Nanoscale characterization of PEGylated phospholipid coatings formed by spray drying on silica microparticles

Phospholipids constitute biocompatible and safe excipients for pulmonary drug delivery. They can retard the drug release and, when PEGylated, also prolong the residence time in the lung. The aim of this work was to assess the structure and coherence of phospholipid coatings formed by spray drying on hydrophilic surfaces (silica microparticles) on the nanoscale and, in particular, the effect of addition of PEGylated lipids thereon. Scanning electron microscopy showed the presence of nanoparticles of varying sizes on the microparticles with different PEGylated lipid concentrations. Atomic force microscopy conﬁrmed the presence of a lipid coating on the spray-dried microparticles. It also revealed that the lipid-coated microparticles without PEGylated lipids had a rather homogenous coating whereas those with PEGylated lipids had a very heterogeneous coating with defects, which was corroborated by confocal laser scanning microscopy. All coated microparticles had good dispersibility without agglomerate formation, as indicated by particle size measurements. This study has demonstrated that coherent coatings of phospholipids on hydrophilic surfaces can be obtained by spray drying. However, the incorporation of PEGylated lipids in a one-step


Introduction
Pulmonary drug delivery is an efficient route of administration for local treatment of respiratory diseases since it can decrease the systemic exposure of the drug and ensure a high drug concentration at the site of drug action [1].Although an increasing number of inhaled drugs are becoming available on the market [2], the therapeutic efficacy of inhaled drugs is often hampered by their rapid release, which can lead to adverse drug reactions and a suboptimal therapeutic efficacy of the drug [3].There is therefore an obvious need to produce formulations with controlled release in the lung; despite tremendous efforts, no such formulation has yet reached the market [4].For controlled-release pulmonary formulations to be effective, they must not only release the drug at the desired rate but also remain sufficiently long at the site of action, i.e. avoid clearance by the mucociliary escalator and alveolar macrophages [5].In addition, the formulation strategies that can be used to modulate the performance of the pulmonary drugs are limited by the small number of excipients authorized for inhalation.Only the use of pharmacopoeial excipients with a well-established history of inhalation use and substances generally recognized as safe is promoted by European Medicines Agency (EMA) and Food and Drug Administration (FDA) [6].
Phospholipids, which constitute the main components of endogenous lung surfactants, are promising multifunctional excipients that are considered as biocompatible and safe for pulmonary drug delivery.They are amphiphilic molecules composed of a hydrophilic head group and (a) hydrophobic tail(s).With surfaceactive properties, the application of phospholipids as excipients can aid the delivery of drugs to the lungs as they are able to improve particle migration to the lung periphery due to the reduction in surface tension [7].Their low aqueous solubility, resulting from the long hydrophobic tails, can help controlling drug dissolution and release.For example, 1,2-dipalmitoyl-sn-glycero-3phosphocholine (DPPC), as the primary lung surfactant, has an aqueous solubility of 5 mg/ml or less [8,9].In addition, PEGylated phospholipids can provide an effective 'stealth' shield to avoid the clearance systems by preventing phagocytosis by macrophages [10][11][12].It has been demonstrated that inhalation of polyethylene glycol (PEG) is safe [13], and PEGylated phospholipids such as distearoyl phosphoethanolamine polyethylene glycol 2000 (DSPE-mPEG-2000) are of low potential pulmonary toxicity [14,15].
Phospholipids are often used in liposome-based formulations, usually referred to as liposomes, lipospheres, and proliposomes.Such liposome-based formulations have been extensively studied for pulmonary drug delivery [16,17] but should, for stability reasons, preferably be converted to powder form [18][19][20][21][22][23]. This involves dehydration-rehydration and freezing-thawing processes [20] that are not only time-consuming but also difficult to scale up.Therefore, solid lipid particles (SLPs), including microparticles (SLmPs) or nanoparticles (SLnPs) depending on whether they present a size above or below 1 mm, have attracted considerable attention.The SLPs can be produced by an easy one-step spray-drying method [24][25][26][27][28]. Spray drying is a particle engineering technique suitable for the efficient production of solid inhalable particles (aerodynamic size <5 mm) with scale-up capability.The technique also allows for the precise tuning of the particle properties by adjusting the processing parameters [29].
Spray-dried SLnPs and SLmPs prepared by spray drying and intended for pulmonary drug delivery as inhalation powders have shown better drug stability and encapsulation efficiency compared to liposomes [30].More importantly, some studies have reported that SLmPs could offer retarded release [28,31].Quite recently, some attempts have been made to include PEGylated phospholipids in order to prepare spray-dried SLmPs with stealth properties [11,12,31,32].However, only a few publications have addressed the effect of PEG/PEGylated lipids on the drug release from the SLmPs.It has been reported that PEGylated excipientcomprising formulations of cisplatin exhibited a higher burst effect, but a lower overall dissolution rate compared to the SLmPs without PEGylated excipients [31].On the other hand, the addition of PEG 4000 in the formulation resulted in accelerated insulin release from the TG 22 (Dynasan D122) SLmPs [32].This was attributed to the pore-forming effect of PEG [32], similar to the one observed for PEG-PLGA microspheres prepared by a double emulsion solvent evaporation [33] and lipid-PEG coated protein particles prepared by a twin-screw extrusion [34].Although the coating quality and surface structure of the coated or encapsulated particle greatly affect its dissolution [9,35], this has never been investigated due to the difficulty of studying the surface structure of a micronized particle on a nanoscale level.
The purpose of this study is to assess the structure and coherence of lipid coatings formed by spray drying on hydrophilic surfaces and, in particular, the effect of addition of PEGylated lipids thereon, which to the best of our knowledge has not been investigated previously.Our result will demonstrate whether it is possible to obtain well-coated microparticles by spray-drying despite the apparent incompatibility between hydrophilic APIs and hydrophobic excipients [32] and will shed light on the seemingly contradictory results obtained in prior studies as described above.To facilitate the analysis, hydrophilic silica microparticles with a smooth surface were employed as hydrophilic 'drug' cores.These were coated with lipid mixtures with different PEGylated lipid content using spray drying.The thus formed lipid coated silica microparticles (LCmPs) could be considered as a special type of SLmPs with a core-shell structure.Since silica microparticles were used, the obtained results are also expected to be relevant for coating of drug-loaded mesoporous silica microparticles, e.g., in order to avoid leakage of drug prior to administration [36].Various techniques, including confocal laser scanning microscopy (CLSM), tapping mode atomic force microscopy (AFM), and intermodulation AFM (ImAFM), were employed in order to obtain a multifaceted characterization.It is worth mentioning that AFM, as a versatile tool for the characterization of micro-and nano-scale surfaces, can provide unique information on surface properties, including topography and local nanomechanical properties [37].In this paper, both tapping mode AFM and ImAFM were employed to study the surface properties of the lipid-coated silica microparticles.The characterization enabled us to have a good insight into the structure and quality of the lipid coating over the silica microparticles with different excipient compositions.

