Strategies to Guide the Manufacturing of Liposomes Using Microfluidics

pharmaceutics

Article The Impact of Solvent Selection: Strategies to Guide the Manufacturing of Liposomes Using Microﬂuidics

1, 1, 1,2 1 Cameron Webb †, Swapnil Khadke †, Signe Tandrup Schmidt , Carla B. Roces , Neil Forbes 1, Gillian Berrie 1 and Yvonne Perrie 1,*

1 Strathclyde Institute of Pharmacy and Biomedical Sciences, University of Strathclyde, 161 Cathedral Street, Glasgow G4 0RE, UK; [email protected] (C.W.); [email protected] (S.K.); [email protected] (S.T.S.); [email protected] (C.B.R.); [email protected] (N.F.); [email protected] (G.B.) 2 Department of Infectious Disease Immunology, Center for Vaccine Research, Statens Serum Institut, Artillerivej 5, 2300 Copenhagen S, Denmark * Correspondence: [email protected]; Tel.: +0141-548-2244 These authors contributed equally to this work. † Received: 26 October 2019; Accepted: 30 November 2019; Published: 4 December 2019

Abstract: The aim of this work was to assess the impact of solvent selection on the microfluidic production of liposomes. To achieve this, liposomes were manufactured using small-scale and bench-scale microfluidics systems using three aqueous miscible solvents (methanol, ethanol or isopropanol, alone or in combination). Liposomes composed of different lipid compositions were manufactured using these different solvents and characterised to investigate the influence of solvents on liposome attributes. Our studies demonstrate that solvent selection is a key consideration during the microfluidics manufacturing process, not only when considering lipid solubility but also with regard to the resultant liposome critical quality attributes. In general, reducing the polarity of the solvent (from methanol to isopropanol) increased the liposome particle size without impacting liposome short-term stability or release characteristics. Furthermore, solvent combinations such as methanol/isopropanol mixtures can be used to modify solvent polarity and the resultant liposome particle size. However, the impact of solvent choice on the liposome product is also influenced by the liposome formulation; liposomes containing charged lipids tended to show more sensitivity to solvent selection and formulations containing increased concentrations of cholesterol or pegylated-lipids were less influenced by the choice of solvent. Indeed, incorporation of 14 wt% or more of pegylated-lipid was shown to negate the impact of solvent selection.

Keywords: liposomes; microﬂuidics; solvents; formulation; particle size; pegylation; alcohol

1. Introduction Liposomes are a versatile group of nanoparticles, which are approved for use in the clinic to improve the delivery of cancer agents, antifungal agents and vaccine adjuvants [1,2]. Currently, most approved liposome products contain already-approved drugs (e.g., amphotericin B, doxorubicin, bupivacaine, verteporﬁn) [3]. Recently, there has been a rise in follow-on ‘nanosimilars’ or generic liposome products being approved for use. Indeed, this has resulted in new recommendations for the naming of liposomal medicines such that the qualiﬁer ‘liposomal’ or ‘pegylated liposomal’ should be added to the name. This recommendation, made jointly by the EMA’s human medicines committee and the Coordination Group for the mutual Recognition and Decentralised Procedures, aims to provide clearer distinctions between liposomal and non-liposomal formulations and hence avoid medication errors [4].

Pharmaceutics 2019, 11, 653; doi:10.3390/pharmaceutics11120653 www.mdpi.com/journal/pharmaceutics PharmaceuticsPharmaceutics 20192019,, 1111,, 653653 2 of 1615

However, despite the increase in their use, the manufacturing processes used in the production of liposomesHowever, remains despite therelatively increase unchanged, in their use, with the manufacturingtime-consuming, processes multi-stage used batch in the production ofprocedures liposomes being remains common. relatively Furthermore, unchanged, in many with time-consuming,liposome research multi-stage studies, small-scale batch production methods proceduresare still used being that common. do not offer Furthermore, realistic line-of-sight in many liposome to a manufacturing research studies, process. small-scale This methodsreduces our are stillability used to thattranslate do not liposome offer realistic technology line-of-sight from bench to a manufacturing to patient. To process. address This this, reduces new manufacturing our ability to translatemethods liposomeare being technology investigated, from benchwhich tocan patient. offer Toscale-independent address this, new manufacturingproduction. In methods particular, are beingmicrofluidics investigated, is being which widely can oinvestigatedffer scale-independent as a scale-independent production. In production particular, microfluidicsmethod (e.g., is [5–7]). being widelyThis method investigated generally as arelies scale-independent on the bottom productionup self-assembly method of the (e.g., liposom [5–7]).es This by methodcontrolled generally mixing reliesof lipids on dissolved the bottom in upan self-assemblyaqueous soluble of theorganic liposomes solvent by with controlled an aqueous mixing buffer of lipids[8,9]. Using dissolved this inmethod, an aqueous we have soluble demonstrated organic solvent the ability with to an entrap aqueous small bu moleculesffer [8,9]. [10,11], Using thisnucleic method, acids we[12] have and demonstratedproteins [13,14] the within ability liposomes. to entrap small We have molecules also demonstrated [10,11], nucleic the acids scale-independent [12] and proteins production [13,14] within of liposomes.liposomal adjuvants We have also using demonstrated this method the [15–18]. scale-independent production of liposomal adjuvants using this methodWhen adopting [15–18]. microfluidics as a production method for liposomes, the lipids are dissolved in an organicWhen adoptingsolvent (normally microfluidics alcohol), as a productionwhich is then method mixed for with liposomes, the aqueous the lipids phase are to dissolved promote innanoprecipitation an organic solvent and (normally liposome alcohol),production. which Therefor is thene, the mixed solvent with selected the aqueous must phasedissolve to promotethe lipid nanoprecipitationcomponent and be and water liposome soluble. production. The water Therefore, miscibility the of solvent organic selected solvents must is directly dissolve related the lipid to componentcarbon chain and length be water and soluble.surface tension The water as miscibilityshown in Figure of organic 1 (adapted solvents from is directly [19]). relatedAs the tolength carbon of chainthe carbon length chain and surface increases, tension the as polar shown OH in Figuregroup1 (adaptedbecomes froma smaller [ 19]). component As the length of of the the solvent carbon chainmolecule. increases, The solubility the polar and OH the group polarity becomes of alcohol a smaller decreases component correspondingly. of the solvent The molecule. solubility The of solubilityorganic solvents and the in polarity water ofis alcoholalso driven decreases by hydrogen-bonding correspondingly. Theinteractions solubility between of organic water solvents and insolvent. water For is also example, driven water by hydrogen-bonding and short-chain solvents interactions such betweenas ethanol water are completely and solvent. miscible For example, due to waterthe ability and short-chainof the ethanol solvents molecules such as to ethanol form hydrogen are completely bonds miscible with water due tomolecules the ability as of well the ethanolas with moleculeseach other to[20]. form However, hydrogen when bonds considering with water solvent molecules selection as wellfor use as in with microfluidics, each other [lipid20]. However,solubility whenand water considering miscibility solvent are not selection the only for factors. use in microfluidics, Whilst in the lipidpurification solubility process and water solvent miscibility is removed, are notworking the only with factors. solvents Whilst that have in the low purification toxicity potential process and solvent defined is removed, as Class working3 in the ICH with Q3C solvents (R6) that[21] (e.g., have ethanol low toxicity and isopropanol potential and (IPA)) defined is preferable as Class 3 followed in the ICH by Q3Cthose (R6) in Class [21] (e.g.,2 (e.g., ethanol methanol) and isopropanol(Figure 1). (IPA)) is preferable followed by those in Class 2 (e.g., methanol) (Figure1).

Figure 1.1. The relationship between aqueous solubility andand the number of carbon atoms inin alcoholsalcohols determined by by surface surface tension tension (adapted (adapted from from [19]) [19 and]) and physical physical properties properties of methanol, of methanol, ethanol, ethanol, and andisopropanol. isopropanol.

Typically, IPAIPA hashas beenbeen usedused asas thethe lipidlipid solventsolvent inin microﬂuidicmicrofluidic processesprocesses [[8,22–24],8,22–24], withwith fewer studies exploring other solvents such as ethanol as a less toxic alternative for medicinal applications, whichwhich wouldwould alsoalso comply with routine industrial processesprocesses [[25].25]. However, thethe choice of solvent used during microfluidics may have an impact; Zook and Vreeland hypothesised that liposomes formed

Pharmaceutics 2019, 11, 653 3 of 15 duePharmaceutics to the alcohol2019, 11 ,and 653 aqueous phase mixing, increasing the polarity of the solvent, which3 of 16caused the lipids to become progressively less soluble and self-assemble into planar lipid bilayers [26]. The authors noted that as the planar bilayers increase in size, to reduce the surface area of hydrophobic during microfluidics may have an impact; Zook and Vreeland hypothesised that liposomes formed due chains exposed to the polar solvent around the edge of the discs, they bend and form spherical to the alcohol and aqueous phase mixing, increasing the polarity of the solvent, which caused the lipids liposomes (Figure 2) [12]. Given that this change in polarity will be dependent on the initial polarity to become progressively less soluble and self-assemble into planar lipid bilayers [26]. The authors of thenoted alcohol that asselected, the planar solvent bilayers polarity increase is a in critical size, to material reduce theattribute surface to area consider of hydrophobic as it will impact chains both on exposedinitial lipid to the solubility polar solvent but also around on the edgenanoprecipitation of the discs, they process. bend and Therefore, form spherical the aim liposomes of this work was(Figure to evaluate2)[ 12]. Giventhe impact that this of change solvent in polarityselection will on be the dependent formulation on the initialof liposomes polarity ofin the terms alcohol of their physicochemicalselected, solvent parameters, polarity is a critical such materialas particle attribute size, to polydispersity consider as it will index, impact liposome both on initial stability lipid and drug/proteinsolubility but loading. also on theTo nanoprecipitationachieve this, we process. initially Therefore, employed the aim a oflow this volume work was high to evaluate throughput microfluidicthe impact system of solvent as selectiona screening on the process, formulation followed of liposomes by more in detailed terms of formulation their physicochemical studies using largerparameters, volume suchmicrofluidic as particle systems. size, polydispersity index, liposome stability and drug/protein loading. To achieve this, we initially employed a low volume high throughput microfluidic system as a screening process, followed by more detailed formulation studies using larger volume microfluidic systems.

