Mechanically-Activated Microcapsules for ‘On-Demand’ Drug Delivery in Dynamically Loaded Musculoskeletal Tissues

Abstract

Delivery of biofactors in a precise and controlled fashion remains a clinical challenge. Stimuli-responsive delivery systems can facilitate ‘on-demand’ release of therapeutics in response to a variety of physiologic triggering mechanisms (e.g. pH, temperature). However, few systems to date have taken advantage of mechanical inputs from the microenvironment to initiate drug release. Here, we developed mechanically-activated microcapsules (MAMCs) that are designed to deliver therapeutics in an on-demand fashion in response to the mechanically loaded environment of regenerating musculoskeletal tissues, with the ultimate goal of furthering tissue repair. To establish a suite of microcapsules with different thresholds for mechano-activation, we first manipulated MAMC physical dimensions and composition, and evaluated their mechano-response under both direct 2D compression and in 3D matrices mimicking the extracellular matrix properties and dynamic loading environment of regenerating tissue. To demonstrate the feasibility of this delivery system, we used an engineered cartilage model to test the efficacy of mechanically-instigated release of TGF-β3 on the chondrogenesis of mesenchymal stem cells. These data establish a novel platform by which to tune the release of therapeutics and/or regenerative factors based on the physiologic dynamic mechanical loading environment, and will find widespread application in the repair and regeneration of numerous musculoskeletal tissues.

INTRODUCTION

Localized and controlled drug delivery has several advantages over systemic delivery, including delivery to specific locations, maintenance of drug concentrations within a therapeutic range (while avoiding systemic effects), and preservation of activity for long-term administration¹. In particular, microcapsule delivery systems are characterized by their small size (diameter of 1–1000 μm) and core-shell formulation. The active agent is encapsulated in the aqueous core, which is surrounded by a solid shell that protects against degradation and environmental factors. Release from such systems depends on the attributes of the capsule shell, and can be controlled as a function of time or stimulus to program release to match a desired profile^2,3. These stimuli-responsive approaches enable self-regulation, wherein physiological feedback may actively controls release through an internal trigger⁴. While several triggers exist for microcapsule formulations^5–11, few have taken advantage of the mechanical environment to initiate release. Recently, Korin et. al. designed microscale aggregates of nanoparticles which disaggregate under the abnormally high shear stresses found in narrowed blood vessels and deliver tissue plasminogen activator to dissolve clots¹². Alternatively, mechanical stretch (strain) has been combined with passive degradation to foster convection of drugs from microdepots within stretchable substrates¹³ or with thermal mechanisms for liquidization of a solid core to enable pressure-induced release in the gastrointestinal tract¹⁴.

While the above illustrates that mechanically-instigated release can be achieved, systems do not exist that are designed to operate in the demanding musculoskeletal environment. Musculoskeletal tissues within the body experience mechanical stimulation across multiple force magnitudes and length scales¹⁵. These forces not only maintain tissue integrity (through modulation of cellular activity), but can also, at supraphysiologic levels, initiate degenerative processes that require surgical intervention to restore load-bearing function^16,17. In articular cartilage for instance, which functions to bear and distribute loads in joints, traumatic lesions on the articular surface show very little endogenous healing, and ultimately increase in size and lead to further degenerative processes in the joint causing pain and disability¹⁸. Osteoarthritis often takes years to develop after cartilage injury; the mechanical demands combined with the inflammatory milieu^19,20 of an injured or osteoarthritic joint present a challenging environment. Therefore, both early and late stage reparative approaches remain an unsolved clinical problem. Advances in biomaterials design for cartilage repair have allowed for greater control over biofactor release in vivo in terms of spatial, temporal, and multi-factor delivery^21–28. While these delivery systems have improved defect fill, matrix deposition, and mechanical properties compared to biofactor-free formulations, native-like cartilage properties have not yet been achieved. Moreover, approaches that deliver multiple factors (e.g. TGF-β and IGF-1) often fail to elicit optimal in vivo responses, despite the promise of staged delivery observed in in vitro studies^25,29,30. These limitations may be due, in part, to passive release from implanted materials; thus, stimuli-responsive approaches may be required to improve in vivo outcomes.

