|Home | About | Journals | Submit | Contact Us | Français|
Spatial and temporal control of bioactive signals in three-dimensional (3D) tissue engineering scaffolds is greatly desired. Coupled together, these attributes may mimic and maintain complex signal patterns, such as those observed during axonal regeneration or neovascularization. Seamless polymer constructs may provide a route to achieve spatial control of signal distribution. In this study, a novel microparticle-based scaffold fabrication technique is introduced as a method to create 3D scaffolds with spatial control over model dyes using uniform poly(D,L-lactide-co-glycolide) microspheres. Uniform microspheres were produced using the Precision Particle Fabrication technique. Scaffolds were assembled by flowing microsphere suspensions into a cylindrical glass mold, and then microspheres were physically attached to form a continuous scaffold using ethanol treatment. An ethanol soak of 1h was found to be optimum for improved mechanical characteristics. Morphological and physical characterization of the scaffolds revealed that microsphere matrices were porous (41.1±2.1%) and well connected, and their compressive stiffness ranged from 142 to 306kPa. Culturing chondrocytes on the scaffolds revealed the compatibility of these substrates with cell attachment and viability. In addition, bilayered, multilayered, and gradient scaffolds were fabricated, exhibiting excellent spatial control and resolution. Such novel scaffolds can serve as sustained delivery devices of heterogeneous signals in a continuous and seamless manner, and may be particularly useful in future interfacial tissue engineering investigations.
Engineering tissues and organs requires combinations of biomaterials, cells, and bioactive signaling molecules.1 Bioactive signals may be exogenously supplied via either the nutrient media (possible in in vitro culture conditions) or polymeric scaffolds (incorporated in a soluble or immobilized form), by utilizing growth factor–secreting natural or genetically modified cells, and/or by gene delivery,2 and are most commonly delivered in a homogeneous manner. However, spatial patterning of biological cues is vital to some of the most fundamental aspects of life, from embryogenesis to wound healing to nerve cell signaling, all involving concentration gradients of signaling molecules. Spatial patterning of bioactive signals may thus be a critical design element in the engineering of tissues or organs.
Various strategies have been developed to create gradients of bioactive signals. As early as the 1960s, diffusion-driven two-dimensional (2D) nonlinear gradients of soluble proteins were developed to identify chemotactic response.3 A few recent studies reported innovative diffusion- or convection-dominated approaches of creating linear or nonlinear protein gradients within three-dimensional (3D) scaffolds.4–6 Using photolithographic and soft lithographic techniques, many innovative methods of protein/cell patterning have been reported that provide micron-level positional accuracy; however, such techniques are largely limited to 2D constructs (reviewed by Park and Shuler7). To fabricate 3D scaffolds with embedded linear gradients, a commercially available gradient maker (Gradient maker; CBS Scientific, Del Mar, CA) has also been utilized in various studies.8–10 A number of other innovative strategies that have been applied to create gradient-based substrates for highly diversified applications have been reviewed recently.11,12
In the areas related to tissue engineering, gradient-based signal delivery systems have by far gained the most attention in the fields of neural tissue engineering4,5,9 and in the study chemotaxis.3,13 Interfacial tissue regeneration is another key area that may benefit from gradients of bioactive signals, as some studies have suggested that signals from a tissue may influence the development and growth of its neighbor. For example, it can be seen during embryonic development and morphogenesis that the fate of one germ layer depends on signals from its neighbor.14 An in vitro culture study reported that only coculture with chondrocytes (as opposed to fibroblasts or osteoblasts) was successful at promoting osteogenic differentiation of mesenchymal stem cells in a selective manner,15 indicating the importance of simultaneous triggering of osteo- and chondroinduction for osteochondral tissue regeneration. An integrated scaffold with embedded gradients of growth factors at the interface, therefore, may trigger simultaneous tissue formation, and may have an adjuvant effect on interfacial tissue regeneration.
