Perlecan domain I gradients establish stable biomimetic heparin binding growth factor gradients for cell migration in hydrogels

Abstract

Growth factor gradients orchestrate many biological processes including organogenesis, wound healing, cancer invasion, and metastasis. Heparin-binding growth factor (HBGF) gradients are established in living systems by proteoglycans including the extracellular matrix heparan sulfate proteoglycan, perlecan/HSPG2. Three potential HBGF-binding glycosaminoglycan attachment sites occur in N-terminal domain I of perlecan’s five domains. Our overarching goal was to form stable, biomimetic non-covalently bound HBGF gradients surrounding cells encapsulated in hyaluronate-based hydrogels by first establishing perlecan domain I (PlnD1) gradients. A versatile multichannel gradient maker device (MGMD) was designed and 3D printed, then used to create desired gradients of microparticles in hydrogels. Next, we used the device to covalently incorporate gradients of PEGylated PlnD1 in hydrogels with high-low-high or high-medium-low concentrations across the hydrogel width. Fluorescently-labeled fibroblast growth factor-2 was delivered to hydrogels in phosphate-buffered saline and allowed to electrostatically bind to the covalently pre-incorporated PlnD1, producing stable non-covalent HBGF gradients. To test cell viability after flow through the MGMD, delicate primary human salivary stem/progenitor cells were encapsulated in gradient hydrogels where they showed high viability and continued to grow. Next, to test migratory behavior in response to HBGF gradients, two cell types, preosteoblastic MC3T3-E1 cell line and breast cancer cell line MDA-MB-231 were encapsulated in or adjacent to PlnD1-modified hydrogels. Both cell lines migrated toward HBGFs bound to PlnD1. We conclude that establishing covalently-bound PlnD1 gradients in hydrogels provides a new means to establish physiologically-relevant gradients of HBGFs that are useful for a variety of applications in tissue engineering and cancer biology.

1. Introduction

Morphogen gradients in extracellular matrices (ECM) direct the formation of complex tissues during embryogenesis and are used throughout the life of an organism to direct cell migration associated with wound healing, cancer, immunity, and homing. In tissue, morphogen gradients of heparin binding growth factors (HBGFs) are formed indirectly by establishing gradients of heparan sulfate proteoglycans (HSPGs). HBGFs, such as fibroblast growth factors (FGFs), then electrostatically bind to their preferred binding sites on the heparan sulfate (HS) chains [1,2]. During embryonic development in Drosophila melanogaster, long- and short-range morphogen gradient signaling was demonstrated elegantly using both genetic and biological approaches, where the patterns of heparin-binding morphogens were seen to follow those of the matrix-deposited HSPGs [1,3]. Similar mechanisms drive biological processes in mammalian systems, for example gradients of fibroblast growth factors (FGFs) −7 and −10 produced in the surrounding mesenchyme coordinate branching morphogenesis in the developing epithelial bud of various tissues including the salivary gland [2]. Gradients also direct branching morphogenesis in a variety of other epithelial organs including lung and breast [2,4–6], and they impact metastasis and migration when cancer in these tissues invades into mesenchyme [7].

Perlecan/HSPG2, a large multidomain HSPG, serves as a natural HBGF reservoir in the ECM [8,9]. Domain I of perlecan (PlnD1) is decorated with three long, linear sulfated GAG pendant chains, typically a combination of HS and chondroitin sulfate (CS) chains determined by the cell type, covalently attached to the core protein [10]. HBGFs, including FGFs−2, −7 and −10, platelet-derived growth factor (PDGF), and heparin-binding epidermal growth factor (HB-EGF), all bind to HS selectively and with different sites and affinities, primarily based upon their preferences for various configurations of 2- and 6-O-sulfate on the HS GAG chains [11–16]. For certain HBGFs, such as basic fibroblast growth factor (FGF2), the binding and activation of the receptor FGFR1 requires HS to stabilize binding or signaling does not occur [17,18]. Although not the major tissue form of HS, soluble protein-free heparin is purified in large quantities from mast cell-rich tissues (i.e. 160–260 mg/kg from porcine mucosa) [19], and thus is routinely used as an economical alternative to HS in vitro to ensure activation of HBGF receptors. On the other hand, tissue HS is protein-bound, produced by many cell types throughout the body, and is only soluble after heparanase digestion that typically occurs locally [20]. Even when soluble, HS is a significantly less potent anticoagulant than heparin and thus is ideal for tissue engineering purposes [21]. Leveraging the noncovalent binding of HBGFs to the HS chains on PlnD1, this work aimed to create three-dimensional (3D) gradients of PlnD1 covalently linked into peptide-modified hyaluronate hydrogels to facilitate the formation of biomimetic HBGF gradients via their electrostatic complexation with GAG chains in the hydrogel.