To that end, we developed a novel class of mechanically-activated microcapsules (MAMCs) as a stimuli-responsive delivery system in which the mechanical loading environment of the regenerating tissue is harnessed to elicit a therapeutic response. We postulate that such in situ mechanical regulation of microcapsule activation can be utilized as a novel means by which to stimulate tissue repair and regeneration. This approach is inspired by self-healing polymer systems used in material science applications, where microcapsules embedded in a polymer matrix rupture upon physical damage, releasing a catalyst for polymerization and repair of the surrounding material³¹. We generated uniform populations of biocompatible and biodegradable microcapsules^32–34 using a custom microfluidic device developed in our laboratory, and established the tunability of their mechano-activation based on the physical attributes of each microcapsule formulation and its composition. From this parameter space, we selected a cohort of MAMCs and evaluated their response in 3D matrices that mimicked regenerating cartilage tissue, and demonstrated that the physical properties of the microenvironment modulated mechano-activation under dynamic loading. Finally, we demonstrated the feasibility of encapsulating biologically active compounds within MAMCs, and showed that mechanically mediated release of TGF-β3, a growth factor important in the maturation of engineered cartilage, could promote tissue formation. These studies establish a novel means by which to tune therapeutic release based on the dynamic mechanical loading environment of musculoskeletal tissues.

RESULTS

Microcapsule Physical Properties Determine Mechano-activation Thresholds

Microcapsules were fabricated using a microfluidic system^32–34 to form a double emulsion template for the evolution of poly(D,L-lactide-co-glycolide) acid (PLGA) microcapsules (Figure S1). To determine how the physical attributes regulated thresholds of MAMC mechano-activation (i.e. empty fraction after rupture), microcapsules were subjected to compression in a 2D setting (parallel plate compression)³⁵. Modulation of microcapsule shell thickness and diameter was achieved by tuning two fabrication parameters: (1) PLGA concentration and (2) fluid phase flow rates. Decreasing the PLGA concentration in the middle phase reduced the shell thickness while maintaining the same outer diameter (Figure 1a-b, 1.8 and 2.7% w/v). The combined effect of a lower polymer concentration (0.9% w/v) and reduced flow rates resulted in a thinner shell as well as a smaller diameter (Figure 1a-b, 0.9% w/v). Thus, the resulting thickness to diameter ratio (t/D) was not different between the two lower PLGA concentration MAMCs (0.9 and 1.8% w/v, t/D~0.006). Conversely, MAMCs fabricated with a higher PLGA concentration had a higher t/D ratio (t/D=0.0096, 2.7% w/v). Under compression (0–1N), the mechano-activation profile depended on the t/D ratio, with both MAMC formulations with a lower t/D ratio (t/D~0.006) showing higher sensitivity to load, despite differences in their outer diameter (Figure 1c-d and Figure S2). In contrast, a higher t/D ratio (t/D=0.0096) rendered the capsules insensitive to load over the same range (0–1N), though all MAMCs ruptured at higher loads (5N).

Figure 1. MAMC physical attributes regulate mechano-activation.

(a) Polymer concentration and fluid flow rates control MAMC shell thickness (t) and diameter (D) (N = 7–8 double emulsion images/batch for shell thickness calculations). (b) Confocal mid-slices of MAMCs with different t/D ratios containing dextran (green) with labeled shells (red). Full image scale bar = 100μm, inset scale bar = 25μm. (c-d) MAMCs fracture with increasing load, releasing fluorescent dextran from the aqueous core, depending on the t/D ratio (N = 3 replicates/load/level/MAMC batch, p values refer to the comparison of slopes for linear regression fits of the mechano-activation response, mean ± SD). Scale bar = 100μm.

The choice of polymer and its degradation also likely regulates the mechano-activation response of MAMCs. Since PLGA is biodegradable via hydrolysis, shell degradation likely influences MAMC structural integrity and therefore, the mechanical response as a function of time. To characterize their time-dependent sensitivity to mechanical loads, MAMCs were fabricated with either fast-degrading (PLGA 50:50, the ratio of lactic and glycolic acids = 50:50) or a slow degrading (PLGA 85:15) PLGA copolymers, and were incubated under physiologic conditions (37°C, PBS, pH ≈ 7.4). At defined time points up to 10 weeks, microcapsules were subjected to compression as above. We demonstrated differences in the degradation profile of each PLGA type by tracking the empty fraction of MAMCs over time under no load (0 N) conditions (Figure S3). MAMCs whose shells rapidly degraded (PLGA 50:50) showed a marked increase in sensitivity to applied load at Day 7 compared to Day 1 (Figure 2b-c). The failure morphology of these MAMCs was confirmed by 3D volume reconstructions of confocal z-stacks and SEM (Figure 2a, Video S1 and S2), which showed a midline rupture in the shell and loss of internal fluorescent contents. By Day 14, degradation of the shell of PLGA 50:50 MAMCs resulted in 80% of the microcapsule population being empty under zero load. With application of load, these microcapsule shells fractured into fragments with 100% rupture observed. In comparison, for the slower degrading MAMC population (PLGA 85:15), the mechanical release profile remained stable over the first 3 weeks, and showed a slightly suppressed mechano-activation response after 28, 42, and 70 days of incubation (Figure 2d-e; weekly images in Figure S4).

