Morphological Comparison of PVA Scaffolds Obtained by Gas

Dec 6, 2012 - Microwave-crosslinked Bio-based Starch/Clay Aerogels. Liang Wang , Miguel Sánchez-Soto , Tobias Abt , Maria Lluisa Maspoch , Orlando O...
2 downloads 0 Views 1MB Size
Download PDF
Article pubs.acs.org/Langmuir

Morphological Comparison of PVA Scaffolds Obtained by Gas Foaming and Microfluidic Foaming Techniques Cristina Colosi,§,† Marco Costantini,§,† Andrea Barbetta,*,† Raffaella Pecci,‡ Rossella Bedini,‡ and Mariella Dentini*,† †

Department of Chemistry, University of Rome “La Sapienza”, P. le A. Moro 5, 00185 Rome, Italy Department of Technology and Health, Istituto Superiore di Sanità, Viale Regina Elena 299, I-00161 Rome, Italy

‡

S Supporting Information *

ABSTRACT: In this article, we have exploited a microfluidic foaming technique for the generation of highly monodisperse gas-in-liquid bubbles as a templating system for scaffolds characterized by an ordered and homogeneous porous texture. An aqueous poly(vinyl alcohol) (PVA) solution (containing a surfactant) and a gas (argon) are injected simultaneously at constant flow rates in a flow-focusing device (FFD), in which the gas thread breaks up to form monodisperse bubbles. Immediately after its formation, the foam is collected and frozen in liquid nitrogen, freeze−dried, and cross-linked with glutaraldehyde. In order to highlight the superior morphological quality of the obtained porous material, a comparison between this scaffold and another one, also constituted of PVA but obtained with a traditional gas foaming technique, was carried out. Such a comparison has been conducted by analyzing electron microscopy and X-ray microtomographic images of the two samples. It turned out that the microfluidic produced scaffold was characterized by much more uniform porous texture than the gas-foaming one as witnessed by narrower pore size, interconnection, and wall thickness distributions. On the other side, scarce pore interconnectivity, relatively low pore volume, and limited production rate represent, by now, the principal disadvantages of microfluidic foaming as scaffold fabrication method, emphasizing the kind of improvement that this technique needs to undergo.

1. INTRODUCTION Tissue engineering is an innovative and multidisciplinary research field whose main purpose is to restore lost tissue functionality. A prominent position in most tissue engineering approaches is held by the scaffold, i.e., a temporary support that aids seeded cells to proliferate and reorganize into tissue-like structures. Many different scaffold fabrication techniques have been developed, each with its pro and con.1 In the course of such research efforts, it has become more and more evident that, as the understanding of the influence of both the scaffold morphology and its mechanical properties on cellular culture outcome progress on, the requirements that a scaffold has to satisfy become more and more severe. As a result, tables listing the optimal range of pores and interconnect dimensions for each cellular type are nowadays available in the literature.2,3 The same holds with regard to scaffold mechanical properties.4 From these premises, it derives the necessity to have a scaffold fabrication method that allows us to exert precise and reproducible control over the morphological characteristics according to a priori design. Unfortunately, most of the scaffold fabrication techniques developed so far (e.g., particulate leaching, freeze−drying, gas foaming, phase separation, etc.) present serious limitations in this respect; in other words, they © 2012 American Chemical Society

do not permit an accurate correlation between processing parameters and desired properties. Furthermore, scaffolds obtained by “conventional techniques” have pores with a wide distribution in sizes, hence making it difficult to carry out systematic studies on the architectural influence of the differences in signaling, gene expression, and organization. To elucidate the effect on cell-to-cell and on cell-to-matrix interactions due to the structure, it is desirable to have highly ordered and uniform structures. For these reasons, a clear trend is taking shape in recent years. Techniques that allow us to exert accurate control over the geometrical and morphological features are becoming preferred over traditional ones. At present, the most advanced scaffold fabrication technology is represented by rapid prototyping. Rapid prototyping is a group of techniques used to fabricate three-dimensional scaffolds from a digital model.5−7 Scaffolds produced with these techniques generally show a very customizable, highly ordered morphology, but each method has limitations concerning the choice of the building materials. In other words, they present limitations regarding the Received: September 20, 2012 Revised: December 4, 2012 Published: December 6, 2012 82

dx.doi.org/10.1021/la303788z | Langmuir 2013, 29, 82−91

Langmuir

Article

applications such as high hydrophilicity, good film forming, and processability.38 Poly(vinyl alcohol) is an ideal component for the preparation of emulsions and foams. The hydrophilic/lipophilic balance between hydrocarbonic (−CH2) and oxidrilic groups confers to this polymer stabilizing properties as witnessed by its wide use in processes such as suspension polymerization and emulsion polymerization and as a coadjutant in the stabilization of emulsions and foams.39,40 Another attractive property of PVA is its high chain flexibility, a feature that translates into the possibility of preparing highly concentrated solutions (5−15% w/v) even when using a polymer characterized by an average molecular weight of 70−100 kg/mol. PVA is already used in areas such as food chemistry, pharmaceuticals, medicine, and biotechnology. In particular, PVA is a polymer of great interest for various pharmaceutical and biomedical applications.41,42 Furthermore, PVA hydrogels can be considered as biocompatible in nature and are nonirritating to soft tissues when in contact with them, making them suitable for many biomedical applications. In order to give a preliminary evaluation of the potentialities of using a microfluidic system to create a porous template, we present a morphological comparison between two different PVA scaffolds: one was produced with the gas foaming technique, and the other with the microfluidic foaming technique. Both samples had the same chemical composition.

