Preparation and evaluation of microfluidic magnetic alginate microparticles for magnetically templated hydrogels

Abstract

Our aim is to develop a hydrogel-based scaffold containing porous microchannels that mimic complex tissue microarchitecture and provide physical cues to guide cell growth for scalable, cost-effective tissue repair. These hydrogels are patterned through the novel process of magnetic templating where magnetic alginate microparticles (MAMs) are dispersed in a hydrogel precursor and aligned in a magnetic field before hydrogel crosslinking and subsequent MAM degradation, leaving behind an aligned, porous architecture. Here, a protocol for fabricating uniform MAMs using microfluidics was developed for improved reproducibility and tunability of templated microarchitecture. Through iron quantification, we find that this approach allows control over magnetic iron oxide loading of the MAMs. Using Brownian dynamics simulations and nano-computed tomography of templated hydrogels to examine MAM chain length and alignment, we find agreement between simulated and measured areal densities of MAM chains. Oscillatory rheology and stress relaxation experiments demonstrate that magnetically templated microchannels alter bulk hydrogel mechanical properties. Finally, in vitro studies where rat Schwann cells were cultured on templated hydrogels to model peripheral nerve injury repair demonstrate their propensity for providing cell guidance along the length of the channels. Our results show promise for a micro-structured biomaterial that could aid in tissue repair applications.

Graphical Abstract

1. Introduction:

A fundamental goal in the field of tissue engineering and regenerative medicine is developing solutions to restore, maintain, or improve function to damaged tissues.[1] Tissue damage can occur as a consequence of traumatic injuries, congenital defects, or disease. In the United States, 22 people die every day waiting for a transplant.[1] With a growing and aging population, shortages of donated organs, and long waitlists for transplants, there is an increasing clinical need for engineered solutions to tissue repair.

A common design strategy for recapitulating biological function is through biomimicry: imitation of native tissue in terms of mechanical properties, biological activity, and physical structure. Mimicking physical structure in particular is a key concern because, for specialized tissues in the body, structure and function are deeply interrelated -- tendon[2], bone[3], muscle (skeletal[4] and cardiac[5]), and nerve[6] all feature hierarchical organization that directly contribute to their function. Emulating this organization can provide physical cues critical for directing cell migration, organization, and differentiation.[7] Furthermore, an open, porous environment can also help improve diffusion of oxygen, nutrients, and growth factors to promote regeneration –particularly important in the repair of thicker tissues in need of vascularization.[8] Thus, mimicking intricate organ microenvironment is a rational design choice for ultimately restoring function to damaged tissues in situ.

Hydrogels are a popular choice for tissue repair applications because of their tunability, hydrophilicity, porosity, polymeric nature, and capacity for incorporation of biological factors – properties that allow them to mimic the extracellular matrix (ECM).[9] A wide variety of strategies have been explored for patterning hydrogels with three-dimensional microarchitecture – examples include injection molding[10], electrospinning[11], template leaching[12, 13], self-assembly[14-16], thermally induced phase separation[17], and direct writing techniques such as photoablation[18], multiphoton-based methods[19, 20], and 3D-printing[21, 22]. Nevertheless, developing cost-effective, versatile methods of patterning three-dimensionally oriented cell-scale structures on clinically relevant length scales is still a challenge.

The application of magnetic fields allows for a novel strategy of creating self-assembled, patterned biomaterials. The diamagnetic properties of ECM components such as collagen[23, 24] and fibrin[25] have been employed to fabricate magnetically aligned scaffolds under very strong magnetic fields. Others make use of superparamagnetic nanomaterials or nano-composites to align cues in the presence of weaker, more easily generated magnetic fields. Kim et al. and Antman-Passig et al. used aligned magnetic nanoparticles embedded in matrices to serve as physical cues in vitro[26, 27]. To form physical cues on longer length-scales, Rose et al. used magneto-responsive poly(ethylene glycol) (PEG) microgels aligned in fibrin-based and PEG-based matrices.[28, 29]

We introduce the novel technique of magnetic templating where sacrificial, aligned magnetic microparticles are used to create three-dimensionally oriented, aligned pores within a hydrogel matrix. Unlike techniques which directly incorporate magnetic materials, the magnetic templating process uses negative subtraction of magnetic materials to produce porous microarchitecture. Magnetic alginate microparticles (MAMs), which are composed of iron oxide (IO) nanoparticles embedded in calcium alginate gels, are first dispersed in a hydrogel precursor solution (Fig 1a). The MAMs are then aligned into chains using a uniform magnetic field (Fig 1b) before the hydrogel is crosslinked (Fig 1c). The magnetically templated hydrogel is then processed with ethylenediaminetetraacetic acid (EDTA) for MAM removal (Fig 1d), leaving behind aligned, porous channels (Fig 1e). This technique allows for the formation of cell-scale microchannels in hydrogels in a manner that is cost-effective, scalable, amenable to the incorporation of cells or bioactive factors, and capable of patterning on clinically relevant centimeter length scales. In this work, we employ magnetic templating to pattern hydrogels with microchannels to mimic the hierarchical structure of nerve to examine its potential for peripheral nerve repair applications. However, based on the magnetic field line patterns used for MAM alignment, magnetic templating also has the capability to create even more complex, porous patterns.

Figure 1.

a) Mold is loaded with precursor hydrogel solution mixed with magnetic alginate microparticles (MAMs). b) Application of a static, uniform magnetic field allows for MAM chain formation. c) Hydrogel is photocrosslinked using UV light at a wavelength of 365 nm. d) MAMs degraded by soaking sample in ethylenediaminetetraacetic acid (EDTA). e) Hydrogel patterned with aligned macroporous channels.