Figure 2. A hypothesised liposome formation mechanism (adapted from Zook and Vreeland [26]). Figure 2. A hypothesised liposome formation mechanism (adapted from Zook and Vreeland [26]). The process starts with the aggregation of lipids in discs (a). It is proposed that the hydrophobic chains The process starts with the aggregation of lipids in discs (a). It is proposed that the hydrophobic chains around the edges are stabilised by alcohol molecules. As the alcohol concentration reduces these lipid arounddiscs the bend edges (b), rapidly are stabilised close (c), by and alcohol form liposomes molecules. (d ).As the alcohol concentration reduces these lipid discs bend (b), rapidly close (c), and form liposomes (d). 2. Materials and Methods

2. Materials2.1. Materials and Methods

2.1. MaterialsSoy bean phosphatidylcholine (SoyPC), hydrogenated soy phosphatidylcholine (HSPC), 1,2-distearoyl-sn-glycero-3-phosphocholine (DSPC), 1,2-distearoyl-sn-glycero-3-phosphoethanolamine- NSoy-[methoxy(polyethylene bean phosphatidylcholine glycol)-2000] (SoyPC), (DSPE-PEG2k) hydrogenated soy were phosphatidylcholine obtained from (HSPC), Lipoid 1,2- distearoyl-(Ludwigshafen,sn-glycero-3-phosphocholine Germany). 1,2-dimyristoyl- (DSPC),sn 1,2-distearoyl--glycero-3-phospho-(1sn-glycero-3-phosphoethanolamine-0-rac-glycerol) (DMPG), N-[methoxy(polyethylene1,2-dioleoyl-sn-glycero-3-phosphoethanolamine glycol)-2000] (DSPE-PEG2k) (DOPE), 1,2-dimyristoyl- were obtainedsn-glycero-3-phosphocholine from Lipoid (Ludwigshafen, Germany).(DMPC), 1,2-dimyristoyl- 1-palmitoyl-2-oleoyl-glycero-3-phosphocholinesn-glycero-3-phospho-(1′-rac-glycerol) (POPC), 1,2-dioleoyl-3-trimethylammonium- (DMPG), 1,2-dioleoyl-sn-glycero- 3-phosphoethanolaminepropane (chloride salt) (DOTAP)(DOPE), were 1,2-dimyristoyl- purchased fromsn Avanti-glycero-3-phosphocholine polar lipids, Alabaster, AL,(DMPC), USA. 1- palmitoyl-2-oleoyl-glycero-3-phosphocholineCholesterol (Chol), chicken egg ovalbumin (OVA), (POP andC), propofol 1,2-dioleoyl-3-trimethylammonium-propane were acquired from Sigma-Aldrich (St. Louis, MO, USA). Phosphate buffered saline was acquired from Oxoid Ltd., Basingstoke, UK. Tris-base (chloride salt) (DOTAP) were purchased from Avanti polar lipids, Alabaster, AL, USA. Cholesterol was obtained from IDN Biomedical Inc. (Aurora, OH, USA). Methanol (MeOH), ethanol (EtOH) and (Chol), chicken egg ovalbumin (OVA), and propofol were acquired from Sigma-Aldrich (St. Louis, isopropanol (IPA) were obtained from Fisher Scientific, Loughborough, UK. All solvents and other MO,chemicals USA). Phosphate were used atbuffered analytical saline grade, was and acquired mQ-water from was Oxoid provided Ltd., by Basingstoke, an in-house system. UK. Tris-base was obtained from IDN Biomedical Inc. (Aurora, OH, USA). Methanol (MeOH), ethanol (EtOH) and isopropanol2.2. Microfluidic (IPA) Production were obtained of Liposomes from Fisher Scientific, Loughborough, UK. All solvents and other chemicals were used at analytical grade, and mQ-water was provided by an in-house system. 2.2.1. Initial Rapid Small-Scale Screening Studies 2.2. MicrofluidicThe liposomes Production were prepared of Liposomes with a phospholipid:cholesterol ratio of 3:1 (SoyPC and POPC) or 2:1 (HSPC, DMPC and DSPC) weight ratio. Pegylated liposomes (DSPC:Chol:DSPE-PEG2k) were prepared 2.2.1.at aInitial 2:1:1 wRapid/w. Anionic Small-Scale liposomes Screening were prepared Studies from DMPC:Chol:DMPG or DSPC:Chol:DMPG at 10:5:4 w/w and cationic liposomes were prepared from DOPE:DOTAP at 1:1 w/w, HSPC:Chol:DOTAP atThe 1:2:4 liposomesw/w and HSPC:Chol:DOTAP:DSPE-PEG2k were prepared with a phospholipid at 1:2:4:0.5:cholesterolw/w. These ratio formulations of 3:1 (SoyPC were and selected POPC) or 2:1 (HSPC, DMPC and DSPC) weight ratio. Pegylated liposomes (DSPC:Chol:DSPE-PEG2k) were prepared at a 2:1:1 w/w. Anionic liposomes were prepared from DMPC:Chol:DMPG or DSPC:Chol:DMPG at 10:5:4 w/w and cationic liposomes were prepared from DOPE:DOTAP at 1:1 w/w, HSPC:Chol:DOTAP at 1:2:4 w/w and HSPC:Chol:DOTAP:DSPE-PEG2k at 1:2:4:0.5 w/w. These

Pharmaceutics 2019, 11, 653 4 of 16 to represent commonly reported formulations. For microfluidic production, the lipids were dissolved in methanol, ethanol or isopropanol at 4 mg/mL. For initial rapid screening, the formulations were prepared using a SparkTM (Precision Nanosystems Inc., Vancouver, Canada). Disposable microfluidic chips were loaded with 31 µL lipid stock and 93 µL PBS or Tris buffer in the reaction chambers, respectively. The receiving chamber was filled with 124 µL PBS or Tris buffer (pH 7.4, 10 mM). With formulations prepared with OVA, the protein was added to the PBS at a concentration of 1 mg/mL, whilst for formulations with propofol, the drug was added to the lipid stock solution at 1 mg/mL. The formulations were processed at setting 8–10, and the product transferred to a glass vial and further diluted with 1000 µL PBS or deionised water. All formulations were prepared at room temperature.

2.2.2. Preparation of Liposomes Using a Bench-Scale System Liposomes were prepared with DSPC:Chol (2:1 to 8:1 w/w ratio; 11 to 33% cholesterol content) and DOPE:DOTAP (1:1 w/w ratio) and 4 mg/mL initial lipid concentration by using a NanoassemblrTM (Precision Nanosystems Inc.). Increasing levels of DSPE-PEG2k (0 to 25 wt%) were also tested with the DSPC:Chol formulation. A range of alcohol mixtures were tested as the organic phases: methanol/ethanol, methanol/IPA and ethanol/IPA at combinations of 100/0, 75/25, 50/50, 25/75 and 0/100 v/v%. Tris-buffer (10 mM, pH 7.4) was used as the aqueous phase at a flow rate ratio of 1:1 for the DOPE:DOTAP. For DSPC:Chol formulations, PBS at a flow rate ratio of 3:1 was used. Again, all formulations were prepared at room temperature.

2.3. Characterization of Particle Size, Polydispersity and Zeta Potential by Using Dynamic Light Scattering The particle sizes, measured as the hydrodynamic diameters, polydispersity indexes (PDI) and zeta potentials were measured by dynamic light scattering using a Zetasizer Nano ZS (Malvern Instruments Ltd., Worcestershire, UK) equipped with a 633 nm laser and a detection angle of 173◦. The samples were measured and the values of water were used for refractive indexes and viscosity. Zetasizer Software v.7.11 (Malvern Instruments Ltd.) was used for the acquisition of data.

2.4. Removal of Free Drug with Tangential Flow Filtration (TFF) The OVA and propofol in the drug loaded samples were removed by tangential flow filtration by using a Krosflo KR2i TFF system (Waltham, MA, USA) equipped with an mPES 300 kD or 750 kD column. The sample volume was 1 mL and the purification process was repeated 12 times. All samples were recovered at the initial volume.