possibility of using biologically valuable biopolymers such as hyaluronic acid, alginate, chitosan, and so forth. In this respect, a technique that has the potential to be a complementary alternative to RP may be represented by microfluidics. So far, this branch of science and technology has been mainly confined to lab-on-chip applications.8−10 The main attractive feature of this technology is the possibility to handle very small amounts of fluids in a very reproducible way. In the recent past, we produced scaffolds using mainly two techniques: polyHIPEs and gas foaming. In the first one, scaffold production takes place via a concentrated oil-in-water HIPE (high internal phase emulsion) in which the continuous phase consists of an aqueous solution of a biopolymer. Crosslinking of the continuous phase and removal of the internal oil phase leave behind a porous material characterized by a highly porous and trabecular morphology.11−16 This technique has been overcome by the gas-in-liquid foaming templating technique because it enjoys intrinsic advantages: it does not need the use of any organic solvents (harmful for cultured cells), and the final morphology best fits the requirements of TE scaffolds.17−22 In such a method, the oil phase is replaced by a hydrophobic gaseous phase (such as Ar, N2, hexafluoro ethane), which is dispersed within the polymeric aqueous solution inside a stirred reactor. The final porous material is characterized by a wide distribution of pore size with a large standard deviation. This translates into an anisotropic environment for seeded cells, which negatively influences cell migration, and generally leads to an incomplete and nonhomogeneous colonization of the scaffold. A natural evolution of the expertise developed so far could be represented by the combination of the traditional gas-in-liquid foaming technique and microfluidics: emulsions and foams characterized by monodisperse droplet/bubble sizes can be readily prepared within the channel of a microfluidic device,23 and these systems could be used to generate porous material with a highly constant pore dimension.24,25 Uniformity allows individual cells grown within the scaffold to experience similar opportunities to occupy space, and they are more likely to behave similarly in the population as a whole rather than differentially react to their alternative surroundings, which could lead to greater cellular inconsistency within a 3D system. Moreover, scaffolds are often modified with bioactive molecules such as growth factors, drug, or adhesion peptides. In a more uniform spatial structure, the levels of chemical stimuli experienced by cells are more likely to be the same. Also, from the mechanical properties and degradation rate point of view, a uniform architecture guarantees constant performance throughout the scaffold. In the literature, two different microfluidic geometries are mainly used and described to create biphasic systems with a monodisperse discrete phase: T-junction26−29 and flow focusing.30−37 In both geometries, the two immiscible phases flow in separated microchannels and meet in a tight junction in which the liquid phase squeezes the gas thread leading to bubble formation. The technical solution that we embraced consists of using a flow focusing microfluidic device (FFD) for generating argon bubbles in a polymeric solution to produce a foam template of porous scaffold characterized by a uniform porous texture. Among organic polymers, poly(vinyl alcohol) (PVA) is one of the very few vinyl polymers soluble in water that has been studied intensively because of its attractive features for medical

2. MATERIALS AND METHODS 2.1. Materials. The polymer used in foam production was poly(vinyl alcohol) (PVA), with an average molecular weight (Mw) of 70−100 kDa and a degree of hydrolysis of 100%. The viscosity of a PVA and cetyl trimethyl ammonium bromide solution (15% and 1% w/v, respectively) at room temperature is 1.57 Pas. Cetyl trimethyl ammonium bromide (CTABr) was added to the solution as the main surface-active agent. During foam formation, argon (Rivoira) was used as the gaseous phase. The produced foams were cross-linked with glutharaldehyde (GLA, 25% v/v aqueous solution). All chemicals were supplied by Sigma-Aldrich and used without further purification. 2.2. Foam Production. An aqueous solution of PVA 15% w/v and CTABr 1% w/v was used as the continuous phase in foam production by both microfluidic and gas foaming methods. The PVA solution was prepared by vigorously stirring with a magnetic stir bar at 80 °C until complete dissolution, then cooled to room temperature. CTABr was then added. 2.2.1. Microfluidic Foaming. In the microfluidic foaming technique, foam is produced in a flow focusing device (FFD). We fabricated the FFD out of polydimethylsiloxane (PDMS) by conventional soft lithography procedures.43−45 The microchannel geometry is shown in Figure 1. In order to control bubble formation, the parameters that are usually changed during experiments are the applied gas pressure (pg) and the liquid flow rate (ql). Increasing pg leads to a corresponding increment in bubble diameter, while an increase in ql implies a decrease in bubble diameter; production frequency grows with the increase of both parameters.37 Repeatability of the process lies in the fine-tuning of these two variables; for these reasons, the liquid phase was injected with a microfluidic pump (Pump 11 Elite, Harvard Apparatus) and the gas flow was regulated with a pressure regulator (Porter, model 8286) to obtain high control of the system. The experimental procedure consists of regulating pg and ql until obtaining a close-packed 2D bubble pattern in the outlet channel. The dispersed phase fraction Φg and the liquid fraction Φl must satisfy the relationship Φg > Φl in order to obtain a final scaffold characterized by high porosity and to facilitate interconnection formation between 83

dx.doi.org/10.1021/la303788z | Langmuir 2013, 29, 82−91

Langmuir

Article

those pixels with a value higher than threshold and into white those with a lower value. The choice of the thresholding partially influences calculation, and a balance between noise reduction and structure’s pixel inclusion must be carefully evaluated. Among the available implemented threshold methods of ImageJ software (section 2.5.1), we used the entropy-based Yen algorithm as it produced the best results for our samples. As mentioned, two different scaffolds were analyzed: one was produced with the gas foaming technique, the other with the microfluidic foaming technique. 2.5.1. ImageJ Software. ImageJ is an open-source software for image analysis developed by Rasband W. S. at National Institute of Health (NIH), Bethesda, Maryland, USA.46 Since its inception, this software has been upgraded by users to become the most used opensource software for image analysis for biomedical applications. Numerous Java plugins are available on the Web. The software can process 2D images and can handle a sequence of images that are spatially or temporally related, collecting them in a stack. For the analysis of our scaffolds, we used the 3D Viewer,47 Local Thickness,48 Color Inspector 3D,49 and 3D Object Counter50 plugins, in addition to the basic functions of the software. 2.5.2. Porosity. To estimate the percentage porosity (P%) of the scaffolds, image stacks were binarized and inverted (so as to show as black the empty volume), and P% was estimated as the ratio of thresholded voxels to total voxels of a specified region of interest (ROI of about 1 mm in height for a number of approximately 150 slices). To better characterize the porosity of the materials, the specific surface area (SSA) was also calculated as the ratio of total surface area to total volume of the sample. 2.5.3. Pore Size Distribution (PSD) Analysis. The pore size distribution (PSD) was estimated from μ-CT images with a “destroy and rebuild” method. This one uses the particle analyzer (PA) basic function of ImageJ, that analyzes horizontal projections of each pore inside the matrix, and an algorithm, written in Visual Basic (Microsoft), which takes the coordinates (Xc; Yc) of the mass centers and the areas of the pore sections obtained with PA to reconstruct a good approximation of the whole porous matrix of the scaffold. In our analysis, we made the assumption that scaffold pores preserve a spherical shape. It can therefore be assumed that horizontal pore sections are circles and their mass centers coincide with the centers of the circles that best fit them. The first step of PSD analysis consists of inverting and thresholding (making binary) a scaffold ROI of μ-CT slices (∼1 mm in height) in order to bring into evidence the empty volume of the matrix. Different voids inside this volume are then separated using a watershed algorithm.51 The PA function is then run, and it returns data of the coordinates of mass centers (Xc; Yc) and the numerical value of areas of horizontal pore sections for each slice. During this procedure, the correspondence between sections and three-dimensional pores is lost. In the second step, our algorithm takes the (Xc; Yc)n (n is the slice number) coordinates of these sections and tries to reconstruct the global pore (Figure 2), grouping horizontal sections which have their coordinates between |(X′c; Y′c)n+1 − (Xc; Yc)n| ≤ δ (δ is an empirical parameter and its value is 3.5% of the mean pore diameter ⟨D⟩). Sections belonging to an already reconstructed pore are not considered anymore. At the end of the second step, a first approximation of the pore structure is obtained. This approximation is postprocessed with a filter function to obtain a better valuation of the PSD. This filter removes from final analysis those pores situated on ROI edges. This is done because these pores are generally broken pores affected by some irregularity arising from the image analysis method. Finally, an equivalent volume and radius are calculated for each pore to construct the PSD. 2.5.4. Wall Thickness Distribution. Wall thickness (WT) was analyzed with the use of the Local Thickness plugin and Color Inspector 3D plugin of ImageJ applied on μCT images of the two samples. The Local Thickness plugin works on image stacks and computes a distance map by the Euclidian distance transformation using pore centers as reference points giving back a colored map of local wall thickness. A quantitative measure of thickness was