For our hydrogel matrix, we have chosen a photocrosslinkable glycidyl methacrylate hyaluronic acid (GMHA). Hyaluronan is a popular material in tissue engineering due in part to its presence as an ECM component and its capability for chemical modification.[30] Furthermore, hyaluronan is upregulated during wound-healing[31] and engages with CD44 receptors which are involved in cell migration and cell-matrix interactions[32]. For improving cell adhesion in in vitro experiments, we have also incorporated collagen I to form an interpenetrating network with GMHA, as has been done previously by Suri et al. for neural tissue applications.[32]

Magnetic alginate beads and particles have been reported mainly for drug delivery[33-36], separation applications[37, 38], and cell spheroid formation.[39] Common fabrication methods include use of emulsions, extrusion, and microfluidics. Emulsions[40, 41] provide a method for producing large amounts of these particles, but this top-down approach can have issues with batch-to-batch variability and non-uniformity in terms of size and IO loading. Extrusion of alginate precursor into crosslinking solution[35, 37] and T-junction microfluidic setups[36] have been used to fabricate large magnetic alginate beads, generally ranging in size on the orders of magnitude of 100 μm to 1 mm; however, smaller magnetic alginate particles (<100 μm) have been formed using microfluidics flow focusing orifices for improved droplet control.[33, 34, 42] For the application of magnetic templating, we are interested in fabricating uniformly-sized MAMs on the order of magnitude of tens of micrometers in size to mimic cell-scale topology; thus, we have opted for the use of a flow-focusing microfluidic device.

In this work, we aim to demonstrate the capability of the magnetic templating process for patterning hydrogels and providing contact guidance for cells in vitro, using peripheral nerve repair as a model application. Furthermore, microfluidic flow-focusing was utilized for MAM fabrication to improve uniformity in physical size and composition as the design of these sacrificial templates has a large impact on chain formation, processing, and ultimate patterned microarchitecture.

2. Materials and Methods

2.1. Iron Oxide Nanoparticle Synthesis

Iron oxide (IO) nanoparticles were prepared through co-precipitation as described previously.[43] Deionized water was deoxygenated with nitrogen for 30 minutes and 0.1 M iron (II) chloride tetrahydrate [Sigma, 236489-500G] and 0.2 M iron (III) chloride hexahydrate [Sigma, 220299-250G, Sigma]. Iron solutions were sonicated for 20 minutes, then deoxygenated once more for 5 minutes, before being mixed in a 500 ml glass reactor. The solution was heated to 75 °C before addition of ammonium hydroxide [Sigma, 4126318-1L] and a one-hour reaction at 85 °C. The resultant iron oxide colloid was cooled to room temperature, poured into conical tubes, centrifuged at 1500 rpm for 10 minutes, then magnetically decanted. To render them stable in water, the IO nanoparticles were peptized with 25% w/w tetramethylammonium hydroxide (TMAOH) [Fisher, A669-212]. TMAOH was added in a 1:1 volumetric ratio to the volume of colloid initially poured to each conical tube. An ultrasonicator with a 1” diameter horn was used to disperse the nanoparticles for 30 minutes at 80% power. The nanoparticles were centrifuged down at 1800 rpm for 10 minutes before repeating the ultrasonication process. The nanoparticles were then centrifuged at 2500 rpm for 10 minutes, collected into a glass jar with a magnet, and then left to air dry overnight. The volume-weighted hydrodynamic diameter of the iron oxide nanoparticles used was approximately 30 nm

2.2. Glycidyl Methacrylate Hyaluronic Acid Synthesis

GMHA was prepared as described previously.[44]A mass of 500 mg of hyaluronic acid, sodium salt [Sigma, 53747] was dissolved at 10 mg/ml in 50% v/v acetone [Fisher, A18-20] in water at room temperature. The sample was incubated with 6.7% triethylamine [Sigma, T0886] for 30 min, then 6.3% glycidyl methacrylate [Sigma, 779342] for 24 hours. The methacrylated hyaluronic acid was then precipitated in a 20X volumetric excess of acetone and re-dissolved into 50 ml of water; precipitation and subsequent dissolution of the product into 50 ml of water was repeated once more. The sample was then purified through dialysis against phosphate buffer saline (PBS) for two days then against water for one day in a 10 kDa molecular weight cutoff dialysis cassette. The methacrylated hyaluronic acid was then sterile filtered and lyophilized for 7 days before use.

2.3. Magnetic Alginate Microparticle Fabrication

A 100 μm etch depth, fluorophilic 3D flow focusing cell [Dolomite, 3200515] was used to produce the droplets that served as precursors to the MAMs. First, the droplet phase was prepared with a composition of 10 mg/ml alginate [Sigma, 71238], 6.25 mM of EDTA-chelated calcium complexes [Fisher C79-500, Sigma EDS-100G], and the desired concentration of ferrofluid. For all experiments, unless otherwise specified, the IO concentration used for the droplet phase was 75 mg/ml. The droplet phase was sonicated for 20 minutes then filtered using a 0.22 μm pore size polyvinylidene fluoride filter. The continuous phase with composition of 1% v/v Pico-surf surfactant [Dolomite, 3200214] in HFE-7500 fluorocarbon oil [Oakwood Chemical, 051243] was prepared. The flow rates used for the droplet and continuous phases were 10 μl/min and 40 μl/min, respectively. Droplets were collected in a vial containing 1 ml of HFE-7500. The collection vessel was mechanically agitated with a small stirrer [Caframo, CG-2037-C-250] operating at 600 rpm throughout droplet production. After droplet production was complete, a volume of fluorocarbon oil with 0.1% v/v acetic acid was added for a final concentration of 0.05% v/v acetic acid [Acros Organics, AC222142500]. Gelation was allowed to proceed for 15 minutes after which fluorocarbon oil was siphoned from the sample using a syringe. The sample was then rinsed with approximately 1.5 ml of acetone and magnetically decanted three times. The sample was additionally crosslinked with 200 mM calcium chloride solution before being re-dispersed in water and stored at 4°C before use. Large MAM agglomerates were manually removed from the final product to obtain a monodisperse product. Losses to the final sample were estimated and accounted for by measuring the lyophilized weight of the separated, agglomerated MAMs.