Gradients have been formed previously using various techniques or devices including: gradient makers, microfluidic devices, micropatterning, and diffusion [22,23]. Gradient makers that mix two solutions pumped at different flow rates to achieve gradients have been used in the casting of gradient acrylamide gels for electrophoresis and western blotting applications [24] as well as for forming gradients of covalently tethered proteins or microspheres releasing proteins in 3D scaffolds for tissue engineering applications [25–28]. These studies utilized a commercially-available gradient maker device that doesn’t allow for incorporation of more than two solutions. Microfluidic devices have been used to form gradients in the x- or y- axis by mixing two solutions of polymer precursor solutions with a molecule of interest along the channels of the device to form a gradient of the molecule of interest upon exiting the device [29–31], or to culture cells within hydrogels containing solute gradients established in the device via diffusion [32–34]. One challenge for the use of microfluidic devices is the small scale that, while extremely useful for basic research, is not practical for tissue regeneration strategies. Therefore, a larger scale multichannel device is necessary to engineer tissue constructs on the order of millimeters to centimeters. Micropatterning has also been used to form gradients by selectively cleaving laser-sensitive groups in a hydrogel structure to expose functional groups beneath the cleaved sites, freeing those groups to react with functional groups of another molecule to attach to the hydrogel [35]. The laser light micropatterning technique can selectively reveal functional groups in higher density on one end of a hydrogel and in lower density on the other side, so that a gradient of functional groups is exposed and able to react with a protein of interest. Another use of lasers is in photoinitiated crosslinking of polymer networks. Photocrosslinking allows for rapid crosslinking of a matrix to stabilize an already-formed gradient in an effort to slow diffusion of the molecule of interest [29,36]. However, cytotoxicity of UV crosslinking [37], attributed to generation of free radicals and DNA damage, remains an issue. The most common way to form gradients of proteins or growth factors in scaffolds is by passive diffusion [32,33,38,39]. One example is the establishment of a solute concentration difference on either side of a hydrogel matrix, where the solute is free to pass through the hydrogel from the higher solute concentration reservoir to the lower concentration reservoir [32,33,38,39]. Combinations of the aforementioned techniques and devices are readily employed, such as the combination of a microfluidic device, diffusion between channels, and photoinitiated crosslinking to establish the final gradient [29]. Gradients of hydrogel crosslinking density [29], bioactive cell-binding motifs [29,34,35], or covalently bound growth factors and proteins [25,26,36,40–42] for use in a range of applications have been generated by other research laboratories. However, gradients of covalently bound proteoglycans to hydrogels for the formation of soluble HBGF gradients have not been reported. We created and 3D-printed a versatile three-inlet multichannel gradient maker device (MGMD) able to create a variety of HBGF gradients that include short or long-range gradients for a variety of applications in tissue engineering or cancer biology. Recombinant PlnD1, produced in milligram quantities in our laboratory, was used to form various configurations of gradients of electrostatically complexed HBGFs across the width of the hydrogel. Three different cell types, representing adult stem cells (human salivary stem/progenitor cells (hS/PCs)), a normal preosteoblastic mesenchymal cell line (MC3T3-E1), and cells from a highly migratory breast cancer line (MDA-MB-231) were used to examine the ability of the MGMD to form gradients of cells and to study cellular responses to HBGF gradients established in hydrogels.