Figure 1. Light micrograph of the PDMS FF device: (a) gas inlet, 300 μm; (b) liquid inlet, 300 μm; (c) orifice, 100 μm;(d) outlet channel, 700 μm. The microfluidic device was about 150 μm in depth. adjacent pores. In PVA foam production, the system reached stability at ql = 2.25 μL/min and pg = 67 kPa. When an acceptable amount of foam is produced, it is frozen in liquid nitrogen, to prevent the further progress of instability phenomena, and is then lyophilized. This is followed by the crosslinking step in which the scaffold structure is irreversibly locked-in. 2.2.2. Gas Foaming. The gas foaming method was developed in our laboratory in the past years and successful results were obtained with many different biopolymers.17−21 This technique consists of stirring an aqueous solution of a biopolymer thermostated glass reactor that has at its bottom a glass porous septum through which an inert gas (generally Ar) is insufflated during mechanical stirring. The gas flow is regulated with an automatic syringe pump while stirring is provided by an overhead stirrer (IKA). The produced foam is frozen in liquid nitrogen immediately after its production to avoid foam’s decay phenomena, and then it is freeze−dried. The lyophilized solid foam is then soaked into a solution containing the cross-linking agent. 2.3. Cross-Linking Reaction. Both lyophilized PVA scaffolds, produced by the two techniques, were cross-linked with a 25% v/v GLA aqueous solution in acidified acetone (acetone/HCl, 74.88/0.12 v/v). Hydrochloric acid is the catalyst of the condensation reaction of the aldehydic groups of GLA with the alcoholic groups of PVA. Scaffolds were left for 4 h at 80 °C under gentle stirring. They were then dialyzed in deionized water and finally lyophilized again for characterization analyses. 2.4. Scanning Electron Microscopy (SEM). The structures of the scaffolds were investigated using scanning electron microscopy (SEM) (LEO 1450VP) operating at 20 kV. Prior to observation, fractured samples were mounted on aluminum stubs using adhesive carbon disks to increase the conductivity. All samples were sputtered with a thin layer of gold (∼10 nm) in argon atmosphere using a SEM coating unit 953, Agar Scientific, to ensure conductivity. 2.5. Image Analysis of the μCT Data. Microcomputed tomography (μCT) is a nondestructive technique that generates cross-sectional images of a sample using an X-ray source. Using these 2D representations, it is possible to generate a 3D reconstruction of the sample and to perform analysis on the material structure. Acquisition was performed using SkyScan 1072 (SKYSCAN, Katruizersweg 3B 2550 Kontich, Belgium) with the following parameters: 40 kV voltage; 248 mA current, no filter material, 0.45° rotation step, and 2 frames as frame averaging. Voxel size was the same for both samples (7 × 7 × 7 μm3). In order to increase the contrast of the obtained images, our scaffolds (two cylinders 5 mm in height and 7 mm in diameter) were dipped into a solution containing CuSO4 for about 15 min. The samples were then washed and lyophilized before being exposed to X-rays. Image analysis of data obtained from μCT was carried out with the use of ImageJ software. Z-projections of the samples are grouped in a stack of images and analyzed to calculate porosity, pore size distribution (PSD), wall thickness (WT), and interconnectivity. The first and most sensible step in image analysis is the binarization of the gray-scale images obtained from μCT data. This is made by choosing a threshold value, and consequently converting into black 84

dx.doi.org/10.1021/la303788z | Langmuir 2013, 29, 82−91

Langmuir

Article

Figure 4. Example of interconnection analysis: (a) binary image of the slice; (b) inverted binary image; (c) separation of pores with watershed algorithm; (d) image difference between (c) and (b). (e) Representation of three-dimensional reconstruction of interconnections performed by Object Counter 3D. stack in the ZY and ZX planes was performed, and the described calculation steps were repeated.

3. RESULTS The mechanistic description of the formation of monodisperse bubbles inside a microfluidic device has been described in detail previously.23 Briefly, within the microchannels of the FFD, the polymeric solution, containing a proper surfactant, flows in two lateral inlet channels and meets the gaseous phase (Ar) in a perpendicular junction. At the beginning of the outlet channel, the two immiscible flows are forced through a bottleneck: here, the liquid stream periodically squeezes the gaseous thread causing the pinch-off phenomenon by which quasi-monodisperse bubbles are formed.31−37 Bubbles produced at the junction travel inside the outlet channel and flow outside the device through a tiny tube; at the end of this tube, the foam is collected inside a Teflon mold. Figure S1 (Supporting Information) shows a micrograph of the liquid foam spontaneously self-assembled into crystal-like structures. The collecting step is crucial with regard to the final solid foam morphology. In fact, the longer the collecting time, the more pronounced the effects of the kinetics instabilities that arise in liquid foams will be, such as drainage of the liquid phase,52,53 coalescence,54,55 and Ostwald ripening.56 These