2.4. Hydrogel Fabrication

Precursor hydrogel solutions with 20 mg/ml GMHA, 0.3% v/v Irgacure I2959 photoinitator (Ciba Specialty Chemicals, 55047962), and the desired volume fraction of MAMs were prepared. For all experiments, unless otherwise specified, a volume fraction of 2.2% MAMs was used to prepare magnetically templated hydrogels. Solution was then injected into silicon molds – molds with dimensions 32 x 3 x 1.7 mm [Grace Bio-Labs, 664201] and 9 mm diameter x 2.5 mm height [Grace Bio-Labs, 622203] have been used throughout this study. Magnetically templated samples were then placed in a uniform magnetic field for 20 to 30 minutes before being photocrosslinked under a 365 nm wavelength UV light for 10 minutes. The MAMs were then cleared from the magnetically templated hydrogels using 0.1 M EDTA [Fisher, SS412-1] at 37°C for one to seven days.

For cell culture experiments, 1.5 mg/ml rat tail collagen I [Corning, 354249] was incorporated into the hydrogel precursor solution in addition to the above listed components. Magnetic alignment was conducted at 4°C for 30 min before 10 minutes of photocrosslinking and finally incubation at 37°C for 35 minutes to complete collagen fibrillogenesis.

2.5. Equilibrium Magnetization Measurements

A Quantum Design magnetic property measurement system 3 (MPMS 3) superconducting quantum interference device (SQUID) magnetometer was used to measure the equilibrium magnetization each microsphere sample, normalized by sample mass. The field was varied from −70,000 Oe to 70,000 Oe, and the temperature was maintained at 300 K. Vacuum dried MAM samples were mounted in gelatin capsules and immobilized with polytetrafluoroethylene tape. The saturation magnetization of each sample was divided by that of bulk magnetite (86 emu/g) to estimate the mass fraction of iron oxide.

2.6. Nano-Computed Tomography Scanning and Reconstruction

A GE |TOME|X M 240 computed tomography scanner equipped with a nano-focus x-ray tube (nanoCT) was used to image magnetically templated hydrogels. Scans were conducted at a voltage of 85 kV, a current of 55 μA for a power at 4.7 W. Voxel sizes for scans performed varied from 4 to 6 μm. Post-processing of reconstructed scans was conducted using VGStudio Max. All reconstructed scans were processed using a Gaussian filter, thresholded to attenuate low density regions, and pseudo-colored green for improved visualization.

2.7. Image Analysis

ImageJ was used to conduct particle analysis on scanning electron and brightfield micrographs of MAMs to obtain size distributions. A total of 720 particles were examined from brightfield micrographs, and a total of 84 particles were manually examined from scanning electron micrographs. Particle counts were also conducted in ImageJ on nanoCT tomograms. Six cross-sectional tomograms evenly spaced throughout the length of the scanned templated hydrogels were chosen and cropped to 0.26 cm by 0.18 cm sized rectangles. The cropped tomograms were subsequently thresholded and manually counted.

2.8. Magnetic Alginate Microparticle Alignment Simulations

MAM chain assembly was studied through conducting Brownian dynamics simulations, accounting for translation and rotation of the MAMs, hydrodynamic drag, thermal fluctuations, and interactions between the MAMs. Since a MAM is loaded with IO nanoparticles which are physically fixed in space and respond to changes in external magnetic field through the Néel relaxation mechanism, they can be simplified to a micron-sized rigid dipole model with an effective magnetic dipole moment that sums up the instantaneous dipole moments of the iron oxide nanoparticles. The Néel relaxation time is much smaller than the characteristic time for MAM translation and rotation, so we assumed that the effective dipole instantaneously aligns with the direction of the local magnetic field. The magnitude m_eff is given by the Langevin function L(α)

$$\frac{m_{\text{eff}}}{m_{\mathrm{s}}}=\text{coth}\alpha -\frac{1}{\alpha }=L(\alpha ).$$ (1)

where m_s = ϕ_NPM_dV_MAM, ϕ_NP represents the volume fraction of the IO nanoparticles in MAM, M_d represents the domain magnetization of iron oxide, the volume of the spherical MAM is V_MAM = 4/3πa³, a represents the MAM radius. The Langevin parameter, α, is defined by α = μ₀m_sH/k_BT where μ₀ represents the permeability of free space, H represents the magnitude of the applied magnetic field, k_B is the Boltzmann constant, and T is the absolute temperature.

When external magnetic fields are applied to the MAMs, the forces and torques acting on each dispersed MAM include those due to hydrodynamic drag F_h and T_h, due to external magnetic field F_m and T_m, due to magnetic dipole-dipole interactions F_dd and T_dd, and force due to inter-MAM repulsion F_rep. Thus, the linear and angular momentum balances reduce to force and torque balances, in the form of

$$\left(\begin{array}{c}\hfill 0\hfill \\ \hfill 0\hfill \end{array}\right)=\left(\begin{array}{c}\hfill {\mathbf{F}}_{\mathrm{h}}\hfill \\ \hfill {\mathbf{T}}_{\mathrm{h}}\hfill \end{array}\right)+\left(\begin{array}{c}\hfill {\mathbf{F}}_{\mathrm{m}}\hfill \\ \hfill {\mathbf{T}}_{\mathrm{m}}\hfill \end{array}\right)+\left(\begin{array}{c}\hfill {\mathbf{F}}_{\text{dd}}\hfill \\ \hfill {\mathbf{T}}_{\text{dd}}\hfill \end{array}\right)+\left(\begin{array}{c}\hfill {\mathbf{F}}_{\text{rep}}\hfill \\ \hfill 0\hfill \end{array}\right).$$ (2)

