume of 25 µl of LPNs was centrifuged at 22,000×g (4°C), a volume of 200 μl CHCl3 was added, and the pellet was resuspended by brief vortexing. A volume of 475 µl of HD solution, composed of 100 µM OG and 1 mg/ml heparin, was added to the CHCl3 mix and vortexed for 1 min. The resulting mixture was rotated end-over-end for 5 min for efficient extraction of siRNA into the aqueous phase. Subsequently, the two phases were
11
separated by centrifugation at 4°C and 22,000×g for 10 min. The supernatant (aqueous phase) was isolated, diluted with TE buffer, and incubated at 37°C to evaporate residual CHCl3. The siRNA concentration in the samples was analysed using the RiboGreen® RNA reagent according to the manufacturer's instructions employing a fluorescence plate reader (FLUOstar OPTIMA, BMG Labtech, DE). The excitation and emission wavelengths were set at 485 nm and 520 nm, respectively. Each sample was assayed in triplicate. The encapsulation efficiency and practical loading were calculated according to Eq. 1 and 2:
Equation 1
Equation 2
The siRNA encapsulation efficiency of the SNALPs was measured according to a previously reported procedure [37] with slight modifications. Briefly, a volume of 25 µl SNALPs was mixed with 100 µl 1% (w/v) Triton X-100, 375 µl HD solution, and 500 µl CH3OH. One part of the resulting solution was used for quantifying the siRNA concentration using the RiboGreen® assay, while the other part was used for lipid quantification. Each sample was assayed in triplicate.
The surface morphology of the formulations was assessed using cryo-transmission electron microscopy (cryo-TEM, Tecnai G2 20 TWIN transmission electron microscope, FEI, Hillsboro, OR, USA). Temperature and humidity conditions were maintained throughout the experiments using an environmental vitrification system. A Pelco Lacey carbon-filmed grid was loaded carefully with a small aliquot of LPNs and spread evenly, resulting in the formation of a thin film
12
of approx. 10-500 nm. The sample-loaded grid was transferred in liquid N2 to an Oxford CT3500 cryo holder connected to the electron microscope. Subsequently, the samples were visualized in the bright field mode at an acceleration voltage of 120 kV, and images were captured with a Gatan Imaging Filter 100 CCD camera (Gatan, Pleasanton, CA, USA).
2.6
Statistical optimization of LPNs
Based on the results of initial experiments, a 32 factorial design with eight augmented points (17run design) was constructed over the grid space of three levels of two critical independent variables. The lipidoid content ranged from 10% - 20% (w/w) while the weight ratio of lipidoid:siRNA ranged from 10:1 to 20:1 (Table S3). The design was augmented to reduce the standard error, which was kept below 0.5 across the region of interest, and to fully explore the design space (Figure S7). The obtained responses were subsequently subjected to model fitting using analysis of variance (ANOVA), and the best model fit was selected on the basis of various statistical parameters, e.g., the p-value, R2, the difference between the adjusted R2 and the predicted R2, and the adequate precision. The statistical data treatment was performed using the Design Expert software (version 10.0.3, StatEase, Minneapolis, MN, USA). Numerical and graphical optimizations were further performed to generate desirability and overlay plots. Finally, a point formulation was selected from the design space and used for validation purposes.