Figure 2. Graphical representation of the PSD calculation method: (Xc; Yc) coordinates representing the centers of pore sections in adjacent slices are used to reconstruct the three-dimensional structure of the pore and to calculate the corresponding volume. extrapolated using the Color Inspector 3D (CI3D) plugin on the colored distance map (Figure 3). CI3D calculates a color-histogram of the image. By creating a calibrated correspondence between a color and a thickness value, a distribution of WT was obtained. 2.5.5. Interconnectivity. To allow interconnection measurement, a binary image of μCT slices was used (Figure 4a): the stack was inverted (Figure 4b) and voids were separated using a watershed algorithm (Figure 4c). To obtain interconnections, the image containing the closed pores (Figure 4c) was subtracted to the inverted one obtaining, for each slice, a set of lines corresponding to the horizontal projections of interconnections (Figure 4d). In order to reconstruct the three-dimensional structure of interconnections, the Object Counter 3D (OC3D) plugin was run; this function joins interconnection projections in adjacent slices and reconstructs the corresponding surface (Figure 4e). OC3D hence provides the areas of interconnections, from which the distribution of interconnection diameters is calculated. To take into account those interconnections that are not visible in the X−Y plane (horizontal plane), a reslice of the

Figure 3. Colored distance maps of wall thickness with calibration bars. (a) Microfluidic Foaming PVA scaffold. (b) Gas Foaming PVA scaffold. 85

dx.doi.org/10.1021/la303788z | Langmuir 2013, 29, 82−91

Langmuir

Article

3.1. Porosity. Porosity is one of the most important parameters characterizing a scaffold: empty space inside the matrix will serve as cellular path during migration and colonization, and it will permit the flow of nutrients and fluids in and out of the scaffold. The specific surface area (SSA), that is the surface area to volume ratio, also influences cell behavior: in general, a large SSA would assist cellular adhesion to the scaffold. The final porosity of a foam-produced scaffold (created either by gas foaming or microfluidic foaming) is determined to a large extent by the gas-in-liquid ratio of the precursor foam. The gas foaming scaffold is characterized by a larger P%, equal to 78%, compared to the microfluidic one, that presents a 63% of empty volume. On the contrary, the gas foaming scaffold exhibits a lower SSA (0.020 against 0.074 μm−1). This suggests that pores inside the gas foaming scaffolds are bigger on average than pores of the microfluidic scaffold. In fact, taking a sphere as an approximation of a pore, a decrease in its diameter (dsphere) leads to a faster decrease of its volume (Vsphere ∝ d3sphere) than of its surface area (SAsphere ∝ d2sphere), making the SSA of a scaffold with bigger pores smaller than the one of a scaffold with smaller pores. 3.2. PSD − Pore Size Distribution. The PSD of the two samples are shown in Figure 6, while in Table 1, the statistical parameters that characterize pore geometry are summarized.

phenomena adversely affect monodispersion of the collected foam. The qualitative inspection of the SEM and 2D μCT micrographs (Figure 5) reveals immediately the remarkable

Figure 5. μCT slices of the (a) PVA scaffold produced with the microfluidic foaming technique and (b) PVA scaffold produced with the gas foaming technique. Scanning electron micrograph of the microfluidic foaming (c) and gas foaming (d) PVA scaffolds. 3D rendering of the microfluidic foaming (e) and gas foaming (f) PVA scaffolds.

Figure 6. Overlapped normalized histograms of pore diameter of the two PVA scaffolds: the green one represents the microfluidic foaming PSD, the red one represents the gas foaming PSD.

homogeneity in pore dimension of the microfluidic scaffold (Figure 5a,c) in comparison to the gas-foaming one (Figure 5b,d). By stacking a series of adjacent μCT slices, a 3D rendering of both the foam and microfluidic scaffolds can be obtained (Figure 5e,f). It must be noted that the 2D representation offered by μCT and the observation of the fracture surface of the SEM micrograph do not permit appreciation of the true morphological features (pores and interconnects) of both the foam-based scaffolds. If we imagine drawing a plane through the scaffold, its intersection with different pores will take place at various heights showing an apparent pore diameter that does not always coincide with the real one (Figure S2, Supporting Information). As a consequence, it is necessary to develop a data treatment (especially for μCT data) that allows us to determine the diameters of the equivalent spheres to which pores can be assimilated. The same considerations hold for interconnects. The aim of the software analysis described in the Methods and Materials (section 2.5) is to obtain a rigorous representation of the true morphological characteristics of the scaffold produced in this work. In the following, we report on the results based on such an analysis.

The microfluidic foaming histogram shows a considerably narrower distribution of pore dimension compared to the gas foaming one. Mean diameter size of the microfluidic-produced scaffold is 140 μm with a σ % of 12. In Figure 6, it is clearly visible that the microfluidic PSD has a bimodal trend, composed of two distinct peaks: the first one is centered at 130 μm, while the second is centered at 200 μm. The existence of this second peak is probably caused by the occurrence to some extent of coalescence phenomena correlated to the long production times that this technique currently requires, favoring to some extent the decay of the foam. The mean pore diameter of the second peak is about 1.5× bigger than the mean pore diameter of the first one; this is close enough to the theoretical 21/3 increment caused by coalescence of two equally sized bubbles. The presence of the second peak negatively influences the statistical parameters of the whole population (in terms of ⟨D⟩ and σ), but it could be removed, or at least reduced, with a faster production rate. 86

dx.doi.org/10.1021/la303788z | Langmuir 2013, 29, 82−91

Langmuir

Article

connectivity, and porosity of the scaffold, and it is therefore an important parameter that deserves evaluation. The two wall thickness distributions are shown in Figure 7. They appear to be comparable in both mean wall size (⟨W⟩) (Table 1) and standard deviation. Even if the microfluidic scaffolds present a lower percentage porosity (P%), the ⟨W⟩ is, on the contrary, slightly smaller with respect to the gas foaming sample. This is probably explainable in terms of higher specific surface area (SSA) exhibited by the microfluidic scaffold. 3.4. Interconnect Size Distribution. Another essential parameter in the characterization of a scaffold is its degree of interconnectivity: it is concerned with the number and dimensions of the windows that connect adjacent pores. This parameter influences cellular response during seeding and culturing. The dimension of interconnects must be sufficiently large to permit cell colonization and migration through the entire scaffold. Using μCT images, evaluation of interconnection mean diameter (⟨d⟩) and of percentage interconnectivity (I%) was conducted. The mean interconnect diameter of the microfluidic-foamed scaffold is about 30 μm, while for the gas-foamed scaffold, it is about 50 μm (Table 1); the two distributions (Figure 8) have the same trend with a tail in the high-diameter sizes. Also in this respect, the width of the distribution referring to the microfluidic scaffold is significantly narrower than the gasfoamed one (Table 1). Both samples present a 100% degree of interconnectivity, although the microfluidic scaffold presents a lower mean number of interconnections per pore (⟨ni⟩): this means that, even if pores within the matrix are totally interconnected, a more tortuous path is needed to join whatever couple of pores inside the scaffold. The lower dimensions of interconnections of the microfluidic foaming sample are strictly related to its lower P% in comparison to the gas foaming one. Large interconnections are, in fact, the result of the thinning of the liquid film surrounding adjacent bubbles in the precursor liquid foam. A higher P% favors the distribution of the liquid phase over a larger volume with the consequence that bubbles deform into polyhedra. As a consequence, the area of maximum approach among adjacent bubbles will be represented by planes. These