Preparation of the lipid coated microparticles
The lipid-coated microparticles or LCmPs were prepared by spray drying according to the formulation composition listed in Table 1.The experiments were carried out by using a standard Büchi Mini Spray Dryer B-290 with an inert loop B295 (Büchi, Flawil, Switzerland) at a laboratory scale.The feed suspensions were prepared by firstly dissolving various amounts of DPPC, DSPC, and DSPC-PEG 2K in methanol to form dilute organic solutions of 0.1% w/v.Then silica microparticles were added to the dilute organic solutions under stirring to get silica suspensions with a concentration of 2% w/v.Thus, the theoretical lipid concentration in the spray-dried LCmPs is around 4.8%.One mole % of dye-labeled lipids Rb-DPPE and CF-PEG2K-DSPE were dissolved in methanol together with the other lipids to prepare the LCmPs used for CLSM measurements.
The suspensions were stirred until the silica microparticles were well dispersed, and no deposit was observed at the bottom of the glass vials.During spray drying, the suspension was pumped into an atomizer and atomized into an aerosol.The aerosol droplets were convected into a cylinder by hot nitrogen.The inlet temperature was set at 70 °C.The solvent was evaporated in the hot gas in milliseconds, and a dry powder can be collected using a cyclone.The outlet temperature at the end of the spray cylinder was around 38 °C to 41 °C, which was not above the phase transition temperature of the lipid excipients.The outlet temperature is used to follow the drying process because it is the highest temperature that the spray-dried product is in contact with, and it should be lower than the phase transition temperature of the lipids to avoid sticking.The pumping speed, aspiration, and atomization gas flow rate were 6 ml/min, 35 m 3 /h, and 600 L/h, respectively.The nozzle diameter of the spray dryer is 0.7 mm.The spray-dried LCmPs were stored at room temperature in a desiccator before they were used in different characterizations.