2.7
In vitro gene silencing
The human non-small lung carcinoma cell line H1299 stably transfected with EGFP (EGFPH1299) was employed for in vitro gene silencing studies as previously reported [24, 38]. The cells were maintained in RPMI 1640 medium containing 10% (v/v) fetal bovine serum (FBS)
13
(Gibco, Grand Island, NY, USA) at 37°C and 95/5% O2/CO2. Upon confluency, the cells were seeded in 24-well tissue culture plates (Corning, Corning, NY, USA) at a density of 1 × 10 5 cells/well and allowed to adhere overnight. Subsequently, the culture medium was aspirated, and new medium (900 µl) was added along with the test formulations (10×, 100 µl) at varying concentrations. The cells were incubated with formulations for 24 h and subsequently washed with PBS pH 7.4 (Sigma-Aldrich) and reincubated with new culture medium for additional 24 h. After incubation, the cells were washed with 1 ml PBS and trypsized using 300 µl trypsin-EDTA solution. Gene silencing was measured by quantifying the EGFP protein expression by flow cytometry using a Gallios flow cytometer (Beckman Coulter, Brea, CA, USA). Data was analysed using FlowJo 7.6.5 (Three Star, Ashland, OR, USA). The in vitro gene silencing effects of selected formulations were also assessed at the mRNA level using the reverse transcription polymerase chain reaction (RT-PCR). Briefly, RNA isolation and purification were performed as previously described [38]. Reverse transcription of total RNA was performed by applying the iScript cDNA synthesis kit (Bio-Rad Laboratories, Hercules, CA, USA), and the PCR reactions were conducted in duplicates using a LightCycler® 480 (Roche) and the Sybr Green® Master mix (Roche). The housekeeping gene glyceraldehyde 3-phosphate dehydrogenase (GAPDH, NCBI ID: NM_002046.3) was used for normalization. The LightCycler® 480 software v1.5.0 (Roche) was used for crossing point (CP) analysis, and the data was normalized to the expression of the reference gene (GAPDH), followed by relative quantification to the untreated cells using the comparative ΔΔCP method [39].
2.8
Cell viability
For cell viability studies, H1299 mEGFP cells were seeded in 96-well plates at a density of 10,000 cells/well and allowed to adhere overnight. The cells were incubated with the
14
formulations (lipoplexes, LPNs and SNALPs) for 24 h. A 10× concentration of test dose of each formulation in 20 µl was added to 180 µl of cell culture media. After incubation, the cell culture medium containing the formulations were aspirated, freshly prepared 3-(4,5-dimethylthiazol-2yl)-2,5-diphenyltetrazolium bromide (MTT, Sigma-Aldrich, 500 µg/ml in PBS, 150 µl/well) was added, and the cells were reincubated for 4 h at 37°C and 5% CO 2 for formation of insoluble MTT formazan. Subsequently, excess MTT solution was carefully removed, and 200 µl DMSO were added per well to dissolve the formazan. The cell viability was assessed by measuring the absorbance of dissolved formazan at 550 nm and normalized to the absorbance of formazan formed by control cells. Two independent experiments were performed in quadruplets, and data was analysed using GraphPad Prism (GraphPad, La Jolla, CA, USA).
2.9
Statistics
Results are expressed as mean values ± standard deviation (SD). Statistical analysis was performed using GraphPad Prism. Statistically significant differences were assessed by ANOVA followed by a Tukey−Kramer multiple comparison test. Significance of the results is indicated according to p-values (*, p < 0.05; **, p < 0.01; and ***, p < 0.001). A p-value below 0.05 was considered statistically significant.