Table 1. Number of Pores Analyzed (n); mean pore diameter size (⟨D⟩); standard deviation (σ); mean wall thickness (⟨W⟩); mean interconnection size (⟨d⟩); percentage of interconnectivity (I%); mean number of interconnections per pore (⟨ni⟩) PSa

WTb

INTc

a

n ⟨D⟩ σ σ% ⟨W⟩ σ σ% ⟨d⟩ σ σ% I% ⟨ni⟩

gas foaming

microfluidic foaming

822 210 μm 95 μm 45 60 μm 28 μm 47 50 μm 25 μm 50 100 8

1406 140 μm 17 μm 12 55 μm 22 μm 40 30 μm 12 μm 40 100 4

Pore size. bWall thickness. cInterconnection size.

From these considerations, we can assert that bubbles inside the microfluidic channels were formed with a diameter of approximately 130 μm (corresponding to the first peak of the PSD), reflecting the dimension of the orifice in the tight junction of the flow focusing device (Figure 1). The gas foaming histogram shows a typical PSD where pore diameters are spread over a large range; mean pore size is about 210 μm with a standard deviation of 45%. As can be seen in Figure 6, the gas foaming PSD has a tail in the higher-diameter values: on one side, this may be caused by some degree of coalescence among bubbles; on the other side, it represents an intrinsic limit of such a production system. In fact, as the volume fraction of the internal phase increases in the course of gas insufflation, the foam experiences a corresponding increase of the overall viscosity. This implies that the process of bubble fragmentation becomes less and less efficient. The final outcome is a rather broad polydispersity in pore size. 3.3. Wall Thickness Distribution. Wall thickness influences mechanical proprieties,57 degradation rate, inter-

Figure 7. Normalized histograms of wall thickness of the two PVA scaffolds. 87

dx.doi.org/10.1021/la303788z | Langmuir 2013, 29, 82−91

Langmuir

Article

Figure 8. Normalized histograms of interconnection diameters of the two PVA scaffolds.

after lyophilization and cross-linking steps will lead to larger interconnections. All μCT data are reported in Table 1. Such data deserve a comment with regard to their level of accuracy. As mentioned in the Image Analysis of μCT Data, the voxel size is about 7 × 7 × 7 μm3. This consequently represents an intrinsic uncertainty value in the mathematical elaboration of tomographic data. Anyway, diameter of pores and interconnections are respectively calculated from volume and area reconstruction, with corresponding spheres/circles containing more than 10 voxels/pixels. In this case, the approximation of a discrete sphere/circle built with square pixels drifts from a real sphere/ circle volume/surface of less than 1%, giving an uncertainty of calculation smaller than the voxel size, and is therefore a useless value. Furthermore, the important value derived from μCT is the shape of distributions for the calculated geometrical features, more than the numerical values (which are acceptable even with a 7 μm uncertainty) to demonstrate the increased homogeneity for the microfluidic foaming scaffold with respect to the gas foamed one. For the above reasons, results in Table 1 are approximated to the nearest integer value.

morphology. The possibility of producing quasi-monodisperse scaffolds represents a great challenge in the regenerative medicine field because it guarantees homogeneous characteristics through space and a uniform environment for seeded cells. For instance, the rate of scaffold degradation should be comparable in each region of the scaffold, and seeded cells have equal opportunity to colonize the matrix. To highlight the superiority of this method over traditional gas-in-liquid foam templating, a detailed comparison of the morphological properties of microfluidic and gas foaming scaffold was carried out. All the analysis used to morphologically characterize the scaffolds was performed through the elaboration of μCT images; computerized methods for the calculation of PSD, WTD, and interconnect size distribution were illustrated in corresponding paragraphs. The microfluidic-obtained scaffold presents a lower percentage porosity in comparison with the gas foaming one; this is primarily caused by the packaging of the microfluidic device that does not guarantee a hermetic seal at high values of gas pressure, this being the origin of our relatively low gas-in-liquid fraction. P% of the scaffold produced by microfluidic foaming actually also suffers from an intrinsic limit: the almost constant dimensions and spherical shape of bubbles inside the FFD (in our geometry, the height of the channels is comparable to the diameter of the bubbles produced) lead to a spatial arrangement that is comparable to the packing of equal spheres. It is known from literature that an upper limit exists for the spatial density of this kind of packing,64 foretold to be 74% in the most ordered case (the so-called Kepler conjecture) and 64% for a random packing.59 An increase in the gas-in-liquid ratio during microfluidic foaming therefore needs the generation of some kind of deformation of the bubble shape inside the device to allow a more efficient packing during foam production. With regard to the PSD, the microfluidic foaming technique leads to a very narrow distribution of pore diameters, just as expected. The results could be improved still with a faster rate of foam production. In the described setup, the foam production rate inside the microfluidic device is on the order of few milliliters per hour; to overcome this limitation, and prevent the destabilization phenomena that lead to partial coalescence of bubbles and the establishment of the bimodal distribution in the derived material, future parallelization of identical microfluidic channels is needed to scale up the production.

