Monodisperse polymeric particles prepared by ink-jet printing: Double emulsions, hydrogels and polymer mixtures

Colloids and Surfaces B: Biointerfaces 79 (2010) 47–52

Contents lists available at ScienceDirect

Colloids and Surfaces B: Biointerfaces journal homepage: www.elsevier.com/locate/colsurfb

Monodisperse polymeric particles prepared by ink-jet printing: Double emulsions, hydrogels and polymer mixtures Marcel R. Böhmer ∗ , Jan A.M. Steenbakkers, Ceciel Chlon Department of Biomolecular Engineering, Philips Research Eindhoven, High Tech Campus 11, 5656 AE Eindhoven, The Netherlands

a r t i c l e

i n f o

Article history: Received 7 January 2010 Received in revised form 11 February 2010 Accepted 22 March 2010 Available online 31 March 2010 Keywords: Monodisperse Microparticle Ink-jet printing Double emulsion Hydrogel

a b s t r a c t Submerged ink-jetting produces a monodisperse emulsion that can be converted into monodisperse particles. As the initial droplet size is known and the final particle size can be easily measured, such a method can be used to quantify the shrinkage and the swelling of polymer particles made from double emulsions, polymer mixtures and hydrogel forming polymers. It is found that at the same starting concentration and initial emulsion droplet size poly-lactide-co-glycolide particles made from an ink-jetted emulsion have the same size as particles ink-jetted from a solution, however with a more porous structure. The total pore volume, however, is negligible compared to the polymer volume of the particle. If polymers containing a poly-ethylene glycol block are included, particles with internal porosity are formed, even if no double emulsion process is applied. Still the final particle size is the same. Only if typical hydrogel forming polymers are used, in which water is distributed more homogeneously, significantly higher particle diameters are found; for a four-arm PEG–poly-caprolacton a degree of swelling of 3.3 is found. © 2010 Elsevier B.V. All rights reserved.

1. Introduction Biodegradable polymer particles that are formed from polymer solutions emulsified under shear followed by solvent removal have, in general [1–5], a wide size distribution. The controlled formation of complex morphologies with the shear-based methods is hindered by coalescence or breaking-up of, already polydisperse, droplets while they are in the process of transformation to particles by solvent dissolving into the aqueous phase followed by its evaporation. The wide size distribution of the particles prepared with traditional methods hampers understanding of the relation between the initial droplet size and the particle size and morphology. In particles formed using double emulsion technology, which is frequently used to include hydrophilic high molecular weight molecules such as proteins, there is a need to understand how the final particle size and morphology are affected by starting from an emulsion rather than a polymer solution. In addition the processing parameters of the emulsion determine the internal and surface porosity which affect the release rates, including the initial or burst release [3]. To offer vulnerable hydrophilic molecules, such as proteins or nucleic acids, a less hostile environment in terms of accumulating acidic degradation products or exposure to organic solvents is required, and therefore, copolymers with poly-ethylene glycol blocks are often used, see Ref. [6–8]. This has

∗ Corresponding author. Tel.: +31 40 2748252; fax: +31 40 2744906. E-mail address: [email protected] (M.R. Böhmer). 0927-7765/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.colsurfb.2010.03.021

consequences for the internal structure and, because the precipitation conditions of such polymers are different, the particle size might be affected, including the consideration that such particles in an aqueous solution may stay swollen due to the uptake of water. The particle composition and morphology have consequences for drug delivery from such particles, the release of active ingredients depends on the size, degradation rate and internal structure including the porosity. As these parameters are likely interdependent, there is a need for techniques to prepare monodisperse particles to study these phenomena. In addition, specific applications such as embolization treatments to locally irradiate liver tumors or arrest the blood supply to uterine fibroids, depend critically on the size. For such applications manufacturing methods that yield high amounts of monodisperse particles are needed [9]. A prerequisite to synthesize monodisperse particles is to start from monodisperse emulsions. Droplets produced in co-flowing liquids can be used to obtain particles with a narrow size distribution after solvent removal [10–13]. The method can be scaled-up to a multi-channel system leading to high throughput as shown by Nisisako and Torii [14]. An alternative technique, operating at high frequencies, is submerged ink-jetting. This method reaches 24,000 drops per second or 8 ml/h for a single nozzle as in our current set-up. It has been used for the preparation of monodisperse solid particles, liquid and gas filled capsules [15,16]. With a similar jetting technique [17], particles derived from double emulsions were prepared [18]. Submerged ink-jet printing with a 30 ␮m nozzle is

