Journal of Colloid and Interface Science 500 (2017) 253–263

Contents lists available at ScienceDirect

Journal of Colloid and Interface Science

journal homepage: www.elsevier .com/locate / jc is

Regular Article

Automated in-line mixing system for large scale production of chitosan-based polyplexes

http://dx.doi.org/10.1016/j.jcis.2017.04.0130021-9797/� 2017 Elsevier Inc. All rights reserved.

⇑ Corresponding author at: Department of Chemical Engineering, PolytechniqueMontreal, Montreal, Quebec, Canada.

E-mail address: michael.buschmann@polymtl.ca (M.D. Buschmann).

Ashkan Tavakoli Naeini a, Ousamah Younoss Soliman a, Mohamad Gabriel Alameh a, Marc Lavertu b,Michael D. Buschmann a,b,⇑a Institute of Biomedical Engineering, Polytechnique Montreal, Montreal, Quebec, CanadabDepartment of Chemical Engineering, Polytechnique Montreal, Montreal, Quebec, Canada

g r a p h i c a l a b s t r a c t

a r t i c l e i n f o

Article history:Received 5 February 2017Revised 5 April 2017Accepted 5 April 2017Available online 07 April 2017

Keywords:In-line mixingChitosansiRNAPlasmidPolyplexNon-viral gene deliveryBioactivity

a b s t r a c t

Chitosan (CS)-based polyplexes are efficient non-viral gene delivery systems that are most commonlyprepared by manual mixing. However, manual mixing is not only poorly controlled but also restrictedto relatively small preparation volumes, limiting clinical applications. In order to overcome these draw-backs and to produce clinical quantities of CS-based polyplexes, a fully automated in-line mixing platformwas developed for production of large batches of small-size and homogeneous CS-based polyplexes.Operational conditions to produce small-sized homogeneous polyplexes were identified. Increasing mix-ing concentrations of CS and nucleic acid was directly associated with an increase in size and polydisper-sity of both CS/pDNA and CS/siRNA polyplexes. We also found that although the speed of mixing has anegligible impact on the properties of CS/pDNA polyplexes, the size and polydispersity of CS/siRNA poly-plexes are strongly influenced by the mixing speed: the higher the speed, the smaller the size and poly-dispersity. While in-line and manual CS/pDNA polyplexes had similar size and PDI, CS/siRNA polyplexeswere smaller and more homogenous when prepared in-line in the non-laminar flow regime compared tomanual method. Finally, we found that in-line mixed CS/siRNA polyplexes have equivalent or highersilencing efficiency of ApoB in HepG2 cells, compared to manually prepared polyplexes.

� 2017 Elsevier Inc. All rights reserved.

1. Introduction

Non-viral gene delivery mainly relies on the use of cationiclipids and cationic polymers to bind and condense negatively

254 A. Tavakoli Naeini et al. / Journal of Colloid and Interface Science 500 (2017) 253–263

charged nucleic acids (NAs). Complexes formed between a nucleicacid and polymers are referred to as polyplexes. Upon mixing,these oppositely charged species bind to each other by means ofattractive electrostatic interactions, a process favored by a con-comitant increase in entropy due to the release of low molecularweight counter-ions. In very dilute solutions, complex formationleads to nanoparticle suspensions at the colloidal level, while mix-ing in a concentrated regime results in macroscopically flocculatedsystems [1–4]. Nonetheless, higher mixing concentrations result inhigher dose of NA in a given volume of the suspension, henceincreasing the deliverable doses.

Chitosan (CS), a naturally derived polycation, is a biocompatibleand biodegradable linear polysaccharide that has gained interestfor safe delivery of NA. It has been demonstrated that CS-baseddelivery systems are efficient for delivery of plasmid DNA (pDNA),[5–12] and siRNA [13–24] both in vitro and in vivo. Manual mixingis the most commonly used method to produce CS-based poly-plexes. However, this manual method varies in implementationwhich raises some concerns regarding comparability of the proper-ties of such complexes. Classical pipetting is a frequently reportedtechnique in the literature, where CS is promptly added into NAsolution followed by pipetting the mixture up and down forhomogenization [5,7–9,11,13,25–27]. One other manual approachis addition of NA into excess CS solution under stirring conditionto allow formation of polyplexes [17–19,24]. Another techniqueis drop-wise addition of CS solution into equal volume of NA solu-tion followed by quick mixing, [28] while similar method wasreported with reverse order of addition [16]. Addition of siRNA intoCS followed by vortexing of the complex for a short period (30 s) isalso a reported technique [14,21]. Properties of final polyplexesdepend upon the mixing technique, experience of the operator,[29] and order of addition of the two polyelectrolytes [30]. Further-more, independent of the adopted approach, conventional manualmixing is not only a poorly controlled method which may result inirreproducibility, [29,31–32] but it is also restricted to relativelysmall preparation volumes, that seriously limit applicability. Smallscale manual preparation of complexes poses risks of batch tobatch and inter-user variability, [29,33–34] thus a controlled andrepeatable production process of polyplexes in large quantities isa prerequisite for their clinical applications [32].

Several mixing techniques and devices have been proposed inthe literature to prepare large volumes of complexes. These tech-niques rely on in-line mixing in Y or T-shaped connectors, [33–36] microfluidics, [29–32,37–40] as well as jet mixing [41–42].To the best of our knowledge, the very first in-line mixing systemfor complex production was designed by Zelphati et al., where syr-inges were used to drive pDNA and cationic lipids into a T-connector [34]. Later, John and Prud’homme invented a ConfinedImpinging Jet mixing (CIJ) apparatus to generate two opposinghigh velocity linear jets of polyelectrolyte solutions, where thetwo streams collide in a chamber for a few milliseconds (rapidmixing takes place in a time less than the nucleation and growthtime of complexes) [42]. Ankerfors et al. applied the CIJ systemto investigate the influence of mixing time on the size of producedPolyallylamine hydrochloride (PAH)/polyacrylic acid (PAA) com-plexes. They reported production of larger particles at reducedspeeds of the polyelectrolyte streams [41]. While establishmentof a jet mixer is sophisticated, mixing in a Y or T-shaped connectoris a simpler design. In 2005, Clement et al. introduced an in-linemixing platform (but not computer controlled) based on two peri-staltic pumps and a Y-connector. They showed no difference intransfection efficiencies of manual and in-line mixed complexes,[35] however, the impact of mixing conditions (e.g., speed andNA concentration) on physico-chemical properties of complexeswas not studied. In 2010 Davies et al. designed a novel mixingdevice based on a static mixer and a cylindrical pneumatic actuator

connected via an adjustable regulator to a compressed air supply[36]. They were the first to reporting an in-line mixing device forthe preparation of polyplexes (PEI/pDNA), however, their devicewas not easily scalable. In 2011, Kasper et al. claimed reproducibleproduction of small and homogeneous PEI-pDNA polyplexes bymeans of an up-scaled micro-mixer system, consisting of two syr-inges and a T-connector [33]. They reported a reduction in the sizeof polyplexes by increasing the mixing speed and reducing the con-centration of pDNA. Finally, a chaotic mixer specified as ‘‘staggeredherringbone microfluidic mixing device” was designed based onmicrofluidic mixing in a laminar regime to produce well-definedlipid nanoparticles (LNPs) [39–40]. This staggered herringbone pat-tern creates transverse flows in the microchannels to induce chao-tic mixing at a low speed of the laminar regime. Belliveau et al.investigated the ability of SHM to produce monodisperse LNP-siRNA complexes [31]. While the size of LNPs remained fairly con-stant over the range of the tested flow rates, the lowest PDI forLNP-siRNA complexes was obtained at the highest tested mixingflow rate. In addition, reducing the concentration of siRNA resultedin a reduction in polydispersity. Although the above-mentionedmethods were introduced for up-scaled mixing platforms, therehas been no systematic study to date on the operational parame-ters of a fully automated in-line mixing system that can repro-ducibly prepare large batches of CS-based polyplexes withdefined physico-chemical properties, and transfection efficiency.