2.5. Characterisation of Drug Loading OVA (43 kDa) and propofol (178 Da) were selected as a water soluble and bilayer soluble drug respectively to consider drug loading. These were selected as we previously investigated drug loading via microfluidics using these two moieties [11,13]. The content of propofol in propofol-loaded formulations after purification with TFF was determined using HPLC as described in [27]. Propofol loading was calculated as the percent of propofol in the liposome formulation after TFF purification to the initial concentration in the untreated sample. The content of OVA in the OVA loaded samples after purification with TFF was quantified using either the bicinchoninic acid assay (BCA) (Thermo Fisher Scientific, Waltham, MA, USA) or via HPLC using a modified published method [28]. A Jupiter 5 µm C5 300A column 4.6 mm i.d. 250 mm length (Phenomenex, Macclesfield, UK) was used with a × gradient flow (0.1% TFA in water (A), 0.1% TFA in acetonitrile (B)). A UV detector was fitted at 210 nm for all OVA loaded samples (retention time 10.6–14 min) whilst 280 nm was used for the protein release study (retention time 8–14 min). Pharmaceutics 2019, 11, 653 5 of 16

2.6. Morphological Characterisation of Liposomes via CryoTEM Samples were prepared by placing 5 µL of liposomes onto a 400-mesh lacey carbon-coated grid using single sided blotting for 2 s then immediately immersing the sample grid into nitrogen cooled ethane (100% ethane). The liposome morphology was then observed using the Joel Jem F-200 microscope (Joel, Tokyo, Japan) at liquid nitrogen temperature and 200 kV.

2.7. Liposome Stability Studies Neutral liposomes (DSPC:Chol 2:1 w/w) were produced using microﬂuidics at a 3:1 FRR and 15 mL/min TFR (4 mg/mL initial lipid concentration) in either MeOH, EtOH or IPA. Solvent was removed by TFF (12 mL wash cycle per mL of sample). Liposome suspensions were then stored at 2–8 ◦C in the fridge and their size and PDI was measured over 7 days with the Zetasizer Nano ZS (Malvern Instruments Ltd., Worcestershire, UK) using a 1/10 dilution with puriﬁed water.

2.8. Drug Release Studies Ovalbumin loaded entrapped within DSPC:Chol liposomes were produced using microfluidics at a 3:1 FRR and 15 mL/min TFR (16 mg/mL initial lipid and 1 mg/mL initial Ovalbumin) using MeOH, EtOH or IPA. These formulations were then purified via TFF as previously described. 1 mL of purified formulation was added into a 300 kD float-a-lyzer™ (Spectrum™, Breda, The Netherlands) in the presence of 20 mL PBS (pH 7.4 0.2). The samples were incubated at 37 C with agitation and at 0, ± ◦ 24, 48 and 72 h 100 µL of the sample was removed and protein retention was quantified using the described RP-HPLC method.

3. Results

3.1. Rapid Pre-Screening of Liposome Manufacture Indicates that Liposome Size Can Be Influenced by Solvent Selection To investigate the impact of solvent selection during microfluidic production on the properties of liposomes, we initially screened a panel of liposome formulations using a small-volume high-throughput microfluidic system (SparkTM; Precision Nanosystems Inc.). Initially a panel of six liposome formulations were tested, which contained a combination of a phosphatidylcholine and cholesterol, and in one formulation we also incorporated a PEG-coating (Figure3). The results showed that the solvent selected could have an impact on liposome size depending on the formulation. In general, the SoyPC and POPC liposome formulations (PC:Chol 3:1 w/w) tended to show a greater impact from the solvent adopted in terms of particle size. For example, with both formulations the liposome particle size approximately doubled when liposomes were formulated using IPA compared to methanol. With the SoyPC:Chol formulation, liposomes increased in size from 70 to 135 nm as we moved from methanol to IPA (Figure3A) and the POPC:Chol liposomes increased from 81 to 161 nm (Figure3B). When the liposomes were prepared from DMPC, DSPC and HSPC (2:1 w/w with cholesterol; Figure3C, D and E respectively), there was less impact on particle size as we decreased the solvent polarity. For example, DMPC:Chol liposomes increased in size from 149 to 190 nm as we switched from methanol to IPA (Figure3C), DSPC:Chol increased from 83 to 104 nm (Figure3D) and HSPC:Chol liposomes increased from 88 to 122 nm (Figure3E). Across all these formulations, the general trend of increasing liposome size in terms of IPA > ethanol > methanol could be seen with the exception of the pegylated formulation. When DSPE-PEG2k was included in the formulation (DSPC:Chol:PEG2k) the liposomes were similar in size (approximately 60 nm) irrespective of the solvent used in their manufacture (Figure3F). Across all formulations, the PDI was in the range of 0.2 to 0.4, with no impact from the solvent used (Figure3). Pharmaceutics 2019, 11, 653 6 of 16 Pharmaceutics 2019, 11, 653 6 of 15

FigureFigure 3. Initial 3. Initial low-volume, low-volume, high-throughput high-throughput screening screening ofof liposomesliposomes formulated formulated using using diff erentdifferent solvents during microfluidic production. Liposomes were prepared using a low volume, high solvents during microfluidic production. Liposomes were prepared using a low volume, high throughput screening system (SparkTM). Different water miscible organic solvents (methanol, ethanol throughput screening system (SparkTM). Different water miscible organic solvents (methanol, ethanol or IPA) were used in combination with the aqueous phase. A flow rate ratio of 3:1 aqueous:organic phase or IPA)was were used withused 4 in mg combination/mL total lipid with concentration the aqueou in thes phase. organic A phase. flow rate Various ratio liposome of 3:1 aqueous:organic formulations phasewere was tested: used ( Awith) soyPC:Chol 4 mg/mL 3:1 wto/talw,( Blipid) POPC:Chol concentration 3:1 w/w ,(inC )the DMPC:Chol organic 2:1phase.w/w,( VariousD) DSPC:Chol liposome formulations2:1 w/w,( Ewere) HSPC:Chol tested: ( 2:1A) wsoyPC:Chol/w,(F) DSPC:Chol:DSPE-PEG2k 3:1 w/w, (B) POPC:Chol 2:1:1 w 3:1/w. w Columns/w, (C) DMPC:Chol represent particle 2:1 w/w, (D) DSPC:Cholsize and open 2:1 circles w/w represent, (E) HSPC:Chol PDI. Results 2:1represent w/w, (F) mean DSPC:Chol:DSPE-PEG2kSD from three independent 2:1:1 batches.w/w. Columns ± represent particle size and open circles represent PDI. Results represent mean ± SD from three To further explore the impact of solvent selection, we also tested a selection of anionic (Figure4) independent batches. and cationic (Figure5) formulations. In terms of the anionic formulations (DMPC:Chol:DMPG and DSPC:Chol:DMPG), again we saw the general trend of increasing particle size (from approximately 100 to 160 nm) for both the DMPC and DSPC formulations (Figure4A,B respectively) as we progressed from methanol to IPA. Again, for all formulations the PDI remained in the range of 0.2 to 0.4 (Figure4). The impact of solvent selection on the zeta potential was also tested; no significant difference was seen across the anionic liposome formulations irrespective of the solvent selected for their production ( 15 − to 25 mV; Figure4C,D). − With the cationic (DOPE:DOTAP) formulations, again we see a trend of increasing liposome size as we decrease the polarity of the solvent, with sizes increasing from approximately 50 nm, to 100 nm to 150 nm when formulated in methanol, ethanol and IPA respectively (Figure5A). When the HSPC:Chol:DOTAP formulation was made in methanol or ethanol, particle size increased from 120 nm to 230 nm, respectively (Figure5B). With this formulation, IPA could not be tested due to precipitation of the formulation. Interestingly, when PEG was added into this formulation, precipitation was not an issue and the solvent adopted in the microfluidic process had no impact on particle size (76–96 nm irrespective of the solvent used; Figure5C) and the PDI remained in the range of 0.2 to 0.4. As with the anionic formulations, the zeta potential of the formulations were not influenced by the solvent selected with DOPE:DOTAP and HSPC:Chol:DOTAP formulations being in the range of 50 to 60 mV (Figure5D,E) and the pegylated cationic formulation being 20–30 mV (Figure5F). TM FigureThe 4. applicabilityThe effect of of solvent using thisselection low volume on the microfluidicparticle size systemof anionic (Spark liposomes.) as a high-throughputLiposomes were screeningprepared toolusing for different liposome water formulations miscible was organic also tested solvents in terms (methanol, of drug loading. ethanol Figureor IPA)6 demonstrates and PBS as that even with low volume formulation testing (down to 100–250 uL) drug loaded liposomes can be aqueous phase on the SparkTM microfluidic system as shown in Figure 3. A flow rate ratio of 3:1 prepared rapidly for initial screening with drug loading of low soluble drugs (propofol) and large aqueous:organic phase was used with 4 mg/mL total lipid concentration in the organic phase. Particle biologicals (OVA) being achieved at levels comparable (40% for propofol [13] and 25–30% for OVA [28]) size and PDI (columns and open circles, respectively) are shown for (A) DMPC:Chol:DMPG 10:5:4 to those previously reported with the larger bench-top microfluidic systems. w/w, and (B) DSPC:Chol:DMPG 10:5:4 w/w formulations. The zeta potential for (C) DMPC:Chol:DMPG 10:5:4 w/w, and (D) DSPC:Chol:DMPG 10:5:4 w/w is also shown. Results represent mean ± SD from three independent batches.