Figure 2. Shell degradation influences MAMC mechano-activation.

(a) Reconstruction of the volume (confocal, scale bar = 50μm) and morphology (SEM, scale bar = 50μm) illustrating rupture mechanism on day 7 (PLGA 50:50). (b-c) Fast-degrading microcapsules (PLGA 50:50) show a marked change in response between Day 1 (D1) and 7 (D7), with nearly all MAMCs rupturing in the absence of load by Day 14 (D14). (d-e) Slow-degrading MAMCs (PLGA 85:15) exhibit a stable or slightly suppressed mechano-activation response compared to day 1 over 70 days of incubation in physiologic conditions (N = 3 replicates/load/time point with the exception of PLGA 85:15 D1 5N, D42 0.5N, and D70 0.25N where N = 2 replicates, p values refer to the comparison of the ‘% empty’ value at 0.5 N for each time point to the Day 1 value, mean ± SD). Fluorescent images scale bar = 100μm.

Microcapsule Mechano-activation in 3D is Determined by Matrix Mechanics

To facilitate the use of mechanically-activated microcapsules for therapeutic delivery in a 3D environment, we evaluated their release characteristics in engineered tissues under physiologically-relevant loading conditions. We hypothesized that microcapsule deformation and release in these 3D environments would depend on the stiffness of the surrounding matrix and its ability to transmit load to the microcapsules. To test this hypothesis, MAMCs (PLGA 75:25, t/D = 0.0046) were embedded in polyethylene glycol diacrylate (PEGDA) hydrogels with equilibrium moduli matching the range of mechanical properties seen in maturing engineered cartilage^36,37. MAMC-hydrogel constructs were subjected to static compression on a confocal microscope-mounted device to track microcapsule deformation as a function of bulk hydrogel deformation (0 to 20% strain, 5% step strain increments). While MAMCs encapsulated within a soft, immature matrix (E_Y = 50 kPa) deformed minimally, those in stiffer hydrogels representative of a mature cartilage-like construct (E_Y = 500 kPa) deformed to greater extents along with the encapsulating material. MAMC deformation occurred in directions parallel (E₁₁) and perpendicular (E₂₂) to that of hydrogel deformation (Figure 3a and Figure S5).

Figure 3. MAMC mechano-activation in 3D depends on the properties of the encapsulating hydrogel.

(a) Schematic of microcapsule compression in PEGDA hydrogels. Quantification of MAMC strain (E₁₁ and E₂₂) in hydrogels of two different stiffnesses demonstrates the effect of encapsulating hydrogel properties on microcapsule deformation, with representative images at each strain step shown (50 kPa hydrogels: N = 40 microcapsules, 500 kPa hydrogels: N = 30 microcapsules, p values refer to comparisons between matrix stiffnesses at each strain step, mean ± SEM). Scale bar = 100μm. (b) Dynamic loading of MAMC-embedded hydrogels resulted in a graded increase in microcapsule rupture as the stiffness of the encapsulating material increased (N = 3 to 6 hydrogels for dynamic loading, N = 8 hydrogels for FS pooled across stiffness values, p values refer to comparison to FS group, mean ± SD). (c) Representative confocal images of MAMCs under free swelling or dynamically-loaded conditions illustrate the mechano-activation as a function of matrix stiffness, as observed by differences in MAMC internal fluorescence (blue: matrix, green: MAMCs, ruptured MAMCs = loss of internal signal). Scale bar = 200μm. (d) Tracking of MAMC mechano-activation with duration of dynamic loading showed an increase in content release over time in stiff matrix microenvironments (N = 4 to 7 hydrogels, p values refer to comparisons between matrix stiffnesses at each time point, mean ± SD).