Microparticles have been long studied as polymeric delivery devices for a variety of drugs due to the ease of fabrication, control over morphology, the ability to discretely control their physicochemical properties, and versatility of controlling the release kinetics of loaded therapeutics.16 Recently, microparticle-based approaches of scaffold design have received much attention in the field of tissue engineering, targeting regeneration/repair of a variety of tissues (e.g., cartilage,17,18 bone,19,20 and neural21,22), where microparticles may act as supporting matrices for cell attachment and/or as carriers of bioactive agents for controlled delivery of exogenous signals. Poly(D,L-lactide-co-glycolide) (PLG), an aliphatic polyester, has been widely used in many of these investigations because the polymer is biocompatible and biodegradable. Moreover, the degradation kinetics of the polymer is flexible, which can be modulated by altering one or more of the factors, such as copolymer ratio, molecular weight, end-group chemistry, crystallinity, glass transition temperature, and the like.23,24 Some recent studies reported fabrication of matrices exclusively made of PLG microspheres utilizing heat-sintering19,25 dichloromethane vapor treatment26,27 or a solvent/nonsolvent sintering method.28,29
It is well known that microsphere size is one of the major determinants of polymer degradation rate, and is a primary factor governing the release kinetics of loaded molecules.16 Unfortunately, microsphere fabrication using traditional methods (e.g., emulsion or spraying technique) generates reproducible, but often poorly controllable, sphere sizes and size distribution.30 To achieve control over local growth factor concentrations in a microsphere-based scaffold, control over microsphere size is critical. Moreover, by controlling the microsphere sizes, precise spatial control over pore sizes and macroporosity may also be achieved.
In the present study, utilizing our ability to create relatively monodisperse microspheres,30 we introduce a novel microparticle-based scaffold fabrication technique to create scaffolds with spatial control over active ingredients using uniform PLG microspheres and an ethanol treatment. As discussed later, such macroscopic gradients can particularly be useful for future interfacial tissue regeneration investigations.
PLG copolymer (50:50 lactic acid:glycolic acid; intrinsic viscosity 0.41dL/g, Mw ~50,000Da, density 1.34g/mL) was purchased from Birmingham Polymers (Pelham, AL). Poly(vinyl alcohol) (PVA; 88% hydrolyzed, 25,000Da) was obtained from Polysciences (Warrington, PA). Rhodamine B base and fluorescein were obtained from Sigma (St. Louis, MO). Dichloromethane (DCM; HPLC grade) was obtained from Fisher Scientific (Pittsburgh, PA). Ethanol (Absolute −200 Proof) was obtained in-house.
Uniform PLG microspheres were prepared using technology from our previous reports.30 Briefly, PLG dissolved in DCM (30% w/v) was sprayed through a small-gauge needle. The polymer stream was acoustically excited using an ultrasonic transducer (Branson Ultrasonics, Danbury, CT) controlled by a waveform generator (model 33220A; Agilent Technologies, Santa Clara, CA), resulting in regular jet instabilities that produced uniform polymer droplets. An annular carrier stream (~0.5% PVA w/v in distilled water) surrounding the droplets was produced using a nozzle coaxial to the needle. The emanated polymer/carrier streams flowed into a beaker containing approximately 1000mL of 0.5% PVA. To extract the solvent, incipient polymer droplets were stirred for 3–4h. Subsequently, the hardened microspheres were filtered and rinsed with distilled water (~1L) to remove residual PVA. Finally, microspheres were lyophilized (Freezone; Labconco benchtop model, Kansas City, MO) for 2 days and stored at −20°C under desiccant. In a similar manner, fluorescent dye–loaded microspheres were prepared for concentration profile assessment (discussed later) by using PLG solution (~30% w/v in DCM) codissolved with rhodamine B or fluorescein.
The size distribution of microsphere preparations was determined using a Coulter Multisizer 3 (Beckman Coulter, Fullerton, CA) equipped with a 560-μm aperture. Freeze-dried particles were suspended in Isoton electrolyte that was stirred at low speeds to prevent particles from settling. A minimum of 5000 microspheres were analyzed for each set of particles.