With the cationic (DOPE:DOTAP) formulations, again we see a trend of increasing liposome size as we decrease the polarity of the solvent, with sizes increasing from approximately 50 nm, to 100 nm to 150 nm when formulated in methanol, ethanol and IPA respectively (Figure 5A). When the HSPC:Chol:DOTAP formulation was made in methanol or ethanol, particle size increased from 120

Pharmaceutics 2019, 11, 653 6 of 15

Figure 3. Initial low-volume, high-throughput screening of liposomes formulated using different solvents during microfluidic production. Liposomes were prepared using a low volume, high throughput screening system (SparkTM). Different water miscible organic solvents (methanol, ethanol or IPA) were used in combination with the aqueous phase. A flow rate ratio of 3:1 aqueous:organic phase was used with 4 mg/mL total lipid concentration in the organic phase. Various liposome formulations were tested: (A) soyPC:Chol 3:1 w/w, (B) POPC:Chol 3:1 w/w, (C) DMPC:Chol 2:1 w/w, (D) DSPC:Chol 2:1 w/w, (E) HSPC:Chol 2:1 w/w, (F) DSPC:Chol:DSPE-PEG2k 2:1:1 w/w. Columns represent particle size and open circles represent PDI. Results represent mean ± SD from three Pharmaceutics 2019, 11, 653 7 of 16 independent batches.

Pharmaceutics 2019, 11, 653 7 of 15

nm to 230 nm, respectively (Figure 5B). With this formulation, IPA could not be tested due to precipitation of the formulation. Interestingly, when PEG was added into this formulation, precipitationFigure was 4. The not e ffanect issue of solvent and selectionthe solvent on the adopte particled sizein the of anionicmicrofluidic liposomes. process Liposomes had no were impact on Figureprepared 4. The using effect di ffoferent solvent water selection miscible organic on the solvents particle (methanol, size of anionic ethanol orliposomes. IPA) and PBS Liposomes as aqueous were particle size (76–96 nm irrespective of the solvent used; Figure 5C) and the PDI remained in the range preparedphase using on the SparkdifferentTM microfluidic water miscible system organic as shown so inlvents Figure 3(methanol,. A flow rate ethano ratio of 3:1l or aqueous:organic IPA) and PBS as of 0.2aqueous to phase0.4. phase As was with usedon the withthe Spark 4anionic mgTM/mL microfluidic totalformulations, lipid concentration system the as zetashown in the potential organic in Figure phase. of 3. theA Particle flow formulations sizerate andratio PDI of were 3:1 not influencedaqueous:organic(columns by the and solvent phase open was circles,selected used respectively) withwith 4 DOPEmg/mL are shown:DOTAP total lip forid (andA concentration) DMPC:Chol:DMPG HSPC:Chol:DOTAP in the organic 10:5:4 wformulations phase./w, and Particle (B) being in thesize range andDSPC:Chol:DMPG PDIof 50 (columns to 60 mV 10:5:4and (Figure openw/w formulations.circ 5D,E)les, respectively) and Thethe zetapegylated are potential shown cationic for for (C ()A DMPC:Chol:DMPG) formulationDMPC:Chol:DMPG being 10:5:4 10:5:420–30 mV w/w, and (D) DSPC:Chol:DMPG 10:5:4 w/w is also shown. Results represent mean SD from three (Figurew/w 5F)., and (B) DSPC:Chol:DMPG 10:5:4 w/w formulations. The zeta potential± for (C) independent batches. DMPC:Chol:DMPG 10:5:4 w/w, and (D) DSPC:Chol:DMPG 10:5:4 w/w is also shown. Results represent mean ± SD from three independent batches.

Figure 5. The effect of solvent selection on the particle size of cationic liposomes. Liposomes were Figuremanufactured 5. The effect as outlinedof solvent in Figureselection3. A flowon the rate particle ratio of 3:1size aqueous:organic of cationic liposomes. phase was Liposomes used with 4 were manufacturedmg/mL total as lipid outlined concentration in Figure in 3. the A organicflow rate phase ratio and of Tris3:1 aqueous:organic buffer (10 mM, pH phase 7.4) as was theaqueous used with 4 mg/mLphase. total Particle lipid concentration size and PDI (columns in the organic and open phas circles,e and respectively) Tris buffer for(10 ( AmM,) DOPE:DOTAP pH 7.4) as the 1:1 waqueous/w, phase.(B )Particle HSPC:Chol:DOTAP size and PDI 1:2:4 (columnsw/w and and (C) open HSPC:Chol:DOTAP:DSPE-PEG2k circles, respectively) for (A 1:2:4:0.5) DOPE:DOTAPw/w. The zeta1:1 w/w, (B) HSPC:Chol:DOTAPpotential for these three 1:2:4 formulations w/w and are (C shown) HSPC:Chol:DOTAP:DSPE-PEG2k in (D), (E) and (F) respectively. Results 1:2:4:0.5 represent w/w. meanThe zeta SD from three independent batches. potential± for these three formulations are shown in (D), (E) and (F) respectively. Results represent mean ± SD from three independent batches.

The applicability of using this low volume microfluidic system (SparkTM) as a high-throughput screening tool for liposome formulations was also tested in terms of drug loading. Figure 6 demonstrates that even with low volume formulation testing (down to 100–250 uL) drug loaded liposomes can be prepared rapidly for initial screening with drug loading of low soluble drugs (propofol) and large biologicals (OVA) being achieved at levels comparable (40% for propofol [13] and 25–30% for OVA [28]) to those previously reported with the larger bench-top microfluidic systems.

Figure 6. Liposome loading efficiencies for the (A) bilayer loaded drug (propofol, 1 mg/mL in MeOH) and (B) aqueous core loaded drug (OVA, 1 mg/mL in PBS) produced using rapid through-put microfluidics (SparkTM). Liposomes were prepared with soyPC:Chol and HSPC:Chol, 3:1 w/w, 10 mg/mL in MeOH. Results represent mean ± SD from three independent batches.

Pharmaceutics 2019, 11, 653 7 of 15 nm to 230 nm, respectively (Figure 5B). With this formulation, IPA could not be tested due to precipitation of the formulation. Interestingly, when PEG was added into this formulation, precipitation was not an issue and the solvent adopted in the microfluidic process had no impact on particle size (76–96 nm irrespective of the solvent used; Figure 5C) and the PDI remained in the range of 0.2 to 0.4. As with the anionic formulations, the zeta potential of the formulations were not influenced by the solvent selected with DOPE:DOTAP and HSPC:Chol:DOTAP formulations being in the range of 50 to 60 mV (Figure 5D,E) and the pegylated cationic formulation being 20–30 mV (Figure 5F).

Figure 5. The effect of solvent selection on the particle size of cationic liposomes. Liposomes were manufactured as outlined in Figure 3. A flow rate ratio of 3:1 aqueous:organic phase was used with 4 mg/mL total lipid concentration in the organic phase and Tris buffer (10 mM, pH 7.4) as the aqueous phase. Particle size and PDI (columns and open circles, respectively) for (A) DOPE:DOTAP 1:1 w/w, (B) HSPC:Chol:DOTAP 1:2:4 w/w and (C) HSPC:Chol:DOTAP:DSPE-PEG2k 1:2:4:0.5 w/w. The zeta potential for these three formulations are shown in (D), (E) and (F) respectively. Results represent mean ± SD from three independent batches.

Pharmaceutics 2019, 11, 653 8 of 15

3.2. The Impact of the Solvent Is Dependent on the Liposome Formulation

To further consider the impact of the liposome formulation in combination with solvent selection,Figure DSPC:Chol 6. Liposome liposome loading formulations efficiencies for were the prepared (A) bilayer with loaded varying drug (propofol,cholesterol 1 mgcontent/mL in from 11 Figureto 33 6.wt%MeOH) Liposome and and the (Bloading bench-top) aqueous efficiencies core NanoAssemblr loaded for drug the (OVA,TM (A was) 1bilayer mg employed./mL inloaded PBS) producedAs drug this (propofol, system using rapid was 1 through-putmg/mL designed in MeOH)for the andformulation (B) microfluidicsaqueous and coredevelopment (Spark loadedTM). Liposomesdrug of nanomedicines, (OVA, were 1 prepared mg/mL it offered with in soyPC:CholPBS) the producedability and for HSPC:Chol, liposomesusing rapid 3:1 tow /bewthrough-put, 10collected microfluidicsmg/mL (Spark in MeOH.TM). Results Liposomes represent were mean preparedSD from threewith independentsoyPC:Chol batches. and HSPC:Chol, 3:1 w/w, 10 during the steady state microfluidic production± and hence produced homogeneous suspensions with mg/mLPDI3.2. commonly in The MeOH. Impact below ofResults the Solvent0.2. represent Figure Is Dependent 7A mean shows on± SDthat the Liposomefrom as we three increased Formulation independent the cholesterol batches. content, the liposome size reduced with all three solvents. The results also showed the trend of liposome size increased as we progressedTo further from consider methanol, the impact to ofethanol, the liposome to IP formulationA irrespective in combination of the cholesterol with solvent content selection, used. However,DSPC:Chol at higher liposome cholesterol formulations concentrations were prepared the with difference varying in cholesterol size between content the from liposomes 11 to 33 formed wt% TM in andthe different the bench-top solvents NanoAssemblr was reduced was(Figure employed. 7A). Furthermore, As this system at low was designedcholesterol for concentrations the formulation (11 and development of nanomedicines, it offered the ability for liposomes to be collected during the steady wt%) liposomes could only be manufactured using methanol (Figure 7A). Attempts to manufacture state microfluidic production and hence produced homogeneous suspensions with PDI commonly this liposome formulation with ethanol or IPA resulted in large aggregates forming. below 0.2. Figure7A shows that as we increased the cholesterol content, the liposome size reduced The addition of PEG-lipid to the DSPC:Chol (2:1 w/w; 33 wt% cholesterol) was then investigated with all three solvents. The results also showed the trend of liposome size increased as we progressed with ethanol and IPA (Figure 7B). At low PEG lipid content (8%), we saw that the choice of solvent from methanol, to ethanol, to IPA irrespective of the cholesterol content used. However, at higher significantlycholesterol concentrations(p < 0.05) impacted the di ffparticleerence in size size (60 between vs. 90 thenm liposomes for ethanol formed and inIPA the respectively; different solvents Figure 7B).was However, reduced (Figureas we increased7A). Furthermore, the PEG-lipid at low cholesterolcontent to concentrations14% or more, we (11 saw wt%) the liposomes effect of could solvent wasonly negated be manufactured (Figure 7B). using methanol (Figure7A). Attempts to manufacture this liposome formulation with ethanol or IPA resulted in large aggregates forming.