Next, to determine how repeated loading cycles impacted MAMC rupture and release in the context of changing matrix stiffness, dynamic compressive loading (2% tare strain, 20% cyclic strain, 5 Hz) was applied over a 90-minute period. The fraction of microcapsules ruptured due to cyclic loading was quantified as a function of matrix stiffness. Results from this assay showed a graded mechano-activation over a range of hydrogel stiffnesses (~25 to 150 kPa), with more microcapsule rupture in stiffer hydrogels (Figure 3b-c). The increasing stiffness of these hydrogels is within the range of maturing cartilage in a repair site^36,37. This mechano-activation did not occur immediately, but rather increased over time with increasing cycle number, indicating a gradual fatigue and rupture of the MAMCs in this dynamic, physiologically-relevant loading environment (Figure 3d). Together, these results support the hypothesis that matrix mechanical properties regulate MAMC release in 3D environments.

Biofactors Encapsulated Within MAMCs Retain Biologic Activity

The clinical application of MAMCs depends on the ability to encapsulate biologically active compounds that modulate disease or accelerate tissue repair upon mechano-activation. In the context of cartilage repair, TGF-β3 is known to stimulate the production of constitutive matrix components (e.g. proteoglycans, type II collagen) that contribute to the functional properties of native and regenerating tissue³⁶. To determine if TGF-β3 is incorporated within the MAMCs and can be measured after release, the supernatant from intact and ruptured MAMCs was evaluated by ELISA. Microcapsules containing only the carrier protein (fluorescently-labelled bovine serum albumin, BSA) or the vehicle solution (BSA and HCl) were also tested as controls. No TGF-β3 was measured in supernatants from control MAMCs (intact or ruptured), and minimal TGF-β3 was measured in the supernatant of intact TGF-β3-containing MAMCs (Figure 4a). Conversely, measuring the supernatant of ruptured TGF-β3-containing MAMCs showed a concentration-dependent release, based on the original number of microcapsules ruptured (Figure 4a). To confirm the bioactivity of released TGF-β3, the supernatant was applied to engineered cartilage produced from mesenchymal stem cells (MSCs), and chondrogenesis was assessed over 7 days. Measures of chondrogenesis in these constructs were evaluated relative to those cultured with and without the exogenous addition of TGF-β3, as positive (CM+) and negative (CM-) controls, respectively. Quantification of sulfated proteoglycan deposition in the engineered construct showed that TGF-β3 released from ruptured MAMCs stimulated matrix synthesis to the same level as exogenous addition of soluble TGF-β3 (CM+, Figure 4b). Minimal TGF-β3 activity was observed in the supernatant of intact TGF-β3-containing MAMCs, though some matrix formation was observed. Similar to media lacking the growth factor (CM-), control MAMCs (BSA and vehicle) did not stimulate chondrogenesis. This biochemical quantification was confirmed by histological analysis using Alcian blue to visualize proteoglycan accumulation, where increased extracellular matrix staining was observed in CM+ and ruptured TGF-β3-containing MAMC conditions compared to CM-media (Figure 4c). These data confirm that MAMCs can effectively be fabricated with clinically relevant molecules and deliver these active biofactors as a result of mechanical perturbation.

Figure 4. Biofactors released from MAMCs retain biological activity.

(a) TGF-β3 released from MAMCs measured by ELISA shows a dose-dependent response (based on the number of microcapsules ruptured). BSA and vehicle controls show no signal (N = 3 replicates/group, *p values indicate comparisons to the intact MAMC group, all other comparisons are indicated by a line, mean ± SD). MSCs undergo chondrogenic differentiation in response to TGF-β3 released from MAMCs as measured by (b) GAG accumulation in the construct (N = 4 replicates/group, *p values indicate comparison to CM-group, all other comparisons are indicated by a line, mean ± SD) and (c) histological evaluation of matrix deposition (Alcian blue staining of proteoglycans, scale bar = 50μm).

DISCUSSION

To advance musculoskeletal tissue repair and regeneration, we developed a novel class of mechanically-activated microcapsules (MAMCs) that promote autonomic healing via the release of anabolic factors in response to the mechanical environment. Clinically, the standard method for delivery of therapeutics to the synovial joint is the direct injection of the substance of interest. However, this results in a short retention time in the joint space^38,39, and there does not yet exist a standard approach for long-term drug release in this setting. Although nano-or micro-spheres are being investigated for delivery in the synovial space^40–54, the continuous drug release profile they provide may not be the optimal strategy for repairing musculoskeletal tissues, given the complexity of the repair processes.