The main objective of this study was to develop a computercontrolled in-line mixing platform for reproducible production ofsmall sized homogeneous CS-based polyplexes, and examine theinfluence of mixing parameters on the properties of polyplexes.We hypothesized that reducing the NA concentration and increas-ing mixing speed would reduce the size and polydispersity of CS-based polyplexes. Another objective was to assess the in vitrosilencing efficiency of the CS/siRNA polyplexes, where we hypoth-esized equal or even better bioactivity for in-line CS/siRNA com-pared to manually prepared polyplexes.

2. Materials and methods

2.1. Materials

Chitosan was obtained fromMarinard. Trehalose dehydrate (Cat#T0167), L-histidine (Cat #H6034) were from Sigma. HepG2 (hep-atocellular carcinoma) cells (Cat# HB-8065), and Eagle’s MinimumEssential Medium (Cat #30-2003) were from American Type Cul-ture Collection (ATCC). Dulbecco’s Modified Eagle Medium, highglucose pyruvate (DMEM-HG, Cat #12800-017) was from Gibco.Fetal bovine serum (FBS, Cat #26140) was from Thermo Fisher Sci-entific. Plasmid EGFPLuc (Cat #6169-1) was from Clontech Labora-tories. siRNA targeting ApoB mRNAs (siRNA ApoB) contains sensesequence of 50-GUCAUCACACUGAAUACCAAU-30 and antisense 50-AUUGGUAUUCAGUGUGAUGACAC-30 was obtained from GE (Cus-tom synthesis, A4 scale). Double stranded oligodeoxynucleotide(dsODN, 21 bp) encoding the same sequences and mimickingsiRNA ApoB physico-chemical properties was obtained from Inte-grated DNA Technologies Inc, Coralville, IO. The commerciallyavailable liposome, DharmaFECTTM2 was from Dharmacon RNAiTechnologies and Diethyl pyrocarbonate (DEPC) from SigmaAldrich (Cat #D5758).

2.2. Preparation of CS, plasmid DNA and siRNA for mixing

The plasmid eGFPLuc stock solution was prepared and charac-terized by UV spectrophotometry, as described in Lavertu et al.[9]. The plasmid DNA (pDNA) stock solution was diluted to 10,50, 100, 200, and 300 lg/mL with Milli-Q water as well as sterile

A. Tavakoli Naeini et al. / Journal of Colloid and Interface Science 500 (2017) 253–263 255

filtered trehalose and histidine (pH 6.5) solutions for having finalconcentration of 0.5% and 3.5 mM, respectively (these are opti-mum concentrations of excipients for freeze-drying and stabilityof complexes over time that have been previously established)[27]. The siRNA ApoB stock solution was prepared and character-ized by UV spectrophotometry, then diluted to 10, 50, 100, 200,300, and 400 mg/mL with RNase/DNase free water as well as theexcipients as described for plasmid DNA. Commercial chitosanwas first heterogeneously deacetylated to 92% using concentratedsodium hydroxide and was depolymerized to 10 kDa and 2 kDausing nitrous acid, as established and previously reported by ourgroup [9]. It has been reported that the 10 kDa chitosan withDDA of 92% (CS92-10) form stable complexes when prepared atmolar ratio of CS amine to NA phosphate ratio of 5 (N:P = 5), andalso dissociates effectively upon release from lysosomes, maximiz-ing the level of transfection [12]. Degree of deacetylation (DDA)and number-average molar mass (Mn) of chitosan was confirmedby 1H NMR, [43] and gel permeation chromatography multi-angle light scattering [44]. A chitosan stock solution at 5 mg/mLwas prepared from dry powder dissolved in 28 mM HCl overnightat room temperature (RT). The stock solution was sterile filtered,then diluted with Milli-Q or RNase free water (for siRNA applica-tions), as well as excipients as described before for N:P = 5, basedon the concentration of nucleic acid (either pDNA or siRNA).

2.3. Preparation of CS/NA polyplexes by manual mixing (classicalpipetteing)

CS/NA polyplexes were prepared by manual addition of CS(100 lL) to equal volume of NA (diluted to maintain N:P = 5), fol-lowing by immediate pipetting up and down repeated 10 times.Samples were then incubated for 30 min at RT before analyses ortransfections.

2.4. Preparation of CS/NA polyplexes by the Automated In-line MixingSystem (AIMS)

In addition to the manual mixing, both types of polyplexes (CS/pDNA and CS/siRNA) were produced via AIMS in the exact sameformulation as for the manual mixing. Both initial and advancedversions of AIMS (configuration a and b of Fig. 1) were tested. Pre-pared samples were incubated in the original collecting vessel for30 min at RT before analyses or transfections.

2.5. Polyplex size, polydispersity, and surface charge analysis

Hydrodynamic size and polydispersity index of the producedpolyplexes were measured using dynamic light scattering (Zeta-sizer Nano ZSP-ZEN5600, Malvern Instruments, Worcestershire,UK). Samples were diluted by 4–5-fold by adding Milli-Q or RNasefree water, then analyzed for three consecutive runs at 25 �C. Zetapotential (ZP) measurements were made with the Malvern Zeta-sizer Nano ZSP using folded capillary cells. ZP value was calculatedfrom the measured electrophoretic mobility uE by applying theHenry equation: ZP ¼ 3guE=2ef ðjaÞ, where UE is the velocity ofthe particle in the applied electric field (referred to as the elec-trophoretic mobility), e and g are the dielectric constant and theviscosity of the medium, respectively. f (ja) is the Henry’s function,where a is the particles radius, and 1/j is the Debye length, thus jais the ratio of particle radius to Debye length. For calculation of theZP from electrophoretic mobility, a value of 3/2 for the Henry func-tion was used so that ZP ¼ guE

e was used. This limiting case isreferred to as the Smoluchowski equation, an approximation thatis valid when the size of the particle is much larger than the Debyelength (ja � 1), a condition generally satisfied in an aqueous

media with moderate to high electrolyte concentration. For ZPmeasurements, each sample was diluted 8-fold to a final volumeof 800 mL by adding 700 mL of buffer (trehalose 0.5%, and 3.5 mMhistidine) for a final ionic strength of about 1 mM. Each samplewas analyzed for three consecutive runs at 25 �C.

2.6. Transmission electron microscopy imaging

Morphology of both inline and manually produced polyplexeswas studied using Transmission Electron Microscopy (TEM). On acarbon coated copper grid (200 mesh, Electron microscopysciences), 5 mL drop of polyplex suspension was pipetted andallowed to evaporate for 30 min. To avoid drying artifacts, remain-ing solvent was removed by blotting using a filter paper. 5 mL of 2%phosphotungstic acid was placed on the grid. After 10 min the stainwas removed as described above. The grid was washed two timeswith deionized water to avoid any salt contamination from thesample and air dried. TEM images were obtained using a TecnaiT12 electron microscope operating at 120 keV. For each samplefour random areas on the grid were imaged at differentmagnifications.

2.7. Sterile and RNase free production via AIMS

The AIMS including the upstream vessels, collecting containers,as well as the tubing set were autoclaved for sterility. In order tomaintain the sterility of AIMS, the entire tubing network was iso-lated from the environment where each vessel was open to atmo-sphere via a 0.22 mm syringe filter in order to maintain pressureduring pumping and mixing. When working with siRNA, the closedsystem was treated with diethylpyrocarbonate (DEPC) in order toinactivate contaminant nucleases.

2.8. Cell culture and transfection

HepG2 cells were cultured in EMEM supplemented with 8%fetal bovine serum (FBS). One day prior to transfection, cells wereseeded in a 24 well plate at 300,000 cells per well to reach �80%confluence on the day of transfection. Prior to transfection, mediaover cells was aspirated and replenished with serum free DMEM-HG media (pH 6.5) supplemented with 0.976 g/L MES and 0.84 g/L NaHCO3. Inline and manually mixed polyplexes were preparedas described above and a specific volume was pipetted into eachwell to reach a target siRNA concentration of 100 nM/well(�0.66 mg/500 mL). FBS was added 5 h post-transfection to reach afinal concentration of 8%. Transfection media was aspirated 24 hpost transfection, replenished with complete EMEM and incubatedfor an extra 24 h. DharmaFECT

�2/siRNA lipoplexes were used as

positive control and were prepared as per manufacturer recom-mendation. Lipoplex and naked siRNA transfection were performedat a final concentration of 100 nM/well. Chitosan only anduntreated cells were used as negative controls.