4. DISCUSSION Recently, several methods have been developed to fabricate 3D ordered scaffolds by solid free-form fabrication techniques,58 such as photolithographic patterning and layering,59 direct writing,60 and two-photon stereolithography.61 Most methods involve expensive robotic control and time-consuming pixel-bypixel writing. Kotov et al.62,63 adopted the self-assembly approach by templating colloidal crystals. Colloidal spheres were organized into crystal-like arrangements by slow evaporation of the solvent. The interstices were then infiltrated with scaffold materials. Finally, the colloidal spheres were removed by organic solvent or calcination. Scaffolds of inverted crystal structures are made. However, the removal step by calcination limits the scaffold materials, and the method that uses organic solvent is slow and may be harmful for cells because of the eventual presence of organic residues. In this article, we have exploited the potential of microfluidics in generating a packed array of monodisperse gas-in-water bubbles as a template of a uniform porous material based on PVA. The goal of the microfluidic foaming technique is to permit the production of scaffolds with a homogeneous 88

dx.doi.org/10.1021/la303788z | Langmuir 2013, 29, 82−91

Langmuir

Article

The two scaffolds were also compared in terms of WTD, but they did not show considerable differences. Another aspect of the microfluidic-obtained scaffold that needs to be enhanced is its interconnectivity: the probability of opening a window connecting two adjacent pores is influenced by the volume fraction (Φg) of the gaseous phase in the precursor foam. In our experience, we have found that foams with a gaseous fraction superior to ∼80% tend to show a high degree of interconnection, due to the deformation of bubbles into polyhedrals. A future target of our research will therefore be focused on increasing Φg during microfluidic foaming process, acting on experimental setup and channel geometry. In spite of the present limitations outlined above, this new scaffold fabrication method enjoys several advantages over free forming and three-dimensional printing, or lithography. First, there is no sophisticated equipment requirement compared to these high-technology fabrication techniques. Second, this new scaffolding has a homogeneous foam skeleton, which is not easily achievable with free forming, three-dimensional printing, or lithography because of their “pixel assembly” nature. Third, the scaffold can be directly fabricated into a specific anatomical shape with a mold. Fourth, the process can be easily expanded or automated for large-scale production. In conclusion, the new processing technique can tailor polymer foams for a variety of potential tissue engineering and other biomedical applications because of the wall-controlled nature in architecture, interpore connectivity, and physical and mechanical properties.

On the material morphological point of view, we have obtained interesting results in this article. A scaffold with 63% porosity by volume that consisted of 140-μm-diameter spherical pores and 30-μm-diameter interconnects was obtained. Additionally, this method is compatible with many biologically valuable biopolymers. Monodispersity of pores and interconnects has been demonstrated, but the degree of interconnectivity was not satisfactory for tissue engineering applications. This issue needs further research, which is currently ongoing. The production of a quasi-crystalline matrix with constant pore dimensions is the final goal of the microfluidic foaming technique. The application implications of this technology in the field of tissue engineering could be relevant and still need to be explored.

■

ASSOCIATED CONTENT

S Supporting Information *

Light micrograph of the wet PVA microfluidic foam selfassembled into a crystalline order and representation of the apparent diameter of pores shown in μCT slices. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*Andrea Barbetta: e-mail [email protected]. Mariella Dentini: e-mail [email protected]. Author Contributions §

These authors have equally contributed to the article.

5. CONCLUSION The use of microfluidics for scaffold synthesis is a very new topic of research. So far, attempts made in this direction are extremely few and the results achieved are not always reliable. This article aims at exploring the potentials and shortcomings of microfluidics in scaffold manufacturing. These are put into evidence when a comparison with a “conventional” and at the same time related technique is made, in this case, gas foaming. We have focused on scaffold production using a single microfluidic channel. This is not a trivial pursuit when a viscous solution (PVA 15% w/v) is used inside the microchannels of a microfluidic chip. While the physics of foam formation from a surfactant aqueous solution and a gas inside a microfluidic device have been the object of intensive research, the same does not hold for viscoelastic aqueous solutions. The set of data we have collected and conclusions reached must not be considered as a point of arrival but as a starting one. For instance, the material used to make the microfluidic chip, i.e., polydimethylsiloxane, was revealed to be not particularly suitable since it prevented us from running the liquid phase at high flow rates, thus limiting the rate of production. In this respect, we have recently fabricated a microfluidic chip out of poly(methylmethacrylate) which is a much more rigid material. The chip made of such material did not suffer any leakage at high flow rates. From our experiments, we could make an estimation of the rate of foam production, which turned out to be quite limited (on the order of few mL/h). Conversely, “conventional” gas foaming allows the production of a potentially limitless amount of scaffold within a single preparation. This induced us, recently, to design a new microfluidic layout which is composed of many microfluidic chips made of poly(methylmethacrylate) and operating in parallel. This allowed an increase by at least 1 order of magnitude of the rate of foam production.

Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS Authors thank MIUR (PRIN 2008) and Sapienza Ateneo for financial support. Authors especially thank Università Campus Bio-Medico di Roma for clean room facilities and Prof. Dino Accoto, Dr. Maria Teresa Francomano, Dr. Caterina Esposito, and Mr. Vittorio Ceccarelli for precious advice received.

■

REFERENCES

(1) Zorlutuna, P.; Annabi, N.; Camci-Unal, G.; Nikkhad, M.; Cha, J. M.; Nichol, J. W.; Manbachi, A.; Bae, H.; Chen, S.; Khademhosseini, A. Microfabricated Biomaterials for Engineering 3D Tissues. Adv. Mater. 2012, 24, 1782−1804. (2) Mandal, B. B.; Kundu, S. C. Cell Proliferation and Migration in Silk Fibroin 3D Scaffolds. Biomaterials 2009, 30, 2956−2965. (3) Lien, S. M.; Ko, L. Y.; Huang, T. J. Effect of Pore Size on ECM Secretion and Cell Growth in Gelatin Scaffold for Articular Cartilage Tissue Engineering. Acta Biomater. 2009, 5, 670−679. (4) McKee, C. T.; Last, J. A.; Russell, P.; Murphy, C. J. Indentation Versus Tensile Measurements of Young’s Modulus for Soft Biological Tissues. Tissue Eng., Part B 2011, 17, 155−164. (5) Rimell, J. T.; Marquis, P. M. Selective Laser Sintering of Ultra High Molecular Weight Polyethylene for Clinical Applications. J. Biomed. Mater. Res. 2000, 53, 414−420. (6) Zein, I.; Hutmacher, D. W.; Tan, K. C.; Teoh, S. H. Fused Deposition Modeling of Novel Scaffold Architectures for Tissue Engineering Applications. Biomaterials 2002, 23, 1169−1185. (7) Nguyen, K. T.; West, J. L. Photopolymerizable Hydrogels for Tissue Engineering Applications. Biomaterials 2002, 23, 4307−4314. (8) Dittrich, P. S.; Manz, A. Lab-on-a-Chip: Microfluidics in Drug Discovery. Nature Rev. Drug Discovery 2006, 5, 210−218. (9) Chin, C. D.; Linder, V.; Sia, S. K. Lab-on-a-Chip Devices for Global Health: Past Studies and Future Opportunities. Lab Chip 2007, 7, 41−57.