48

M.R. Böhmer et al. / Colloids and Surfaces B: Biointerfaces 79 (2010) 47–52

very suitable for the preparation of particles in the size range of 3–30 ␮m, and therefore complements the recently introduced jetting approach using a needle and a glass capillary in a PVC tube by Choi et al. [19] suitable for particles larger than 30 ␮m. A similar system with a glass tube is used by Utada et al. [20] also for fairly large droplets and particles. These jetting methods have the advantage that poly-dimethyl siloxane (PDMS) microstructures can be avoided. As pointed out by Choi et al. [19], PDMS-based structures can cause complications due to swelling and contamination. Both in microfluidic structures and in jetting systems, monodisperse emulsion droplets containing a polymer solution or emulsion can be dispersed in an aqueous continuous phase, which is followed by conversion of the liquid droplets into particles with minimal agitation. In this work we focus on submerged ink-jet technology using a 30 ␮m nozzle applied to the formation of particles of biodegradable polymers. With larger nozzles, larger initial droplet sizes can be prepared, yielding, depending on the polymer concentration, for instance particles up to 85 ␮m [18]. The aim of this work is to obtain insight in the factors determining the particle size and morphology starting from the same initial droplet size for a variety of polymer compositions, including mixtures with pegylated polymers. We will compare final particle sizes of particles prepared from an ink-jetted polymer solution and from an ink-jetted emulsion to determine if the presence of an additional aqueous phase leads to a change in the particle size because of the inclusion of pores. Further analysis also allows to obtain information on the internal and surface porosity of these particles. We present sizes and particle morphologies obtained for polylactide-co-glycolic acid (PLGA) particles prepared from a solution and from an emulsion, followed by the experiments in which part of the PLGA was replaced by a block-copolymer of poly-lactide and poly-ethylene glycol (PDLLA–PEG). Finally, we choose a polymer which is expected to form hydrogel particles. Results for hydrogel particles obtained with a drop-by-drop method have been presented [21] but also in this case we want to demonstrate the relation between initial drop size and structure and particle size explicitly. Therefore we show results for a four-arm PEG–poly-caprolacton, as a pure component but also mixed with PLGA.

Fig. 1. (A) Schematic representation of the submerged ink-jet printing set-up for the preparation of monodisperse particles. The liquid to be printed (the ink) is in reservoir (1) from which it is transported to the ink-jet nozzle (2) under an assistant pressure (not shown). The printed droplets that are formed arrive on a slope which transports them away from the nozzle to a collection vessel. The droplet formation is monitored using a camera (4). (B) Stroboscopic image of the jetted droplets.

2.3. Ink-jet printing

Poly-lactic-co-glycolide (75:25), PLGA Mw 66,000–107,000, was obtained from Sigma. Other polymers were obtained from Polymersource: poly(ethylene oxide-b-lactide) PDLLA–PEG, with a Mw /Mn of 1.15 and a PDLLA and PEG block having a Mw of 1800 and 2500 respectively and four-arm poly(ethylene oxide-b-␧-caprolactone) with pentaerythritol core, abbreviated as 4A-PEG–PCL, with a Mw /Mn of 1.09 and each PEG and PCL branch having a Mw of 2500 and 11,500 respectively. Dichloromethane (DCM) was obtained from Merck, nile red from Invitrogen, Dextran-FITC (Mw = 40,000) from Sigma and poly-vinyl alcohol (PVA) Mw 13,000–23,000 from Aldrich.