2.9. Assessment of silencing efficiency

Polyplex bioactivity was assessed using MIQE compliant quan-titative real time PCR (qPCR) [45]. All primers and probes wereintron spanning to avoid amplification of potential genomic DNAcontamination and in silico validated using the NCBI blast tool forspecificity (Supp info). Total RNA extraction was performed 48 hpost transfection using the EZ-10 Spin column Animal total RNAextraction kit (BioBasic). Total RNA was extracted as per manufac-turer protocol and subjected to in-column DNase digestion (Qia-gen, Cat#79254). The purity of the eluate was assessed using UV/VIS spectrometry at 260, 280, 230 and 340 (Nanodrop 8000). Theintegrity of extracted RNA was determined using the Agilent

Fig. 1. Schematic of Automated In-line Mixing System (AIMS): initial version (configuration a), and advanced version (configuration b).

256 A. Tavakoli Naeini et al. / Journal of Colloid and Interface Science 500 (2017) 253–263

2100 Bioanalyzer (Agilent technologies). Total RNA concentrationwas determined using the Qubit reagent (Life technologies, Cat#Q10210) as per the manufacturer’s protocol. For each conditiontested, a total RNA quantity of 200 ng/20 mL reaction was reversetranscribed using the SuperScript VILO cDNA synthesis kit (Lifetechnologies, Cat#11754250). Primer and probe efficiency wasexperimentally determined using the standard curve method onthe same plate where target gene knockdown is being assessed(Supp info). Reference gene stability was validated using the GeN-orm statistical package (Biogazelle NV) (Supp info) cDNA wasamplified using TaqMan Fast Advanced Master Mix (Life technolo-gies, Cat# 4444557) on a QuantStudio 12 K flex system (Thermo-Fisher Scientific). All reactions were performed in a 384 wellplate (ThermoFisher Scientific, Cat# 4309849) using a final volumeof 10 mL (10 ng of cDNA) and the following cycling conditions:2 min hold at 55 �C, then 10 min hold at 95 �C followed by 40cycles at 95 �C for 15 s and 60 �C for 1 min. Data was exported inRDML and analyzed using the Biogazelle qBase + software package.

2.10. Statistical analyses

All experiments were repeated once with technical replicateswithin each for the transfections. Values are given as aver-age ± standard deviation. For statistics and plotting, SigmaPlot13.0 package was used. One way ANOVA (at a significance levelof 0.05) was applied and normality and other assumptions werevalidated.

3. Results and discussion

3.1. Design and establishment of the AIMS

The in-line mixing setup uses peristaltic pumps to drive fluidsbased on pulsatile motion, where the natural oscillation of the flu-ids provides a wavy and stretched interface between the twostreams, thus enhancing the mixing [46–47]. A schematic of thein-line mixing is shown in Fig. 1. Digital peristaltic pumps are pro-

grammed via LabVIEWTM to drive CS and NA solutions througheither silicon or PharmaPure tubings that were connected to eachother with a connector (Y, T or cross connector). The mixing systemalso comprises digital balances that record the masses of buffers(e.g., water) for calibration of the pump(s), and to monitor the vol-ume/mass of payloads during mixing process. The whole closedtubing network was designed to be primed in order to removeair pockets trapped in streams as well as to pre-wet the inner wallsof the tubings prior to the actual mixing. Pinch valves were pro-grammed to switch from one stream to the other (i.e., in the down-stream: from waste line to the collecting vessel, as well as in theupstream: from buffer line to the payloads after priming, and viceversa). Calibration of the pump(s) was done automatically via asub-program developed in house by LabVIEWTM. Configuration(a) of Fig. 1 displays an overview of the initial version of AIMSwhich features: (1) two separate but inter-connected peristalticpumps where one drives NA and the other dispenses the CS solu-tion, simultaneously, at the same flow rate; (2) priming of the tub-ings is done using the NA and CS solutions; (3) mixing takes placeboth during and after the acceleration phase of the pumps (accel-eration phase refers to the period of time during which the pumphead starts from stationary and reaches the target speed). Config-uration (b) of Fig. 1 displays the advanced version of the mixingsystem that features some technical, hardware and softwareimprovements: (1) only one peristaltic pump with a dual channelhead drives both CS and NA, simultaneously; (2) priming of thetubings is done with the buffer solution rather than the NA andCS solutions in order to minimize consumption of NA; (3) mixingtakes place after the acceleration phase of the pump, hence, poly-plexes are mixed only at the target speed by switching the pinchvalve fromwaste to the collection vessel once the final target speedwas reached.

3.2. Influence of mixing speed on the properties of CS/NA polyplexes

In order to examine the impact of mixing speed on the size andpolydispersity of final complexes, CS/NA polyplexes were mixed

A. Tavakoli Naeini et al. / Journal of Colloid and Interface Science 500 (2017) 253–263 257

via AIMS using flow rates ranging from 2 to 300 mL/min (1.7 to253 cm/s given 1/1600 as the tube inner diameter, ID). 300 mL/min is the maximum flow rate achievable using tubingID = 1/1600. Equal volumetric flow rates between the inlets wasused in all mixing experiments. Production of polyplexes was donein duplicates for every mixing flow rate tested. The dimensionlessReynolds number (Re) was calculated in the outlet as Re = UD/m,where U is fluid velocity, D is the tube inner diameter and m repre-sents the fluid kinematic viscosity. The Re range covered in thisstudy was 26–4000. In a cylindrical/tubular geometry with smoothwalls, Re � <2000 and Re � >4000 are indicative of laminar, andturbulent flow regime, respectively [48]. 2000 < Re < 4000 corre-sponds to transitional flow regime, which has characteristics ofboth laminar and turbulent regime. However, these values maynot accurately translate the flow regimes in our system, due tothe presence of peristaltic motions.

Literature reveals that the size and polydispersity of manuallymixed CS/pDNA and CS/siRNA polyplexes both increase signifi-cantly at NA concentrations higher than about 0.1 mg/mL, and0.2 mg/mL, respectively, [9,27,49–50] hence, these concentrationswere taken to optimize the speed of mixing in AIMS. Both initialand advanced versions of AIMS were used for producing polyplexesby mixing 2 mL of CS and 2 mL of NA with either LS14 (ID = 1/1600)or LS13 (ID = 1/3200) tubing. For comparison purposes, manual mix-ing was performed through classical pipetting by addition of100 mL of CS into 100 mL of NA followed by pipetting up and downten times (Fig. 2).

3.2.1. Faster mixing results in smaller and more homogenous CS/siRNApolyplexes

As shown in Fig. 2a, a marked reduction in size and polydisper-sity of CS/siRNA polyplexes was observed when Re increased. Z-average diameter and PDI dropped from 235 to 75 nm, and from0.40 to 0.17, respectively, using the initial version of AIMS in theRe range from 26 to 4000. Similar trend was observed using theadvanced version of AIMS in the same range of Re where Z-average diameter and PDI were reduced further from 230 to47 nm, and from 0.44 to 0.12, respectively.

In a purely laminar flow regime with smooth walls (e.g., usingsyringe pumps), where the streams move in parallel with no trans-verse movements or vortex, mixing relies solely on diffusion

Fig. 2. Influence of mixing speed on size and polydispersity of CS/siRNA polyplexes (a) anand siRNA concentration prior to mixing were 0.1 mg/mL, and 0.2 mg/mL, respectively,done using 2 mL of CS and 2 mL of NA. Manual mixing was done by addition of 100 mLrepresent standard deviation between the duplicates.