It should be noted that thermal fluctuations were neglected due to the relatively large size of the MAMs. For a MAM suspended in a motionless hydrogel precursor solution, the hydrodynamic force F_h and torque T_h exerted on the it can be related to its velocity and angular velocity through

$$\left(\begin{array}{c}\hfill \mathbf{U}\hfill \\ \hfill \omega \hfill \end{array}\right)=-\mathcal{M}\cdot \left(\begin{array}{c}\hfill {\mathbf{F}}_{\mathrm{h}}\hfill \\ \hfill {\mathbf{T}}_{\mathrm{h}}\hfill \end{array}\right).$$ (3)

where $\mathcal{M}$ is the symmetric and positive defined mobility matrix Since the applied magnetic field has no field gradients and the MAMs cannot physically rotate due to the magnetic relaxation, it is justified that F_m = 0, T_m = 0 and T_dd = 0. The magnetic dipole-dipole force exerted by MAM j on MAM i is given by

$${\mathbf{F}}_{\text{dd},ij}=\frac{3{\mu }_0{m}_i{m}_j}{4\pi r_{ij}^4}[{\widehat{\mathbf{r}}}_{ij}({\widehat{\mathbf{m}}}_i\cdot {\widehat{\mathbf{m}}}_j)+({\widehat{\mathbf{m}}}_j\cdot {\widehat{\mathbf{r}}}_{ij}){\widehat{\mathbf{m}}}_i+({\widehat{\mathbf{m}}}_i\cdot {\widehat{\mathbf{r}}}_{ij}){\widehat{\mathbf{m}}}_j-5({\widehat{\mathbf{m}}}_i\cdot {\widehat{\mathbf{r}}}_{ij})({\widehat{\mathbf{m}}}_j\cdot {\widehat{\mathbf{r}}}_{ij}){\widehat{\mathbf{r}}}_{ij}].$$ (4)

where the vacuum permeability μ₀ = 4π × 10⁻⁷N/A², m_i and m_j represent the magnitude of effective dipole moment of MAMs j and i, respectively, r_ij represents the separation between two MAM centers, ${\widehat{\mathbf{r}}}_{ij}$ represents the unit vector of center distance, and ${\widehat{\mathbf{m}}}_i$ and ${\widehat{\mathbf{m}}}_j$ are the unit vectors that represent the orientations of MAMs j and i, respectively. In current work, we assumed that ${\widehat{\mathbf{m}}}_i$ and ${\widehat{\mathbf{m}}}_j$ are always in the direction of the magnetic field and m_i = m_j = m for the uniform MAM models. The repulsive force between the MAMs was modeled in terms of the repulsive part of the Lennard-Jones potential, which is given by

$${\mathit{u}}_{\mathrm{L}-\mathrm{J},\text{repulsive}}(r_{ij})=4\epsilon {\left(\frac{\sigma }{r_{ij}}\right)}^{12}.$$ (5)

where ε is the depth of the potential well, and σ represents the MAM separation that gives zero potential energy, i.e., the effective MAM diameter. The force exerted by MAM j on i due to the repulsive Lennard-Jones potential then is obtained from

$${\mathbf{F}}_{\text{rep},ij}=-\nabla u_{\mathrm{L}-\mathrm{J},\text{repulsive}}(r_{ij}).$$ (6)

By conducting a similar derivation and non-dimensionalization as in Zhao’s work[45], the translational and rotational variance of MAM i within the dimensionless minimum time step $\Delta \tilde{t}$ can be obtained. The strengths of the magnetic dipole-dipole interactions and repulsive Lennard-Jones potential are tuned through β_dd = μ₀m²/(a³πk_BT) and $\tilde{\epsilon }={}^{\epsilon }∕(k_{\mathrm{B}}T)$, respectively, where k_B is the Boltzmann constant and T represents the absolute temperature.

Simulations were made for uniform spherical MAMs that were in the diameter of D_MAM = 60 μm and dispersed in a hydrogel precursor solution. The volume fraction of iron oxide nanoparticles in the MAM was ϕ_NP = 14 %. The simulation box is 1.5 mm × 1.5 mm × 6 mm with periodic boundaries. The temperature in the simulations corresponds to 300 K. Corresponding to the balance between the attractive and repulsive forces, we chose σ = 1.8a and $\tilde{\epsilon }=3.82$. Runs were carried out starting from random MAM configurations and using a minimum time interval Δt = 1.64 × 10 ⁻³ s. The total simulation time was 16.39 s. The MAM number per slice was calculated by averaging the MAM number within the sample slices, which were perpendicular to the long side and in the volume 1.5 mm × 1.5 mm × 0.03 mm. The areal density was obtained by dividing the MAM number per slice by the slide area 1.5 mm × 1.5 mm. It should be noted that in this method the single MAMs, which were not in any chains, might also be counted to calculate the areal density.

2.9. Iron Quantification

MAM samples were vacuum-dried, weighed, then degraded with 0.1 M EDTA. Hydrogel samples of uniform volume were weighed and degraded in 50 units/mL hyaluronidase solution in PBS. Iron quantification was conducted through an o-phenanthroline assay, as described previously.[41] Briefly, For each of these aqueous samples, 10-100 μl aliquots were taken in triplicate and further digested in 1 ml of 70% nitric acid [Fisher, A467-2] at 101°C for 12 hours. After digestion, 10-100 μl aliquots were taken from each sample, placed into 1.5 ml glass vials, and heated to 115°C for 1 hour to evaporate all liquid. Samples were re-suspended in 46 μl of deionized water, and to reduce the iron from Fe⁺³ to Fe⁺², 30 μl of 8.06 M hydroxylamine [Sigma, 431362-50G] was added to each vial. After 1 hour, aliquots of 49 μl of 1.22 M sodium acetate [Sigma, S8750-500G] and 75 μl of 13 mM 1,10-phenanthroline [Sigma, 77500-25G-F] were added to each vial, allowing for the formation of iron (II)-orthophenanthroline complexes. The absorbance values were read at 508 nm using a SpectraMax® M5 Multimode plate reader. Using a calibration curve prepared from a dilution series of iron standards [Sigma, 56209-100ML] as the stock, the concentration of the iron was related to the absorbance value, allowing the iron concentrations of the unknown samples to be quantified.

