|Home | About | Journals | Submit | Contact Us | Français|
To examine the relationships between biochemical composition and mechanical properties of articular cartilage explants during interleukin-1 (IL-1)-induced degradation and post-exposure recovery.
Bovine articular cartilage explants were cultured for up to 32 days with or without 20 ng/mL interleukin-1. The dynamic shear modulus |G*dyn| and equilibrium and dynamic unconfined compression moduli (Eequil and |E*dyn|) were measured at intervals throughout the culture period. In a subsequent recovery study, explants were cultured for 4 days with or without 20ng/mL IL-1 and for an additional 16 days in control media. The dynamic moduli |E*dyn| and |G*dyn| were measured at intervals during degeneration and recovery. Conditioned media and explant digests were assayed for sulfated glycosaminoglycans (sGAG) and collagen content.
Continuous IL-1 stimulation triggered progressive decreases in Eequil, |E*dyn|, and |G*dyn| concomitant with the sequential release of sGAG and collagen from the explants. Brief IL-1 exposure resulted in a short release of sGAG but not collagen, followed by a gradual and incomplete repopulation of sGAG. The temporary sGAG depletion was associated with decreases in both |E*dyn| and |G*dyn| which also recovered after removal of IL-1. During IL-1-induced degradation and post-exposure recovery, explant mechanical properties correlated well with tissue sGAG concentration.
As previously shown for developing cartilages and engineered cartilage constructs, cytokine-induced changes in sGAG concentration (i.e., fixed charge density) are coincident with changes in compressive and shear properties of articular cartilage. Further, recovery of cartilage mechanical properties can be achieved by relief from proinflammatory stimuli and subsequent restoration of tissue sGAG concentration.
Articular cartilage is comprised of chondrocytes embedded within an extracellular matrix (ECM) consisting primarily of the large, aggregating proteoglycan (PG) aggrecan, type II collagen, and water. The high density of negatively-charged sulfated glycosaminoglycans (sGAG) attached to the aggrecan core protein gives rise to an osmotic swelling pressure that resists compression and is balanced by tensile stresses carried by the collagen fiber network. Due to low tissue permeability, dynamic physiologic loads are carried primarily through pressurization of entrapped fluid 1. The resident chondrocytes synthesize and remodel the ECM 2-5 but can also contribute to tissue destruction in various degenerative conditions.
Quantitative relationships between ECM molecule content and cartilage mechanical properties have been explored through regression analyses in healthy, developing, and degenerating articular cartilage 6-13. These studies demonstrated statistically significant correlations between the concentration of sGAG and the compressive stiffness, and elevated sGAG concentrations were inversely correlated with the hydraulic permeability. The collagen concentration was also well-correlated with cartilage properties, indicating that the mechanical function of cartilage hinges on contributions from PGs and collagen. In addition, the interaction of PGs and the collagen network under cyclic and steady-state loads has been extensively characterized by the dynamic and equilibrium moduli in compression, shear and tension 17-20. There are no reports, however, of composition-function regression analyses for cartilage undergoing cytokine-induced degradation.
Interleukin-1 (IL-1) cytokines have demonstrated roles in promoting cartilage matrix resorption in vitro and mediating inflammation in vivo. Within hours of exposure to exogenous IL-1, chondrocytes in monolayer and explants upregulate and activate aggrecanases, leading to proteolysis and release of PGs 14, 23-25. Following PG release, the collagen network undergoes MMP-mediated degradation 14, 24-26. Furthermore, IL-1 decreases sGAG 14, 27-29 and protein synthesis 14, 30-32. IL-1-induced depletion of PGs and collagen degradation are accompanied by increases in permeability and decreases in equilibrium and dynamic compressive moduli and compression-induced streaming potential 14, 24. Autocrine IL-1 expression is believed to play a role in cartilage matrix remodeling as part of homeostasis 33, and chondrocytes have been reported to restore PG content and PG synthesis following transient IL-1 exposure in vitro and in vivo 25, 34-37. However, the extent to which this PG repopulation restores the mechanical properties of recovering cartilage has not been examined.
The studies described here present detailed time courses of the biochemical and biophysical changes associated with IL-1-induced cartilage degradation. The loss and subsequent recovery of matrix constituents in response to transient IL-1 exposure are also described, along with detailed measurements of the physical properties of these recovering explants. Based on linear regression analysis, the relationships between cartilage composition and mechanical properties in both the degrading and recovering explants are described. The results of these experiments, which further illuminate the roles of PGs and collagen in cartilage mechanics during tissue degradation and recovery lend insight into targets for diagnosis and treatment.