89

dx.doi.org/10.1021/la303788z | Langmuir 2013, 29, 82−91

Langmuir

Article

(31) Ward, T.; Faivre, M.; Abkarian, M.; Stone, H. A. Microfluidic Flow Focusing: Drop Size and Scaling in Pressure versus Flow-RateDriven Pumping. Electrophoresis 2005, 26, 3716−3724. (32) Tan, Y. C.; Cristini, V.; Lee, A. P. Monodispersed Microfluidic Droplet Generation by Shear Focusing Microfluidic Device. Sens. Actuators, B 2006, 114, 350−356. (33) Anna, S. L.; Bontoux, N.; Stone, H. A. Formation of Dispersions Using “Flow Focusing” in Microchannels. Appl. Phys. Lett. 2003, 82, 364−366. (34) Raven, J.-P.; Marmottant, P.; Graner, F. Dry Microfoams: Formation and Flow in a Confined Channel. Eur. Phys. J. B 2006, 51, 137−143. (35) Dollet, B.; Hoeve, W. V.; Raven, J.-P.; Marmottant, P.; Versluis, M. Role of the Channel Geometry on the Bubble Pinch-off in FlowFocusing Devices. Phys. Rev. Lett. 2008, 100, 034504. (36) Hashimoto, M.; Garstecki, P.; Whitesides, G. M. Synthesis of Composite Emulsions and Complex Foams with the Use of Microfluidic Flow-Focusing Devices. Small 2007, 3, 1792−1802. (37) Garstecki, P.; Gitlin, I.; DiLuzio, W.; Whitesides, G. M.; Kumacheva, E.; Stone, H. A. Formation of Monodisperse Bubbles in a Microfluidic Flow-Focusing Device. Appl. Phys. Lett. 2004, 85, 2649− 2651. (38) Mansur, H. S.; Costa, H. S. Nanostructured Poly(vinyl alcohol)/ Bioactive Glass and Poly(vinyl alcohol)/Chitosan/Bioactive Glass Hybrid Scaffolds for Biomedical Applications. Chem. Eng. J. 2008, 137, 72−83. (39) Kim, N.; Sudol, E. D.; Dimonie, V. L.; El-Aasser, M. S. Poly(vinyl alcohol) Stabilization of Acrylic Emulsion Polymers Using the Miniemulsion Approach. Macromolecules 2003, 36, 5573−79. (40) Cohen, N.; Silverstein, M. S. One-Pot Emulsion-Templated Synthesis of an Elastomer-Filled Hydrogel Framework. Macromolecules 2012, 45, 1612−21. (41) Takeuchi, H.; Kojima, H.; Yamamoto, H.; Kawashima, Y. Polymer Coating of Liposomes with a Modified Polyvinyl Alcohol and their Systemic Circulation and RES Uptake in Rats. J. Controlled Release 2000, 68, 195−205. (42) Pal, K.; Banthia, A. K.; K.Majumdar, D. Preparation and Characterization of Polyvinyl Alcohol−Gelatin Hydrogel Membranes for Biomedical Applications. AAPS PharmSciTech 2007, 8 (1), E1−E5. (43) Xia, Y.; Whitesides, G. M. Soft Lithography. Annu. Rev. Mater. Sci. 1998, 28, 153−184. (44) Whitesides, G. M.; Ostuni, E.; Takayama, S.; Jiang, X.; Ingber, D. E. Soft Lithography in Biology and Biochemistry. Annu. Rev. Biomed. Eng. 2001, 3, 335. (45) McDonald, J. C.; Duffy, D. C.; Anderson, J. R.; Chiu, D. T.; Wu, H.; Schueller, O. J. A.; Whitesides, G. M. Fabrication of Microfluidic Systems in Poly(dimethylsiloxane). Electrophoresis 2000, 21, 27−40. (46) Rasband, W. S. ImageJ; U. S. National Institutes of Health (NIH), Bethesda, Maryland, USA; available at http://imagej.nih.gov/ ij/. (47) Schmid, B. 3D Viewer; available at http://rsbweb.nih.gov/ij/ plugins/3d-viewer/. (48) Dougherty, B. Local Thickness; available at http://www.optinav. com/Local_Thickness.htm. (49) Barthel, K. U. Color inspector 3D; available at http://imagej. nih.gov/ij/plugins/color-inspector.html. (50) Bolte, S.; Cordelières, F. P. A Guided Tour into Subcellular Colocalization Analysis in Light Microscopy. J. Microsc. 2006, 224, 213−232 (3D Object counter plugin available at http://rsbweb.nih.gov/ ij/plugins/track/objects.html). (51) Couprie, M.; Bertrand, G. Topological Grayscale Watershed Transformation. In SPIE Vision Geometry VI Proceedings, 1997; Vol 3168, pp 136−146. (52) Sun, Q.; Tan, L.; Wang, G. Liquid Foam Drainage: An Overview. Int. J. Mod. Phys. B 2008, 22, 2333−2354. (53) Kruglyakov, P. M.; Karakashev, S. I.; Nguyen, A. V. Foam Drainage. Curr. Opin. Colloid Interface Sci. 2008, 13, 163−170. (54) Derjaguin, V. B.; Obukov, E. V. Anomalien Dü nner Flüssigkeitss- Chichten III. ActaPhusicochim.URSS 1936, 5, 1.