2.10. Scanning Electron Microscopy

A Hitachi SU-5000 FEG scanning electron microscope (SEM) equipped with an ultra-variable pressure detector was used to image hydrated MAM and hydrogel samples in low vacuum mode with a voltage of 5.0 kV and a pressure of 10 Pa at x70 magnification.

2.11. Microchannel visualization with Fluorescent Dextran

Templated hydrogels were soaked for overnight in a solution of 1 mg/ml 500 kDa fluorescein isothiocyanate-dextran (FITC-dextran) [Sigma, 46947] in PBS. The samples were then soaked in PBS for 1 hour to rinse away excess FITC-dextran before being imaged using a Zeiss LSM 710 laser-scanning confocal microscope.

2.12. Oscillatory Rheology

Oscillatory rheological measurements were made through an Anton Paar MCR 302 rheometer using an 8 mm diameter parallel plate geometry with sandblasted surfaces to minimize slipping and an environmental chamber to prevent dehydration of the samples. An amplitude sweep from 0.01 to 100% strain was performed on hydrogel samples at a constant frequency of 6.3 rad/s (~1 Hz) (n = 3 from the experimental group of interest). For all measurements, temperature was maintained at 37°C.

2.13. Stress Relaxation Measurements

Stress relaxation measurements were conducted using a Bruker BioSoft in situ indenter. Samples were indented with a 3 mm spherical glass tip 150 μm into the surface and held at a constant strain (up to 9.03%) for 90 s while force relaxation data were collected. Three different locations on each of the four samples per experimental group (n = 4) were indented. As described by Stewart et al.,[46] the relaxation data were converted into a time-dependent relaxation modulus using the Hertz contact model accounting for a parabolic contact region. The standard linear solid model for viscoelasticity was then used to fit the relaxation data and ultimately obtain the reported steady-state relaxation modulus.

2.14. In Vitro Culture of Rat Schwann Cells on Hydrogels

Hydrogel samples were cut with 6 mm biopsy punch [Integra, 3336] before being placed into 24-well plate Transwell® inserts [Corning, CLS3470]. All samples were subjected to a 5-day soak in 0.1 M EDTA. The EDTA was subsequently removed by equilibrating the samples in PBS for 3 hours. The samples were finally equilibrated in complete rat Schwann cell (RSC) media (10% fetal bovine serum [Life Technologies, 10438026], 1% v/v penicillin-streptomycin-amphotericin B stock [MP Biomedicals, 091674049], 20 μg/ml bovine pituitary extract [Gibco, 13028014], 4 μM forskolin [Sigma-Aldrich, F3917], and 10 ng/ml fibroblast growth factor [Gibco, PHG0264] in Dulbecco’s Modified Eagle’s medium [Corning, MT10013CV]).

Passage 5 RSCs (Sciencell, R1700) seeded on the hydrogel samples at a seeding density of 1.25 × 10⁴ cells/cm². Complete RSC media was used in the top compartment of the Transwell® setup while complete RSC media with varying concentration of nerve growth media (NGF) [R&D Systems, 556-NG] was used in the lower compartment to create a concentration gradient of NGF. Media changes were conducted every two days, and the RSCs were cultured for seven days total before fixation with 4% paraformaldehyde for 1 hour.

Samples were incubated with blocking buffer (0.3% goat serum, 0.3% Triton X-100 in PBS) for 1 hour at room temperature. Primary antibody (1:400 dilution of rabbit anti-S100 [Sigma-Aldrich, S2644] in PBS) was applied for two days at 4°C. Samples were soaked in 0.05% v/v Tween 20 in PBS to remove excess primary Next, samples were incubated with secondary antibody (1:500 dilution of goat anti-rabbit Alexa Fluor 568 [Abcam, 175473]) for 1 day; excess secondary antibody rinsed away with 0.05% v/v Tween 20 in PBS. Nuclei were stained with a 1:1000 dilution of Hoescht 33342 [Life Technologies, H1399] in PBS for 15 minutes before being washed with PBS for 30 minutes.

Samples were placed cell-side-down in a 35 mm diameter glass-bottom dish and imaged using a Zeiss LSM 710 laser-scanning confocal microscope with image stacks 500 μm deep with pitch of 7.5 μm at 20x magnification.

2.15. Statistics

Analysis of variance (ANOVA) was conducted on steady-state relaxation moduli data to determine differences between experimental groups, followed by post-hoc Tukey’s honestly significant difference analyses in JMP Pro 14 using a significance level of α = 0.01. All data are presented as averages with standard deviations.

3. Results and Discussion

3.1. Control of Size and Iron Oxide Content in Magnetic Alginate Microparticles

The use of microfluidic flow-focusing was motivated by its ability to allow fabrication of MAMs that are highly uniform in terms of size and IO content: two parameters that have a strong impact MAM chain formation and microchannel formation.

Representative images taken with low vacuum SEM (Fig. 2a) and brightfield microscopy (Fig. 2b) demonstrate their rounded morphology. Image analysis on these micrographs was used to further characterize the size distribution of the MAMs (Figs. 2c-d). The average diameter measured from scanning electron micrographs was 54.4 ±10.1 μm (as measured from 84 MAMs), and the average diameter measured from brightfield images was 68.6 ± 15.4 μm (as measured from 720 MAMs). We believe that the difference from these two measured values arises from lower levels of hydration with consequent drying and shrinking in the low-pressure environment of the SEM – in contrast, the MAMs were fully hydrated and swollen for brightfield imaging. To confirm that the microfluidic process also enabled control over IO loading, MAMs with varying concentrations of IO in the precursor droplet phase (18.75 mg/ml, 37.5 mg/ml, and 75 mg/ml of iron oxide) were prepared. These measured for their saturation magnetization using a SQUID magnetometer and their iron content with the o-phenanthroline assay.