Scanning electron microscopy
The shape and morphology of the LCmPs were evaluated by scanning electron microscopy (SEM), using a Zeiss 1550 SEM (Oberkochen, Germany).Samples were scattered onto doublesided adhesive carbon tapes, which were previously adhered to aluminum stubs.The samples were then coated with a gold/palladium alloy thin film using a Thermo SC7640 (Waltham, MA) sputter coater at 20 mA for 40 s under argon gas.The electron beam with an accelerating voltage of 2 kV was used at a working distance of 2.3 mm.Images were captured at two magnifications, 5000Â, and 25,000Â respectively.

Tapping mode atomic force microscopy
Tapping mode AFM, also known as amplitude-modulation AFM, is typically used to measure surface topography and other material parameters on the nanometer scale, and it is primarily used for imaging in air and liquid.This technique has been thoroughly explained in the literature [38].In tapping mode, a cantilever is excited externally at a constant frequency near its resonance (around 300 kHz in our work).During the imaging process, the tip intermittently contacts the sample, and the tip-sample interactions change the amplitude and phase of the oscillating cantilever.Thus, the amplitude and phase serve as the experimental observation channels, which offer topography or height image and phase image of the scanned sample [38].The phase image shows contrast for variations in material properties, such as surface viscoelasticity and tip-sample adhesion [39].
A Dimension Icon AFM (Bruker, Santa Barbara, USA) was used to acquire the tapping mode images, and the image analysis was performed in the NanoScope Analysis software (Version 1.50, Bruker).A third-order polynomial-flattening algorithm was employed to remove surface tilt from the height images.All other images were unaltered.

Intermodulation atomic force microscopy
ImAFM is a multi-frequency dynamic AFM technique developed quite recently, and the detailed description of this technique can be found in the literature [40,41].In brief, in ImAFM measurements, the cantilever is simultaneously excited with two drive frequencies centered around its resonance (around 300 kHz in our work).When the nonlinear tip-sample forces perturb the freely oscillating cantilever, there will be a frequency mixing of the two drive tones, giving rise to new frequency components.Those new frequency components are called intermodulation products (IMPs) [42].The IMPs contain information about tip-surface forces, which can be reconstructed from intermodulation spectra.Furthermore, the tip-surface force curves can be used to extract local surface nanomechanical properties [37,43].
The Dimension Icon AFM, as used in the tapping mode AFM measurements, was employed to perform the ImAFM measurements, and it was additionally connected to a multi-frequency lock-in amplifier (Intermodulation Products AB, Sweden).The amplifier generates the drive signals and records the intermodulation spectra.The amplitude and phase of each IMP are recorded at every image pixel in real-time while scanning, providing a multitude of amplitude and phase images.The IMP software suite (Version 2.7.1, Intermodulation Products AB) was used to analyze the data.All experiments were performed in ambient air.
Rectangular cantilevers Tap300DLC (BudgetSensors, Bulgaria) with a spring constant %40 N/m and a tip radius <15 nm, as specified by the manufacturer, were used in both tapping mode AFM and ImAFM experiments.Those cantilevers have a diamond-like-carbon coating (15 nm thick) on the tip side of the cantilever and an aluminum coating (30 nm thick) on the detector side of the cantilever.The spring constant of each cantilever was determined with the non-invasive thermal noise method as implemented in the ImAFM Software Suite (Intermodulation Products AB, Sweden) [44,45].