Figure 7. The effect of increasing cholesterol content and PEG-lipid in liposomes manufactured using Figuremicrofluidics 7. The effect and di offf increasingerent solvents. cholesterol (A) Cholesterol content wasand PEG-lipid increased fromin liposomes 11 to 33 wt%manufactured in DSPC:Chol using microfluidicsliposomes and and (B different) PEG-lipid solvents. content was(A) Cholesterol increased from was 0 toincreased 25% within from DSPC:Chol:PEG2k 11 to 33 wt% in liposomes. DSPC:Chol liposomesLiposomes and were (B prepared) PEG-lipid at acontent FRR of 3:1was and increased TFR of 15 from mL/min 0 to for 25% DSPC:Chol within liposomesDSPC:Chol:PEG2k and 12 liposomes.mL/min for Liposomes DSPC:Chol:DSPE-PEG2k were prepared liposomes.at a FRR of Initial 3:1 and lipid TFR concentration of 15 mL/min for for both DSPC:Chol formulations liposomes was 4 mg/mL. Results represent mean SD from three independent batches. and 12 mL/min for DSPC:Chol:DSPE-PEG2k± liposomes. Initial lipid concentration for both formulationsThe addition was of 4PEG-lipid mg/mL. Results to the represent DSPC:Chol mean (2:1 ± wSD/w from; 33 wt% three cholesterol) independent was batches. then investigated with ethanol and IPA (Figure7B). At low PEG lipid content (8%), we saw that the choice of solvent 3.3.significantly Varying the ( pSolvent< 0.05) Mixture impacted Composition particle size Impacts (60 vs. 90on nmLiposome for ethanol Particle and Size IPA respectively; Figure7B). However,To further as westudy increased the link the between PEG-lipid solvent content selection to 14% and or more, particle we size, saw Figure the effect 8 explores of solvent in wasdetail thenegated impact (Figure of solvent7B). mixtures on the particle size attributes of liposomes formulated from DSPC:Chol. From Figure 8A, we can see that as we increased the ethanol content within the methanol/ethanol solvent mix from 0 to 100%, the particle size increased only moderately from 45 to 55 nm with the PDI remaining below 0.2 and the suspensions were unimodal in nature (Figure 8B). When we consider the impact of IPA on the formulations, the increase in size was more notable. DSPC:Chol liposomes prepared using mixtures of methanol/IPA showed an increase in size from 45 nm up to 93 nm as we increased the IPA content from 0 to 100% (Figure 8C), again with the PDI remaining below 0.2 and the particle size being unimodal (Figure 8D). When these liposomes were prepared using mixtures of ethanol/IPA, we could similarly see an increase in size (from 55 to 93 nm as we go from 0 to 100% IPA; Figure 8E). Again, the PDI of all formulations remained <0.2 and were unimodal in nature (Figure 8F). These results show that we can control the liposome particle size via controlling the solvent mixture.

Pharmaceutics 2019, 11, 653 9 of 16

3.3. Varying the Solvent Mixture Composition Impacts on Liposome Particle Size

Pharmaceutics 2019To, further 11, 653 study the link between solvent selection and particle size, Figure8 explores in detail 9 of 15 the impact of solvent mixtures on the particle size attributes of liposomes formulated from DSPC:Chol. To testFrom if Figure similar8A, wesize can control see that could as we increasedbe achieved the ethanol with contentcationic within formulations, the methanol again/ethanol a range of solvent mix from 0 to 100%, the particle size increased only moderately from 45 to 55 nm with the PDI methanol/ethanol solvent mixtures were tested (Figure 9). Once, again the particle size was controlled remaining below 0.2 and the suspensions were unimodal in nature (Figure8B). When we consider through thethe impact polarity of IPA of on the the solvent formulations, mixture. the increase By in increasing size was the more ratio notable. of ethanol, DSPC:Chol the liposomes liposome size increasedprepared from 33 using to 58 mixtures nm (Figure of methanol 9A) with/IPA showedthe PDI an remaining increase in sizebelow from 0.2 45 and nm up the to formulations 93 nm as we being unimodalincreased in nature the (Figure IPA content 9B,C). from Again 0 to 100% formulating (Figure8C), these again withformulations the PDI remaining with IPA below increased 0.2 and the size (up to 120the particlenm; Figure size being 9A) unimodal but did (Figure increase8D). Whenthe po theselydispersity liposomes were of the prepared formulation using mixtures (Figure of 9B,C). ethanol/IPA, we could similarly see an increase in size (from 55 to 93 nm as we go from 0 to 100% IPA; However, across all the formulations, the cationic zeta potential remained high (40–60 mV; Figure Figure8E). Again, the PDI of all formulations remained <0.2 and were unimodal in nature (Figure8F). 9D). These results show that we can control the liposome particle size via controlling the solvent mixture.

Figure 8. Investigating the eﬀect of solvent mixtures on physical liposome characteristics. Particle size Figure 8.and Investigating PDI recorded the with effect lipids dissolvedof solvent in (mixturesA) Methanol:Ethanol on physical mix liposome 0–100%, (C )characteristics. Methanol:Isopropanol Particle size and PDImix 0–100%,recorded (E) Ethanol:Isopropanolwith lipids dissolved mix 0–100% within intensity(A) Methanol:Ethanol plots for each respective mix solvent 0–100%, mix (C) Methanol:Isopropanolshown in (B), (D) mix and (0–100%,F). Liposomes (E) Ethanol:Isopropanol were composed of DSPC:Chol mix 0–100% (2:1 w/w ),with which intensity were produced plots for each using a FRR of 3:1; 15 mL/min TFR and initial lipid concentration of 4 mg/mL. Results represent mean respective solvent mix shown in (B), (D) and (F). Liposomes were composed of DSPC:Chol (2:1 w/w), SD from three independent batches. which were± produced using a FRR of 3:1; 15 mL/min TFR and initial lipid concentration of 4 mg/mL. Results representTo test if similarmean ± size SD controlfrom three could independent be achieved batches. with cationic formulations, again a range of methanol/ethanol solvent mixtures were tested (Figure9). Once, again the particle size was controlled through the polarity of the solvent mixture. By increasing the ratio of ethanol, the liposome size increased from 33 to 58 nm (Figure9A) with the PDI remaining below 0.2 and the formulations being unimodal in nature (Figure9B,C). Again formulating these formulations with IPA increased the size (up to 120 nm; Figure9A) but did increase the polydispersity of the formulation (Figure9B,C). However, across all the formulations, the cationic zeta potential remained high (40–60 mV; Figure9D).

Figure 9. Controlling the vesicle size of cationic liposomes using solvent mixtures. DOPE:DOTAP (1:1 w/w) liposomal formulations were prepared using the NanoAssemblrTM bench-scale system. A flow rate ratio of 1:1 aqueous:organic phase was used with 4 mg/mL total lipid concentration in the organic phase and Tris buffer (10 mM, pH 7.4) as aqueous phase. Solvent mixtures of methanol/ethanol and IPA were tested. Results are shown as (A) Particle size, (B) PDI, (C) Intensity plots and (D) Zeta potential of the formulations. Results represent mean ± SD from three independent batches.

Pharmaceutics 2019, 11, 653 9 of 15

To test if similar size control could be achieved with cationic formulations, again a range of methanol/ethanol solvent mixtures were tested (Figure 9). Once, again the particle size was controlled through the polarity of the solvent mixture. By increasing the ratio of ethanol, the liposome size increased from 33 to 58 nm (Figure 9A) with the PDI remaining below 0.2 and the formulations being unimodal in nature (Figure 9B,C). Again formulating these formulations with IPA increased the size (up to 120 nm; Figure 9A) but did increase the polydispersity of the formulation (Figure 9B,C). However, across all the formulations, the cationic zeta potential remained high (40–60 mV; Figure 9D).