Under aseptic conditions, 3mm diameter full thickness cartilage explants were harvested with a biopsy punch (Miltex, York, PA) from the femoral condyles and patellar grooves of both stifles of an immature calf (Research 87, Boylston, MA). A custom cutting block was used to remove the most superficial ~300μm and the deep zone tissue, resulting in middle zone explants with a thickness of 1.78 ± 0.020mm as measured with digital calipers. Forty explants were randomly assigned to either control or IL-1-stimulated groups. To allow for equilibration to culture conditions, explants were precultured for 72 hours at 37°C and 5% CO2 in serum-free control medium consisting of high glucose Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 50μg/mL gentamicin, 0.1mM non-essential amino acids (Invitrogen, Carlsbad, CA) and 50μg/mL ascorbate (Sigma, St. Louis, MO). Explants were placed 2 per well in 48 well plates (Becton Dickinson, Franklin Lakes, NJ) in 0.5mL of either control medium or control medium supplemented with 20ng/mL recombinant human IL-1α (Peprotech, Rocky Hill, NJ). This IL-1 dosage has been shown to induce aggressive cell-mediated degradation, with complete depletion of sGAGs within two weeks 24, 38. Samples were cultured for up to 32 additional days, with media changed and collected every 48 hours. Four explants per group were removed from culture at days 4, 8, 12, 16, 24, and 32 and stored in 0.15M PBS (Invitrogen) with protease inhibitors (PI Cocktail Set I, Calbiochem, San Diego, CA) at −20°C for subsequent compression testing and biochemical analysis.
A second degradation study was conducted specifically to examine changes in shear properties due to IL-1 stimulation. Sixty-five explants (4mm diameter × 2mm thick) were isolated as described above from middle zone cartilage of a second calf. Following a 3-day preculture, explants were cultured in individual wells for up to 24 additional days in 0.5mL of control medium or control medium supplemented with 20ng/mL IL-1α. Five explants per group were removed from culture at days 0, 4, 8, 12, 16, 20, and 24 and stored in 0.15M PBS with protease inhibitors at −20°C for subsequent dynamic shear testing and biochemical analysis.
To study recovery from transient IL-1 exposure, thirty-six explants (4mm diameter × 2mm thick) were isolated as described above from middle zone cartilage of a third calf. Following a 3-day preculture, the explants were cultured for 4 days in 0.5mL of either control medium or control medium supplemented with 20ng/mL IL-1α, followed by up to 16 days of culture in control medium. Four explants per group were removed at days 0, 4, 8, 12 and 20 and stored in 0.15M PBS with protease inhibitors at −20°C for subsequent mechanical testing and biochemical analysis.
To examine whether IL-1 exposure decreased cell viability, explants (4mm diameter × 2mm thick) were isolated as described above from a fourth calf, precultured for 3 days and then cultured for up to 14 days in 0.5mL of either control medium or control medium supplemented with 20ng/mL IL-1α. Analysis of mitochondrial activity as an indirect, quantitative measure of viability was performed using the WST-1 assay kit (Biovision, Mountain View, CA) according to the manufacturer's instructions. Briefly, at day 2, 8 or 14, explants (n=3/condition) were incubated with the WST-1 reagent in DMEM for 2h at 37C. The conditioned media were then transferred to a 96-well plate and the abundance of formazan product was measured as the absorbance at 450nm. Viability of one additional explant per group was examined qualitatively by fluorescent staining (Live/Dead®; Invitrogen). Explants were rinsed briefly with DPBS, stained according to the manufacturer's instructions, and bisected along the axial plane. Images of calcein AM (live cells) and ethidium homodimer-1 (dead cells) staining were captured using the appropriate filters on an Zeiss LSM510 confocal microscope with a 20x objective.
Prior to mechanical testing, explants were thawed at room temperature. Thicknesses were measured with digital calipers (Mitutoyo USA, Aurora, IL) and wet masses were determined. Compression testing was performed using a SMT-S 5.6lbf load cell (Interface, Scottsdale, Arizona) and an ELF 3230 testing frame (Enduratec; Minnetonka, MN) at room temperature in 0.15M PBS with protease inhibitors, with displacements corrected for load cell compliance. Torsional shear testing was performed using a CVO 120HR stress-controlled rheometer with strain feedback (Bohlin; East Brunswick, NJ).