3
RESULTS AND DISCUSSION
Ideal materials for safe and efficacious delivery should confer the following attributes to the delivery system: (i) protection of siRNA against degradation in vivo (e.g. by nucleases), (ii) permeation across biological barriers, (iii) incorporation of siRNA within the carrier matrix for sustained release over a period of time to provide prolonged therapeutic effect, (iv) rapid cellular
15
internalization into target cells, (v) endosomal escape capability, and (vi) release of siRNA from the delivery system in the cytosol, from where it can enter the RNAi pathway [32]. Lipidoids represent a new class of lipid-like synthetic materials that meet the aforementioned criteria. The presence of multiple secondary and tertiary amines in lipidoids allows for efficient interaction with the highly anionic siRNA while the lipophilic character counter-balances the high hydrophilicity of siRNA arising from the sugar-phosphate backbone. PLGA is a biodegradable polymer with a proven track record via FDA-approved formulations. Hence, the aim of the present work was to engineer and optimize siRNA-loaded LPNs using lipidoids as the lipid component, and the biodegradable and biocompatible PLGA as the polymer component. This constitutes a continuation of our previous efforts to explore the potential of LPNs for efficient intracellular delivery of siRNA [23, 24, 40-44]. However, formulation design of such complex formulations is highly challenging due to the strong interplay between several factors, e.g. cationic lipids increase the encapsulation of the siRNA and gene silencing effect. Nevertheless, such PLGA-lipidoid hybrid LPNs might be relatively more toxic, albeit the expected concomitant increased siRNA loading may neutralize the excess cationic charge conferred by the high lipidoid content thereby reducing toxicity. However, this might be achieved at the cost of reduced efficacy. Furthermore, second-order and beyond interaction among the above-mentioned factors might be anticipated. Therefore, a highly systematic quality-by-design (QbD) approach was applied to define the optimal operating space (OOS), eventually resulting in the identification of a robust, highly efficacious and safe formulation by simultaneously varying the levels of critical factors. The initial experiments were performed by changing one factor at a time to identify their criticality at various levels in influencing the overall quality attributes of the LPNs.
16
3.1
Inclusion of lipidoids in LPNs remarkably increases the siRNA encapsulation efficiency
An array of lipidoids with varying degree of amine alkylation [i.e. lipidoid 4 (L4), lipidoid 5 (L5), lipidoid 6 (L6) and lipidoid mix (Lmix)] was synthesized, purified and employed for the preparation of LPNs. The inclusion of lipidoids did not interfere with the nanoparticle formation process, as statistically insignificant differences (p-value > 0.05) in the average hydrodynamic diamenter (z-average) and PDI were observed when compared to non-modified PLGA nanoparticles (Table S4). However, significantly higher zeta potential values were found for the lipidoid-modified LPNs as compared to the non-modified PLGA nanoparticles. Interestingly, a significant increase in encapsulation efficiency was measured, from 3% in the case of PLGA NPs to >60% for all types of lipidoid-modified LPNs (Table S4). At 10% (w/w) lipidoid content, no statistically significant difference in encapsulation efficiency was noted for L 4-modified LPNs, L6-modified LPNs and Lmix-modified LPNs when compared to that of L5-modified LPNs. However, the encapsulation efficiency was significantly lower (p-value < 0.05) in case of L6modified LPNs as compared to that of Lmix-modified LPNs.
3.2
The type of PLGA core affects the encapsulation of siRNA in LPNs
A number of PLGA polymers with different properties, i.e., molecular weight, lactic/glycolic acid molar ratio and type of end group modification, were tested to assess the influence of the PLGA type on the siRNA encapsulation efficiency at a fixed lipidoid content and lipidoid:siRNA ratio. A complex correlation between the siRNA encapsulation efficiency and the properties of the LPNs was recognized (Table S5): A significantly higher encapsulation efficiency (74.6%) was measured for LPNs prepared with high-molecular weight PLGA (20 kDa) as compared to
17
that of LPNs prepared using low-molecular weight PLGA (10 kDa, 32.5%) at a constant lactic/glycolic acid molar ratio (75/25). Interestingly, no significant differences (p-value > 0.05) in particle size, PDI and zeta potential were found (Table S5). Furthermore, with PLGAs of a 20 kDa molecular weight, a two-fold increase in encapsulation efficiency was measured for LPNs prepared with PLGA glycolic acid content higher than 25 mol% as compared to LPNs prepared with PLGA with a glycolic acid content below 15%. Interestingly, the zeta potential in the latter case was markedly higher (+27.9 mV) as compared to that of the high-glycolic acid content formulation, which displayed an almost neutral zeta potential. Significantly higher zeta potential and lower encapsulation efficiency were also observed for LPNs prepared with PLGA with different end group modifications (acid or amine), in contrast to those of LPNs prepared with PLGA having esters as end groups (Table S6).