Confocal laser scanning microscopy
Two batches of LCmPs, L1 and P1, with dye-labeled lipids, were prepared for CLSM measurements to investigate the distribution of lipids and PEGylated lipids on the surface of the LCmPs.The LCmPs were scattered onto a glass slide (refractive index 1.5) covered with a thin layer of the standard immersion oil (Immersol 518F, fluorescence free, Carl Zeiss, Germany) and then they were covered by another glass slide.The samples were subsequently visualized with a Zeiss confocal laser scanning microscopy (LSM 700, Jena, Germany).The excitation and emission wavelength was 560 nm and 583 nm for Rb-DPPE; and 485 nm and 523 nm for CF-PEG2K-DSPE, respectively.The distribution of the dye-labeled The collected images were firstly analyzed by Huygens (Scientific Volume Imaging B.V., Netherlands) using the deconvolution function to obtain images with an optimal signal-to-noise ratio.Those images were then processed by Fiji (Image J) to make stack montage and stack Z project images.

Particle size distribution
Particle size was measured with a Malvern Mastersizer 2000 Ò laser diffractometer, with a dry sampling system (Scirocco 2000, Malvern, UK).The Mie theory was applied for calculations of the scattering matrix as it provides less biased particle-size distributions for small particles (diameter <50 mm) than the Fraunhofer approximation.The refractive index values 1.544 (real part) and 0 (imaginary part), were used for both silica and the LCmPs.The dispersive air pressure used in the measurements was 0.1 bar.The size of the particles was quantified in terms of the 10th, 50th and 90th percentile (d 10 , d 50 , and d 90 respectively), defined such that the indicated percentage of the total particle volume resided in particles smaller than the corresponding percentile.The span ((d 90 -d 10 )/ d 50 ), which reflects the size distribution width, was calculated.The reported values are the average of two determinations.

Results
Three types of LCmPs, L1, P1 and P2, were prepared by a single step spray-drying method.Their morphology, topography, and surface nanomechanical properties were characterized by SEM and AFM, respectively.The distribution of the dye-labeled lipids over the L1 and P1 LCmPs was observed by CLSM, and the particle size distribution was analyzed by laser diffraction.The results for all these characterizations are present in this section.

Morphology and topography
The macroscopic morphology and topography of the LCmPs were observed by SEM and tapping mode AFM.The SEM images of silica and the LCmPs are shown in Fig. 1.As we can see, the silica microparticles are spherical and well dispersed.They generally have quite smooth surfaces with some bumps scattered on the surface.After coating with lipids by spray-drying, the LCmPs are still spherical.There are no obvious agglomerates present for all the three types of LCmPs, L1, P1 and P2 observed at 5000Â.At higher magnifications 25,000Â, we observe that the surface of the LCmPs has both smooth and rough areas.The rough area consists of many tiny nanoparticles of varying sizes.The presence of tiny particles on the spray-dried LCmPs was also observed previously by other researchers [46].During the spray-drying process, aerosols of different sizes will be atomized, and the tiny nanoparticles tend to fall onto larger ones during the drying process in the chamber.The nanoparticles on P1 and P2 look more spherical and slightly larger than those on L1, as shown in Fig. 1 at 25,000Â, which should be due to the presence of PEGylated lipids.These long PEG chains covalently bonded to lipid molecules may interact with each other, resulting in aggregates during spray drying.
The topography of the silica microparticle, as well as the LCmPs, L1, P1 and P2, was characterized by tapping mode AFM.The height and phase images are shown in Fig. 2. The height image of silica looks quite smooth with a root mean square roughness (R q ) of 2.7 nm, despite the presence of bumps at the surface with a height variation from 20 nm to 40 nm.The height profiles along the lines (as shown in the height images) across the nanoparticles are shown in Fig. S1 in the supporting information.
The topography of the LCmPs, L1, P1 and P2, is also shown in Fig. 2. As presented in the height images, the scanned areas of the LCmPs are much rougher than the bare silica, with larger lumps on the surfaces for all three types of LCmPs.The height variation over the scanned areas of these four microparticles is in this order, silica (%18 nm) < L1 (%85 nm) < P1 (%569 nm) < P2 (%663 nm).With such large height variation, it is not meaningful to compare the roughness of the LCmPs.Instead, the height profiles of those lumps on the LCmPs present in the height images are provided (Supporting Information).As shown in Fig. S1, the height variation for the lump on L1 (around 70 nm) is much smaller than that on P1 and P2 (between 200 nm and 400 nm).The larger nanoparticles present on the surface of the P1 and P2 might be due to the formation of small aggregates of PEGylated lipids induced by the long PEG chains.
The corresponding phase images of all four types of microparticles are also shown in Fig. 2. It is clearly seen that the phase images of the LCmPs with more fine features are different from that of silica.This observation indicates the presence of a lipid coating on the surface of the LCmPs, L1, P1 and P2 since the phase image contrast is due to changes in material properties, such as surface viscoelasticity and tip-sample adhesion [39].Furthermore, the phase variation over the surfaces of those LCmPs signifies a clear variation or heterogeneity in surface properties.However, it is difficult to detect if there is any defect or pore in the lipid coating because the phase images show some contrast between different areas, especially for P1 and P2.In summary, there is a lipid coating on the spray-dried LCmPs, L1, P1 and P2; but the presence of the PEGylated lipids in the excipients resulted in larger nanoparticles or PEG rich domains on P1 and P2 compared to L1, giving rise to more heterogeneous coating films on P1 and P2.