Two sets of freeze-dried microspheres were separately loaded into two syringes in the form of suspensions, prepared by suspending microspheres (~1% w/v) in distilled water/PVA solution (volume ratio PVA:distilled water 1:20 [PVA 0.5% w/v]). The syringes were then installed in the scaffold fabrication apparatus (Fig. 1). The suspensions were pumped through the attached tubing to a cylindrical glass mold (6mm diameter) in a controlled manner using programmable syringe pumps (PHD 22/2000; Harvard Apparatus, Holliston, MA). Through the bottom of the mold, the distilled water/PVA solution was filtered, while the microparticles stacked in the mold. The suspensions in the syringes were constantly stirred magnetically to keep them homogeneous. To prevent microspheres from rapid settling or sticking to the walls of the mold, a constant level of distilled water was maintained in the mold, controlled by an additional infusion syringe pump (Harvard Apparatus) and a vacuum pump. The stacked microspheres were washed with distilled water (~100mL) and were allowed to soak in ethanol (100%) for 50min, and then the ethanol was pulled out using a vacuum pump. Ethanol soak resulted in physical attachment of adjacent microspheres, resulting in the formation of a matrix. To compare the effects of the duration of ethanol soak on scaffold morphology, porosity, and mechanical characteristics, additional scaffolds with various ethanol soak times (i.e., 30min, 1h, 2h, or 4h) were prepared. The molds (containing the scaffolds) were freeze-dried (Freezone, Labconco benchtop model) for a minimum of 2 days, and then the scaffolds were retrieved from the molds. In some cases, the scaffolds were prepared using suspension(s) of dye-loaded microspheres with predefined distinct flow profiles, which were later used in concentration profile assessment studies.
Both microspheres and scaffolds were freeze-dried overnight and cryofractured with a razor blade, and then sputter coated with gold. The imaging was performed using a Leo 1550 field emission scanning electron microscope at an accelerating voltage of 5kV under a high vacuum.
Differential scanning calorimetry (DSC) (Q100; TA Instruments, New Castle, DE) was used to measure the change in glass transition temperature (Tg) of the PLG following the microparticle and scaffold preparations. Prior to the analysis, raw PLG and one set of microspheres were lyophilized for 1 day, and a scaffold (prepared by ethanol soak of 50min) was lyophilized for 2 days. The experiments were carried out in triplicate on the samples (~15–20mg each) packed in sealed aluminum pans. For each sample, a nonisothermal scan was performed from −10°C to +80°C at a heating rate of 10°C min−1 under nitrogen atmosphere, and the corresponding Tg was recorded.
Porosity measurements were performed by imaging the scaffolds (prepared by 50min ethanol soak) using microCT (μCT 40; Scanco Medical, Southeastern, PA). Using 3D MicroCT reconstruction and a segmentation value of 75, porosities and degrees of anisotropy were directly determined from scaffold sections (height 0.44–0.47mm) (n=4). Using NIH ImageJ software, 2D microCT images were also analyzed. In addition, theoretical porosities of the scaffolds were calculated using the density of the raw PLG and the apparent densities of the scaffolds prepared by 30min, 50min, 1h, 2h, and 4h of ethanol exposure.31 The diameter (d), thickness (h), and mass (m) of the cylindrical scaffolds were measured, and porosities were calculated as:
where ρapp is the apparent density of the scaffold, given by , and ρ is the density of the stock PLG.
Unconfined compression tests were performed using a uniaxial testing apparatus (Instron Model 5848, Canton, MA) with a 50N load cell. A custom-made stainless steel bath and compression-plate assembly were mounted in the apparatus. Cylindrical scaffold samples prepared with an ethanol soak duration of 30min, 1h, 2h, and 4h (2.7–6mm height, ~6mm diameter) were tested at a strain rate of 5mm/min under simulated physiological conditions (i.e., under phosphate-buffered saline [PBS—0.138M sodium chloride, 0.0027M potassium chloride] at 37°C). A strain rate in this range, that is, ~1% to 2%/s, is considered to be a moderate value for compressive testing of cartilaginous tissues.32–35 Moduli of elasticity were obtained from the initial linear regions of the stress–strain curves (rationale explained later).36 The stress was defined as the ratio of the load to the initial cross-sectional area, and the strain was defined as the ratio of the change in the length to the original length.
Porcine chondrocytes were harvested from a hog ankle (Duroc breed, 6 months old, female) obtained from a local slaughterhouse in a manner described previously.37 The cells were then plated for expansion in monolayer and incubated at 37°C in 5% CO2, with media changed every 2–3 days. The cell culture medium consisted of Dulbecco's modified Eagle's medium, 10% fetal bovine serum (ES cell quantified), 1% penicillin–streptomycin–fungizone, 1% nonessential amino acids (all from Invitrogen Life Technologies, Carlsbad, CA), and 50μg/mL L-ascorbic acid (tissue culture grade; Fisher Scientific). The cells were expanded and passaged twice before being seeded onto the scaffolds.