2. Materials and methods

2.1. 3D printing the multichannel gradient maker device

Previous work demonstrated the use of herringbone features to facilitate chaotic mixing within microfluidic channels to homogeneously mix two solutions without turbulent flow [30,43,44]. Flow of two solutions over the herringbone features causes the two solutions to mix by folding or braiding. Previous microfluidic devices contained two inlets and channel dimensions on the order of micrometers [30,44]. The current work aimed to produce constructs relevant in size for tissue engineering applications, increase the number of achievable gradient configurations by incorporating three precursor solution inlets, and allow for cell encapsulation from single- to multi-cellular by increasing channel dimensions up to one millimeter (Fig. 1A–C). Computer-aided design (CAD) models of MGMDs and syringe holders were generated using Solidworks software (Dassault Systèmes Solidworks Corporation, Waltham, MA). The largest cross-sectional dimensions of channels within the device were 1 mm in width and 1 mm in height and the smallest dimensions at the herringbone features measured 1 mm in width and 0.5 mm in height. The CAD models of the gradient maker devices were printed using two 3D printers, the Pegasus Touch FSL3D printer (Full Spectrum Laser, Las Vegas, NV) with Full Spectrum Laser’s Universal Clear Resinor the Objet260 Connex1™ 3D printer (Stratasys, Eden Prairie, MN) with Stratasys’ VeroWhite-Plus resin. The 3D printed parts provided negative templates for the MGMD to be made using PDMS cured directly on top of the 3D printed parts. After the generation of the 3D printed molds, polydimethylsiloxane (PDMS) was mixed and poured over the mold to generate the MGMD. Finally, to readily visualize a gradient across a viscous solution using the MGMD, 70% glycerol was mixed with red, yellow, and blue food dyes and then pumped through the MGMD prior to immediate imaging.

Fig. 1.

Development of Multichannel Gradient Maker Device (MGMD). (A) Solidworks 3D computer-aided design software used to design the MGMD with channels of cross-sectional dimensions 1 mm by 1 mm and with staggered herringbone features (inset) with cross-sectional dimensions of 1 mm by 0.5 mm to facilitate chaotic mixing in channels. (B) Pegasus Touch 3D printer (Full Spectrum Laser 3D) produced the mold of the MGMD from (A). PDMS MGMDs were generated using 3D printed molds and a syringe pump was used to flow solutions through MGMD channels. (C) Colored dyes were mixed with 70% glycerol and pumped through the MGMD to visually display gradient generation.

2.2. PDMS mold of multichannel gradient maker device

Resin release was sprayed onto the surface of 3D printed parts to prevent the PDMS from sticking to the 3D printed part. Dow Corning Sylgard 184 Silicone Elastomer (Dow Chemical Company, Midland, MI) silicone elastomer kit was mixed according to the product sheet (mixed in a ratio of 10:1, base to curing agent) and the mixture was poured over the surface of 3D printed parts, placed inside a desiccator, and vacuum was applied for 30 min. After air bubbles were removed from the PDMS solution, the PDMS-covered 3D printed pieces were placed inside of an oven and heated to 60°C for 2 h. After cooling, the PDMS was carefully peeled away from the 3D printed part and washed extensively using labware soap to remove excess resin release spray.