3.3
Identification of critical quality attributes (CQAs) for optimization of LPNs
The results of initial experiments (based on one-factor-at-a-time experiments) revealed that for a constant siRNA:lipidoid ratio, a significantly higher (p-value < 0.001) loading of siRNA was achieved, and that this was proportional to the lipidoid content (Table S7). An approx. five-fold increase in the loading was measured when the lipidoid content was increased from 5% to 20% (w/w). However, no statistically significant differences (p-value > 0.05) in the particle size, PDI, and encapsulation efficiency were found (Table S7). Furthermore, at a fixed lipidoid content of 15% (w/w), a significantly higher siRNA loading was obtained upon increasing the siRNA:lipidoid ratio. The siRNA loading was increased from 3.9 µg/mg for nanoparticles with an siRNA:lipidoid ratio of 1:30 (w/w), to as high as 10.6 µg/mg for nanoparticles at a ratio of 1:10 (w/w) with no significant differences (p-value > 0.05) in encapsulation efficiency and zeta
18
potential (Table S8). However, a significant increase in particle size and PDI was observed for LPNs prepared with an siRNA:lipidoid ratio of 1:10 (w/w) (p-value < 0.001), as compared to that of LPNs prepared at ratios of 1:15 and 1:20 (w/w). Hence, these two factors, i.e. the lipidoid content and the siRNA:lipidoid ratio (w/w), were included for a systematic QbD optimization of the LPNs.
3.4
Statistical optimization of LPNs
A quadratic model and an I-optimal approach were employed for the design of experiments (DoE), and a 17 experimental run design was executed across the formulation design space (Table S9). The obtained data sets were compared to the preset quality target product profile (QTPP, Table 1).
Table 1. Quality target product profile (QTPP) of optimized LPNs Response
Target
z-average (nm)
< 250 nm
Polydispersity index (PDI) Zeta potential (mV) Encapsulation efficiency (%) siRNA loading efficiency (µg/mg NPs) Transfection efficiency (IC50, nM) Cell cytotoxicity (IC50, nM)
< 0.3 > 0 mV > 60% > 5 µg/mg NPs < 5 nM > 50 nM
3.5
Reason Rapid cellular internalization and sustained release characteristics Narrow particle size distribution and stability Efficient interaction with cell membranes Pharmacoeconomic considerations Reduces the particle burden Enhanced therapeutic efficacy Improved safety profile
Contour profiling of the response variables
Contour plots for each response variable, as a function of the two independent variables, were constructed (Figure 2). The particle size and PDI were marginally affected by the lipidoid content and the siRNA:lipidoid weight ratio (Figure 2A and 2B). A slight increase in the particle size and PDI was measured at higher siRNA:lipidoid ratios, i.e. 1:12.5 to 1:10, and at a lipidoid
19
content of 13-18% (w/w). This could be attributed to an influence on the particle formation process due to the higher siRNA concentrations and the concurrent reduced amount of lipidoid. Evaluation of experimental runs suggested that the particle sizes for all formulations in the explored region were within the range from 200 nm to 260 nm. The initial experiments with L5modified LPNs revealed that the ratio of siRNA:lipidoid, and not the lipidoid content, is a determining factor for the particle size and PDI (Table S7 and S8). Interestingly, the DoE revealed that both the lipidoid content and the siRNA:lipidoid ratio contribute equally to reaching the overall CQAs of the formulation. The particle size was kept as one of the QTPP parameters, as it is widely known to influence pharmacokinetics, tissue distribution, tissue extravasation, uptake and/or accumulation within the clearance organs [45]. Considering that the developed LPNs were intended for pulmonary delivery, the upper limit of particle size was set to be 60%, while the QTPP for siRNA loading was set to >5 µg/mg NPs.