Cylindrical scaffolds (~6mm diameter, 1mm height) were prepared using a 50min ethanol soak, as mentioned earlier. Cells were seeded onto these scaffolds at a density of 3×106 cells per scaffold using the orbital shaker method as described previously,38 and cultured for 18 days with half of the media refreshed every other day. Following this incubation period, the scaffolds were stained with LIVE/DEAD reagent (dye concentration 2mM calcein AM, 4mM ethidium homodimer-1; Molecular Probes, Carlsbad, CA) and incubated for 45min, before being subjected to fluorescence microscopy (Olympus/Intelligent Innovations Spinning Disk Confocal Microscope, University of Kansas, Lawrence, KS).
Four specific scaffolds were prepared, as mentioned earlier, using two different microsphere types (rhodamine-loaded and fluorescein-loaded microspheres, or rhodamine-loaded and blank microspheres). The syringes were loaded individually with one microsphere type and attached to the scaffold fabrication apparatus. Microspheres were pumped in a predefined manner using specific flow profiles (Fig. 2), and then the scaffolds were prepared by physically attaching the microspheres together with an ethanol soak of 50min. The resulting scaffolds were imaged under UV light using a UV lamp (254/365nm; UVGL-25; UVP, Upland, CA) and a high-resolution camera, and images were analyzed using NIH ImageJ software to assess spatial control over the composition of the scaffolds.
The effects of microparticle preparation and ethanol soak on the Tg of PLG were statistically analyzed using a three-level single factor analysis of variance (ANOVA) and a Tukey's Honestly Significant Difference post hoc test when significance was detected (n=3). Moduli of elasticity and theoretical porosities of scaffolds prepared using varying ethanol soak times were analyzed in a similar manner using a four-level single factor ANOVA, followed by a Tukey's Honestly Significant Difference post hoc test when significance was detected (n=5).
Microspheres having a uniform diameter were created using the previously reported Precision Particle Fabrication method30 and were characterized for their size and morphology. Four sets of microparticles were produced; (a) blank-220μm, (b) blank-160μm, (c) rhodamine B–loaded (10% w/w)-150μm, and (d) fluorescein-loaded (10% w/w)-150μm. Uniform solid PLG microspheres of 220μm diameter were used to form scaffolds for all substudies, except the scaffolds used in scanning electron microscopic imaging (160μm diameter was also used) and the flow profile assessment study (dye-loaded microspheres were also used). The microspheres had relatively monodisperse size distribution and a solid interior morphology (Fig. 3). Specifically, microsphere sizes in the range of 100–300μm were chosen, as it is our hypothesis for future in vitro tissue engineering investigations that this range may provide optimal pore sizes to allow cellular infiltration and cell-to-cell interaction.
A novel scaffold fabrication apparatus was designed, as mentioned earlier, to construct microsphere-based scaffolds with spatial control over their structure. Microsphere matrices were constructed using blank microspheres of 160 or 220μm diameter with 50min ethanol soak time (denoted by Smicrosphere-size-time in minutes; S160-50 group and S220-50, respectively). In addition, to observe the effect of ethanol soak times on scaffold morphology and mechanical properties, scaffolds were prepared using microspheres of 220μm diameter with varying ethanol soak times of 30min, 1h, 2h, and 4h (S220-30, S220-60, S220-120, and S220-240, respectively). The ethanol soak time was selected based on preliminary scaffold fabrication results, which indicated that a minimum of a 30min ethanol soak was required to attain some partial integration between the microspheres. It was also observed that exceeding a soaking time of 70min resulted in reduced porosities. In general, the optimum range of ethanol soak time was a function of the polymer properties (copolymer ratio, molecular weight, etc.). Microspheres composed of PLG with lower molecular weights are expected to require shorter ethanol soak times or dilution of ethanol for production of mechanically integrated porous matrices.