Figure 8. Investigating the effect of solvent mixtures on physical liposome characteristics. Particle size and PDI recorded with lipids dissolved in (A) Methanol:Ethanol mix 0–100%, (C) Methanol:Isopropanol mix 0–100%, (E) Ethanol:Isopropanol mix 0–100% with intensity plots for each respective solvent mix shown in (B), (D) and (F). Liposomes were composed of DSPC:Chol (2:1 w/w), which were produced using a FRR of 3:1; 15 mL/min TFR and initial lipid concentration of 4 mg/mL. Results Pharmaceuticsrepresent2019 mean, 11, 653 ± SD from three independent batches. 10 of 16

Figure 9. Controlling the vesicle size of cationic liposomes using solvent mixtures. DOPE:DOTAP (1:1 Figure 9. Controllingw/w) liposomal the formulations vesicle size were of prepared cationic using liposomes the NanoAssemblr using solventTM bench-scale mixtures. system. DOPE:DOTAP A flow (1:1 w/w) liposomalrate ratio formulations of 1:1 aqueous:organic were prepared phase was usedusing with the 4 mg NanoAssemblr/mL total lipid concentrationTM bench-scale in the organic system. A flow rate ratio ofphase 1:1 aqueous:organic and Tris buffer (10 mM, phase pH 7.4) was as aqueous used wi phase.th 4Solvent mg/mL mixtures total oflipid methanol concentration/ethanol and IPA in the organic were tested. Results are shown as (A) Particle size, (B) PDI, (C) Intensity plots and (D) Zeta potential of phase and Tristhe formulations. buffer (10 ResultsmM, pH represent 7.4) meanas aqueousSD from phase. three independent Solvent mixtures batches. of methanol/ethanol and IPA were tested. Results are shown as (A)± Particle size, (B) PDI, (C) Intensity plots and (D) Zeta 3.4. Solvent Choice Influences Liposome Morphology and Initial Protein Loading but Does Not Impact Liposome potentialStability of the nor formulations. Release Attributes Results represent mean ± SD from three independent batches. To further investigate the impact of solvent choice during microfluidic manufacturing, DSPC:Chol (2:1 w/w) were prepared incorporating protein (OVA) (Figure 10). From the results in Figure 10A, we can see that there was no significant difference in protein loading for liposomes prepared with methanol and ethanol (35–40%; Figure 10A), but protein loading was significantly (p < 0.05) reduced when IPA (20%; Figure 10A) was used for the manufacturing process. Interestingly, when considering the morphology (Figure 10B), we can see that liposomes formed as small unilamellar vesicles when prepared using methanol or ethanol, whilst those formed with IPA were larger and had one or two bilayers. In terms of stability, all three liposome formulations showed good short-term stability when stored at 2–8 ◦C (Figure 10C) with no significant difference in size being noted over the length of the study. Similarly, there was no significant difference in protein release profiles of the liposomes prepared using methanol, ethanol or IPA with approximately 80% protein release after 72 h (Figure 10D). Pharmaceutics 2019, 11, 653 10 of 15

3.4. Solvent Choice Influences Liposome Morphology and Initial Protein Loading but Does Not Impact Liposome Stability nor Release Attributes To further investigate the impact of solvent choice during microfluidic manufacturing, DSPC:Chol (2:1 w/w) were prepared incorporating protein (OVA) (Figure 10). From the results in Figure 10A, we can see that there was no significant difference in protein loading for liposomes prepared with methanol and ethanol (35–40%; Figure 10A), but protein loading was significantly (p < 0.05) reduced when IPA (20%; Figure 10A) was used for the manufacturing process. Interestingly, when considering the morphology (Figure 10B), we can see that liposomes formed as small unilamellar vesicles when prepared using methanol or ethanol, whilst those formed with IPA were larger and had one or two bilayers. In terms of stability, all three liposome formulations showed good short-term stability when stored at 2–8 °C (Figure 10C) with no significant difference in size being noted over the length of the study. Similarly, there was no significant difference in protein release profiles of the liposomes prepared using methanol, ethanol or IPA with approximately 80% protein Pharmaceutics 2019, 11, 653 11 of 16 release after 72 h (Figure 10D).

FigureFigure 10. Solvent 10. Solvent selection selection and and its its impactimpact on on OVA OVA encapsulation, encapsulation, morphology, morphology, liposome liposome stability and stability protein release. (A) Protein entrapment eﬃciency of liposomes loaded with 0.25 mg/mL OVA with and protein release. (A) Protein entrapment efficiency of liposomes loaded with 0.25 mg/mL OVA an initial lipid concentration of 4 mg/mL produced using methanol, ethanol or isopropanol. (B) the with anmorphology initial lipid of liposomesconcentration obtained of 4 by mg/mL cryoTEM produced in the respective using methanol, solvents. (ethanolC) Stability or isopropanol. of ‘empty’ (B)

the morphologyliposomes stored of liposomes at 2–8 ◦Cover obtained 7 days by at ancryoTEM initial lipid in concentrationthe respective of 4solvents. mg/mL. ( D(C)) Protein Stability release of ‘empty’ liposomesprofiles stored of liposomes at 2–8 °C (initial over concentration 7 days at an of initial 16 mg lipid/mL loaded concentration with 1 mg /ofmL 4 OVA)mg/mL. was ( investigatedD) Protein release profilesover of 72liposomes h at 37 ◦C (initial to access concentration if solvent selection of 16 impactedmg/mL loaded membrane with permeability. 1 mg/mL OVA) Liposomes was investigated were over 72composed h at 37 of°C DSPC:Chol to access (2:1 if solventw/w) and selection produced atimpacted a 3:1 FRR membrane and 15 mL/min. permeability. Results represent Liposomes mean were SD from three independent batches. composed± of DSPC:Chol (2:1 w/w) and produced at a 3:1 FRR and 15 mL/min. Results represent mean ±4. SD Discussion from three independent batches. Microfluidic production offers new manufacturing strategies for the production of liposomes 4. Discussionand other nanoparticles. This method can be scale-independent [16,29] and thus gives formulation Microfluidicscientists a direct production route to translateoffers new their manufacturing research off the benchstrategies into thefor clinic. the production When considering of liposomes liposomes, there are a range of parameters that are critical quality attributes. Amongst these, vesicle and other nanoparticles. This method can be scale-independent [16,29] and thus gives formulation size is particularly important given its impact on biodistribution, which is a key feature in the ability of scientistsliposomes a direct to improve route drugto translate delivery, enhancetheir resear efficacych andoff reducethe bench off target into toxicity. the clinic. It is widely When reported considering liposomes,that critical there processare a range parameters of parameters that influence that the are size critical of nanoparticles quality attributes. produced Amongst via microfluidics these, vesicle size isinclude particularly the micromixer important design given (e.g., its [30 impact]) and the on ratebiodistribution, of mixing (e.g., which [13,16,17 is,31 a ]).key Within feature this in study, the ability of liposomeswe demonstrated to improve that drug solvent delivery, selection enhance was also aeffic criticalacy process and reduce parameter off target in the manufacturetoxicity. It is of widely reportedliposomes that critical using microfluidics process parameters as it could impact that onin particlefluence size. the Furthermore, size of nano weparticles demonstrated produced that via microfluidicsalcohol mixtures include could the bemicromixer used to manipulate design liposome(e.g., [30]) size. and However, the rate the sensitivityof mixing of (e.g., the liposome [13,16,17,31]). particle size to solvent selection was dependent on the given liposome formulation. This can be tested using rapid low volume screening protocols (as outlined in Figures3–6), which is particularly useful in early studies where lipids and active pharmaceutical ingredients may be limited. However, changes in particle size when switching between methanol and ethanol were generally low (Figures7–9). When produced by microfluidics, liposome sizes tended to increase with decreased solvent polarity across all the formulations tested and size control of the liposomes could be achieved using solvent mixtures (Figures8 and9). Furthermore, increasing concentrations of cholesterol or pegylated-lipid formulations tended to reduce the impact of the solvent selections, and formulations containing over Pharmaceutics 2019, 11, 653 12 of 16