The ink-jet set-up has been described previously [15]. Experiments were conducted using piezo-driven Microdrop ink-jet nozzles (MK-140H) with a 30 ␮m diameter. An external pulse generator was used (Fluke PM 5139) to obtain frequencies in the range of 24 kHz. The liquid volume ink-jetted at this frequency is 7.5 ml/h or 86 million particles per hour. The Microdrop driver (MD-E-201H) also allows for taking pulse-triggered images, which was used to follow the drop formation process. A schematic overview of the ink-jet set-up is given in Fig. 1A which shows an ink-jet nozzle submerged in a poly-vinyl alcohol (PVA) solution in a rectangular 80 mm × 30 mm × 100 mm (l × w × h) glass container. The solution or first emulsion with dichloromethane (DCM) as the continuous phase to be jetted is led through the nozzle with an assistant pressure of 0.40–0.45 bar. Stroboscopic images are taken throughout the process and Fig. 1B shows an array of droplets ejected from the nozzle. The inkjetted droplets with a size of 55 ␮m slide down the glass slope, made out of a glass tube with an inner diameter of 6 mm, which was opened on one side, and are collected in a 50 ml vessel, in which they sediment. The sedimentation occurs rapidly because the droplets predominantly consist of DCM, which has a density of 1.33 g/ml. As the set-up has two slides, and the nozzle can be switched from one slide to the other, semi-continuous production of monodisperse droplets is possible. The only force exerted on the droplets after ink-jetting is the gravitational force. The liquid level in the jetting cell is maintained by a connection to a 1 l PVA solution in a large flask connected to the rectangular glass container and the sedimentation vessel. The PVA solution is circulated by means of peristaltic pump. This allows for a start of the removal of DCM, by which the droplets become less vulnerable to breakage in subsequent processing. Meanwhile the PVA in the aqueous phase acts as a stabilizer and prevents coalescence of the droplets [22]. After jetting for typically 30 min and sedimentation in a collection vessel, the droplets are further processed under gentle stirring to remove DCM and washed to remove excess PVA.

2.2. Solutions

2.4. Analysis

Polymer solutions were prepared at 2% (w/w) in DCM. Polymer mixtures were used in a 1:1 (w/w) ratio adding up to 2% (w/w). First emulsions were prepared by adding 2% (w/w) water to the DCM phase, optionally containing Dextran-FITC or nile red and emulsifying with an ultrasonic agitator, while cooling on ice, before ink-jetting. The PVA solution was prepared at 80 ◦ C and left for minimally 24 h before being used.

Particle size distributions were measured in isoton II (Beckman Coulter) on a Beckman Coulter Counter Multisizer III with a 50 ␮m aperture tube, allowing quantification between 1.1 and 30 ␮m. A Leica Z16APO microscope was used to obtain bright field images. SEM images were recorded on a Philips SEM XL 40 FEG system at an acceleration voltage of 1 kV. To obtain fluorescent images a Leica TCS SP5 confocal microscope is used. Images were captured with

2. Methods 2.1. Materials

M.R. Böhmer et al. / Colloids and Surfaces B: Biointerfaces 79 (2010) 47–52

49

Table 1 Overview of particle size distributions.