Figure 2.

Representative images of MAMs fabricated via microfluidic flow focusing taken with a) SEM and b) brightfield microscopy alongside histograms of MAM diameters collected from c) SEM and d) brightfield microscopy along with average MAM diameter measured from each (D_avg).

Fig. 3a shows the theoretical mass fraction of the samples based on the masses of alginate and IO concentrations used in the droplet phase, the mass fraction of iron as measured by the o-phenanthroline assay, and the mass fraction of IO as estimated by using the saturation magnetization from magnetometer measurements. The magnetization curves for MAMs prepared with varying IO loadings are plotted in fig. 3b-d. Discrepancies between theoretical and experimentally measured values from the o-phenanthroline assay may be attributed to potential non-homogeneity in the digestion process. The IO mass fractions as measured by the SQUID magnetometer trend lower than the theoretical mass fraction – this could potentially be due to particle-particle interactions in the samples with higher IO loading. The low standard deviations for these measurements are indicative of the reproducibility of the process. Overall, we observe a trend where increases in IO loading correspond to increases in measured IO content, demonstrating the ability of the MAMs to appropriately encapsulate IO nanoparticles during the fabrication process.

Figure 3.

a) Theoretical mass fractions, mass fractions of iron determined by the o-phenanthroline absorbance assay, and mass fractions of IO determined by using the saturation magnetization from SQUID measurements for three IO loadings. b-d) Magnetization vs. field strength curves measured using a SQUID magnetometer, for three samples each, for the three IO loadings studied: 18.75 mg/ml, 37.5 mg/ml, and 75 mg/ml, respectively. e-g) Pseudo-colored, high density thresholded nanoCT scans for three hydrogels magnetically templated with MAMs with varying IO loadings (18.75 mg/ml, 37.5 mg/ml, and 75 mg/ml, respectively) prior to MAM degradation. As IO loading increases, MAM alignment qualitatively improves.

We studied the effect IO loading on the magnetic templating process by templating hydrogels with MAMs prepared with varying IO concentrations. NanoCT scans, using the high-density IO nanoparticles embedded in the MAMs as contrast agents against the surrounding low-density hydrogel, allowed us to create three-dimensional reconstructions of the full-thickness of magnetically templated hydrogels still containing MAMs, providing a complete visualization of MAM chain organization. Thus, we took nanoCT scans of templated hydrogels prepared with MAMs of varying IO density while maintaining a constant volume fraction of 2.2% v/v MAMs. Longitudinal maximum projections are shown in fig. 3e-g, demonstrating that as IO loading increases, alignment also visibly increases. This trend is to be expected as increased IO loading correlates with a larger net magnetic moment for each individual MAM – this improves the ability of the MAMs to respond and align to magnetic field lines via dipole-dipole interactions. For our current application of peripheral nerve repair, we were interested in mimicking the long, well-aligned, and highly organized, anisotropic structure of native nerve.[6] Furthermore, a high degree of alignment could also be desirable for other tissue engineering applications—for instance, skeletal muscle repair.[4] Consequently, for all following experiments, the highest IO loading tested (75 mg/ml) was used for magnetically templating hydrogels to obtain MAM chains and consequently microchannels that were long and well-aligned.

Thus, we were able to demonstrate that microfluidic flow-focusing permits formation of MAMs uniform in size and composition. Such uniformity in both the magnetic templating process and final product is vital for future clinical translation and performance.

3.2. Characterization of Magnetic Alginate Microparticle Chain Assembly

Fig. 4 shows longitudinal and cross-sectional brightfield microscopy images of magnetically templated hydrogels prior to MAM clearance at three representative volume fractions. As volume fraction of MAMs in the sample increases, qualitatively so does chain length and areal density. While large agglomerates of MAMs were removed from samples before use in magnetic templating, smaller aggregates such as doublets and triplets were still present. While these smaller aggregates did disrupt the linearity and uniformity of the MAM chains, for the most part, non-agglomerated MAMs visibly aligned in single file chains.

Figure 4.

Light microscopy images of longitudinal and cross-sectional views of templated hydrogels with varying volume fractions. As volume fraction increases, areal density qualitatively increases. (Scale bar, 500 μm)

To gain more information regarding the density of MAM chains, we used both nanoCT scans of magnetically templated hydrogels along with dynamics simulations of the MAM chain formation process. Fig. 5 shows both longitudinal maximum projection images of high-density-thresholded nanoCT scans and cross-sectional tomograms of magnetically templated hydrogels prepared with varying volume fractions of hydrogels side-by-side with the results of MAM dynamics simulations. From conducting a particle count on cross-sectional tomograms from each sample, we estimated an areal density of chains. As MAM volume fraction increases, so does areal density of chains. The areal density of MAM chains measured from tomograms closely matches those calculated from the simulations, demonstrating that these simulations can be used to better understand how the many parameters involved in the magnetic templating process affects MAM chain and channel formation.

Figure 5.

Shown are dynamics simulations of MAM alignment, maximum projection longitudinal views of pseudo-colored, high density thresholded CT scans of three templated hydrogels, and pseudo-colored, high density thresholded cross-sectional tomograms from the three representative CT scans for three different volume fractions of MAMs (1.1%, 2.2%, and 4.4%). The dynamics simulations agree with the CT scans in terms of areal density and the trend of increasing areal density with increasing MAM volume fraction.

3.3. Magnetic Alginate Microparticle Degradation and Clearance from Templated Hydrogels

Since the MAMs are composed of calcium-crosslinked alginate, the chelating agent EDTA was used to process the templated hydrogels – EDTA breaks down the MAMs by chelating the calcium ions, allowing the alginate and IO nanoparticles to clear the templated hydrogels through diffusion. The EDTA also likely chelates the iron in the IO nanoparticles themselves, further accelerating the process. To confirm that the iron is being cleared from the templated hydrogels, the o-phenanthroline iron quantification assay was used to track iron content throughout the EDTA degradation process.