Samples from the first degradation study were tested in unconfined compression by applying a 0.1N tare load followed by four steps (5% each, 1mm/s ramp rate) of stress relaxation with a 10 minute relaxation after each strain step. As 10 minutes may not reflect complete relaxation, the force vs. time data F(t) for each step were fit to an analytical solution for the unconfined compression stress relaxation of a linear biphasic material described by Armstrong et al. 17:
where F∞ is the equilibrium (relaxed) force, , αn are the roots to the characteristic Bessel function equation J1(x) − (1−v)x J0 (x) / (1−2v) = 0, a is the sample radius, HA is the aggregate modulus and k is the permeability. For each relaxation step, a two parameter least squares fit was performed using Matlab 7.1 (Mathworks, Natick, MA) to determine the value of F∞ and the product HAk, assuming a Poisson's ratio of 0.1 and using the first five terms of the analytical solution. A linear regression of the equilibrium stresses ( F∞ /πa2 ) against the applied strain over the range of 10-20% was used to determine Eequil, the unconfined compression equilibrium modulus of each sample.
After relaxation at the second (10%) step, sinusoidal compression (±1.5% strain) was applied at 0.005, 0.05, 0.5, and 5Hz. For each frequency, the magnitude of the dynamic compressive modulus, |E*dyn|, was calculated by WinTest DMA software (Enduratec) as the ratio of the fundamental stress and strain magnitudes determined using a Fast Fourier Transform. Samples from the recovery study were similarly tested after a single 10% stress relaxation step to determine the dynamic compressive moduli at 0.01, 0.1, 1.0 and 10Hz.
Samples from the second degradation study and the recovery study were tested in torsional shear by applying a 10% compressive offset and, after relaxation, applying a nominal 0.5% sinusoidal shear strain at 0.01, 0.1, 1, and 10Hz. The magnitude of the complex shear modulus |G*dyn| was determined by the rheometer software from the ratio of the measured stress to the applied strain.
Following mechanical testing, explants were lyophilized overnight and dry masses were determined. Explants were then solubilized with 4 mg/mL proteinase K (Calbiochem) in 100mM ammonium acetate (pH 7.0) (Sigma). The digested explants and conditioned media were assayed for sGAGs via the dimethylmethylene blue (DMMB) assay 39, using shark chondroitin sulfate (Calbiochem) as a standard. Collagen was assayed via the pDAB/chloramine-T assay for hydroxyproline, using trans-4-hydroxy-Lproline (Sigma) as a standard and assuming a 1:8 mass ratio of hydroxyproline:collagen 40. As the compressive properties were expected to be related to the matrix fixed charge density (FCD), the sGAG and (for consistency) collagen contents were normalized by water volume in the composition-function analyses.
All statistical analyses were performed using Minitab Release 12 (Minitab, Inc., State College, PA) with significance at p<0.05. Collinearity among matrix components or among mechanical properties was determined via linear correlation analyses. Differences among time points in a treatment group or between treatments at a time point were examined via one-way ANOVA with Tukey's test for posthoc pairwise comparisons. The equilibrium modulus and dynamic shear modulus data were log transformed prior to ANOVA. Due to loss of sample integrity for IL-1-stimulated explants, sample sizes were insufficient to perform some comparisons at later time points. Regressions of mechanical parameters against matrix constituents were compared between treatment groups as described by Zar 41 using custom Minitab macros. Briefly, the groups were first compared to test the null hypothesis of equal regression slopes. In the case of a common slope, the groups were then compared to test the null hypothesis of equal intercepts. A common regression was used if neither null hypothesis was rejected.