Our MAMC system represents a new method for delivery that is specifically tailored and responsive to the physical forces experienced by tissues in the joint space, expanding the range and specificity of therapeutic application. Release from MAMCs is initiated when microcapsule mechanical failure thresholds are met. We demonstrated that these failure thresholds are governed by physical attributes (t/D ratio) and the polymer degradation rate, as well as the relationship between MAMC and matrix mechanical properties. We also showed that the stability (or ‘lifetime’) of mechanical activation of the current suite of microcapsules is relevant in the context of current repair strategies, such as matrix-assisted cartilage implantation (MACI). With a microcapsule stability of >70 days, delivery in this context would enable mechanoactivation and biofactor release when patients begin load bearing activities at 8 to 10 weeks post-surgery. For this, stable microcapsules would be injected into the synovial space or embedded within the repair site, and activate upon demand during tissue regeneration. Towards this, we have also shown that osmotically annealed microcapsules³² are stable in the synovial fluid environment (Figure S7).

MAMC mechano-activation need not be a one-time event, but rather, could represent an orchestrated delivery over time and circumstances. This would be achieved by combining different populations of microcapsules with tunable thresholds to establish a sequence of release events programmed into the tissue (Figure 5). Biofactor release would occur in a temporal fashion to regulate the complex biological transitions required for robust tissue repair. Agents that instigate differentiation may be programmed to release first, in order to stimulate early cartilage maturation, in response to deformation of the weaker, nascent matrix that forms early in repair. As the matrix continues to mature and stiffen, release of additional matrix-promoting or cell-protective agents could be triggered by cohorts of microcapsules with higher mechanical thresholds. Alternatively, additional release signatures could be programmed to maintain cartilage or combat inflammation after a subsequent joint injury.

Figure 5. MAMC repair of cartilage defects.

A proof-of-concept construct containing two different classes of MAMCs (green), along with chondrocytes (blue), is shown. MAMC physical attributes can be tuned to program release in maturing repair tissue exposed to dynamic loading within the joint. Biofactors released from MAMCs promote cell differentiation and matrix deposition in the cartilage defect, resulting in a stiffer tissue environment. As the surrounding matrix stiffens, it then engages and initiates release of therapeutics from additional MAMC populations. A number of biofactors can be encapsulated within this system, including anabolic factors to stimulate nascent matrix synthesis and anti-inflammatory factors to attenuate catabolic processes and control pathologic changes. Chondro-protective small molecules could also be included to maintain long-term cartilage stability and protect against future injury.

This suite of mechano-activation profiles can be achieved by additional modification of the fabrication system as well as tuning interactions with the surroundings. For instance, further characterization of the deformation behavior and release thresholds in various matrices (over a wide range of microcapsule physical and matrix mechanical properties) could inform finite element models to predict and refine MAMC therapeutic benefit in these complex and time evolving environments. Additional factors, such as microcapsule-matrix adhesion⁵⁵ could also be tuned to influence MAMC rupture behavior. Interfacial adhesion could also be modulated by changing physical characteristics (e.g. surface roughness of the shell wall⁵⁶) or the nature of the interaction (e.g. electrostatic interactions⁵⁷). These and other modifications can be used to generate a suite of mechano-responsive microcapsules, whose release is governed by physical inputs.

In terms of payload, our initial focus here was on an anabolic growth factor; however, the delivery of other factors including small molecules and disease modifying compounds (e.g. anti-inflammatory drugs, steroids, antibiotics,)^48,58,59 could likewise be included to enhance regeneration and intervene in disease progression⁶⁰. Importantly, improving the longevity of activity (Figure S6) will be necessary for effective in vivo cartilage repair. Loss of activity of biologics encapsulated within polymer systems is largely attributed to local acidification due to polymer degradation products, protein aggregation, and polymer-protein interactions due to hydrophobic or electrostatic interactions^61,62. Addition of excipients within the aqueous MAMC core, including pH buffering additives, sugars (e.g. sucrose, trehalose), liquid polymers (e.g. PEG, cyclodextrins), or free amino acids (e.g. arginine, glutamate), might also improve the long term stability of growth factors^61–63. Inclusion of these stabilizing factors may also extend the shelf life of MAMCs and help maintain biologic activity within the in vivo environment⁶⁴.

Taken together, this work introduces for the first time mechanically-activated microcapsules as a programmable drug delivery system to advance and promote musculoskeletal tissue repair and regeneration. As we work towards clinical translation of this delivery platform, it will be critical to expand our understanding of MAMC property-function (release) relationships in the context of in vivo mechanical loading. Future studies will evaluate the release of biofactors from microcapsules embedded within hydrogels cultured in dynamically loaded in vitro environments, as well as in 3D in vivo load-bearing environments to validate the therapeutic potential of this MAMC technology. This new mechanically-activated delivery system is particularly noteworthy in that it can be deployed in the context of already existing cartilage repair and regeneration technologies (such MACI), where rehabilitation regimens (passive motion and return to ambulation) can be tailored to instigate sequential MAMC release as tissue formation progresses. Finally, while this mechanically activated release system was specifically designed for the promotion of cartilage repair, it has the potential for widespread application in the repair of commonly injured musculoskeletal tissues where mechanical loading plays an important role.