14% PEG-lipid did not show a notable sensitivity to the alcohol used in their manufacture (Figure7). Despite differences in morphology, the liposomes formulated using methanol, ethanol or IPA all showed good stability and similar protein release profiles (Figure 10). A range of models have been proposed for the self-assembly of liposomes due to the mixing of water miscible solvents and recently Zook and Vreeland proposed a non-equilibrium model [26]. Their model proposed that liposome size was determined by two factors: 1) the growth rate of planar bilayer discs and 2) the rate the discs close into spherical vesicles (Figure2). In their studies, Zook and Vreeland [26] tended to focus their investigations on considering the impact of the rate of disc closure, and the authors showed that the radius of the liposomes formed during their microfluidic process was proportional to the ratio of membrane bending elasticity modulus to the hydrophobic edges of the lipid discs formed. In general, the membrane elasticity modulus of a bilayer is higher with longer alkyl chain length/high transition temperature lipids (i.e., DSPC > DPPC > DMPC in terms of bending elastic modulus, and a higher elastic modulus indicates a more rigid membrane). Therefore, bilayer discs formed during the microfluidic process (as shown in Figure2A) should be more rigid and less able to bend when the membrane is at or below the bilayer transition temperature [26]. Thus, at room temperature high transition temperature bilayer formulations would form larger liposomes. Indeed, in discussing their model, Zook and Vreeland note that temperature is a key consideration and increasing the temperature above the lipid transition temperature will increase elasticity and hence reduce particle size [26]. These findings were not in line with the results shown in Figure3 (e.g., DMPC vs. DSPC formulations) nor with our previous findings [13]; we show that increasing the carbon chain length of the PC lipid results in smaller liposomes being formed. Similarly, we have previously shown that during the microfluidic production, liposomes produced at room temperatures or at a temperature above the main lipid transition temperature were similar in size; we were able to prepare DSPC:Chol liposomes of the same size at production temperatures from 20 to 60 ◦C. This demonstrates that there was no requirement to work above the lipid transition temperature during the microfluidic manufacturing process [13]. Therefore, in our studies, the impact of temperature on membrane elasticity is not a key contributing factor controlling liposome size. However, it can be useful to use elevated temperatures to improve the solubility of some lipids in solvents during the processes irrespective of their Tc (e.g., [16]). When considering bilayer rigidity, cholesterol content is also a key factor as the transition temperature of liposome bilayers (Tc) varies when mixtures of different hydrocarbon chain length phospholipids are compared to pure lipids [32]. In the absence of cholesterol, the hydrocarbon chains of the lipids in the bilayer crystallize into the rigid-crystalline phase which results in, for example, the Tc for DSPC liposomes being 55 ◦C[32]. However, with the addition of 33 mol% cholesterol, this transition temperature is no longer detectable [33]. Similarly, the large head-group component of charged lipids or pegylated lipids can reduce/negate the transition temperature of lipid bilayers by inhibiting the packaging of the lipids into the rigid-crystalline phase. Moreover, it has also been shown that adding cholesterol to liposome bilayers tends to increase the elasticity of the membrane [34,35]. From Figure7, we can see that increasing the cholesterol content within liposomes resulted in reduced vesicle sizes, and this may be attributed to the increased elasticity and more rapid closure of the discs into smaller liposomes. The presence of higher levels of cholesterol and PEG-lipid was also shown to negate the effect of solvent selection on particle size. This suggests that the geometric packaging of these lipids within a bilayer and subsequently into a liposome may overrule the impact of solvent choice. Alcohols are known modulators of bilayer properties and this may explain why temperature and lipid bilayer transition temperature are not factors in our process. Whilst the alcohol is removed from the liposome formulations after production (via tangential flow filtration within our studies), during the microfluidic mixing phase this is a consideration. There are various studies which demonstrate the ability of alcohols to change the lipid bilayer free volume and promote bilayer disorder, and it has been reported that the bilayer-modifying potency of short-chain alcohols scales linearly with their bilayer partitioning [36]. Within a bilayer, alcohols have their –OH group in the bilayer interfacial region and Pharmaceutics 2019, 11, 653 13 of 16 their hydrophobic methyl groups in the hydrophobic core of the bilayer and thus they disrupt the bilayer packing [37]. Thus, the impact of the alcohol on the bilayer will depend both on the alcohol and the composition of the bilayer [36]. Indeed, the ability of ethanol to enhance the bilayer permeability has been exploited to improve pH gradient loading of cholesterol-free liposomes [38]. Normally, for pH gradient loading, the liposomes are incubated at temperatures above their phase transition temperature to promote drug loading. However, in cholesterol free bilayers, increased temperatures can result in the collapse of the pH gradient. Therefore, Dos Santos et al. [38] circumvented this problem by the addition of ethanol as a permeability enhancer. Ethanol is also reported to reduce the transition temperature of lipid bilayers (through promoting bilayer disorder) and at high concentrations can cause lipid interdigitation [39,40]. This disordering effect is not restricted to ethanol, and all three of the alcohols used within our studies had a bilayer disordering effect [41]. Thus, during the production of liposomes, the miscible alcohols present during the mixing process may reduce the transition temperature of the bilayer discs, hence increasing their elasticity and promoting vesicle formation. Yet, given that the release properties of the three different liposome formulations were similar (Figure 10), this suggested that on purification of the liposomes via TFF, the residual solvent was removed (as shown previously [13]) and there was no difference in fluidity/permeability of the different liposome products. Given that in our process the membrane elasticity and rate of membrane closure (Figure2C,D) did not appear to be a key factor, the link between solvent selection and liposome particle size may be linked to the process outlined in Figure2a and the initial size of the lipid discs formed [ 26]. During the mixing process of the alcohol and buffer, an increase in the polarity results in the formation of the discs. As we reduce the polarity of the alcohol used (going from methanol to IPA), the rate of change in polarity during the mixing process will be reduced. This may result in larger discs being formed, which subsequently close into larger liposomes. Higher concentrations of cholesterol or PEG-lipids may inhibit the packaging of lipids into larger discs. The longer chain IPA may also provide additional stabilization for the lipid discs formed, which could also contribute to the formation of larger liposomes, as suggested by Zook and Vreeland [26]. However, within their studies only IPA was used and therefore the impact of solvent selection on this stabilizing effect was not clear.

5. Conclusions From our studies, we demonstrate that solvent selection is a key consideration in developing a microfluidic manufacturing process for liposome production. In terms of solvent selection, ethanol offers a range of advantages including its suitability for large scale manufacture and its Class 3 status. However, a range of solvents can be considered (e.g., octanol [42], IPA [8], methanol [30]) and solvent selection can be used to control liposome size. Furthermore, upon appropriate dilution/solvent removal the efficacy of the produced liposomes can be equivalent (e.g., [43]). Within our current studies, we outline a rapid high-throughput method for initially screening the impact of solvent selection on liposome attributes. We also show that alcohol mixtures can be used to fine-tune the liposome size. We have previously shown that buffer concentration can be used to offer size control [18]. Therefore, liposome size is under the influence of alcohol and the salt concentration of the aqueous buffer and both are key process parameters that should be considered in the development of microfluidic manufacturing processes.

Author Contributions: Conceptualization, Y.P., S.K. and S.T.S.; methodology, C.W., S.K., C.B.R., N.F., S.T.S.; validation, C.W., S.K. C.B.R., N.F., S.T.S.; formal analysis, C.W., S.K., C.B.R., N.F., S.T.S.; investigation, C.W., S.K., C.B.R., N.F., S.T.S. G.B.; resources, Y.P.; writing, review and editing, Y.P., S.K., C.W., C.B.R., N.F., S.T.S., G.B.; visualization, Y.P., C.W., S.K., C.B.R., N.F., S.T.S. G.B.; supervision, Y.P.; project administration, Y.P.; funding acquisition, Y.P., S.T.S. Funding: This research was funded by Innovate UK (KTP 10793) (SK), Independent Research Fund Denmark (7026-00027B) (SS) Microsun (NFC-01 Catapult) (CR), EPSRC Centre for Innovative Manufacturing in Emergent Therapies (EPSRC) (EP/I033270/1) (NF, CW). Pharmaceutics 2019, 11, 653 14 of 16

Data Access: All data underpinning this publication is openly available from the University of Strathclyde KnowledgeBase at https://doi.org/10.15129/81b22c47-c133-41e9-86e5-8f8a45a4d443. Conﬂicts of Interest: The authors declare no conﬂicts of interest.