Mean Median Mode SD CV

PLGA

PLGA DE

PLGA/PDLA–PEG

PLGA/PDLA–PEG DE

4A-PEG–PCL

4A-PEG–PCL + PLGA

15.27 15.36 15.40 0.74 4.85%

15.37 15.15 15.02 1.101 7.17%

14.82 14.79 14.83 0.908 6.12%

14.48 14.38 14.44 1.094 7.55%

22.15 22.39 22.65 1.829 8.26%

19.24 19.45 19.64 1.518 7.89%

a 60× objective lens using the 561 nm line was used to excite the nile red and the 488 nm line to excite Dextran-FITC. 3. Results 3.1. PLGA particles The PLGA particles produced have a particle diameter of 15.4 ␮m and the coefficient of variance (CV) is 4.8% as shown in Table 1. The initial droplet size measured from the stroboscopic image is 55 ± 2 ␮m which demonstrates that the size can indeed be accurately predicted from the polymer concentration and the initial droplet size as already proven in Ref. [7]. Batch to batch variations showed differences in the modal diameter of about 0.3–0.4 ␮m and variations in the CV between 4 and 8%. An overview image of the particles prepared is given in Fig. 2A, this bright field microscopic image shows homogeneous particles, with a smooth surface as can be deduced from the SEM pictures, see Fig. 2B. Ink-jetting was repeated in the presence of the hydrophobic fluorescent dye nile red. A homogeneous distribution of nile red is obtained (Fig. 2C) with the exception of a few (5 in this confocal plane) pores which are smaller than 10% of the particle diameter, typical for PLGA particles, including the PLGA particles prepared by shear-based methods [3]. When 2% water was added to the solution the emulsion that was formed by ultrasonic agitation stayed turbid for more than

1 h, which is sufficient for jetting experiments. The resulting particles are presented in Fig. 2D–F. In Fig. 2D the bright field overview image shows a distinct change in the internal structure of the PLGA particles made from this emulsion as compared to the particles made from the PLGA solution. The particles do no longer appear with a white, homogeneous, interior but are greyish at the same microscope settings. Not all particles in the sample have the same structure; some show a distinct white mostly spherical spot indicating a large pore, while for the majority of such spots are not visible. Therefore although the final particle size is the same for the prepared particles, there is variation in the internal structure. The outer surface of the particles is still smooth as shown by the SEM image (Fig. 2E). Some particles display pores on their surface as pointed out by the arrows in Fig. 2E. The particles have a more internal porosity as appears from the confocal image of a preparation made in the presence of nile red, see Fig. 2F, showing a confocal plane with 7 pores larger than 2 ␮m. The increased surface porosity and internal porosity agree with observation for other jetted emulsions [18] as well as particle made from double emulsions by shear-based methods [3]. Although these marked changes in the morphology are present, the particle size has not changed, the modal diameter is 15.02 mm with an insignificant increase in the CV from 4.9 to 7.2%. The impact of the residual internal pores on the particle diameter is very limited since the diameter scales with the volume1/3 . For

Fig. 2. Particles obtained by submerged ink-jet printing and subsequent processing. (A–C) PLGA particle from solution: (A) bright field image, (B) SEM picture, and (C) confocal plane with the hydrophobic dye nile red. (D–F) Particles formed by ink-jetting water in DCM emulsion with PLGA: (D) bright field image, (E) SEM picture with arrows showing distinct surface pores, and (F) confocal plane of these particles with the hydrophobic dye nile red. The bar indicates 15 ␮m.

50

M.R. Böhmer et al. / Colloids and Surfaces B: Biointerfaces 79 (2010) 47–52

Fig. 3. Effect of incorporating pores on the particle diameter, starting from a 15 ␮m dense polymer particle and inclusion of pores of 2–12 ␮m.

instance if 4 pores with a diameter of 4 mm are added to a 15 mm solid polymer particle, the volume of the particle is increased with 4 times the volume of a 4 ␮m spherical pore which is 0.134 pl. This is less than 10% of the volume (1.77 pl) and consequently negligible considering the change in the particle radius. A guide to demonstrate the impact of the pores is presented in Fig. 3, where the effect of the number and size of the pores on the particle radius are presented. The largest pores we observe in the particles are about 4–5 ␮m, so we can indeed not expect a drastic increase in size due to the obtained pore sizes. If, however, the pores would contain the total water content present in the first emulsion, a 55 ␮m droplet would shrink to 4% of its original volume which corresponds to a particle diameter of 19 ␮m (instead of 15 mm corresponding to 2% of the original droplet volume). It is therefore obvious that the majority of the water present in the first aqueous phase has left the particle. The particle surface is still smooth but does contain a few pores the shape of the surface pores,