Surface nanomechanical properties
The results obtained from the tapping mode AFM confirmed the presence of a lipid coating over the LCmPs.Herein, ImAFM was employed to study the quality of the film coatings.This AFM-based technique has the ability to produce a complete force curve at each image pixel while capturing the height and phase images.The force curves can be used to extract local surface nanomechanical properties, such as elastic modulus, indentation, and adhesive force minimum, by fitting the Derjaguin-Muller-Toporov (DMT) contact mechanics model to parts of the measured force curves [43].The DMT model was originally introduced to describe the elastic deformation of a sphere indenting a plane in the presence of adhesion [47].The DMT model has been widely used to assess the contact mechanics from AFM data [48][49][50] although its applicability has been criticized for being overly simplistic [49,51].In any case, the extracted surface nanomechanical properties of this relatively simple model enable the comparison between the bare silica and the coated particles.Since parameters, such as tip-sample distance and the surface topography variations in the scanned area, affect the response in ImAFM [42], relatively smooth areas were chosen in all the ImAFM measurements.
The height images of the scanned microparticles in ImAFM measurements are shown in Fig. 3.The silica surface is much smoother than that of the LCmPs, L1, P1 and P2, which was also seen in the tapping mode AFM images.Since relatively smooth areas were chosen in the measurements, the height variation of the scanned areas for P1 and P2 (around 40-60 nm), without the presence of large nanoparticles, is much smaller than that observed in the tapping mode AFM height images.
The DMT model was used to extract the surface nanomechanical properties, including elastic modulus (E) and indentation (z), of the scanned areas following the procedure described previously [43].A tip radius of 13 nm was assumed in all modeling.The elastic modulus maps of silica and the LCmPs are also shown in Fig. 3.We note that the bare silica with a higher elastic modulus is generally stiffer than the LCmPs.The elastic modulus map of L1 is relatively more homogeneous, and it shows a larger dark area with low elastic modulus compared to P1 and P2, with the same scale bar.This means that the coating film on L1 is more homogeneous and softer than that on P1 and P2.On the other hand, the elastic modulus maps of P1 and P2 show clear variation, indicating heterogeneity in the surface property due to an uneven coating film.
The elastic modulus distribution of silica, L1, P1 and P2, extracted from the elastic modulus maps, are shown in Fig. 4. The bare silica shows a wide distribution of elastic modulus, rang- ing from 0.5 GPa to 5 GPa.In contrast, the elastic modulus of L1 has a very narrow distribution, with a majority of the values <0.5 GPa.The contrast confirms the presence of a more or less uniform lipid coating on L1.On the other hand, we observe that the elastic modulus of P1 and P2 has a wider distribution compared to L1, which indicates that those two microparticles are coated less homogeneously than L1.In addition, some of the elastic modulus values of P1 and P2 are in the range of 0.5 to 1 GPa, overlapping with the elastic modulus for silica, which may hint that some spots on P1 and P2 are coated with an extremely thin layer of lipid or not even coated.
The average values of the modeling results, E and z, obtained by fitting Gaussian distribution to the experimental data, are presented in Table 2.As shown in the table, the elastic modulus of silica is around 2.18 ± 1.2 GPa.This value is similar to what has been reported for the stiff silica particles used in the preparation of silica-reinforced rubbers, whose elastic modulus values are as high as 2 GPa.[52] With a lipid coating, the elastic modulus of the LCmPs, L1, P1 and P2, decreased dramatically, suggesting softer surfaces.On the other hand, the indentation was almost doubled for all the LCmPs compared to silica with an average indentation of about 3.1 ± 0.7 nm, which is close to the measured surface roughness (2.7 nm).The high indentation values obtained for those LCmPs further confirm the presence of a lipid coating on the three types of LCmPs.Herein, we would like to point out that the indentation values are not equivalent to the coating thickness.It is necessary to point out that the force curves measured between the tip and the sample are affected by many factors, such as tip radius, tipsample distance, and the height variation of the sample surface, which will also induce variation to the extracted surface nanomechanical properties.