(10) Craighead, H. Future Lab-on-a-Chip Technologies for Interrogating Individual Molecules. Nature 2006, 442, 387−393. (11) Cameron, N. R.; Kimmins, S. D. Functional Porous Polymers by Emulsion Templating: Recent Advances. Adv. Funct. Mater. 2012, 21, 211−225. (12) Zhang, H. F.; Cooper, A. I. Thermoresponsive “Particle Pumps”: Activated Release of Organic Nanoparticles from Open-Cell Macroporous Polymers. Adv. Mater. 2007, 19, 2439−2444. (13) Christenson, E. M.; Soofi, W.; Holm, J. L.; Cameron, N. R.; Mikos, A. G. Biodegradable Fumarate-Based PolyHIPEs as Tissue Engineering Scaffolds. Biomacromolecules 2007, 8, 3806−3814. (14) Barbetta, A.; Massimi, M.; Di Rosario, B.; Nardecchia, S.; De Colli, M.; Conti Devirgiliis, L.; Dentini, M. Emulsion Templated Scaffolds that Include Gelatin and Glycosaminoglycans. Biomacromolecules 2008, 9, 2844−2856. (15) Barbetta, A.; Massimi, M.; Conti Devirgiliis, L.; Dentini, M. Enzymatic Cross-Linking versus Radical Polymerization in the Preparation of Gelatin PolyHIPEs and Their Performance as Scaffolds in the Culture of Hepatocytes. Biomacromolecules 2006, 7, 3059−3068. (16) Barbetta, A.; Dentini, M.; De Vecchis, M. S.; Filippini, P.; Formisano, G.; Caiazza, S. Scaffold Based on Biopolymeric Foams. Adv. Funct. Mater. 2005, 15, 118−124. (17) Chimenti, I.; Rizzitelli, G.; Gaetani, R.; Angelici, F.; Ionta, V.; Forte, E.; Frati, G.; Schussler, O.; Barbetta, A.; Messina, E.; Dentini, M.; Giacomello, A. Human Cardiosphere-Seeded Gelatin and Collagen Scaffolds as Cardiogenic Engineered Bioconstructs. Biomaterials 2011, 32, 9271−9281. (18) Barbetta, A.; Carrino, A.; Costantini, M.; Dentini, M. Polysaccharide Based Scaffolds Obtained by Freezing the External Phase of Gas-in-Liquid Foams. Soft Matter 2010, 6, 5213−5224. (19) Barbetta, A.; Rizzitelli, G.; Bedini, R.; Pecci, R.; Dentini, M. Porous Gelatin Hydrogels by Gas-in-Liquid Foam Templating. Soft Matter 2010, 6, 1785−1792. (20) Barbetta, A.; Barigelli, E.; Dentini, M. Porous Alginate Hydrogels: Synthetic Methods for Tailoring the Porous Texture. Biomacromolecules 2009, 10, 2328−2337. (21) Barbetta, A.; Gumiero, A.; Pecci, R.; Bedini, R.; Dentini, M. Gasin-Liquid Foam Templating as a Method for the Production of Highly Porous Scaffolds. Biomacromolecules 2009, 10, 3188−3192. (22) Dehghani, F.; Annabi, N. Engineering Porous Scaffolds Using Gas-Based Techniques. Curr. Opin. Biotechnol. 2011, 22, 661−666. (23) Garstecki, P.; Gañań -Calvo, A. M.; Whitesides, G. M. Formation of Bubbles and Droplets in Microfluidic Systems. Bull. Pol. Acad. Sci.: Tech. Sci.. 2005, 53, 361−372. (24) Sun, Y. S.; Peng, S. W.; Lin, K. H.; Cheng, J. Y. Electrotaxis of Lung Cancer Cells in Ordered Three-Dimensional Scaffolds. Biomicrofluidics 2012, 6, 014102. (25) Wang, C. C.; Yang, K. C.; Lin, K. H.; Liu, H. C.; Lin, F. H. A Highly Organized Three-Dimensional Alginate Scaffold for Cartilage Tissue Engineering Prepared by Microfluidic Technology. Biomaterials 2011, 32, 7118−7126. (26) Lin, J. Y.; Lin, W. J.; Hong, W. H.; Hung, W. C.; Nowotarski, S. H.; Montenegro Gouveia, S.; Cristo, I.; Lin, K. H. Morphology and Organization of Tissue Cells in 3D Microenvironment of Monodisperse Foam Scaffolds. Soft Matter 2011, 7, 10010−10016. (27) Choi, S. W.; Zhang, Y.; Xia, Y. Three-Dimensional Scaffolds for Tissue Engineering: the Importance of Uniformity in Pore Size and Structure. Langmuir 2010, 26, 19001−19006. (28) Link, D. R.; Anna, S. L.; Weitz, D. A.; Stone, H. A. Geometrically Mediated Breakup of Drops in Microfluidic Devices. Phys. Rev. Lett. 2004, 92, 054503. (29) Garstecki, P.; Fuerstman, M. J.; Stone, H. A.; Whitesides, G. M. Formation of Droplets and Bubbles in a Microfluidic T-junction Scaling and Mechanism of Break-up. Lab Chip 2006, 6, 437−446. (30) Xu, J. H.; Li, S. W.; Tan, J.; Wang, Y. J.; Luo, G. S. Preparation of Highly Monodisperse Droplet in a T-junction Microfluidic Device. AIChE J. 2006, 52, 3005−3010. 90

dx.doi.org/10.1021/la303788z | Langmuir 2013, 29, 82−91

Langmuir

Article

(55) Derjaguin, V. B. Theory of Stability of Colloids and Thin Liquids Films; Plenum/Consultants Bureau: New York, 1989. (56) Taylor, P. Ostwald Ripening in Emulsions. Adv. Colloid Interface Sci. 1998, 75, 107−163. (57) Gibson, L. J.; Ashby, M. F. Cellular Solids: Structure and Properties, 2nd ed.; Cambridge University Press: Cambridge, U.K., 1997. (58) Hollister, S. J. Porous Scaffold Design for Tissue Engineering. Nat. Mater. 2005, 4, 518−524. (59) Gallego, D.; Ferrell, N.; Sun, Y.; Hansford, D. Multilayer Micromolding of Degradable Polymer Tissue Engineering Scaffolds. J. Mater. Sci. Eng. C 2008, 28, 353−358. (60) Simon, J. L.; Michna, S.; Lewis, J. A.; Rekow, E. D.; Thompson, V. P.; Smay, J. E.; Yampolsky, A.; Parsons, J. R.; Ricci, J. L. In Vivo Bone Response to 3D Periodic Hydroxyapatite Scaffolds Assembled by Direct Ink Writing. J. Biomed. Mater. Res., Part A 2007, 83A, 747−758. (61) Hahn, M. S.; Miller, J. S.; West, J. L. Three-Dimensional Biochemical and Biomechanical Patterning of Hydrogels for Guiding Cell Behavior. Adv. Mater. 2006, 18, 2679−2684. (62) Kotov, N. A.; Liu, Y. F.; Wang, S. P.; Cumming, C.; Eghtedari, M.; Vargas, G.; Motamedi, M.; Nichols, J.; Cortiella, J. Inverted Colloidal Crystals as Three-Dimensional Cells scaffolds. Langmuir 2004, 20, 7887−7892. (63) Zhang, Y. J.; Wang, S. P.; Eghtedari, M.; Motamedi, M.; Kotov, N. A. Inverted-Colloidal-Crystal Hydrogel Matrices as Three-Dimensional Cell Scaffolds. Adv. Funct. Mater. 2005, 15, 725−731. (64) Aste, T.; Weaire, D. Hard Problems with hard balls, in The pursuit of perfect packing; CRC Press Taylor & Francis Group: New York, 2000; pp 20−34. (65) Bernal, J. D.; Mason, J. Co-ordination of randomly packed spheres. Nature 1960, 385, 910−911.

91

dx.doi.org/10.1021/la303788z | Langmuir 2013, 29, 82−91