METHODS

Experimental Design.

The objective of these studies was to establish the feasibility of fabricating mechanically-activated microcapsules. The study design focuses on characterization of MAMC release profile in 2D and 3D loading environments.

Microcapsule Fabrication.

Microcapsules were fabricated using a glass capillary microfluidic device as previously described^32,33 to form monodisperse Water/Oil/Water (W/O/W) double emulsions from three fluid phases (Figure 1). The inner aqueous phase was maintained at pH 7.4 and, for all experiments, included a fluorescent compound to visualize microcapsule integrity. The middle oil phase consisted of poly(D,L-lactide-co-glycolid) acid (PLGA) dissolved in chloroform or dichloromethane with the addition of Nile Red (100μg/mL, Sigma, N3013) to fluorescently visualize the shell wall. PLGA of various molar ratios of lactic:glycolic acid were used: ester-terminated PLGA 50:50 (MW 38,000–54000, ester-terminated, Sigma 739944), PLGA 75:25 (0.55–0.75 dL/g, MW 76,000–119,900, ester-terminated, Lactel B6007–1), PLGA 85:15 (0.55–0.75 dL/g, MW 76,000–119,900, ester-terminated, Lactel B6006–1). The outer aqueous phase contained 2% wt poly(vinyl alcohol) (PVA). All double emulsions were collected in 0.1–1% w/v bovine serum albumin (BSA, Sigma A7906) in phosphate buffered saline (PBS, P4417), with the pH tuned to the PLGA composition (PLGA 50:50 ≥ pH 7, PLGA 85:15 and 75:25 > pH 12) to maintain stability during microcapsule evolution (SI Figure 1c). The microcapsules’ size and the shell thickness were tuned by controlling the flow rates of the three fluid phases and the PLGA concentration (% w/v) in the middle phase. Average shell thickness was calculated based on conservation of mass and assuming uniform shell thickness as previously described²⁹, where the inner and outer radii of the double emulsions were quantified from images taken during fabrication using an upright optical microscope. The formed shells may have a slight non-uniformity in thickness due to the density difference between the inner and middle phases of the double emulsion. The average outer diameter of fully formed microcapsules was measured from the maximum intensity projection of a confocal z-stack image (Nikon A1R+ confocal microscope, 20X magnification, NIS Elements AR software). Shell thickness to outer diameter (t/D) ratio was calculated as a metric to assess effect of microcapsule geometry on mechano-activation. MAMC concentration (as defined by the number of microcapsules/μL, n=3–4 aliquots per fabrication batch) was also measured by confocal microscopy within the first three days (considered ‘Day 0’) following microcapsule condensation prior to conducting mechano-activation experiments.

2D Mechano-Activation.

A single layer of microcapsules (~500 MAMCs) was seeded between two glass coverslips and uniaxially compressed at a controlled strain rate ($\dot{\epsilon }$=0.5/s) to defined loads (0.1 to 5N). This assay was based on a parallel plate compression testing method previously described³¹. Unloaded (0N) microcapsules seeded between coverslips served as negative controls. Microcapsules in these experiments contained a model drug, fluorescein isothiocyanate-dextran (0.01 to 0.05% w/v, FITC-dextran, 2MDa, Sigma FD2000S) with 1% w/v BSA in PBS. Following compression, microcapsules were collected into PBS for overnight incubation at 37C, 5% CO₂ to allow for complete release of encapsulated contents. Microcapsules were imaged on a confocal microscope (4X magnification, microcapsule mid-plane) before and after load. Images were analyzed to quantify the number of intact microcapsules based on a threshold of >50% of ‘Day 0’ fluorescence signal intensity (FITC signal), as well as the total number of microcapsules (Nile Red signal). The fraction of empty microcapsules (% empty) was then calculated at each load level. Representative confocal z-stack images (20X magnification) were also obtained for volume reconstructions. In select experiments, scanning electron micrographs (SEM, FEI Quanta 600 ESEM) were obtained to visualize the failure morphology of microcapsules. Mechano-activation was assessed on Day 1 for microcapsule geometry experiments and at weekly time points for up to 10 weeks for polymer degradation experiments, where microcapsules were incubated at 37^oC and 5% CO₂.