2.3. Hydrogels, flow rate variation, and microparticle gradients

To form gradients using the PDMS multichannel gradient maker, three one-mL syringes were used simultaneously to pump three hydrogel precursor solutions through the device. The three syringes were attached to the device using 16-gauge blunt-end needles. Syringes were filled with identical hydrogel precursor solutions described below except that microparticles were added in the solutions that filled the two outer syringes. The hydrogel backbone utilized in these studies was thiolated hyaluronic acid (HA-SH) (HyStem®, GS311, ESI-BIO, Alameda, CA). HA-SH, molecular weight was 240 kDa and degree of thiolation was 1 μmol/mg HA-SH [45], was reconstituted according to protocol to 1% (w/v) solution using degassed deionized water supplied with the hydrogel kit and adjusted to pH 8.0 using 1 N NaOH. Poly(ethylene glycol)diacrylate (PEGDA) crosslinker (Extra-Link, ESI-BIO) was reconstituted in degassed deionized sterile water to 0.5% (w/v) solution. For the outer syringes, yellow-green 10.0 μm fluorescent microparticles (Fluoresbrite® YG, #18140–2, Polysciences Inc., Warrington, PA) were added to the HA-SH solutions at 1 × 10⁶ microparticles/mL hydrogel solution. In the middle syringe, no microparticles were added to the HA-SH precursor solution. The PEGDA crosslinker solution was mixed with HA-SH or HA-SH/microparticles at a 1:4 ratio (vol:vol). Upon mixture of the precursor solutions, the well-described primary reaction occurs between thiols on the HA-SH backbone and acrylates on the PEGDA crosslinker to form covalent bonds via Michael addition within thirty minutes [46,47]. The secondary reaction of thiols on adjacent HA chains to form disulfide bonds occurs subsequent to the primary reaction [46,48,49]. Solutions were allowed to cross-link for three minutes prior to loading into syringes to prevent microparticle settling in the precursor solutions. With only three minutes of crosslinking, the solution remains liquid with a slight increase in viscosity to prevent settling. A syringe pump was propped up to a 45-degree angle from horizontal predetermined by the angle of entry into the MGMD. The filled syringes then were secured to the syringe pump and the blunt-end needles were inserted into the PDMS multichannel gradient maker device and the pump was turned on to run at three different constant flow rates: 0.9 mL/min, 0.3 mL/min, or 0.1 mL/min. Once the hydrogel solution reached the end of the device, the needles were removed from the PDMS mold and the gradient maker device with hydrogel precursor solution was placed into a 37°C incubator. The hydrogel solutions were crosslinked for 1 h at 37°C prior to careful removal of the PDMS from the hydrogel, and then the hydrogel was sectioned and placed into separate wells of a 12-well plate using scalpel and spatulas into PBS solution (additional details in Supplemental Information). The microparticle gradients were imaged using the A1 Nikon Confocal microscope (Nikon, Tokyo, Japan) across the width of the hydrogel and MATLAB software was utilized to analyze the fluorescence intensity across the hydro-gel at the various flow rates.