[40,46]. The speed of molecular diffusion is a constant and intrin-sically slow process; hence relying on diffusion results in incom-plete mixing before fluid enters the collecting vessel, and non-homogenous complexes due to concentration gradient. As men-tioned previously, the natural peristaltic actions of the pump(s)promote(s) diffusion by stretching the interfacial area [46–47].The stretched contact area is associated with shorter diffusion pathso that at each snapshot of the mixture throughout the outlet tub-ing, the two polyelectrolytes are brought closer to each other tobind quickly and form the complex [42]. Increasing the frequencyof rotation of the peristaltic pump (faster mixing), augments theintensity of pulsatile effect which enhances the overall mixing pro-cess by further stretching the interface area which favors produc-tion of smaller and more homogenous polyplexes. Reduction insize and PDI of CS/siRNA polyplexes could be also due to introduc-tion of chaotic advection (irregular/random transport of matter byflow) in the mixing process at high mixing flow rates. Chaoticadvection can be introduced to a system by either implementingobstacles in the flow channel, [40] or non-aligned inputs that gen-erates vertical flow in the mixer, [51] or turbulence and vortex byhigh speed streams [41–42]. By folding the streams, chaotic advec-tion offers enhanced mixing performance due to transverse com-ponents of flow. Not only intertwinement of the two streams bychaotic advection generates additional contact area, but alsoquickly alters and refreshes the interfacial area that accelerates dif-fusion. Here, it was revealed that by increasing the speed(Re > 2140), mixing efficiency increases and translates into smallerandmore homogenous CS/siRNA polyplexes in non-laminar regime(Fig. 2a).

AIMS led to better reproducibility in production of CS/siRNApolyplexes versus the manual method except for the very low Reof 26. AIMS produced much smaller and more homogenous CS/siRNA polyplexes in the non-laminar regime (Re between 2140and 4000) compared to manual polyplexes, with Z-average diame-ter and PDI from 84 nm to 75 nm and, 0.18 to 0.17 in the initial ver-sion of AIMS, and 65 nm to 47 nm and, 0.18 to 0.12 using theadvanced version of AIMS, versus 101 nm and 0.2, respectivelyfor manual mixing. The largest improvement was at the highestRe = 4000 with the advanced version of AIMS to produce CS/siRNApolyplexes with Z-average of 47 nm, and PDI of 0.12, while manualmixing led to Z-average of 101 nm, and PDI of 0.2. This suggests

d CS/pDNA polyplexes, (b) produced by initial and advanced versions of AIMS. pDNAat molar ratio of CS amine to NA phosphate ratio of 5 (N:P = 5). In-line mixing wasof CS into 100 mL of NA, followed by pipetting up and down ten times. Error bars

258 A. Tavakoli Naeini et al. / Journal of Colloid and Interface Science 500 (2017) 253–263

that in-line mixing in a non-laminar regime is more efficient thanmanual mixing, which is likely due to the role of chaotic advectionwhich offers enhanced mixing performance due to transverse com-ponents of flow.

Moreover, as shown in Fig. 2a, smaller and more homogenousCS/siRNA polyplexes were obtained using the advanced versionof AIMS. In order to better understand the source of this reductionin size and polydispersity of polyplexes, each of the three modifica-tions of the advanced version of AIMS (as described above) was iso-lated and studied separately, where oligonucleotide (ODN) wasused to mimic siRNA as a lower-cost alternative. It was found thatmixing during the acceleration phase of the pump was the sourceof increase in size and PDI of CS/ODN polyplexes (see SupportingInformation, Fig. S1). This can be attributed to the fact that CS/siRNA polyplexes are sensitive to the speed (Fig. 2a) and mixingin the acceleration phase of the pump, where the mixing takesplace at different speeds results in the formation CS/siRNA poly-plexes with various sizes.

3.2.2. CS/pDNA polyplexes properties are not very sensitive to mixingflow rate

For 26 < Re < 4000, Z-average diameter and PDI of CS/pDNApolyplexes varied only slightly between 186 and 112 nm, andbetween 0.25 and 0.23, respectively. As shown in Fig. 2b, sizeand PDI of CS/pDNA polyplexes change very smoothly throughoutthe tested range of Re. Davies et al. reported the same behavior forPEI/pDNA polyplexes in the moderate range of 568 < Re < 2840when a static mixer was applied in the downstream [36]. This sug-gests that the pulsatile action of the peristaltic pumps producessufficient perturbation and interface stretching for efficient mixingand production of homogeneous small sized pDNA/CS polyplexes,even at very low speeds. Due to its high MW (�4 MDa, with6.4 kbp), pDNA has a low diffusivity, hence, while pDNA has lim-ited mobility, mostly CS molecules diffuse to contribute to complexformation. Once association starts on a pDNA supercoil, the chanceof collision with subsequent CS is much higher than colliding withanother pDNA molecule, which limits random bridging of poly-plexes and heterogeneity. Rapid saturation of pDNA binding sitesby CS molecules limits the heterogeneity of the CS/pDNA poly-plexes, even when advection is limited as pDNA does not diffusequickly.

On the other hand, the stronger dependence of CS/siRNA poly-plexes size upon the mixing speed is likely due to high diffusivityof siRNA molecules. siRNA as a much smaller molecule(MW � 14 kDa with only 21 bp duplex) has a characteristic diffu-sion time scale much shorter than that of pDNA. Hence, in additionto CS, siRNA diffusion significantly contributes to the complex for-mation. Thus, in order to achieve efficient mixing, stronger advec-tion is required to eventually saturate siRNA with excess CSmolecules rather than colliding with another siRNA.

It was also observed that in-line CS/pDNA polyplexes (regard-less of the AIMS configuration) have similar properties and similarreproducibility to the manually prepared polyplexes (Fig. 2b). Thiscould be again explained by low diffusivity of pDNA and limitedsensitivity of CS/pDNA polyplexes to the mixing speed, suggestingthat even the low advection provided by manual mixing is suffi-cient for having an efficient mixing.

3.3. Mixing patterns and tube length have no impact on the propertiesof CS/NA polyplexes

Two other mixers were tested at relatively low speed (Re = 266),as well as in the highest mixing speed (Re = 4000): Y-connectorwas replaced by either a T-connector or a cross connector. More-over, five mixing lengths with the output downstream after theconnector, were tested (25, 35, 45, 55, and 65 cm) at the highest

speed, Re = 4000. We found that the mixing pattern and mixinglength have no effect or a very slight effect on polyplex properties(data not shown), suggesting that pulsatile actions and transverseadvection are sufficient for mixing over 25 cm.

3.4. Influence of NA concentration on the properties of CS/NApolyplexes

Based on the results of Fig. 2, the highest Re (4000) was chosenfor further mixings and investigation of the impact of NA concen-tration on properties of polyplexes. Here, NA concentration wasvaried from 0.01 to 0.4 mg/mL. Equal flow rate between the inletswas maintained. Both configurations of AIMS were tested. For com-parison purposes, manual mixing was done at every concentrationby pipetting, as described in the method section. Polyplexes wereprepared in duplicate for each mixing condition.

3.4.1. Mixing in dilute regime results in small and homogeneouspolyplexes

As shown in Fig. 3, regardless of the type of NA (siRNA or pDNA),mixing method (manual or in-line), and configuration of AIMS (ini-tial or advanced), concentration of NA prior to mixing increases thesize and PDI of polyplexes. The increase in polyplex size and PDIwith increasing NA concentration has been reported in the litera-ture [31,33–34]. Z-average diameter for CS/siRNA polyplexes pro-duced with AIMS varied from 53 to 95 nm in the concentrationrange of 0.01–0.4 mg/mL (Fig. 3a), while for CS/pDNA Z-averagediameter increased from 80 to 261 nm in the concentration rangeof 0.01–0.3 mg/mL (Fig. 3b), respectively. The small increase in Z-average and PDI of polyplexes prepared at the very lowest NA con-centration (0.01 mg/mL) compared to 0.05 mg/mL can beexplained by the low intensity of the scattering signal. When thenumber of particles is very low, the particle count (particles recog-nized by the DLS instrument) is too low for the instrument toderive a precise size distribution. Independent of the mixingmethod, macroscopic flocculation in CS/pDNA polyplexes wasobserved at 0.4 mg/mL, however, such behavior was not foundfor CS/siRNA polyplexes up to the highest concentration tested(0.4 mg/mL). It is already reported in the literature that for produc-tion of stable and small size homogenous complexes, the mixing oftwo polyelectrolytes has to be done in dilute regime [9–10,19,52].In fact, there is a significant difference between the characteristicsof a dilute polymer solution (where polymer chains are completelyseparated) and a semi-dilute to concentrated solution (where poly-mer chains are overlapped). Generally, the transition concentrationof polyelectrolyte solutions from dilute to semi-dilute regimes isdefined as the overlap concentration (C⁄) [53]. The closer the start-ing concentration of polyelectrolytes to their C⁄, the bigger the sizeand increased polydispersity of the final complexes.