References

1. Allen, T.M.; Cullis, P.R. Liposomal drug delivery systems: From concept to clinical applications. Adv. Drug Deliv. Rev. 2013, 65, 36–48. [CrossRef][PubMed] 2. Del Giudice, G.; Rappuoli, R.; Didierlaurent, A.M. Correlates of adjuvanticity: A review on adjuvants in licensed vaccines. Semin. Immunol. 2018, 39, 14–21. [CrossRef][PubMed] 3. Bulbake, U.; Doppalapudi, S.; Kommineni, N.; Khan, W. Liposomal formulations in clinical use: An updated review. Pharmaceutics 2017, 9, 12. [CrossRef][PubMed] 4. Agency, E.M. Names of Liposomal Medicines to be Changed to Avoid Medication Errors; European Medicines Agency, News, 2019. Available online: https://www.ema.europa.eu/en/news/names-liposomal-medicines- be-changed-avoid-medication-errors (accessed on 31 July 2019). 5. Bochicchio, S.; Dalmoro, A.; Bertoncin, P.; Lamberti, G.; Moustafine, R.I.; Barba, A.A. Design and production of hybrid nanoparticles with polymeric-lipid shell–core structures: Conventional and next-generation approaches. RSC Adv. 2018, 8, 34614–34624. [CrossRef] 6. Dalmoro, A.; Bochicchio, S.; Nasibullin, S.F.; Bertoncin, P.; Lamberti, G.; Barba, A.A.; Moustafine, R.I. Polymer-lipid hybrid nanoparticles as enhanced indomethacin delivery systems. Eur. J. Pharm. Sci. 2018, 121, 16–28. [CrossRef] 7. Bochicchio, S.; Dalmoro, A.; Recupido, F.; Lamberti, G.; Barba, A.A. Nanoliposomes production by a protocol based on a simil-microfluidic approach. In Advances in Bionanomaterials; Springer: Berlin/Heidelberg, Germany, 2018; pp. 3–10. 8. Jahn, A.; Vreeland, W.N.; DeVoe, D.L.; Locascio, L.E.; Gaitan, M. Microfluidic directed formation of liposomes of controlled size. Langmuir 2007, 23, 6289–6293. [CrossRef] 9. Maeki, M.; Fujishima, Y.; Sato, Y.; Yasui, T.; Kaji, N.; Ishida, A.; Tani, H.; Baba, Y.; Harashima, H.; Tokeshi, M. Understanding the formation mechanism of lipid nanoparticles in microfluidic devices with chaotic micromixers. PLoS ONE 2017, 12, e0187962. [CrossRef] 10. Joshi, S.; Hussain, M.T.; Roces, C.B.; Anderluzzi, G.; Kastner, E.; Salmaso, S.; Kirby, D.J.; Perrie, Y. Microfluidics based manufacture of liposomes simultaneously entrapping hydrophilic and lipophilic drugs. Int. J. Pharm. 2016, 514, 160–168. [CrossRef] 11. Kastner, E.; Verma, V.; Lowry, D.; Perrie, Y. Microfluidic-controlled manufacture of liposomes for the solubilisation of a poorly water soluble drug. Int. J. Pharm. 2015, 485, 122–130. [CrossRef] 12. Kastner, E.; Kaur, R.; Lowry, D.; Moghaddam, B.; Wilkinson, A.; Perrie, Y. High-throughput manufacturing of size-tuned liposomes by a new microfluidics method using enhanced statistical tools for characterization. Int. J. Pharm. 2014, 477, 361–368. [CrossRef] 13. Forbes, N.; Hussain, M.T.; Briuglia, M.L.; Edwards, D.P.; Ter Horst, J.H.; Szita, N.; Perrie, Y. Rapid and scale-independent microfluidic manufacture of liposomes entrapping protein incorporating in-line purification and at-line size monitoring. Int. J. Pharm. 2019, 556, 68–81. [CrossRef][PubMed] 14. Hussain, M.T.; Forbes, N.; Perrie, Y.; Malik, K.P.; Duru, C.; Matejtschuk, P. Freeze-drying cycle optimization for the rapid preservation of protein-loaded liposomal formulations. Int. J. Pharm. 2019.[CrossRef] [PubMed] 15. Perrie, Y.; Kastner, E.; Khadke, S.; Roces, C.B.; Stone, P. Manufacturing methods for liposome adjuvants. In Vaccine Adjuvants; Springer: Berlin/Heidelberg, Germany, 2017; pp. 127–144. 16. Roces, C.B.; Khadke, S.; Christensen, D.; Perrie, Y. Scale-independent microfluidic production of cationic liposomal adjuvants and development of enhanced lymphatic targeting strategies. Mol. Pharm. 2019, 16, 4372–4386. [CrossRef][PubMed] 17. Khadke, S.; Roces, C.B.; Cameron, A.; Devitt, A.; Perrie, Y. Formulation and manufacturing of lymphatic targeting liposomes using microfluidics. J. Control. Release 2019, 307, 211–220. [CrossRef][PubMed] 18. Lou, G.; Anderluzzi, G.; Woods, S.; Roberts, C.W.; Perrie, Y. A novel microfluidic-based approach to formulate size-tuneable large unilamellar cationic liposomes: Formulation, cellular uptake and biodistribution investigations. Eur. J. Pharm. Biopharm. 2019, 143, 51–60. [CrossRef][PubMed] Pharmaceutics 2019, 11, 653 15 of 16

19. Kinoshita, K.; Ishikawa, H.; Shinoda, K.O.Z.O. Solubility of alcohols in water determined the surface tension measurements. Bull. Chem. Soc. Jpn. 1958, 31, 1081–1082. [CrossRef] 20. Forster, S.; Buckton, G.; Beezer, A.E. The importance of chain length on the wettability and solubility of organic homologs. Int. J. Pharm. 1991, 72, 29–34. [CrossRef] 21. ICH Harmonised Tripartite Guideline. Impurities: Guideline for residual solvents Q3C (R5). Curr. Step 2016, 4, 1–25. 22. Hong, J.S.; Stavis, S.M.; DePaoli Lacerda, S.H.; Locascio, L.E.; Raghavan, S.R.; Gaitan, M. Microfluidic directed self-assembly of liposome—Hydrogel hybrid nanoparticles. Langmuir 2010, 26, 11581–11588. [CrossRef] 23. Wi, R.; Oh, Y.; Chae, C. Formation of liposome by microfluidic flow focusing and its application in gene delivery. Korea Aust. Rheol. J. 2012, 24, 129–135. [CrossRef] 24. Mijajlovic, M.; Wright, D.; Zivkovic, V.; Bi, J.; Biggs, M.J. Microfluidic hydrodynamic focusing based synthesis of POPC liposomes for model biological systems. Colloids Surf. B Biointerfaces 2013, 104, 276–281. [CrossRef] [PubMed] 25. Church, A.S.; Witting, M.D. Laboratory testing in ethanol, methanol, ethylene glycol, and isopropanol toxicities. J. Emerg. Med. 1997, 15, 687–692. [CrossRef] 26. Zook, J.M.; Vreeland, W.N. Effects of temperature, acyl chain length, and flow-rate ratio on liposome formation and size in a microfluidic hydrodynamic focusing device. Soft Matter 2010, 6, 1352–1360. [CrossRef] 27. Khadke, S.; Stone, P.; Rozhin, A.; Kroonen, J.; Perrie, Y. Point of use production of liposomal solubilised products. Int. J. Pharm. 2018, 537, 1–8. [CrossRef][PubMed] 28. Hussain, M.T.; Forbes, N.; Perrie, Y. Comparative analysis of protein quantification methods for the rapid determination of protein loading in liposomal formulations. Pharmaceutics 2019, 11, 39. [CrossRef][PubMed] 29. Carugo, D.; Bottaro, E.; Owen, J.; Stride, E.; Nastruzzi, C. Liposome production by microfluidics: Potential and limiting factors. Sci. Rep. 2016, 6, 25876. [CrossRef][PubMed] 30. Phapal, S.M.; Sunthar, P. Influence of micro-mixing on the size of liposomes self-assembled from miscible liquid phases. Chem. Phys. Lipids 2013, 172, 20–30. [CrossRef] 31. Zizzari, A.; Bianco, M.; Carbone, L.; Perrone, E.; Amato, F.; Maruccio, G.; Rendina, F.; Arima, V. Continuous-flow production of injectable liposomes via a microfluidic approach. Materials 2017, 10, 1411. [CrossRef] 32. Taylor, K.M.; Morris, R.M. Thermal analysis of phase transition behaviour in liposomes. Thermochim. Acta 1995, 248, 289–301. [CrossRef] 33. Moghaddam, B.; Ali, M.H.; Wilkhu, J.; Kirby, D.J.; Mohammed, A.R.; Zheng, Q.; Perrie, Y. The application of monolayer studies in the understanding of liposomal formulations. Int. J. Pharm. 2011, 417, 235–244. [CrossRef] 34. Hodzic, A.; Rappolt, M.; Amenitsch, H.; Laggner, P.; Pabst, G. Differential modulation of membrane structure and fluctuations by plant sterols and cholesterol. Biophys. J. 2008, 94, 3935–3944. [CrossRef][PubMed] 35. Henriksen, J.; Rowat, A.C.; Brief, E.; Hsueh, Y.; Thewalt, J.; Zuckermann, M.; Ipsen, J.H. Universal behavior of membranes with sterols. Biophys. J. 2006, 90, 1639–1649. [CrossRef][PubMed] 36. Ingólfsson, H.I.; Andersen, O.S. Alcohol’s effects on lipid bilayer properties. Biophys. J. 2011, 101, 847–855. [CrossRef][PubMed] 37. Barry, J.A.; Gawrisch, K. Direct NMR evidence for ethanol binding to the lipid-water interface of phospholipid bilayers. Biochemistry 1994, 33, 8082–8088. [CrossRef] 38. Dos Santos, N.; Cox, K.A.; McKenzie, C.A.; van Baarda, F.; Gallagher, R.C.; Karlsson, G.; Edwards, K.; Mayer, L.D.; Allen, C.; Bally, M.B. pH gradient loading of anthracyclines into cholesterol-free liposomes: Enhancing drug loading rates through use of ethanol. Biochim. Biophys. Acta BBA Biomembr. 2004, 1661, 47–60. [CrossRef] 39. Zeng, J.; Smith, K.E.; Chong, P. Effects of alcohol-induced lipid interdigitation on proton permeability in L-alpha-dipalmitoylphosphatidylcholine vesicles. Biophys. J. 1993, 65, 1404–1414. [CrossRef] 40. Zeng, J.; Chong, P.L.G. Interactions between pressure and ethanol on the formation of interdigitated DPPC liposomes: A study with Prodan fluorescence. Biochemistry 1991, 30, 9485–9491. [CrossRef] 41. Paxman, J.; Hunt, B.; Hallan, D.; Zarbock, S.R.; Woodbury, D.J. Drunken membranes: Short-chain alcohols alter fusion of liposomes to planar lipid bilayers. Biophys. J. 2017, 112, 121–132. [CrossRef] Pharmaceutics 2019, 11, 653 16 of 16

42. Deshpande, S.; Caspi, Y.; Meijering, A.E.; Dekker, C. Octanol-assisted liposome assembly on chip. Nat. Commun. 2016, 7, 10447. [CrossRef] 43. Meisel, J.W.; Gokel, G.W. A simpliﬁed direct lipid mixing lipoplex preparation: Comparison of liposomal-, dimethylsulfoxide-, and ethanol-based methods. Sci. Rep. 2016, 6, 27662. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).