As previously demonstrated 14, 23, 24, 26, 42, 43, treatment with 20ng/mL IL-1α resulted in substantial matrix depletion over the 32 day culture period. sGAG release for IL-1-stimulated samples peaked at day 8 with 59% of total sGAG released (Figure 1A) and was essentially complete by day 14. Collagen released from IL-1-stimulated explants was significantly elevated over controls at day 6 (p<0.05), peaked near day 14, and persisted through day 24, at which point 74% of total collagen release had occurred (Figure 1B). Control explants displayed a steady increase in sGAG content, indicating a basal level of PG synthesis (Figure 2A). The sGAG content of IL-1-stimulated explants decreased during the first 16 days in culture and reached the detection limit of the DMMB assay by day 24. The collagen content of control explants did not vary significantly during the culture period, while the collagen content of the IL-1-stimulated explants gradually decreased following day 8 (Figure 2B). No differences in cell viability or mitochondrial activity were noted between control and IL-1-stimulated explants (Figure 3), indicating that cell death did not contribute substantially to the decreases in matrix content with IL-1 exposure. Explant water content showed little variation with culture in either control or IL-1 groups (Figure 2C). Increases in the thickness of control explants accompanied increases in sGAG (Figure 2D). In contrast, the thickness of IL-1-stimulated explants did not change significantly during culture.
The ANOVA indicated a significant effect of time on Eequil for controls although no pairwise comparisons were significant (Figure 4A). |E*dyn| did not vary significantly during culture for control explants (Figure 4B). In contrast, IL-1-stimulated explants exhibited substantial changes in mechanical properties reflecting the ECM degradation. Both Eequil and |E*dyn| decreased as sGAG was released (Figures 4A-4B). Eequil reached detection limits by day 16, whereas |E*dyn| remained measurable through day 24. IL-1 stimulation for 32 days resulted in sufficient matrix loss that explants could not sustain handling for mechanical testing. |E*dyn| exhibited similar frequency dependence across time points and treatment groups (Figure 5).
For control explants, |G*dyn| remained fairly steady throughout the culture period, as indicated by the non-significant effect of time in the ANOVA (Figure 4C). In contrast, |G*dyn| for IL-1-stimulated explants exhibited an initial rapid decrease through day 8 and continued to decrease until it reached detection limits at day 20. As with |E*dyn|, |G*dyn| exhibited similar frequency dependence across time points and treatment groups (Figure 6).
Consistent with the degradation studies, transient IL-1 exposure lead to significant early loss of sGAG. Control explants released low levels of sGAG throughout the 20 day culture, whereas substantial amounts of sGAG were released to the culture medium during and immediately following IL-1 stimulation (Figure 7A). The peak in sGAG release at day 4 coincided with the end of IL-1 exposure. Withdrawal of IL-1 from the culture media was followed by a gradual decrease in sGAG release that reached control levels by day 8. When normalized to account for differences in explant size, the levels of sGAG released are consistent between the degradation and recovery experiments, although a slightly accelerated response to IL-1 occurred in the recovery study. Collagen release was low and comparable in control and IL-1-stimulated explants (Figure 7B), indicating that transient IL-1 exposure did not initiate substantial damage to the collagen network. As in the degradation studies, the sGAG content of control explants gradually increased (Figure 8A). The IL-1-stimulated explants lost sGAG content over the first 8 days of culture, indicating that the effects of IL-1 exposure persisted beyond the 4 days of stimulation. Between days 8 and 12, explants in the IL-1 group recovered sGAG to the level of the day 4 group, but through day 20 the sGAG content remained significantly below the baseline level of the day 0 group. The collagen content did not vary during culture (Figure 8B). The water content of control explants did not vary between groups over the first 12 days of culture, but the control group was significantly higher than the IL-1 group at day 20 (Figure 8C). The difference in water content at day 20 is particularly significant because when the residual sGAG content is normalized by the water content (mL H2O), there is no difference in sGAG/H2O at day 20 between control and IL-1 groups (data not shown). The thickness of control explants increased with time, consistent with the accumulation of sGAG, while the thickness of IL-1-stimulated explants did not vary significantly.
Changes in the mechanical properties of explants reflect the observed differences in explant composition. Control explants showed no significant differences in |E*dyn| or |G*dyn| with culture (Figures 9A-9B). |E*dyn| and |G*dyn| for IL-1-stimulated explants decreased to a minimum at day 8 before recovering to control levels by day 20, despite significantly lower levels of sGAG content. |E*dyn| and |G*dyn| exhibited similar frequency dependence across time points and treatment groups (data not shown).