2.4. Recombinant perlecan domain I generation

HEK293-EBNA cells transfected to express mPlnD1 have been in our laboratory for many years since they were generously donated by the Timpl Lab [10,50]. HEK293-EBNA-PlnD1 cells were expanded in selection medium containing high glucose DMEM (#11965, Gibco/Invitrogen, Carlsbad, CA), 10% (v/v) heat-inactivated fetal bovine serum (FBS, Atlanta Biologics, #S11150), 2 mM L-glutamine (Gibco/invitrogen #25030), 1% (v/v) penicillin/streptomycin (Gibco/Invitrogen #15140, 100 units/mL penicillin and 100 μg/mL streptomycin), 225 μg/mL geneticin (G418, Gibco/Invitrogen #10131), and 10 μg/mL puromycin dihydrochloride (Sigma #P8833). Hyperflask™ M Cell Culture Vessels (1720 cm², Corning #10030) were utilized for cell conditioned medium collection. Hyperflasks™ were seeded at 3.6 × 10⁷ HEK293-EBNA cells per flask. Three days after seeding, the medium in the Hyperflasks™ was changed to 2% (v/v) FBS, DMEM, L-glutamine, penicillin/streptomycin, G418, and puromycin. The collection of 2% (v/v) FBS-containing cell-conditioned medium began on Day 6. Protease inhibitors and an antibacterial agent were added to conditioned medium prior to filtering: 0.5 mM benzamidine, 0.1 mM phenylmethysulfonyl fluoride (PMSF), 0.5 mM EDTA, and 0.1% (w/v) EDTA. Conditioned medium was collected and filtered with a 0.22 mm vacuum filtration system and frozen at [C0]20 °C. HEK293-EBNA conditioned medium was concentrated from 10 L down to 250 mL using a 10 K MWCO spiral wound membrane cartridge (Millipore, Amicon S34010) and a Micron LP-1 Pump (Masterflex Cole-Parmer model 70–1821). PlnD1 was purified out of the conditioned medium using a diethylaminoethyl (DEAE) Sepharose CL-6B (GE Healthcare #17-0710-01) anion exchange column. The equilibration and elution buffers were identical except for the sodium chloride (NaCl) content of 0.25 M and 0.75 M in the equilibration and elution buffers, respectively. The contents of the equilibration and elution buffers were as follows: 0.05 M HEPES/HCl pH 8.0, 2 M urea, 2.5 mM EDTA, 0.5 mM PMSF, 0.5 mM benzamidine, 0.02% (w/v) sodium azide, and either 0.25 M or 0.75 M NaCl. Protein concentration in the elution buffer of eluted fractions was measured using the NanoDrop 2000 Spectrophotometer Protein A280 method and peak fractions pooled. After DEAE purification, the total PlnD1 measured using A280 was 0.55 mg/mL for a total of 22.7 mg. Next, the eluted solution was added to dialysis membrane with 3500 MWCO (Spectrum Laboratories Inc., Rancho Dominguez, CA) and dialyzed against 28 L (seven exchanges of four-liter volumes) of 20 mM HEPES 150 mM NaCl solution, completed in 4°C cold room. After dialysis, the total PlnD1 measured using A280 was 0.38 mg/mL for a total of 18.4 mg. Finally, purity of the PlnD1 sample is always confirmed via silver staining of the sample in an electrophoresed PAGE gel (Supplemental S1) [50].

2.5. PEGylation of PlnD1 and western blot analysis

PlnD1 was PEGylated using acrylate-PEG-succinimidyl valerate (Ac-PEG-SVA, 3.4 kDa, Laysan Bio) at 1 mg PlnD1 per 0.642 mg Ac-PEG-SVA. The ratio of Ac-PEG-SVA to PlnD1 was calculated based on available primary amines on core protein of PlnD1 and overestimated to 10:1 (Ac-PEG-SVA: PlnD1). PlnD1 was dissolved in HEPBS solution (N-(2-hydroxyethyl)piperazine-N’-(4-butanesulfonic acid); 20 mM HEPBS [Santa Cruz Biotechnology Inc.], 100 mM NaCl, 2 mM CaCl₂, and 2 mM MgCl₂) at 100 μg/mL and the solution was adjusted to pH 8.5 using 1 N NaOH. Ac-PEG-SVA was dissolved in HEPBS to 1 mg/mL. The solution of Ac-PEG-SVA was carefully added to the PlnD1 solution while gently vortexing. The solution then was placed on a rocker set to highest agitation in a 4°C cold room. The pH of the solution was periodically tested and adjusted to pH 8.0. Ac-PEG-SVA was reacted with PlnD1 for 12–16 h. Then, the solution was transferred to 15,000 MWCO dialysis membrane tubing (Spectrum Laboratories Inc.) and dialyzed against 20 L milliQ water (five exchanges of 4-liter volumes every 4 h). After dialysis, the solution of PlnD1-PEG-Ac was sterile filtered, measured for protein content using the NanoDrop A280, aliquoted into 20 μL aliquots in sterile 200 μL tubes, and stored at −80 °C until use. The protein content of the PlnD1-PEG-Ac solution as measured using A280 was 3.1 mg PlnD1-PEG-Ac/mL.