The maximum concentration of pDNA for the preparation ofsmall size uniform polyplexes is reported as �0.1 mg/mL, [50]while production of small sized lipoplexes using relatively highconcentration of siRNA (0.59 mg/mL) was reported by Belliveauet al. [31]. As shown in Fig. 3b, regardless of the mixing method(manual vs in-line) and configuration of AIMS (initial vs advanced),the onset pDNA concentration for increasing size and PDI of CS/pDNA polyplexes is clearly 0.1 mg/mL. However, with siRNA AIMSallowed production of relatively small and homogeneous CS/siRNApolyplexes at increased concentrations of 0.3 mg/mL and even0.4 mg/mL. The possibility of producing CS/siRNA polyplexes athigher concentrations could be attributed to the smaller size ofsiRNA (higher C⁄) as compared to pDNA. Several techniques areused to measure the overlap concentration C⁄ of polymeric solu-tions, such as small angle light scattering, osmometry, viscometry,light scattering and nuclear magnetic resonance [54–55]. Accord-ing to our calculations, while C⁄ for CS 92-10 and pDNA are close

Fig. 3. Influence of NA mixing concentration on size and polydispersity of CS/siRNA polyplexes (a) and CS/pDNA polyplexes, (b) produced manually and by AIMS (using bothinitial and advanced versions), at molar ratio of CS amine to NA phosphate ratio of 5 (N:P = 5). In-line mixing was done using 2 mL of CS and 2 mL of NA at Re = 4000. Manualmixing was done by addition of 100 mL of CS into 100 mL of NA, followed by pipetting up and down for ten times. CS/pDNA polyplexes severely aggregate at �0.4 mg/mL. Errorbars represent standard deviation between the duplicates.

A. Tavakoli Naeini et al. / Journal of Colloid and Interface Science 500 (2017) 253–263 259

to each other and fairly low (C⁄ � 3.6 mg/mL and �1.36 mg/mL forchitosan 92-10 and pDNA, respectively), C⁄ for siRNA is relativelyhigh at �690 mg/mL, which permits increased siRNA mixing con-centration before reaching flocculation (see Supporting Informa-tion for C⁄ calculations).

In order to further investigate the impact of C⁄ on the highestpossible mixing concentrations, a smaller CS molecule (Mn = 2kD)was used to prepare CS/siRNA polyplexes. Being a smaller chain,CS92-2 has a smaller radius of gyration (Rg), hence, has a higherC⁄. It was shown that CS92-2 allows for production of small sizehomogeneous polyplexes at increased concentrations of siRNA,up to 0.6 mg/mL (see Supporting Information, Fig. S2), confirmingthe direct impact of chain length on the mixing concentrations.Using manual method, the maximum siRNA concentration for pro-ducing relatively small size polyplexes is clearly 0.2 mg/mL, wherea sharp increase in size of polyplexes was observed at higher con-centrations. However, using the advanced version of AIMS smalland monodisperse polyplexes were successfully prepared at ele-vated concentration of 0.4 mg/mL.

3.4.2. CS/siRNA polyplexes are smaller than CS/pDNA polyplexesAt a given concentration, CS/siRNA polyplexes were generally

smaller than CS/pDNA polyplexes. For example, at an NA mixingconcentration of 0.1 mg/mL, CS/siRNA and CS/pDNA polyplexeshave Z-average of 40 nm and 112 nm, respectively. This is mostprobably due to the fact that pDNA is a larger molecule comparedto siRNA, which naturally leads to larger complexes. Yet, the largersize of CS/pDNA polyplexes could be attributed to the slower diffu-sion of pDNA molecules that leads to broader distribution of pre-complexes which therefore induces formation of polyplexes withmore than one pre-complex [41]. As shown in Fig. 3b, the configu-ration of AIMS has no significant influence on the properties of CS/pDNA polyplexes. This was also observed previously in Fig. 2bwhere the properties of CS/pDNA polyplexes are independent ofthe AIMS configuration. However, in the case of CS/siRNA poly-plexes (Fig. 3a), the advanced version of AIMS produced smallerand more monodisperse CS/siRNA polyplexes at the given concen-trations, which again confirms their sensitivity to mixing speed asdescribed in Fig. 2a. Compared to manual mixing, production of CS/siRNA polyplexes with AIMS at Re = 4000 (regardless of the AIMSconfigurations) provided better reproducibility, smaller size, lower

PDI, and production of small size homogeneous polyplexes at ele-vated concentrations.

3.5. Surface charge density of CS/NA polyplexes increases with NAmixing concentration

Independent of the type of NA, concentration of polyelectrolytesprior to mixing also influenced the zeta potential (ZP) of poly-plexes: the higher the NA concentration prior to mixing, the higherthe electrophoretic mobility/calculated ZP of final complexes(Fig. 4). In-line CS/siRNA had ZP varying between +10.5 and+21 mV in the concentration range of 0.01–0.4 mg/mL Fig. 4a),whereas for in-line CS/pDNA the recorded ZP was from +5 to+16.5 mV in the concentration range of 0.01–0.3 mg/mL (Fig. 4a).As shown previously in Fig. 3, polyplex size increases with increas-ing NA concentration, so both ZP and size increase with mixingconcentration. It is worth mentioning that for the polyplexes testedand analysis conditions used in this study, the ratio of the particlesize to the Debye length changes with particle size and in all con-ditions, does not strictly meet the criteria for Smoluchowski modelto apply precisely (i.e. ja � 1, and the Henry function is notexactly 1.5 as explained in the methods). In order to examine if sizevariations could account, at least partially, for the observed ZP vari-ations, calculations using the complete Henry function were alsoperformed (data not shown). These calculations reveal that evenif the calculated ZP changes slightly versus that obtained with theSmoluchowski equation, the general trend observed is unchanged,namely, ZP increases as mixing concentration increases. Therefore,although ZP is not a direct measurement of surface charge, [56] ourresults indicate that the surface charge density of the polyplexesincreases with mixing concentration, suggesting that more chi-tosan is incorporated in polyplexes when mixing concentration ishigher. This could be the result of faster association kinetics athigher mixing concentrations.

3.6. Morphology of the CS/pDNA and CS/siRNA polyplexes

Morphology of polyplexes was assessed by Transmission Elec-tron Microscopy (TEM). CS/pDNA, and CS/siRNA polyplexes wereprepared both manually and in-line using the advanced versionof AIMS at NA starting concentrations of 0.1 mg/mL and 0.2 mg/

Fig. 4. Influence of NA mixing concentration on zeta potential of CS/pDNA polyplexes (a) and CS/siRNA polyplexes, (b). In-line mixing was done using the advanced version ofAIMS: 2 mL of CS and 2 mL of NA at Re = 4000. Manual mixing: pipetting 100 mL of CS into 100 mL of NA. Zeta potential measurements were done in 0.5% trehalose and 3.5 mMhistidine (Ionic strength = 1 mM). Error bars represent standard deviation between the duplicates.