The distribution of lipids/PEGylated lipids over the lipid-coated microparticles
CLSM is usually used to study the distribution of the encapsulated drug in the solid microparticles or hydrogel [32,53].Herein, this technique is employed to study the distribution of dyelabeled lipids and PEGylated lipids over the LCmPs using a Zstack experiment procedure.The scanning in the z direction was performed with a step of 0.2 mm from the top/bottom to the bottom/top of the selected particles, which also means that the planes within the steps (0.2 mm) were not scanned.Depending on the height range of the scanned particles, different slices of images were collected.Those images are present at the top row in Fig. 5 as stack montage images for L1 and P1, respectively.We can observe the presence of dye-labeled lipids at each measured focusing plane.Two Z project images of L1 and P1 were produced based on those stack montage images, respectively, as shown in the bottom row in Fig. 5.The Z project images accumulate the laser intensity from all the focusing planes shown in the stack montage images and convert the signal into 3-D images.From the 3-D images, we can not only see the distribution of dye-labeled lipids but also identify the intensity at different spots.As shown in the scale bar, the brighter the color, the higher the intensity.On the other hand, a black spot does not mean that this area is uncoated but rather indicates the absence of dye-labeled lipids.Such spots could occur since the concentration of the dye-labeled lipids is only 1 mol % of the total amount of lipids used for spray drying.Furthermore, in the Z-stack experiments, the images were taken at a 0.2 mm interval, which also caused the loss of intensity.As shown in the Z project images, the intensity varies at different spots as a result of the uneven distribution of the dye-labeled lipids.The dark area in the Z project image of P1 is larger than that of L1, which indicates that P1 is less evenly coated compared to L1.This phenomenon may be due to the formation of PEG rich domains, present as large nanoparticles on P1 (as shown in both SEM image and AFM height image), resulting in a lower amount of lipid molecules available for coating.Furthermore, it might also indicate that the thickness of the lipid coating for some areas on P1 is smaller than that on L1.

Particle size distribution
The particle size distribution of the bare silica and the LCmPs were analyzed using a dry sampling system.The purpose of this characterization was to test if there were any agglomerates in the SLmP powder products; thus, a very low dispersive air pressure, 0.1 bar, was applied in the measurements so that it would be unlikely to break any existing agglomerates.The laser diffraction results, as presented in Fig. 6 and Table 3, show that the particle size distributions of the LCmPs (L1, P1 and P2) are unimodal and narrow (the span is < 1), ranging from 3 to 11 mm (more than 90% of the particles having a diameter below 10 mm), which is similar to the bare silica microparticles.The similarity in the particle size distribution between silica and the LCmPs indicates that no agglomerates were formed during the spray-drying process for the preparation of the LCmPs.However, the diameter of the silica microparticles is slightly larger than the coated ones, with an exception for d 50 and d 90 of P1.The slightly larger particle size of silica microparticles might be due to the drying of the silica during the spray-drying process, which may cause a slight reduction of the size of the silica microparticles since amorphous materials are usually in equilibrium with the surrounding moisture.Another possible reason might be due to the measurement procedure.It is worth noting that the particle size distribution for all three types of LCmPs shows very small variation, which may suggest that the lipid coating did not affect the dispersibility of the powder.Since this is a pre-formulation study and the silica microparticles used have a diameter larger than 5 mm, it is not meaningful to compare the size of the LCmPs with the upper limit required for an optimal deep lung deposition, the aerodynamic diameter of which should be <5 mm [4].