Cylindrical scaffolds, 6mm in diameter and 2.7–6mm in height, were prepared, and their morphology was analyzed (Fig. 4). Scanning electron micrographs of a representative scaffold from the S160-50 group revealed that the scaffolds were porous, having interconnected pores. In addition, to investigate the effect of the duration of ethanol soak on the microspheres, S220-30, S220-60, S220-120, and S220-240 scaffolds were analyzed using SEM (Fig. 5). Microsphere morphology and pore sizes in the scaffolds were found to be a function of the duration of ethanol soak, where soak times longer than 1h resulted in visible distortion from spherical morphology and pore closure at several places (Fig. 5A–D). Representative pore size range from one representative time group (1h ethanol soak) measured from the SEM images was found to be ~20–120μm. Higher magnification images (Fig. 5E–H) show the points of contact between the microspheres. Surface film layers formed as a result of plasticization of PLG with ethanol. Integration of these surface films led to the formation of a well-connected matrix.
The mechanical integrity of the scaffolds was analyzed by unconfined compression testing at simulated physiological conditions. A characteristic plot obtained from the testing of a scaffold sample demonstrated that the curve had a typical nature: an initial linear region (0% to ~25% strain), a middle nonlinear region (~25% to 70% strain), and a final linear region of considerably higher linear slope (post 70% strain) (Fig. 6A). The hypothesized mechanism of compression is somewhat analogous to the compression of closed-foam cellular solids.36 Accordingly, the initial linear region (0% to ~25% strain) was used to determine the stiffness of the scaffolds (142–308kPa) (Fig. 6B). The middle nonlinear region (~25% to 70% strain), the onset of which begins after achieving a yield stress, signifies the phenomenon of pore collapse, causing a continuous increase in scaffold density and stiffness. The final linear region (post 70% strain) corresponds to the material densification regime due to the absence of pores or any further changes in pore volume. Unlike the observations in cellular solids,36,39 a collapse plateau region was absent from the stress–strain curve, probably owing to the differences in testing conditions (i.e., simulated physiological condition as opposed to testing in air at ambient temperature). Mechanical test results indicated that the average modulus of elasticity of S220-60 was significantly higher than the moduli of the S220-30 and S220-240 scaffolds (p<0.05). While it may be expected that an increase in ethanol soak time would result in an increased stiffness and reduced porosity of the scaffolds, no such trends were seen in the range of ethanol soak times examined. Visual inspection of the scaffolds revealed that scaffolds prepared by a 30min soak did not have well-integrated microspheres, and microspheres were falling off of the ends of the scaffolds. When the ethanol soak time exceeded 1h (i.e., for S220-120 and S220-240 groups), the reduction in mechanical integrity might be a result of increased morphological distortion of the microspheres from a spherical shape that may have resulted in a poor packing of the microspheres. Thus, an ethanol soak of 1h was found to provide optimal mechanical properties.
The effect of microsphere preparation and scaffold fabrication on Tg of PLG was analyzed by DSC (Table 1). Microsphere preparation led to a small drop (~1.4%) in the Tg of the raw polymer (p<0.005). However, the ethanol treatment during the fabrication of scaffolds resulted in reduced Tg of the PLG, where the average Tg dropped by 14% compared to raw PLG (p<10−6).
Changing the ethanol soak time resulted in the slight variation in the overall morphology of the scaffold. The mean theoretical porosities of S220-30, S220-60, S220-120, and S220-240 scaffolds were 41.1%, 38.8%, 32.8%, and 40.4%, respectively (n=5). Unexpectedly, the theoretical porosity for the S220-240 group was found to be higher than that for the S220-60 and S220-120 groups; however, no statistically significant differences were noticed among the groups. For the measurement of porosities of the scaffolds experimentally, scaffolds prepared using a 50min ethanol soak time were imaged using microCT (Fig. 7), and average porosities determined directly from scaffold sections using 3D reconstruction were found to be 41.1%. The values were found to be considerably similar to the corresponding porosity value obtained using 2D ImageJ analysis (41.5%) and similar to the average theoretical porosities (44.9%) (Table 2). MicroCT analysis also confirmed that the scaffolds were isotropic; the average degree of anisotropy was 1.06±0.1. In addition, interconnectivity of the pore structure was confirmed by microCT.