260 A. Tavakoli Naeini et al. / Journal of Colloid and Interface Science 500 (2017) 253–263

mL, respectively. TEM images show CS/pDNA polyplexes withmixed morphology (spherical, toroidal and rod like) when pro-duced both manually and using the AIMS (Fig. 5a and b). This mor-phology of CS/pDNA polyplexes was reported in literature [57–58].On the other hand, CS/siRNA polyplexes were generally sphericalregardless of the method of preparation (Fig. 5c and d). The spher-ical shape of CS/siRNA polyplexes was also reported previously[16–18]. However, CS/siRNA polyplexes produced using AIMS

Fig. 5. Transmission Electron Microscopy (TEM) images of CS/pDNA polyplexes (a and b)advanced version of AIMS (b and d). pDNA and siRNA concentration prior to mixing wereCS and 2 mL of NA at 150 mL/min. Manual mixing was done by pipetting 100 mL of CS i

(Fig. 5d) were smaller and more uniform than those prepared man-ually (Fig. 5c), in agreement with DLS results.

3.7. CS/siRNA polyplexes produced in-line are as bioactive as manualpolyplexes

In order to test the suitability of AIMS to produce bioactive CS/siRNA polyplexes, the hepatocellular carcinoma cell line HepG2

and CS/siRNA polyplexes (c and d), produced both manually (a and c) and using the0.1 mg/mL and 0.2 mg/mL, respectively. In-line mixing was done by mixing 2 mL ofnto 100 mL of NA.

Fig. 6. Bioactivity of Inline and manually mixed CS/siRNA polyplexes. HepG2 cells were transfected with either inline or manually mixed anti-ApoB CS/siRNA polyplexes at afinal siRNA concentration of 100 nM/well. Apob mRNA knockdown was assessed by qPCR using geometric averaging of the most stable reference genes (GAPDH and B2 M)and expressed relative to non-treated cells. Data represent the mean of two independent experiments (N = 2) with two technical replicate per experiment (n = 2). Analysis ofVariance (ANOVA) was performed using the SigmaPlot 13.0 statistical package.

A. Tavakoli Naeini et al. / Journal of Colloid and Interface Science 500 (2017) 253–263 261

was transfected in vitro and Apolipoprotein B (ApoB) mRNA knock-down assessed using quantitative real time polymerase chain reac-tion (qPCR). Cells were transfected with polyplexes prepared usingthe advanced version of AIMS at Re = 4000, and different siRNAmixing concentration (0.05–0.4 mg/mL). Manually prepared poly-plexes were also transfected for comparison purposes.DharmaFECT

�2, a commercially available lipoplex, was used as

positive control. Naked siRNA, mock chitosan and non-treated cellswere used as negative controls.

ApoB mRNA knockdown reached 40–55% when produced viaAIMS (Fig. 6). Statistical analysis revealed no significant effect of

Fig. 7. Correlation between bioactivity and physico-chemical parameters of inline and(average) and polyplexe size (average). (b) Correlation between ApoB mRNA knockdownanalysis was performed with SigmaPlot 13.0 statistical package.

mixing siRNA concentrations on gene knockdown. Compared tothe inline mixed polyplexes, manually prepared polyplexesresulted in slightly lower ApoB mRNA knockdown (�30–45%).These results are consistent with a previous study where CS/siRNApolyplexes achieved ApoB mRNA knockdown of �50% in HepG2when prepared manually at N:P 5 and siRNA concentration of0.05 mg/mL [5]. Although manually prepared polyplexes at siRNAconcentration of 0.2 mg/mL showed the lowest transfection effi-ciency (�30%), analysis of variance did not show statistical signif-icance compared to other siRNA mixing concentrations. CS-basedpolyplexes achieved comparable knockdown efficiency to

manually prepared polyplexes. (a) Correlation between ApoB mRNA knockdown(average) and polyplexe zeta potential (average). Regression and Pearson correlation

262 A. Tavakoli Naeini et al. / Journal of Colloid and Interface Science 500 (2017) 253–263

DharmaFECT�2 (Fig. 6), independent of the mixing method. We

have previously shown that manually prepared polyplexes achievecomparable knockdown efficiencies to DharmaFECT

�2 but with

lower toxicity [7,13,59]. Nanoparticle size and surface charge arecited in the literature to be important parameters for nanoparticleuptake and bioactivity [60–61]. In our study, polyplex size corre-lated moderately (r2 = 0.376) with knockdown efficiency (Fig. 7a),with smaller sized polyplexes inducing higher knockdown effi-ciency. This trend can be seen from results presented inFigs. 3a and 6 where inline mixed polyplexes demonstrate asmooth reduction in size with decreasing siRNA concentrationand a generally better efficiency at knocking down ApoB mRNAin vitro. The slight correlation suggests that uptake of smaller poly-plexes could be higher, but determining if it is really the casewould require further investigation and is beyond the scope of thisstudy. Interestingly, our results also indicate that ZP does not cor-relate (r2 = 0.060) with improved bioactivity (Fig. 7b). The slightcorrelation between size and knock-down as well as the lack ofcorrelation between ZP and knock-down could be explained byserum dependent size stabilization of polyplexes occurringthrough rapid protein corona formation, which will also rapidlyalter their surface charge and ZP [11,62]. The aggregation due toprotein binding upon transfection, might either inhibit, or facilitatethe cellular uptake [63]. It is worth mentioning that ZP was mea-sured in low ionic strength (1 mM) buffer and not at �150 mMas in culture medium, and it could contribute to the lack of corre-lation between ZP and knock-down. Additionally, the somewhatlimited range of both size and ZP values tested could also explainthe limited correlations observed.

Altogether our data show that AIMS is able to produce smallsized polyplexes that retain their physico-chemical integrity andbioactivity as demonstrated by equal or superior performance tomanually prepared polyplexes.

4. Conclusion

A fully automated in-line mixing platform was developed forproduction of large batches of CS-based polyplexes to overcomethe drawbacks of conventional manual mixing methods that arenot only poorly controlled but also restricted to small volumes.This scaled-up platform addresses the risk of batch to batchinter-user and intra-user variability, resulting in a controlled qual-ity of large quantities of polyplexes. While manual mixing cannotpractically deliver more than 2 mL of product, polyplexes preparedvia AIMS are expected to have similar physico-chemical propertiesand bioactivity when produced in batch sizes up to a litre, 10 L, and100 L. Size and polydispersity of CS/siRNA polyplexes can be sim-ply controlled via the mixing flow rate and mixing concentrations.We found that by increasing NA concentration, Z-average diameterand the PDI of the polyplexes increase. Moreover, AIMS is capableof production of smaller and more homogeneous CS/siRNA poly-plexes using a non-laminar flow regime compared to manual poly-plexes. The established AIMS allowed production of well-definedCS/siRNA polyplexes with maintained bioactivity at every siRNAmixing concentration tested, demonstrating the suitability of thissystem for large scale production of polyplexes intended for pre-clinical and clinical applications.

Acknowledgment

The authors thank Dr. R. K. Panicker for help with TransmissionElectron Microscopy, and M. Bail for assistance in qPCR experi-ments. This research was supported by the Natural Sciences andEngineering Research Council (NSERC) and ANRis PharmaceuticalsInc.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.jcis.2017.04.013.

References

[1] H. Dautzenberg, J. Hartmann, S. Grunewald, F. Brand, Stoichiometry andstructure of polyelectrolyte complex particles in diluted solutions, Ber. Bunsen.Phys. Chem. 100 (6) (1996) 1024–1032.

[2] M. Muller, Sizing, shaping and pharmaceutical applications of polyelectrolytecomplex nanoparticles, in: Z.A. Rogovin (Ed.), Advances in Polymer Science,Wiley, 1974.

[3] M. Muller, B. Kessler, J. Frohlich, S. Poeschla, B. Torger, Polyelectrolyte complexnanoparticles of poly(ethyleneimine) and poly(acrylic acid): preparation andapplications, Polym.-Basel 3 (2) (2011) 762–778.

[4] A.F. Thunemann, M. Muller, H. Dautzenberg, J.F.O. Joanny, H. Lowen,Polyelectrolyte complexes, Adv. Polym. Sci. 166 (2004) 113–171.

[5] M. Alameh, M. Jean, D. Dejesus, M.D. Buschmann, A. Merzouki, Chitosanase-based method for RNA isolation from cells transfected with chitosan/siRNAnanocomplexes for real-time RT-PCR in gene silencing, Int. J. Nanomed. 5(2010) 473–481.