3D Mechano-Activation.

Microcapsules were embedded in poly(ethylene glycol) diacrylate (PEGDA, MW 508 kDa) and cast between two glass plates to create a uniform hydrogel sheet. The polymer matrix was cross-linked using a free-radical initiation method as previously described^65,66 using ammonium persulfate (APS, 1mg/mL, BioRad 1610700) and tetramethylethylenediamine (TEMED, 0.4% v/v, BioRad 1610801). A methacrylated dye (9-anthracenylmethyl methacrylate, Sigma 578207, 0.1mg/mL) was included in some cases to visualize the hydrogel matrix in which the microcapsules were embedded. Cylindrical constructs (⊘: 4mm, H: 2.25mm) were cored via biopsy punch from the hydrogel sheet. Unconfined compression testing of blank hydrogels across a range of PEGDA concentrations (5–20% w/v) was used to determine the equilibrium modulus of the encapsulating matrix. For static-loading experiments, microcapsule-laden hydrogel constructs (0.1% v/v) were evaluated using a custom confocal-mounted compression device⁶⁷. MAMC deformation was tracked over 20% compressive strain applied to the hydrogel surface, including four strain steps of 5% each, which were followed by compression until hydrogel failure. At each strain step, confocal z-stack images were collected (10X magnification, depth: 300 μm), processed to acquire maximum intensity projections of the microcapsule shell, and thresholded to determine the bounding box lengths (x-and y-axis) for each identified microcapsule object in the binary image. Microcapsules in contact with one another or void of internal fluorescent contents were excluded from analysis. Deformation was quantified according to the MAMC strain in the direction of loading (E₁₁, x-direction) and perpendicular to (E₂₂, y-direction) loading. MAMC strain was computed as the difference in bounding box length between each strain step and the initial, un-loaded (ε=0) state, normalized to the initial, un-loaded state. For dynamic loading experiments, constructs were cyclically loaded in unconfined compression (Instron Electropuls) under physiologic conditions (PBS, 37°C). Free-swelling constructs maintained under the same conditions served as negative controls. Constructs were compressed to a 2% tare strain, followed by 20% cyclic strain applied at 5Hz for 1.5 hours (shorter time intervals were used for temporal experiments). After loading, constructs were incubated at 37°C overnight and imaged the following day. Maximum projection images of confocal z-stacks of constructs collected prior to dynamic loading (10X, depth: 300μm) were compared to post-dynamic loading images of the same microcapsule population in order to compute the number of ruptured microcapsules (dynamic % empty).

Biofactor Activity.

Microcapsules were fabricated containing activated Transforming Growth Factor-β3 (TGF-β3, R&D Systems, 243-B3/CF, 10μg/mL) with BSA (1mg/mL) and AlexaFluor 488-BSA (100μg/mL, Life Technologies, A13100) in PBS. Microcapsules containing only AlexaFluor 488-BSA (protein control) or the TGF-β3 activation solution (vehicle control, 1mg/mL BSA, 4mM HCl) served as negative controls. All solutions were filtered for sterility prior to fabrication for use in cell culture experiments. TGF-β3 content was assessed in the supernatant of ruptured and intact microcapsules via ELISA (R&D Systems, DY243). All microcapsules were completely ruptured in PBS using a TissueLyser LT system (Qiagen at 50Hz for 60s) in PBS. Both intact and ruptured microcapsules were diluted with an additional volume of PBS and centrifuged to separate the shell pellet from the supernatant. In a separate set of studies, microcapsules were stored at 4°C, with activity regularly assessed over three weeks to determine the shelf life of the growth factor within the MAMC.

Functional activity of TGF-β3 released from microcapsules was measured using an engineered cartilage model previously established in the literature⁶⁸. Briefly, mesenchymal stem cells were isolated from femoral and tibial bone marrow from juvenile bovine knees (Research 87, Bolyston, MA) and were expanded in basal medium (high glucose DMEM, 10% Fetal Bovine Serum, 1% penicillin/streptomycin/fungizone). MSCs (passage 2, 20 million cells/mL) were then embedded in agarose (2% w/v, Type VII, Sigma) to generate cylindrical constructs (⊘: 4mm, H: 2.25mm). Microcapsules were ruptured as described above and diluted in chemically defined medium (CM)⁶⁹. To test activity, the number of microcapsules ruptured into the CM was tuned such that 100% release would equate to a media concentration of 10ng/mL of TGF-β3⁴⁹. Supernatant from intact TGF-β3 microcapsules, as well as from ruptured and intact ‘vehicle’ and ‘protein’ microcapsules served as delivery controls. CM with or without the addition of exogenous TGF-β3 (10ng/mL) served as positive and negative aqueous controls, respectively. Cell-seeded constructs were cultured under the above media conditions for a period of 7 days. All media was prepared at the time of construct fabrication, with a separate aliquot frozen at −20°C for a media change on Day 4. Chondrogenesis was assessed by quantification of sulfated glycosaminoglycan (GAG) content (DMMB assay⁷⁰) and visualization of matrix deposition by histological evaluation (Alcian blue, Nikon Eclipse Ni).