Fig. 6a shows sections of templated hydrogels at various time points during the EDTA degradation process: prior to placement in EDTA (0 hours), 3 hours, 6 hours, 18 hours, and 24 hours. As seen in the image, the aligned MAMs appear dark but turn lighter brown as they eventually break down and clear out of the hydrogel. Through the images, we can see that the IO content of the templated hydrogels drops significantly in the first six hours of the degradation process. This is confirmed through iron quantification for the mass of iron in each templated hydrogel using the o-phenanthroline assay. More than 93% of the original iron content of the templated hydrogels is removed within 24 hours of the MAM dissolution process. Based on the limit of detection for this study and an assumed volume of 100 μl for each hydrogel section, less than 25 μg/ml of iron (the calculated limit of detection for the o-phenanthroline assay performed here) is present in the templated hydrogels after 24 hours of MAM degradation (fig. 6b). According to Pisanic et al., 830 μg/ml of iron was found to be neurotoxic while 83 μg/ml of iron was not.[47] Dunning et al. observed that incubating dextran-coated IO nanoparticles with Schwann cells at concentrations up to 4 mg/ml had no effect on cell viability.[48] Huang et al. observed statistically significant Schwann cell death for concentrations of poly-L-lysine-coated IO nanoparticles only at 50 μg/ml and higher in vitro.[49] For more general biomedical applications, iron oxide nanoparticles are widely considered biocompatible and safe.[50] Hence, the residual iron in the templated hydrogels is not expected to cause toxicity to cells during PNI or broader tissue repair applications.

Figure 6.

a) Images of a magnetically templated hydrogel throughout the MAM clearance process. b) Mass of iron per 1-cm-length hydrogel section as measured using the o-phenanthroline iron quantification assay as a function of time. Over 93% of the original iron content is removed through the EDTA clearance process, leaving less than 25 μg/ml (limit of detection for the assay) of iron in the templated hydrogels.

3.4. Characterization of Microchannel Formation

The outlines of templated channels are faintly visible when imaged using brightfield microscopy (fig. 7a). To better visualize the microarchitecture within the templated hydrogel, the hydrogel was soaked in high molecular weight FITC-dextran and subsequently imaged using confocal microscopy (fig. 7b). The visible channels maintain the distinct shape and contours of the sacrificial MAM chains. Fig. 7c shows a low vacuum SEM image of a magnetically templated hydrogel with the channels oriented into the page. Visible are distinctly hollow openings in the surface of the hydrogel, further indicating that the channels maintain a macro-porous structure after magnetic templating and MAM clearance.

Figure 7.

a) Light microscopy image of magnetically templated hydrogel after MAM degradation. b) Confocal microscopy image of FITC-dextran backfilled templated hydrogel. c) An SEM image of a magnetically templated hydrogel with the channels going into the page, demonstrating that the channels are hollow and maintain shape after the magnetic templating process and MAM clearance.

3.5. Oscillatory Rheology

Oscillatory rheology measurements were used to probe the viscoelastic properties of the hydrogels as a function of strain. The experimental groups examined were non-templated hydrogels (hydrogels formed without incorporation of MAMs), magnetically templated hydrogels prior to MAM clearance, and magnetically templated hydrogels after MAM clearance The non-templated hydrogels were fabricated using 20 mg/ml GMHA and 0.3% 12959, whereas templated hydrogels were fabricated using 20 mg/ml GMHA, 0.3% 12959, and 2.2% v/v MAMs. The magnetically templated hydrogels were fabricated such that chain/channel orientation was parallel to the axis of rotation in the oscillatory rheology setup.

Fig. 8 shows the storage modulus for each of the experimental groups as a function of strain. In general, the processed, templated hydrogels have storage moduli that are lower than the non-templated and unprocessed, templated hydrogels. This decrease in elastic, solid-like properties can be attributed to the introduction of porous channels into the hydrogel.

Figure 8.

Storage modulus as a function of strain measured through oscillatory rheology for three different experimental groups: 1) non-templated hydrogels, 2) templated hydrogels before MAM removal, and 3) templated hydrogels after MAM removal. Templated hydrogels after degradation are characterized by a decreased storage modulus.

3.6. Stress Relaxation Measurements

Stress relaxation measurements allowed further characterization of the mechanical properties of the hydrogels under non-oscillatory conditions. The experimental groups examined are illustrated in fig. 9: non-templated hydrogels (fig. 9a), hydrogels with templating parallel to the indentation tip (fig. 9b), hydrogels with templating parallel to the indentation tip (fig. 9c), and hydrogels with unaligned templating that were not subjected to a magnetic field (fig. 9d). As done for the oscillatory rheology measurements, the non-templated hydrogels were fabricated using a precursor solution of 20 mg/ml GMHA and 0.3% 12959, while templated hydrogels were fabricated using a precursor solution of 20 mg/ml GMHA, 0.3% 12959, and 2.2% v/v MAMs. All templated hydrogel groups were examined before and after MAM removal. A comparison of the steady-state relaxation modulus across the different experimental groups is shown in fig. 9e. ANOVA and post-hoc Tukey’s tests were conducted to analyze the data – a table laying out relationships of significance for each of the experimental groups can be found in fig. S1.

Figure 9.

a-d) Pictured are schematics depicting the experimental groups examined in stress relaxation experiments: hydrogels that are non-templated, hydrogels templated parallel to the direction of loading with the indenting tip, hydrogels templated perpendicular to the direction of loading, and hydrogels with unaligned templating. e) Steady-state relaxation modulus data for hydrogels templated under varying conditions obtained from fitting stress relaxation data -- All groups that are labeled with different letters are significantly different at α = 0.01. The templating process results in a significant drop in steady-state relaxation modulus. There is also a significant difference between the steady-state relaxation modulus of hydrogels templated parallel to the indenting tip and those with templating perpendicular to the indenting tip and those with unaligned templating.