Porcine ankle chondrocytes, dynamically seeded and cultured on S220-50 group scaffolds, were assessed for their viability. The majority of the cell population was identified as viable after a total of 18 days in culture (Fig. 8).
Images of scaffolds that were prepared using specific flow profiles with dye-loaded or blank microsphere suspensions (described in Fig. 2) are shown in Figure 9A–D. Images of the scaffolds captured under UV light were modified by pseudocoloring them to create black and white images. Each image was divided in five equal parts, and particle distribution in the direction perpendicular to the interface was analyzed using ImageJ software to create relative intensity versus relative distance plots. The plots demonstrated successful fabrication of bilayered, multilayered, and gradient scaffolds (Fig. 9E–H). Irrespective of the scaffolds, standard deviations were higher at the interface, probably due to imprecise settling of the microspheres in the mold and/or wetting effects on the walls of the mold. The characteristic nature of each plot, however, was similar to the corresponding flow profile applied during the scaffold fabrication. The plots demonstrated the ability of the scaffold fabrication set-up to create scaffolds of various predefined profiles with spatial control. In addition, the orientation of the interface may also be varied (compare Fig. 9B and 9C), which can be controlled by manipulating the vertical orientation of the cylindrical mold. Note that similar flow profiles were used to prepare these two scaffolds (Fig. 2B, C), the only difference being the pitch of the mold.
Spatial control over the release of the bioactive molecules is a critical aspect that, along with the temporal control, may provide the possibility of mimicking complex signal patterns, such as those during embryonic development.2 In the present study, we designed a novel scaffold fabrication apparatus and demonstrated our ability to produce microsphere-based scaffolds with spatial control over molecular composition. Proof of concept was provided using blank or dye-loaded microspheres, which were used as building blocks to fabricate bilayered, multilayered, and gradient scaffolds.
In comparison to traditional microsphere preparation methods, our ability to synthesize monodisperse microspheres may lead to improved systems to explore the effects of microparticle size on microsphere-based scaffolds. Scaffolds made of uniform microspheres are ideal to study the influence of microparticle size on the degradation patterns and rates within scaffolds. In addition, as observed in the case of colloidal crystal-templated gel-based scaffolds,40,41 uniform microspheres can pack closely compared to randomly sized microspheres, providing better control over the pore sizes and porosity of the scaffold, and may considerably aid the mechanical integrity of the scaffolds. Moreover, local release of molecules from the microspheres in a bulk scaffold is related to individual microsphere size and polymer properties. Reproducibility and predictability associated with uniform microsphere-based scaffolds may make them suitable for a systematic study of physical and chemical effects in order to achieve control over local release of growth factors within such a scaffold.
Integrated microsphere matrices have been created in the past by employing a heat-sintering technique, which requires heating of microspheres above their Tg. Heat sintering is suitable for the preparation of bioconductive microsphere-based scaffolds; however, inclusion of growth factors in the microspheres before exposing them to heat may severely affect protein activity. The sintering temperatures and durations of heat exposure used in some previous studies were 160°C for 4h (PLG; 85:15 lactic acid:glycolic acid),19 65°C for 4h (PLG/bioactive glass),25 70°C for 4h (poly(D,L-lactide)/poly(ethylene glycol)),42 and 62°C for varied times of 4, 24, 48, and 72h (PLG; 58:42 lactic acid:glycolic acid).43 Such elevated temperatures for extended durations may lead to reduction in the bioactivity or complete denaturation of encapsulated proteins.42 In the present study, we introduced an ethanol treatment technique as an alternative for the production of microsphere-based matrices, which may alleviate such concerns. For example, the proprietary process used by Alkermes (Cambridge, MA) utilizes low temperature casting of microspheres with ethanol as the antisolvent.44 Recently, another technique for creating microsphere-based scaffolds was reported that utilized dichloromethane vapor treatment, a benign process that was shown not to affect the activity of bioactive signals.26,27 Dichloromethane is an organic solvent commonly used in microsphere preparation methods for dissolving polymeric materials. Ethanol, on the other hand, is an organic solvent of higher polarity, which is arguably a more common chemical in tissue engineering protocols (applied as a sterilizing agent) and possesses an ability to physically attach aliphatic polyesters. As predicted by the Hildebrand theory involving the polymer/solvent solubility,45 ethanol is a poor solvent for PLG in comparison to dichloromethane. When ethanol came in physical contact with PLG, dissolution of a thin surface layer of PLG microspheres was previously observed,46 typically resulting in the formation of a “skin layer” or a “thin film” around the microsphere, as can also be evidenced from the SEM images (Fig. 5). Due to ethanol treatment, a drop in Tg and subsequent plasticization of PLG scaffolds were reported.47 Based on these reports and our observations, the mechanism of microsphere attachment appears to be that the treatment with ethanol leads to a drop in Tg, resulting in softening of