Fig. 4. Confocal scan of Dextran-FITC containing 15 ␮m PLGA particles, showing the distribution. The fluorescence intensity varies from pore to pore, as a consequence of the presence of a number of pores that were filled from the outside of the particle.

with the inward curvature, demonstrates that the last process that occurred before they were completely solidified was inflow of liquid into the particle. This indicates a process in which PLGA is largely solidified before all solvent has left the particle and that upon disappearance of the DCM pores are formed that are finally filled with water from the outside. This was confirmed by inclusion of the hydrophilic dye Dextran-FITC, dissolved in the first aqueous phase. Many pores of the resulting particles showed fluorescence but some pores were hardly fluorescent, therefore did not contain Dextran-FITC, or water, from the first aqueous phase, see Fig. 4.

Fig. 5. (A–C) PLGA:PDLLA–PEG particles from solution: (A) bright field image, (B) SEM picture showing surface pores, and (C) confocal plane of these particles with the hydrophobic dye nile red. (D–F) Particles formed by ink-jetting water in DCM emulsion with PLGA and PDLLA–PEG: (D) bright field image, (E) SEM picture of particles, and (F) confocal plane of these particles with the hydrophobic dye nile red. The bar indicates 15 ␮m.

M.R. Böhmer et al. / Colloids and Surfaces B: Biointerfaces 79 (2010) 47–52

3.2. Polymer mixtures PLGA with PDLLA–PEG To study differences in particle structure and size, a solution containing two polymers in a 1:1 weight ratio of PLGA and polyDL-lactic acid-co-poly-ethylene glycol, PDLLA–PEG, is jetted (at a total polymer concentration of 2% (w/w)). After shrinking by solvent removal 15 ␮m particles are formed, which is the same size as obtained for pure PLGA particles. The modal size is 14.8 ␮m and the CV 6% which is still very close to the pure PLGA particles, see Table 1. The initial emulsion drop size might have been slightly smaller due to a somewhat lower viscosity of the polymer mixture containing also the low molecular weight block-copolymer (replacing the high molecular weight PLGA) but this was below the detection accuracy of the analysis of the stroboscopic image. Fig. 5A shows that these particles appear internally inhomogeneous, contrary to the pure PLGA particles prepared from solution (Fig. 5A–C). In the overview (Fig. 5A) it appears that no particles with large pores are present; no particles with a white center as in Fig. 2D are present. The surface (Fig. 5B) is smooth with small (<1 ␮m) indentations, 2–5 per particle as observed from this direction. Confocal images demonstrate that the internal porosity of these particles (Fig. 5C) is much more pronounced than for pure PLGA. The PEG blocks of the PDLLA–PEG polymer tend to hydrate and therefore a limited amount of water will be taken up in the particle, leading to internal phase separation. Contrary to the pores of PLGA, the pores formed in the mixed polymer particle do not all appear spherical. In the slice shown two pores are larger than 2 ␮m and many smaller pores are present. If, instead of the solution, an emulsion is made by the addition of 2% water to the solution of the two polymers followed by ultrasonic mixing and ink-jetting into the second aqueous phase, the