The effect of PEG on the surface properties and performance of the lipid-coated microparticles
Both the morphology images obtained with SEM and height images obtained with tapping mode AFM demonstrated that the size of the nanoparticles on the LCmPs increased with the presence of PEGylated lipids.The formation of larger nanoparticles on P1 and P2 compared to that on L1 is probably due to the interactions between the long PEG chains, which facilitated the formation of aggregates during the mixing and spray-drying processes.These hydrophilic PEG chains would preferentially end up on the outer shell of the LCmPs as a result of their higher solubility in methanol compared to DPPC or DSPC [31].It has been reported that the interaction between the hydrophilic PEG chains of the PEG-DSPE molecules regulated the formation of domains rich in PEG-DSPE when a supported lipid bilayer was prepared by the vesicle fusion method on a SiO 2 /Si substrate.Those domains rich in bulky PEG chains were detected by tapping mode AFM [54].The formation of PEG rich nanoparticles or domains is likely to reduce the surface coverage and local lipid coating thickness on P1 and P2 compared to L1, as evidenced by the Z project confocal images of P1 and L1, as shown in Fig. 5.The reduction in surface coverage and local coating thickness would inevitably affect the surface nanomechanical properties of P1 and P2, resulting in heterogeneously coated surfaces with a wide distribution of elastic modulus.Since there is some overlap between the elastic modulus of bare silica and P1, as well as P2, it is possible that some spots or areas on P1 and P2 are coated with a very thin layer or even not coated at all, presenting as defects in the lipid coating.In contrast, the SLmP without the presence of PEGylated lipids, L1, shows a narrow distribution of elastic modulus, the values of which are much lower than that of silica, suggesting an almost homogeneously coated surface.One should bear in mind that the silica microparticles employed in this study have smooth surface features, which might ease the formation of lipid coating and improve the coating quality on L1.However, in reality, the API microparticles usually have a rough surface feature, which may pose some difficulty in achieving high-quality lipid coating during the spray-drying process.In conclusion, the incorporation of PEGylated lipids in the excipients for the production of spray-dried LCmPs is likely to introduce some defects in the coating.Herein, it should be pointed out that the weight concentration of the lipid excipients in the LCmPs is only 4.8%.It might be possible to improve the coating quality of the PEGylated LCmPs by increasing the lipid concentration.
The morphology of the particle and its surface properties affect important powder properties such as cohesiveness or dispersibility because the surface asperities of the particles could lower the true contacting area between the particles, and thus reduce the van der Waals force or powder cohesiveness [55,56].Therefore, one of the advantages of those nanoparticles observed on the LCmPs is that they may reduce the inter-particulate van der Waals forces by decreasing the contact area, and thus promote drug dispersibility, which explains the good dispersibility of all three types of LCmPs as evidenced by the absence of agglomerates detected by laser diffraction.Excellent aerosol dispersion performance of spray-dried microparticles consisting of DPPC and dipalmitoylphosphatidylethanolamine PEG (DPPE-PEG) was also reported by Mansour et al., which was evaluated using the Next Generation Impactor TM (NGI TM ) coupled with the HandiHaler Ò dry powder inhaler device [12].