the wetted microspheres near the surface. The surface films that form in the process adjoin, and then the subsequent freeze-drying step removes the ethanol from the integrated microsphere-based matrix. In the ethanol-free dry state, a recovery in mechanical properties of the scaffold is expected.47 The duration of the ethanol soak was an important process parameter, as longer durations were expected to lead to increased thicknesses of the surface film layer by affecting the diffusion of ethanol into the microspheres. Figure 5 provides supporting evidence, where increased durations of ethanol exposure were observed to lead to increased deformation from a spherical morphology. In addition, the extent of interconnection between the microspheres was also increased with increasing ethanol soak. Plasticization of PLG with ethanol is also a function of polymer properties, such as molecular weight and crystallinity. For example, Perugini et al.48 reported softening of lower molecular weight PLG (~12,000Da) with an ethanol wash, while the same treatment did not affect the physical integrity of higher molecular weight PLG (~34,000Da). When contrasted to a solvent/antisolvent sintering technique,28,29 the ethanol treatment process is an antisolvent treatment method, which requires a longer time exposure comparatively, however, to a relatively mild organic solvent.
Because a highly porous substrate may provide more surface area for cellular attachment, high porosity and interconnectivity is a desired feature for tissue engineering scaffolds, which often comes at the expense of mechanical integrity. In this regard, an ethanol soak of 1h was found to be the optimum that led to acceptable porosities and improved mechanical characteristics. The average porosity of the ethanol-treated scaffolds, as determined in the present study by 3D microCT analysis, was found to be 41.1% for an ethanol soak of close to 1h. These values were found to be similar to the corresponding theoretical porosity measurement and porosity values determined using 2D image analysis (Table 2). The degradable nature of the substrate material suggests that the porosity of such a construct may eventually increase with time; thus, porosities in this range (~38–43%) may be acceptable. Moreover, a previous in vivo study utilizing microparticle-based scaffolds provides supporting evidence regarding the suitability of such scaffolds for in vivo defect repair.19 However, to enhance the porosity as well as interconnectivity of the pores in the scaffolds, one possible future approach would be to utilize hollow or porous microspheres in the preparation of the scaffolds.49,50 Our group has demonstrated the ability to create porous microspheres in the past,51 and scaffolds made of such entities will be investigated in the future. In addition, future work will be directed to improve the cell seeding efficiency and cell distribution within such matrices.
Unidirectional compression testing was performed in a simulated physiological environment, since these scaffolds are designed for in vitro/in vivo tissue engineering. Mechanical properties are direction independent due to the isotropic nature of the scaffolds. The moduli of elasticity of the scaffolds were found to be ~140–300kPa, after which the scaffold material began to deform plastically. The primary reason for such a low range of elastic moduli was the testing conditions (hydrated conditions, 37°C), known to affect the mechanical properties of similar polymeric materials.47 Due to the degradable nature of the used polymer, one may expect a further decrease in the mechanical integrity of the scaffold with degradation. Although the degradation effects may be counteracted by tissue regeneration in the pores, which could enhance the mechanical properties, a few strategies can be investigated in the future in this regard. First, selection of an adequate PLG polymer (e.g., a higher lactic acid content, higher molecular weight, ester-end group chemistry, and a higher Tg) can be utilized in microsphere preparation that may significantly decrease the reduction in mechanical properties with hydration and temperature, and may allow tailoring the degradation rate of the scaffold for specific needs. Second, a bimodal distribution of spheres can be utilized in scaffold preparation that will allow for a better mechanical integrity and closer packing of the microspheres (although it may lead to a reduced pore volume that can potentially reduce cellular infiltration). Finally, one may encapsulate nanophase materials in the microspheres to create composite spheres that may improve the mechanical characteristics of the building units. For example, one may encapsulate nanophase biocompatible calcium salts as a filler material in the microspheres that may improve the mechanical characteristics, and also act as a buffering agent to reduce pH changes that occur with the degradation of PLG (due to the formation of acidic by-products).52 This approach may also enhance the overall cellular viability over a long-term culture. Finally, effects of adding haptotactic cues, such as Arg-Gly-Asp (RGD) and Tyr-Ile-Gly-Ser-Arg (YIGSR), can be explored to improve the overall cell viability.53
The effect of ethanol treatment on the Tg of the raw PLG was found to be significant. It may indicate a possibility that residual levels of ethanol may be present in the scaffold. More importantly, the Tg dropped below 37°C, which is not desired as it may considerably affect the mechanical properties of the scaffolds when placed in vivo. To keep the Tg of the scaffolds above the limit of 37°C, some possible strategies could be to use PLG with higher molecular weights, or with a higher lactic acid–to–glycolic acid ratio.