51

internal structure, surface structure, and the size of the particles do not change a lot, see Fig. 5D–F. No significant swelling of the particle occurs; the particle size is still 14.4 ␮m (CV = 7.5%), see also Table 1. Closer inspection of the overview (Fig. 5D) shows that, contrary to the particle made from the solution of the two polymers, some particles with a large pore are present, as shown by the particle with a white interior, so variation in the internal structure does still occur. The internal structure of the slice shown for the double emulsion particle, Fig. 2F shows small, mostly smaller than 2 ␮m pores and just as in Fig. 5C not all pores appear spherical. The apparent closing of surface pores, witnessed by the indentations on the surface rather than pores and the presence of non-spherical internal pores indicates different mechanical properties of particles made with additional PEG–PDLA compared to pure PLGA particles, allowing for more morphological changes in the final hardening phase. With this low molecular weight PDLLA–PEG alone no monodisperse particles could be obtained by ink-jetting and subsequent solvent removal. Experiments with Dextran-FITC failed because no emulsion with sufficient stability could be made in the presence of PEG–PDLLA. In summary, addition of a PDLLA–PEG leads to particles that contain an increased number of small pores for single emulsions. For double emulsions the particles with PDLLA–PEG seem to have more small pores but may still contain large pores as well. The total pore volume of all four systems studied, PLGA, PLGA-double emulsion, PLGA + PDLLA–PEG and PLGA + PDLLA–PEG-double emulsion is small compared to the total amount of polymer. Only the starting concentration of polymer and the initial drop size dominate the particle size. The inclusion of high molecular weight hydrophilic compounds into such particles is based on trapping in closed pores that are formed during solidification.

Fig. 6. (A and B) Particles originating from initial droplets containing 2% 4A-PEG–PCL: (A) bright field image and (B) SEM picture. (C and D) Particles originating from 1% PLGA and 1% 4A-PEG–PCL solutions: (C) bright field image and (D) SEM picture. The bar indicates 15 ␮m.

52

M.R. Böhmer et al. / Colloids and Surfaces B: Biointerfaces 79 (2010) 47–52

4. Conclusions This study demonstrates that submerged ink-jet printing for the formation of monodisperse polymeric particles from monodisperse emulsions can be used to obtain detailed information on the porosity and degree of swelling of polymeric particles. Material combinations producing microscopic pores do not yield particles with a larger diameter as the majority of the water incorporated in the preparation flows out during the solidification process. Inflow of water into the particles occurs, leaving surface pores, which are largest for PLGA and can be reduced in size by the addition of PEG–PDLLA. Only for polymers that form hydrogels a significant degree of swelling is obtained. Fig. 7. Size distribution of PLGA, 4A-PEG–PCL and 1:1 mixed particles of these polymers as measured on a Coulter Multisizer III. All particles are made with the same initial droplet size of 55 ␮m.

3.3. Hydrogel particles Particles made from a four-arm poly(ethylene oxide-b-␧caprolactone) (4A-PEG–PCL) that are formed under same ink-jet conditions appear to have a sponge-like structure (Fig. 6) and a modal particle size of 22.6 ␮m, see Table 1. This is (in volume) a degree of swelling of 3.3 compared with the reference 15 ␮m solid polymer particles. The size distribution appears somewhat broader as can be deduced from the Coulter data as presented in Fig. 7 but, because the modal diameter is higher, the increase in the coefficient of variance to 8% is only moderate. Quantification of the degree of swelling has been performed on polymer films, giving swelling ratio’s depending on the amount of hydrophilic material and chain lengths [23]. For polymer particles, however, a dropby-drop method like ink-jet printing provides a straightforward means to quantify swelling of particles, which is the form in which these materials are often used. When 1:1 mixing 4A-PEG–PCL with PLGA, the particles show, contrary to the PLA–PDLLA–PEG mixtures, internal phase separation, as shown in Fig. 6C and D, with a solid spherical PLGA particle and a separated gel structure of the 4A-PEG–PCL. The degree of swelling of the total particle, derived from the size measurement (Fig. 7) using 15 ␮m as the reference, is 1.46. It is striking that for this polymer, even though it has a similar amount of poly-ethylene glycol as the amount of PEG in the PEG–PDLA molecules used, a significant swelling is found. The specific architecture of these molecules with the more hydrophilic core and surrounded by four more hydrophobic arms enforces the formation of very small crystalline poly-caprolacton nuclei. The PEG can stay hydrated and therefore create much more homogeneous swelling than two block-copolymers such as PDLLA–PEG. An environment suitable for the incorporation of high molecular weight hydrophilic molecules may therefore be better created using a hydrogel forming polymer.