Implication for controlled drug delivery
In theoretical studies, microencapsulated particles are often considered as reservoir systems [57,58].Hence, drug release from such particles is often modeled as a progressive dissolution of the internal solid drug core (due to solvent penetration through the coating) that gives a liquid solution in the region between the coating and the dissolving solid core.The existence of a concentration gradient between the inner solution and the outer dissolution medium determines drug diffusion through the coating [28,59].This process would be greatly enhanced if there were any defect or pore in the coating.On the other hand, a thicker coating would decrease the diffusion rate [28].There are only a few studies on the effect of PEG on the dissolution of the LCmPs.Among these, Levet et al. reported that PEGylated excipient-comprising formulations exhibited a higher burst effect than formulation without PEGylated excipient [31].They suggested that the surfactant effect of the lipids and PEGylated lipids might enhance the solubilization of unencapsulated cisplatin microcrystals on the SLmP surface, giving rise to the higher burst effect [31].In addition, Mu and her co-workers reported that the addition of PEG 4000 in the formulation resulted in the accelerated release of insulin from the TG 22 (Dynasan D122) microparticles.They proposed that the reason could be due to the pore-forming effect of PEG, which created hydrophilic channels for the release medium to enter and hence accelerated the drug release [32].In our study, we found out that the lipid coating on L1 type LCmPs, without PEGylated lipids, is more homogeneous than that on P1 and P2.On the other hand, the presence of PEGylated-lipids might reduce the surface coverage and local coating thickness of P1 and P2, inducing some defects in the coating, which would certainly affect the dissolution if some fraction of APIs were encapsulated instead of silica.Furthermore, the PEG rich domains present on P1 and P2 may be dissolved very fast when they are in contact with an aqueous solution (the solubility of DSPE-PEG2K in water is 8 g/L, as provided by Avanta safety data sheet), resulting in hydrophilic channels which would accelerate the drug release.Therefore, the presence of PEGylated lipids in the excipients is expected do result in a burst release and/or higher dissolution rate of the LCmPs prepared by spray drying compared to the ones without PEGylated lipids.

Conclusions
This study has demonstrated that coherent coatings of phospholipids on hydrophilic surfaces can be obtained by spray drying, despite the apparent incompatibility of hydrophilic surfaces and hydrophobic excipients [32], indicating the viability of phospholipids for controlled pulmonary drug delivery.Indeed, lipid-coated silica microparticles prepared by a single step spray-drying method exhibited a relatively uniform coating, as inferred from tapping mode atomic force microscopy (AFM), intermodulation AFM and confocal laser scanning microscopy.PEGylated phospholipids can prolong the residence time of drug delivered to the lung by prevention of macrophage uptake [10][11][12].However, according to our results, the use of such lipids is problematic in a controlled-release context.Incorporation of PEGylated lipids lead to the formation of larger nanoparticles on the particle surfaces, resulting in a more heterogeneous lipid coating with an extremely thin local thickness or even uncoated areas, inducing defects in the coating.The defects, together with the presence of PEG rich domains, will result in the formation of hydrophilic channels through the coating, which will be detrimental for the ability of the coating to control the controlled drug release.The observed defects in the coating caused by the presence of PEGylated lipids could contribute to the acceleration of the drug release often seen when such molecules are incorporated in solid lipid microparticles [31,32], something that hitherto has been hypothesised as being due to the pore-forming effect of PEG [32].However, the presence of PEGylated lipids in the coating does not affect the dispersibility of the coated microparticles since no agglomerates were detected in any of the three types of powder despite that the PEG chains are prone to attract water.For future studies, it is noted that a layer-by-layer method, as proposed by Luo et al. for preparation of redispersible solid lipid nanoparticle by nano spray drying technology [60], might offer a possibility to produce solid lipid or lipid coated microparticles with both controlled-release and stealth properties.

Fig. 1 .
Fig. 1.Scanning electron microscopy images of silica and the LCmPs, L1, P1 and P2.Magnification 5000Â for the images on the left; magnification 25,000Â for the images on the right.

Fig. 2 .Fig. 3 .
Fig. 2. The height (left) and phase (right) images of the microparticles, silica, L1, P1 and P2, measured by tapping mode AFM in air.The scanned area for all the samples is 2 mm Â 2 mm.

Table 1
Composition of the coating recipes for the preparation of LCmPs.

Table 2
Surface nanomechanical properties extracted from the ImAFM data.

Table 3
The particle size distribution of the silica microparticles and the LCmPs.