Regression analyses indicated substantial differences between control and IL-1-stimulated explants in the dependence of mechanical properties on tissue composition (Table 1). The dependence of Eequil on sGAG concentration was not significantly different for control and IL-1-stimulated explants, but the regressions were significantly different between treatment groups for all other mechanical properties. The compressive properties of control explants were significantly dependent on sGAG/H2O (Figures 10A-10B) and collagen/H2O (Figures 10D-10E). In contrast, for IL-1-stimulated explants, all compressive properties were strongly dependent on sGAG/H2O (Figures 10A-10B) but exhibited no dependence on collagen/H2O (Figures 10D-10E). Both for control and IL-1-stimulated explants, |G*dyn| was strongly dependent on sGAG/H2O (Figure 10C) and weakly dependent on collagen/H2O (Figure 10F). Regression results of |E*dyn| and |G*dyn| at 0.5Hz and 1.0Hz, respectively, were consistent with analysis at all frequencies tested.
Collinearities among ECM components and among the measured mechanical properties were examined to aid in interpreting the composition-function regression analyses. Data from all time points were pooled for control and IL-1-stimulated explants. Significant, negative correlations were found between percent water and both collagen and sGAG concentrations for control and IL-1 groups (Table 2). A significant, positive correlation was found between collagen and sGAG concentrations for control explants but not for IL-1-stimulated explants, reflecting the decoupled sGAG and collagen release profiles. Significant, positive correlation coefficients were found among all compressive mechanical properties for both the control and IL-1 groups (Table 3). In control and IL-1-stimulated explants, high correlations (0.877<R<0.929) were found between |Eequil| and |E*dyn|, suggesting that the measurements offer similar predictions of the mechanical integrity of the ECM. Regressions of |E*dyn| at 0.5Hz and |G*dyn| at 1.0Hz were consistent at the other tested frequencies.
Consistent with the IL-1 degradation studies, regression analysis of the IL-1 recovery data indicated the differential dependence of mechanical properties on matrix composition (Figure 11, Table 4). No dependence on collagen content was found for IL-1-stimulated explants for either |E*dyn| or |G*dyn|, reflecting the stable collagen contents. The depletion and subsequent recovery of sGAG in the IL-1 group was reflected in a strong dependence of |E*dyn| and |G*dyn| on sGAG/H2O. For control explants, sGAG/H2O and collagen//H2O were significant predictors of |G*dyn| but not of |E*dyn|.
The relationships between articular cartilage matrix composition and tissue mechanical properties were investigated in immature bovine explants during exhaustive IL-1-induced degradation and during transient IL-1 exposure and recovery. Detailed time courses of the biochemical and biophysical changes associated with persistent and transient proinflammatory stimulation were generated in a common model for studying cartilage matrix catabolism, and regression analyses revealed similar composition-function relationships in degrading and recovering cartilage. Significantly, |E*dyn| and |G*dyn| for the transiently stimulated IL-1 explants recovered to levels similar to control explants despite incomplete repopulation of the matrix with sGAG. The mechanical properties of untreated explants were generally dependent on both sGAG and collagen concentration, which is consistent with previous reports 2, 4, 6, 20, 45, 46. The physical properties of cartilage continuously and transiently stimulated with IL-1 were strongly dependent on sGAG content, with only the shear modulus of continuously stimulated cartilage significantly related to collagen content. These results underscore the importance of sGAG concentration in the maintenance of tissue mechanical function and lend support to therapeutic strategies aimed at promoting sGAG synthesis and diagnostic approaches to monitor degradation and recovery by tracking changes in sGAG/H2O.
The concurrent loss of sGAG and mechanical properties following IL-1 stimulation likely accounts for the strong correlation between cartilage sGAG concentration and mechanical properties. Further, it is not surprising that the mechanical properties exhibited minimal dependence on collagen content, since the loss of biomechanical function preceded significant collagen depletion in these experiments. Likewise, the recovery of explant mechanical properties following transient IL-1 exposure correlated well with increases in sGAG and water content while there was no change in collagen content. Direct enzymatic digestion of collagen with collagenases has been used to investigate the specific dependence of mechanical properties on collagen 46, but the results of this model are difficult to interpret given the nonspecific activity of collagenases on aggrecan47-49 and any role of the collagen network in retaining PG aggregates. The results presented here highlight the functional implications of sGAG depletion in the early stages of cytokine-induced cartilage degradation, but due to the experimental model yield little direct insight into the dependence on collagen content in normal or degraded cartilage.