References [1] Y.Y. Yang, T.S. Chung, N.P. Ng, Biomaterials 22 (2001) 231. [2] S. Freitas, H.P. Merkle, B. Gander, Journal of Controlled Release 102 (2005) 313. [3] S. Mao, J. Xu, C. Cai, O. Germershaus, A. Schaper, T. Kissel, International Journal of Pharmaceutics 334 (2007) 137. [4] S. Mao, Y. Shi, L. Li, J. Xu, A. Schaper, T. Kissel, European Journal of Pharmaceutics and Biopharmaceutics 68 (2008) 214. [5] R.C. Mundargi, V.R. Babu, V. Rangaswamy, P. Patel, T.M. Aminabhavi, Journal of Controlled Release 125 (2008) 193. [6] B. Bittner, C. Witt, K. Mader, T. Kissel, Journal of Controlled Release 60 (1999) 297. [7] K.F. Pistel, B. Bittner, H. Koll, G. Winter, T. Kissel, Journal of Controlled Release 59 (1999) 309. [8] J.H. Kim, Y.H. Bae, European Journal of Pharmaceutical Sciences 23 (2004) 245. [9] F. Nijsen, D. Rook, C. Brandt, R. Meijer, H. Dullens, B. Zonnenberg, J. de Klerk, P. van Rijk, W. Hennink, F.V. Schip, European Journal of Nuclear Medicine 28 (2001) 743. [10] L. Martin-Banderas, M. Flores-Mosquera, P. Riesco-Chueca, A. Rodriguez-Gil, A. Cebolla, S. Chavez, A.M. Ganan-Calvo, Small 1 (2005) 688. [11] S.Q. Xu, Z.H. Nie, M. Seo, P. Lewis, E. Kumacheva, H.A. Stone, P. Garstecki, D.B. Weibel, I. Gitlin, G.M. Whitesides, Angewandte Chemie-International Edition 44 (2005) 724. [12] W. Engl, R. Backov, P. Panizza, Current Opinion in Colloid & Interface Science 13 (2008) 206. [13] Q.B. Xu, M. Hashimoto, T.T. Dang, T. Hoare, D.S. Kohane, G.M. Whitesides, R. Langer, D.G. Anderson, Small 5 (2009) 1575. [14] T. Nisisako, T. Torii, Lab on a Chip 8 (2008) 287. [15] M.R. Böhmer, R. Schroeders, J.A.M. Steenbakkers, S. de Winter, P.A. Duineveld, J. Lub, W.P.M. Nijssen, J.A. Pikkemaat, H.R. Stapert, Colloids and Surfaces A: Physicochemical and Engineering Aspects 289 (2006) 96. [16] B. Schmidt, I. Souza, A.A. van Beek, M.R. Böhmer, Journal of Controlled Release 131 (2008) 19. [17] C. Berkland, K. Kim, D.W. Pack, Journal of Controlled Release 73 (1) (2001) 59–74. [18] C. Berkland, E. Pollauf, C. Raman, R. Silverman, K. Kim, D.W. Pack, Journal of Pharmaceutical Sciences 96 (2007) 1176. [19] S.W. Choi, I.W. Cheong, J.H. Kim, Y.N. Xia, Small 5 (2009) 454. [20] A.S. Utada, E. Lorenceau, D.R. Link, P.D. Kaplan, H.A. Stone, D.A. Weitz, Science 308 (2005) 537. [21] E. Tumarkin, E. Kumacheva, Chemical Society Reviews 38 (2009) 2161. [22] K.M. Shakesheff, C. Evora, I.I. Soriano, R. Langer, Journal of Colloid and Interface Science 185 (1997) 538. [23] R. van Dijkhuizen-Radersma, J.R. Roosma, P. Kaim, S. Metairie, F. Peters, J. de Wijn, P.G. Zijlstra, K. de Groot, J.M. Bezemer, Journal of Biomedical Materials Research Part A 67A (2003) 1294.