The biological responses, i.e. the in vitro gene silencing and cell viability for the 17 formulations in the experimental design were also measured (Figure 3A-B). A dose-dependent gene silencing vis-à-vis reduction of cell viability was noted for all formulations. The concentrations corresponding to 50% (IC50 values) for each curve were calculated using curve fitting algorithms (GraphPad Prism) and used for modeling because the full profile could not be used. Subsequently, contour plots for each response representing the IC50 values for transfection efficiency and cell viability were constructed. Although siRNA-based therapeutics are highly potent, the therapeutic efficacy is often compromised because of numerous challenges associated with siRNA stability and delivery [8]. Therefore, the in vitro gene silencing was also included in the QTPP. The IC50 values for transfection efficiency of all the formulations screened for responses ranged from 0.8 to 28.8 nM. Notably, the in vitro gene silencing depended largely on the lipidoid content up to 15% (w/w), however, it remained unaffected beyond this concentration, irrespectively of the siRNA:lipidoid ratio (Figure 3C). Similarly, a favorable correlation of the siRNA:lipidoid ratio on the transfection efficiency was also noted only up to 15% (w/w). The transfection efficiency, measured as the IC50 value, was in the range of 4-10 nM in the majority of the region, however, considering the overall response across the design space, a value of 5 nM was chosen as a cut off value for the QTPP.
23
Figure 3. Dose-response curves based on 17 formulations mediating in vitro gene silencing (A) and cell viability (B). The 50% gene silencing and cell viability (IC50 values) is marked with a dotted line. Contour plots for IC50 values of transfection efficiency (C) and cell viability (D). The in vitro gene silencing experiments were performed in duplicate. The in vitro gene silencing for three independent batches of a representative formulation is also provided as supplementary data (Figure S8): the standard deviation was generally below 10%. The cell viability experiments was performed in quadruplicate, and the standard deviation was also below 10%.
Potential sources for toxicity of siRNA-based therapeutics have been identified and include: (i) off-target effects, (ii) innate immune activation, (iii) adaptive immune responses, and (iv)
24
toxicity arising from delivery systems, per se [33]. Hence, to address the toxicity concerns associated with lipidoid-modified LPNs, cell viability was also considered as part of the QTPP. Very complex correlations were observed for the impact of independent variables on cell viability of LPNs (Figure 3D). At an siRNA:lipidoid ratio of 1:10, the LPNs were found to influence cell viability to a lower extent, as compared to that of LPNs prepared at a higher siRNA:lipidoid ratio, while only a marginal effect of lipidoid content was noted. However, the effect of lipidoid content on cell viability was much more prominent for formulations with a siRNA:lipidoid ratio of 1:20 with a strong direct correlation. Overall, the IC50 values for cell viability remained in the range of 40-70 nM at siRNA:lipidoid ratios of 1:10 – 1:15, and 20-40 nM at siRNA:lipidoid ratios of 1:15 – 1:20. The results of the DoE analysis suggested that the siRNA:lipidoid ratio dominated over the lipidoid content in determining the overall safety of the formulations. The formulations remained relatively well-tolerated at higher concentrations of siRNA (or with lower siRNA:lipidoid ratios), whereas they exhibited marginal toxicity at lower concentrations of siRNA (or higher siRNA:lipidoid ratios). Similar results were noted previously for DOTAP/DOPE-pDNA lipoplexes for which the cell viability increased with a decrease in N/P ratio when tested in macrophage cell lines [50]. In vitro safety studies of polyplexes is often carried out as a measure of erythrocyte aggregation, and an identical trend was also found for PEGylated PEI and siRNA complexes [51]. However, to the best of the authors’ knowledge, till now no systematic studies have been performed where both the lipid content and N/P ratio are varied to assess their influence on transfection efficiency and cell viability.
25
3.6
Mathematical modeling of the responses
The obtained response variables were subsequently model-fitted using an ANOVA approach. A partial sum of squares method was used to calculate model statistics wherein each effect for a model was adjusted upon correcting other effects. Design Expert software (version 10, Stat-Ease Inc, USA) was employed for these algorithms, and the highest order model for each response variable was selected, provided that it was not aliased (Table 2). As evident, the model selected for each response variable was significant with a p-value