[6] M.D. Buschmann, A. Merzouki, M. Lavertu, M. Thibault, M. Jean, V. Darras,Chitosans for delivery of nucleic acids, Adv. Drug Deliv. Rev. 65 (9) (2013)1234–1270.

[7] M. Jean, M. Alameh, D. De Jesus, M. Thibault, M. Lavertu, V. Darras, M. Nelea, M.D. Buschmann, A. Merzouki, Chitosan-based therapeutic nanoparticles forcombination gene therapy and gene silencing of in vitro cell lines relevant totype 2 diabetes, Eur. J. Pharm. Sci. 45 (1–2) (2012) 138–149.

[8] M. Jean, F. Smaoui, M. Lavertu, S. Methot, L. Bouhdoud, M.D. Buschmann, A.Merzouki, Chitosan-plasmid nanoparticle formulations for IM and SC deliveryof recombinant FGF-2 and PDGF-BB or generation of antibodies, Gene Ther. 16(9) (2009) 1097–1110.

[9] M. Lavertu, S. Methot, N. Tran-Khanh, M.D. Buschmann, High efficiency genetransfer using chitosan/DNA nanoparticles with specific combinations ofmolecular weight and degree of deacetylation, Biomaterials 27 (27) (2006)4815–4824.

[10] H.Q. Mao, K. Roy, V.L. Troung-Le, K.A. Janes, K.Y. Lin, Y. Wang, J.T. August, K.W.Leong, Chitosan-DNA nanoparticles as gene carriers: synthesis,characterization and transfection efficiency, J. Control. Release 70 (3) (2001)399–421.

[11] S. Nimesh, M.M. Thibault, M. Lavertu, M.D. Buschmann, Enhanced genedelivery mediated by lowmolecular weight chitosan/DNA complexes: effect ofpH and serum, Mol. Biotechnol. 46 (2) (2010) 182–196.

[12] M. Thibault, S. Nimesh, M. Lavertu, M.D. Buschmann, Intracellular traffickingand decondensation kinetics of chitosan-pDNA polyplexes, Mol. Ther. 18 (10)(2010) 1787–1795.

[13] M. Alameh, D. Dejesus, M. Jean, V. Darras, M. Thibault, M. Lavertu, M.D.Buschmann, A. Merzouki, Low molecular weight chitosan nanoparticulatesystem at low N: P ratio for nontoxic polynucleotide delivery, Int J Nanomed. 7(2012) 1399–1414.

[14] C. Corbet, H. Ragelle, V. Pourcelle, K. Vanvarenberg, J. Marchand-Brynaert, V.Preat, O. Feron, Delivery of siRNA targeting tumor metabolism using non-covalent PEGylated chitosan nanoparticles: identification of an optimalcombination of ligand structure, linker and grafting method, J. Control.Release 223 (2016) 53–63.

[15] S. Gao, S. Hein, F. Dagnaes-Hansen, K. Weyer, C. Yang, R. Nielsen, E.I.Christensen, R.A. Fenton, J. Kjems, Megalin-mediated specific uptake ofchitosan/siRNA nanoparticles in mouse kidney proximal tubule epithelialcells enables AQP1 gene silencing, Theranostics 4 (10) (2014) 1039–1051.

[16] P. Holzerny, B. Ajdini, W. Heusermann, K. Bruno, M. Schuleit, L. Meinel, M.Keller, Biophysical properties of chitosan/siRNA polyplexes: profiling thepolymer/siRNA interactions and bioactivity, J. Control. Release 157 (2) (2012)297–304.

[17] K.A. Howard, U.L. Rahbek, X. Liu, C.K. Damgaard, S.Z. Glud, M.O. Andersen, M.B.Hovgaard, A. Schmitz, J.R. Nyengaard, F. Besenbacher, J. Kjems, RNAinterference in vitro and in vivo using a novel chitosan/siRNA nanoparticlesystem, Mol. Ther. 14 (4) (2006) 476–484.

[18] X. Liu, K.A. Howard, M. Dong, M.O. Andersen, U.L. Rahbek, M.G. Johnsen, O.C.Hansen, F. Besenbacher, J. Kjems, The influence of polymeric properties onchitosan/siRNA nanoparticle formulation and gene silencing, Biomaterials 28(6) (2007) 1280–1288.

[19] J. Malmo, H. Sorgard, K.M. Varum, S.P. Strand, SiRNA delivery with chitosannanoparticles: molecular properties favoring efficient gene silencing, J.Control. Release 158 (2) (2012) 261–268.

[20] S. Mao, W. Sun, T. Kissel, Chitosan-based formulations for delivery of DNA andsiRNA, Adv. Drug Deliv. Rev. 62 (1) (2010) 12–27.

[21] H. Ragelle, R. Riva, G. Vandermeulen, B. Naeye, V. Pourcelle, C.S. Le Duff, C.D’Haese, B. Nysten, K. Braeckmans, S.C. De Smedt, C. Jerome, V. Preat, Chitosannanoparticles for siRNA delivery: optimizing formulation to increase stabilityand efficiency, J. Control. Release 176 (2014) 54–63.

[22] H. Ragelle, G. Vandermeulen, V. Preat, Chitosan-based siRNA delivery systems,J. Control. Release 172 (1) (2013) 207–218.

A. Tavakoli Naeini et al. / Journal of Colloid and Interface Science 500 (2017) 253–263 263

[23] M.A. Raja, H. Katas, T. Jing Wen, Stability, intracellular delivery, and release ofsiRNA from chitosan nanoparticles using different cross-linkers, PLoS One 10(6) (2015) e0128963.

[24] C. Yang, L. Nilsson, M.U. Cheema, Y. Wang, J. Frokiaer, S. Gao, J. Kjems, R.Norregaard, Chitosan/siRNA nanoparticles targeting cyclooxygenase type 2attenuate unilateral ureteral obstruction-induced kidney injury in mice,Theranostics 5 (2) (2015) 110–123.

[25] M. Jean, M. Alameh, M.D. Buschmann, A. Merzouki, Effective and safe gene-based delivery of GLP-1 using chitosan/plasmid-DNA therapeuticnanocomplexes in an animal model of type 2 diabetes, Gene Ther. 18 (8)(2011) 807–816.

[26] F.C. MacLaughlin, R.J. Mumper, J.J. Wang, J.M. Tagliaferri, I. Gill, M. Hinchcliffe,A.P. Rolland, Chitosan and depolymerized chitosan oligomers as condensingcarriers for in vivo plasmid delivery, J. Control. Release 56 (1–3) (1998) 259–272.

[27] D. Veilleux, M. Nelea, K. Biniecki, M. Lavertu, M.D. Buschmann, Preparation ofconcentrated chitosan/DNA nanoparticle formulations by lyophilization forgene delivery at clinically relevant dosages, J. Pharm. Sci. 105 (1) (2016) 88–96.

[28] H. Katas, H.O. Alpar, Development and characterisation of chitosannanoparticles for siRNA delivery, J. Control. Release 115 (2) (2006) 216–225.

[29] M. Lu, Y.P. Ho, C.L. Grigsby, A.A. Nawaz, K.W. Leong, T.J. Huang, Three-dimensional hydrodynamic focusing method for polyplex synthesis, ACS Nano8 (1) (2014) 332–339.

[30] I.V. Zhigaltsev, N. Belliveau, I. Hafez, A.K. Leung, J. Huft, C. Hansen, P.R. Cullis,Bottom-up design and synthesis of limit size lipid nanoparticle systems withaqueous and triglyceride cores using millisecond microfluidic mixing,Langmuir 28 (7) (2012) 3633–3640.

[31] N.M. Belliveau, J. Huft, P.J. Lin, S. Chen, A.K. Leung, T.J. Leaver, A.W. Wild, J.B.Lee, R.J. Taylor, Y.K. Tam, C.L. Hansen, P.R. Cullis, Microfluidic synthesis ofhighly potent limit-size lipid nanoparticles for in vivo delivery of siRNA, Mol.Ther. Nucleic Acids 1 (2012) e37.