From the perspective of osteochondral tissue engineering, there are only a few previous reports of scaffold designs having heterogeneous distribution of growth factors embedded within an integrated scaffold. Holland et al. created bilayered oligo(polyethylene glycol) hydrogel–based scaffolds, designed to release transforming growth factor-β1 (TGF-β1) and/or insulin-like growth factor-I (IGF-I) in the cartilage-forming layer, while no growth factors were added to the bone forming layer.54 In a similar previous study from the same group, single growth factor (TGF- β1) release from the cartilage-forming layer was investigated.55 The concept of zonal release for multiple growth factor delivery was demonstrated by Suciati et al.42 In their work, poly(D,L-lactic acid) microsphere-based scaffolds were created using poly(ethylene glycol) as a plasticizer, and subsequently zonal release of horseradish peroxidase and bone morphogenic protein-2 (BMP-2) from trilayered and bilayered scaffolds, respectively, was demonstrated. As the release kinetics of growth factors from such carriers is a diffusion-controlled phenomenon, a step transitioning of the carriers of signaling molecules in bilayered scaffolds, as applied in the aforementioned studies, may be sufficient to promote the interfacial tissue regeneration successfully. The approach described here offers the potential advantage of a seamless transition at the osteochondral interface using spatially controlled macroscopic gradients of bioactive signal molecules at the interface. Although a single cell per se will not be able to sense such a macroscopic gradient (unless it is a migrating cell), spatial variations in the signals will cause an individual cell A at point 1 to experience different signals than cell B at point 2, and individual cell responses combined with cell-to-cell signaling may lead to a commensurate cell response that may benefit interfacial tissue regeneration. In addition, microsphere-based gradient scaffolds made of subcellular-sized microspheres may provide an alternative to generate stable gradients of soluble bioactive signals over a long time periods by acting as sustained delivery vehicles, and thus may prove helpful in the study of chemotaxis.
We have demonstrated a novel approach to creating microparticle-based gradient scaffolds, which may be designed to release opposing gradients of bioactive signals at the interface of a biphasic scaffold. The methodology may also be extended to create biphasic scaffolds with more than two growth factors or multiphasic scaffolds with more than one interface. In addition, utilizing microspheres made of two different materials, one may employ this technique to create transversely isotropic substrates that contain a macroscopic gradient in composition and stiffness. We reported use of ethanol treatment to create interconnected microsphere-based matrices, as an alternative to heat sintering and solvent treatment methods. For example, growth factor–loaded microspheres may be used to create similar heterogeneous 3D scaffolds to deliver growth factors with predefined spatial and temporal release profiles. Future studies will be directed to quantify the release of active factors from the scaffolds, improve the mechanical properties, investigate the effect of microsphere size, and evaluate the performance of seamless gradient scaffolds, compared to bilayered scaffolds, for osteochondral tissue regeneration.
The present work was supported by NIH/NIDCR grant 1 R21 DE017673-01A1, the Juvenile Diabetes Research Foundation, and the University of Kansas General Research Fund. We are also grateful to Dr. Eric Munson for providing the DSC facility. We would also like to acknowledge Junming Luo of the University of Kansas Medical Center and Rasesh Kapadia of Scanco Medical for the microCT scanning and analysis.