For each of the three templating orientations, there is a significantly different decrease in steady-state relaxation modulus after MAM clearance. This demonstrates a softening of mechanical properties through the introduction of a microporous architecture in the bulk hydrogel, as expected. Significant differences in relaxation moduli are also observed between hydrogels templated parallel to the indenting tip and hydrogels templated in different orientations – this is observed both before and after MAM clearance. Hydrogels templated parallel to the indenting tip likely have a higher relaxation modulus because the bulk hydrogel surrounding the MAM chains/microchannels provides support. In contrast, hydrogels with unaligned templating or templating perpendicular to the indenting tip are characterized by voids due to microchannels or discontinuities in the bulk hydrogel from the presence of MAMs that could more easily buckle under compression. Therefore, magnetic templating induces anisotropy not only in microarchitecture but also in bulk mechanical properties under steady-state conditions.

3.7. In vitro Evaluation of Cell Penetration into Magnetically Templated Hydrogels:

We designed our in vitro model with the peripheral nerve repair process in mind. Schwann cells were chosen because of their role in peripheral nerve regeneration in preceding axonal growth and providing support.[6] Furthermore, we wanted to generate a growth factor gradient along the length of the templated channels to mimic the chemotactic cues used in regenerating peripheral nerve to guide reconnection to the distal target. This was accomplished by using Transwell® permeable supports as seen in fig. S2. By placing media with varying concentrations of nerve growth factor (NGF) in the bottom compartment and media without growth factor in the upper compartment, a growth factor gradient was formed across the hydrogel placed at the bottom of the upper compartment.

In GMHA/collagen-based hydrogels, we observed evidence of cellular infiltration after culturing RSCs on the surface of hydrogels in a Transwell® setup for seven days. As seen in fig. 10, no red S100 protein signal or blue nuclear signal appears to be visible below the surface of the non-templated hydrogel sample. This is true both in the absence (fig. 10) and in the presence of a growth factor gradient (fig. S3). Visible on the surface of templated hydrogels are what appear to be circular, red artifacts – we attribute this to potential accumulation of secondary antibody near the openings of the templated channels (fig. S4). However, distinct red S100 protein signal and blue nuclear signal are both visible below the surface of the hydrogel for all magnetically templated samples seeded with RSCs. Furthermore, for the samples in which non-zero concentrations of NGF were placed in the lower compartment (50 ng/ml and 100 ng/ml), S100 protein and nuclei signals were visible at least 500 μm deep below the hydrogel surface. Since the depths at which S100 protein and nuclei signals are visible appear to be dependent on NGF concentration, we believe this serves as potential evidence of RSC infiltration into the templated hydrogels during culture. Thus, magnetically templated hydrogels appear to provide guidance to RSCs, and this is likely enhanced in the presence of a growth factor gradient. This suggests that the patterned microchannels have the capability of providing cellular guidance and perhaps also facilitate transport of nutrients and growth factors. These two qualities demonstrate that our magnetically templated hydrogels are promising materials for peripheral nerve regeneration and likely other tissue repair applications as well.

Figure 10.

In the top row are maximum projection images of the hydrogel surface for non-templated and templated hydrogels. In the bottom row are maximum projection side-views of the image stacks, showing signal below the surface. The red stain represents S100 protein, and the blue stain represents nuclei. For the non-templated hydrogel, signal is only visible on the surface of the hydrogel – only in the templated hydrogels is signal visible below the surface. With no NGF in the lower compartment of the Transwell, RSC penetration does not appear to extend past 150 μm below the hydrogel surface. With use of 50 ng/ml and 100 ng/ml of NGF in the lower compartment, penetration improves to at least 500 μm below the hydrogel surface.

Conclusions

In this paper, we introduce a method for developing MAMs using a microfluidic platform for magnetically templating hydrogels. Using this bottom-up approach allows the fabrication of MAMs that are uniform in size and magnetic properties: two parameters critical in microchannel formation, size, and alignment. We have also demonstrated that simulations can be used to model MAM chain formation and can serve as a predictive tool in exploring how modulating different variables in MAM affect chain formation. These MAMs clear rapidly from templated hydrogels, introducing anisotropic microstructure and mechanical properties to the material. Finally, we have also demonstrated the capability of magnetically templated microchannels to provide contact guidance for RSCs in vitro.

The magnetically templated hydrogels presented in this paper feature channels with diameters on the order of one hundred microns. However, for different tissue engineering applications, microchannels of varying diameters may be of interest. Through modulation of the flow rate conditions in the microfluidic droplet production process, potential future work could parametrically examine the effect of varying diameter and consequently templated microstructure on cell guidance and ultimately regeneration outcomes in vivo. Furthermore, it may be beneficial to magnetically template hydrogels with a higher void fraction to further improve cellular infiltration and transport properties. Currently, increasing the void fraction the templated hydrogels is limited by how high MAM volume fractions prevent sufficient UV light penetration for uniform hydrogel crosslinking. Thus, utilization of an alternative hydrogel crosslinking chemistry that does not depend on UV light may be beneficial in improving areal density of channels.

One aspect of the magnetic templating process is that while the length scales over which MAMs can be aligned is dictated by the size of the uniform magnetic field applied (which can be centimeters in scale), the individual length of the MAM chains and channels are only millimeters long in scale. Chain/microchannel length could be improved by tuning a number of parameters during the magnetic alignment step of the process – for instance: increasing the magnetic field strength, time placed in the magnetic field, and MAM IO loading. Nevertheless, we hypothesize that despite the discontinuity of the microchannels, infiltrating cells should be able to traverse down a channel in an implant and remodel the biodegradable hydrogel if “dead ends” are encountered.

Future work in both MAM and hydrogel design can allow better control over hydrogel microarchitecture, including both physical and biological cues. Finally, the magnetically templated hydrogels described here have potential for PNI repair and broader regenerative medicine and tissue engineering applications.