[32] P.M. Valencia, O.C. Farokhzad, R. Karnik, R. Langer, Microfluidic technologiesfor accelerating the clinical translation of nanoparticles, Nat. Nanotechnol. 7(10) (2012) 623–629.

[33] J.C. Kasper, D. Schaffert, M. Ogris, E. Wagner, W. Friess, The establishment of anup-scaled micro-mixer method allows the standardized and reproduciblepreparation of well-defined plasmid/LPEI polyplexes, Eur. J. Pharm. Biopharm.77 (1) (2011) 182–185.

[34] O. Zelphati, C. Nguyen, M. Ferrari, J. Felgner, Y. Tsai, P.L. Felgner, Stable andmonodisperse lipoplex formulations for gene delivery, Gene Ther. 5 (1998)1272–1282.

[35] J. Clement, K. Kiefer, A. Kimpfler, P. Garidel, R. Peschka-Suss, Large-scaleproduction of lipoplexes with long shelf-life, Eur. J. Pharm. Biopharm. 59 (1)(2005) 35–43.

[36] L.A. Davies, G.A. Nunez-Alonso, H.L. Hebel, R.K. Scheule, S.H. Cheng, S.C. Hyde,D.R. Gill, A novel mixing device for the reproducible generation of nonviralgene therapy formulations, Biotechniques 49 (3) (2010) 666–668.

[37] T.A. Balbino, A.R. Azzoni, L.G. de la Torre, Microfluidic devices for continuousproduction of pDNA/cationic liposome complexes for gene delivery andvaccine therapy, Colloids Surf. B Biointerfaces 111 (2013) 203–210.

[38] C.L. Grigsby, Y.P. Ho, C. Lin, J.F. Engbersen, K.W. Leong, Microfluidic preparationof polymer-nucleic acid nanocomplexes improves nonviral gene transfer, Sci.Rep. 3 (2013) 3155.

[39] A.K. Leung, I.M. Hafez, S. Baoukina, N.M. Belliveau, I.V. Zhigaltsev, E.Afshinmanesh, D.P. Tieleman, C.L. Hansen, M.J. Hope, P.R. Cullis, Lipidnanoparticles containing siRNA synthesized by microfluidic mixing exhibitan electron-dense nanostructured core, J. Phys. Chem. C Nanomater. Interfaces116 (34) (2012) 18440–18450.

[40] A.D. Stroock, S.K. Dertinger, A. Ajdari, I. Mezic, H.A. Stone, G.M. Whitesides,Chaotic mixer for microchannels, Science 295 (5555) (2002) 647–651.

[41] C. Ankerfors, S. Ondaral, L. Wagberg, L. Odberg, Using jet mixing to preparepolyelectrolyte complexes: complex properties and their interaction withsilicon oxide surfaces, J. Colloid Interface Sci. 351 (1) (2010) 88–95.

[42] B.K. Johnson, R.K. Prud’homme, Chemical Processing and Micromixing inConfined Impinging Jets, AIChE J. 49 (9) (2003) 2264–2282.

[43] M. Lavertu, Z. Xia, A.N. Serreqi, M. Berrada, A. Rodrigues, D. Wang, M.D.Buschmann, A. Gupta, A validated 1H NMR method for the determination ofthe degree of deacetylation of chitosan, J. Pharm. Biomed. Anal. 32 (6) (2003)1149–1158.

[44] S. Nguyen, F.M. Winnik, M.D. Buschmann, Improved reproducibility in thedetermination of the molecular weight of chitosan by analytical size exclusionchromatography, Carbohyd. Polym. 75 (3) (2009) 6.

[45] S.A. Bustin, V. Benes, J.A. Garson, J. Hellemans, J. Huggett, M. Kubista, R.Mueller, T. Nolan, M.W. Pfaffl, G.L. Shipley, J. Vandesompele, C.T. Wittwer, TheMIQE guidelines: minimum information for publication of quantitative real-time PCR experiments, Clin. Chem. 55 (4) (2009) 611–622.

[46] W.C. Jackson, F. Kuckuck, B.S. Edwards, A. Mammoli, C.M. Gallegos, G.P. Lopez,T. Buranda, L.A. Sklar, Mixing small volumes for continuous high-throughputflow cytometry: performance of a mixing Y and peristaltic sample delivery,Cytometry 47 (3) (2002) 183–191.

[47] R.A. Truesdell, P.V. Vorobieff, L.A. Sklar, A.A. Mammoli, Mixing of a continuousflow of two fluids due to unsteady flow, Phys. Rev. E: Stat., Nonlin, Soft MatterPhys. 67 (6 Pt 2) (2003) 066304.

[48] D. Green, R. Perry, Perry’s Chemical Engineers’ Handbook, eighth ed., McGraw-Hill Education, 2007.

[49] A.F. Thunemann, Polyelectrolytes with Defined Molecular Architecture II,Springer, Berlin Heidelberg, 2004.

[50] L. Xu, T. Anchordoquy, Drug delivery trends in clinical trials and translationalmedicine: challenges and opportunities in the delivery of nucleic acid-basedtherapeutics, J. Pharm. Sci. 100 (1) (2011) 38–52.

[51] M.A. Ansari, K.Y. Kim, K. Anwar, S.M. Kim, Vortex micro T-mixer with non-aligned inputs, Chem. Eng. J. 181 (2012) 6.

[52] E.A. Klausner, Z. Zhang, R.L. Chapman, R.F. Multack, M.V. Volin, Ultrapurechitosan oligomers as carriers for corneal gene transfer, Biomaterials 31 (7)(2010) 1814–1820.

[53] P.G. de Gennes, Scaling Concepts in Polymer Physics, Cornell University Press,1979.

[54] C. Wandrey, Concentration regimes in polyelectrolyte solutions, Langmuir 15(12) (1999) 4069–4075.

[55] X.S. Zhu, K.H. Choo, A novel approach to determine the molecular overlap ofpolyelectrolyte using an ultrafiltration membrane, Colloid Surf. A 312 (2–3)(2008) 231–237.

[56] T.L. Doane, C.H. Chuang, R.J. Hill, C. Burda, Nanoparticle zeta -potentials, Acc.Chem. Res. 45 (3) (2012) 317–326.

[57] S. Danielsen, K.M. Varum, B.T. Stokke, Structural analysis of chitosan mediatedDNA condensation by AFM: influence of chitosan molecular parameters,Biomacromolecules 5 (3) (2004) 928–936.

[58] S.P. Strand, S. Danielsen, B.E. Christensen, K.M. Varum, Influence of chitosanstructure on the formation and stability of DNA-chitosan polyelectrolytecomplexes, Biomacromolecules 6 (6) (2005) 3357–3366.

[59] M.G. Alameh, M. Lavertu, N. TranKhanh, C.-Y. Chang, F. Lesage, V. Darras, M.D.Buschmann, siRNA delivery with chitosan: influence of chitosan molecularweight, degree of deacetylation and amine to phosphate ratio on in vitrosilencing efficiency, hemocompatibility and in vivo biodistribution, J. Control.Release (2017) (submitted for publication).

[60] A.E. Nel, L. Madler, D. Velegol, T. Xia, E.M. Hoek, P. Somasundaran, F. Klaessig,V. Castranova, M. Thompson, Understanding biophysicochemical interactionsat the nano-bio interface, Nat. Mater. 8 (7) (2009) 543–557.

[61] K.K. Son, D. Tkach, D.H. Patel, Zeta potential of transfection complexes formedin serum-free medium can predict in vitro gene transfer efficiency oftransfection reagent, Biochem. Biophys. Acta. 1468 (1–2) (2000) 11–14.

[62] A. Albanese, W.C. Chan, Effect of gold nanoparticle aggregation on cell uptakeand toxicity, ACS Nano 5 (7) (2011) 5478–5489.

[63] S.R. Saptarshi, A. Duschl, A.L. Lopata, Interaction of nanoparticles withproteins: relation to bio-reactivity of the nanoparticle, J. Nanobiotechnol. 11(2013) 26.