Statistical Analysis.

All analysis was conducted using Graphpad Prism (v5) or SPSS statistical software packages. Linear regression was applied to ‘% empty’ vs. load plots and the slopes compared between microcapsule batches (t/D ratios) to determine 2D mechano-sensitivity. One-way ANOVA with Bonferroni’s post-hoc test to compare ‘% empty’ values of a given time point against Day 1 values at 0.5N load to evaluate the effect of polymer degradation. Independent t-tests (two-tailed) were used to compare microcapsule E₁₁ and E₂₂ strain between 50 and 500 kPa hydrogels at each strain step. One-way ANOVA with Bonferroni’s post-hoc test was used to compare ‘dynamic % empty’ values across all hydrogel matrix stiffness values against free-swelling controls hydrogels. An independent t-test (two-tailed) was used to compare the ‘dynamic % empty’ value between 54 and 150 kPa hydrogels at each time point. For TGF-β ELISA and GAG measurements, independent t-tests were used to compare intact vs. released conditions within a MAMC subset. One-way ANOVA with Bonferroni’s post-hoc test was used to analyze the dose response pattern for released TGF MAMCs, to compare all BSA, vehicle, TGF intact, and TGF released groups to CM-for GAG quantification, and to evaluate differences in TGF activity at each time point compared to ‘Day 0’, for a given storage temperature.

Supplementary Material

Supplemental Figure 1

Figure S1. MAMC Fabrication. (a) Schematic and (b) image of water/oil/water (W/O/W) double emulsion generation using a capillary microfluidic device. (c) Schematic showing the time evolution of PLGA microcapsules following evaporation of the oil phase.

Figure 2

Figure S2. MAMC mechano-sensitivity depends on the shell thickness to diameter ratio (t/D) across a range of applied loads (0 to 5N). Data for three distinct formulations are shown. Scale bar = 100μm.

Figure 3

Figure S3. Shell degradation rate impacts retention of mechano-activation capacity. PLGA 50:50 MAMCs show an accelerated degradation rate (as determined by empty microcapsules under zero load) as compared to PLGA 85:15 MAMCs with incubation at 37°C in PBS over 10 weeks (N=3 replicates/time point, mean ± SD).

Figure 4

Figure S4. Full image set of PLGA 85:15 MAMCs under direct compression (0 and 0.5) following incubation under physiologic conditions over a period of 70 days. Scale bar = 100μm.

Figure 5

Figure S5. Tracking of MAMC deformation from 0–20% strain at 5% strain increments. Quantification of MAMC deformation perpendicular to the axis of applied hydrogel deformation (E₂₂) correlates with E₁₁ data showing that microcapsule deformation is only observed within the stiffer hydrogel matrix (50 kPa hydrogels: N = 40 microcapsules and 500 kPa hydrogels: N = 30 microcapsules, p values refer to comparisons between matrix stiffnesses at each strain step, mean ± SEM). Scale bar = 100μm.

Figure 6

Figure S6. Bioactivity of TGF-β3 within MAMCs is sensitive to the storage temperature (4°C vs. 37°C in PBS). ELISA measurements over 22 days showed a more rapid decline in TGF-β3 content when MAMCs were stored at 37°C (within 4 days), compared to storage at 4°C (within 8 days, N = 3 replicates/time point/storage temperature, mean ± SD). R = ruptured, I = intact.

Figure 7

Figure S7. MAMC stability in a simulated intra-articular environment. Improvements in MAMC fabrication via osmotic annealing³² resulted in microcapsules that remain stable in media and synovial fluid environments.

Supplemental Video 2

Video S2. Confocal 3D volume reconstructions of ruptured MAMCs on Day 7 of incubation for PLGA 50:50 microcapsules.

Supplemental Video 1

Video S1. Confocal 3D volume reconstructions of intact MAMCs on Day 7 of incubation for PLGA 50:50 microcapsules.