Previous studies have described the loss of ECM and mechanical properties during persistent IL-1 stimulation of cartilage explants 14, 23, 24, 26, 42, 43, and PG content has been restored in cartilage explants following IL-1 insult in vitro and in vivo 34-37, 50. The results of this study demonstrate the chondrocyte's potential (albeit under in vitro culture conditions) to reestablish the tissue's mechanical function following partial matrix degradation and indicate that functional recovery of |E*dyn| and |G*dyn| can be achieved with incomplete repopulation of sGAG content. Recovered explants contained 40% of the sGAG of the control explants after 16 days of recovery (day 20) and equal levels of collagen content, yet |E*dyn| and |G*dyn| were not statistically different from those of the control explants. However, when sGAG concentration was calculated from the explant water content (to approximate FCD), no difference was noted between control and IL-1-stimulated explants at day 20. This finding demonstrates the utility of FCD (approximated by sGAG concentration) as an indicator of cartilage health and function.
Given the complex ECM organization and nonlinear mechanical behavior of articular cartilage, the linear regression analyses used here may not capture the true composition-function relationships. Indeed, Donnan and Poisson-Boltzmann models 51, 52 predict that osmotic swelling is a nonlinear function of PG concentration. The correlations described here may also be biased by the range of PGs and collagen content examined, particularly given the heteroscedasticity over a wide range of matrix density. The composition-function relationships may be further influenced by the choice of biochemical parameters (i.e., collagen, sGAG, water). Other studies describing composition-function relationships have included collagen cleavage products 15 and collagen crosslinks as ECM components 6, 7, 15, 53. Collagen crosslinks and cleavage products may be critical parameters to examine since the integrity of the collagen network cannot be evaluated by the hydroxyproline assay used in the present study. Despite the additional parameters and models that could be used to analyze the data, the significance and strength of the correlations reported here demonstrate the fundamental importance of sGAG to the compressive and shear properties of articular cartilage.
The composition-function relationships observed in untreated and IL-1-stimulated immature bovine cartilage may be specific to this tissue's age, species, and anatomic origin. Mature articular cartilage contains a significantly higher collagen fiber and collagen crosslink density 6, 54, lower PG content 54, shorter sGAG chain length 55-58, and fewer cells 59 than immature cartilage. IL-1-induced matrix degradation may be more aggressive in immature than mature cartilage due to the higher cell density, although the sensitivity of human articular cartilage to IL-1 stimulation appears to be independent of age 60. Variations in articular cartilage ECM composition and mechanical properties have also been reported with tissue depth 61, 62 and topographical location 7, 45. While the present study did not examine variations in composition-function relationships with topographical location, the use of immature middle zone cartilage minimizes variations due to tissue inhomogeneity through the tissue depth.
The strong dependence of all measured physical properties on sGAG concentration during IL-1-induced degeneration but prior to the onset of major collagen depletion suggests that measurements of sGAG concentration may be a useful surrogate for direct mechanical measurements in early stage degeneration. For example, imaging techniques such as dGEMRIC (delayed gadolinium-enhanced MRI of cartilage) 63 and EPIC-μCT (Equilibrium Partitioning of an Ionic Contrast agent via microcomputed tomography) 64 produce maps of tissue FCD that may provide sufficient information to noninvasively estimate the mechanical properties of injured or diseased cartilage in vivo. The application of composition-function relationships with these quantitative imaging techniques may also have utility in evaluating the competency of engineered cartilage grafts and resurfaced joints.
We thank Dr. Elsie Eugui and Dr. Fengrong Zuo for helpful conversations regarding these studies. This work was funded by an Arthritis Foundation Arthritis Investigator grant, the ERC program of the NSF under award number EEC-9731643 (Georgia Tech/Emory Center (GTEC) for the Engineering of Living Tissues), a graduate fellowship under NSF IGERT award number 0221600 (AWP), a graduate fellowship from the Cellular and Tissue Engineering Training Grant Program under NIH award number 5 T32 GM008433-13 (CGW), and by Roche Palo Alto. The study sponsors were not involved in the study design or execution or in the manuscript preparation.
Sources of Support: Arthritis Foundation Arthritis Investigator grant, the ERC program of the NSF under award number EEC-9731643 (Georgia Tech/Emory Center (GTEC) for the Engineering of Living Tissues), a graduate fellowship under NSF IGERT award number 0221600 (AWP), a graduate fellowship from the Cellular and Tissue Engineering Training Grant Program under NIH award number 5 T32 GM008433-13 (CGW), by Roche Palo Alto.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.