Western blot analysis was performed using PlnD1 and PlnD1-PEG-Ac. Because PlnD1 is decorated with HS/CS chains of variable length, PlnD1 and PlnD1-PEG-Ac appeared as broad, smeared bands in a western blot. Therefore, heparinases and chondroitinases were used to digest the HS and CS chains from PlnD1 and PlnD1-PEG-Ac for better resolution of the protein band. First, PlnD1 and PlnD1-PEG-Ac samples were digested using 10 mU each of heparinases I, II, and III and chondroitinase ABC per 5 μg protein in enzymatic digestion buffer (50 mM HEPES, 200 mM NaCl, 3 mM CaCl₂, pH 7.4) for 12 h at 37°C. Then, four samples were loaded onto and separated in a NuPAGE™ 4–12% Bis-Tris Gel (ThermoFisher Scientific, Waltham, MA). The four samples were: undigested PlnD1 (2 μg per lane), digested PlnD1 (2 μg per lane), undigested PlnD1-PEG-Ac (1 μg per lane), and digested PlnD1-PEG-Ac (0.5 μg per lane). The proteoglycan samples were transferred from the gel to a 0.45 μm pore size nitrocellulose membrane (Bio-Rad, 162–0094) and then blocked with 3% (w/v) bovine serum albumin (BSA) in Tris Buffered Saline Tween-20 (TBST: 20 mM Tris, pH 7.6, 150 mM NaCl, 0.05% (v/v) Tween-20). Primary and secondary antibodies were diluted in 3% BSA in TBST. Primary antibody to the N-terminus of perlecan (N-20) (1:2000, Santa Cruz Biotechnology, Dallas, TX) and secondary antibody donkey anti-goat IgG-HRP (1:50,000, Santa Cruz Biotechnology) were used. Protein blot bands were visualized using SuperSignal™ Western Blot Enhancer (ThermoFisher Scientific, #46640). Densitometry analysis was performed on the western blot bands in lane 4 using ImageJ software.

2.6. Protect and label FGF2

To protect the HS binding groups from derivatization, FGF2 was bound to heparin during fluorescence labeling with amine-reactive fluorophores by modifying a previously reported protocol [51]. First, 35 μg FGF2 (Biolegend, #571504, San Diego, CA) was added to a solution of heparin resin (Toyopearl, TOSOH) according to a previous protocol [51]. Briefly, 20 μL heparin resin was added to a 200 μL PCR tube and washed in 180 μL phosphate buffer (PB) containing 1 M NaCl (PB 1 M pH 7.8: 17.9 mM Na₂HPO₄, 2.1 mM NaH₂PO₄, 1 M NaCl pH 7.8). Then the resin was equilibrated by pipetting 4 × 175 μL of PB 150 mM pH 7.8 (17.9 mM Na₂HPO₄, 2.1 mM NaH₂PO₄, 150 mM NaCl, pH 7.8). Centrifugation was used to settle beads between all steps. Next, 35 μg FGF2 was bound to the resin for 60 min on a rocker table. The resin was washed using 3 × 50 μL PB 150 mM pH 7.8 solution, then, 20 μL PB 150 mM pH 8.3 was used to equilibrate the column to prepare it for conjugation. FGF2 was reacted with 27.46 nM tetrafluorophenyl-Alexa Fluor® 488 (TFP-AF488; ThermoFisher Scientific, A30006) in 20 μL PB 150 mM pH 8.3 for 15 min at room temperature to ensure complete conjugation of exposed lysines to TFP. The resin next was washed five times with PB 150 mM pH 7.8 solution. Fluorescent FGF2 (FGF2–488) was eluted using elution buffer containing 2 M NaCl (44.75 mM Na₂HPO₄, 5.25 mM NaH₂PO₄, 2 M NaCl pH 7.8). Degree of labeling was quantified using NanoDrop Spectrophotometer and measuring absorbance at 280 nm and 494 nm. Eluted fractions were kept separate and only fractions of high protein content and greatest degree of labeling were used. A buffer exchange was performed on eluted fractions of FGF2–488 to decrease the NaCl content from 2 M NaCl to 140 mM NaCl because NaCl interferes with the electrostatic binding of HBGFs to HS chains.

2.7. FGF2 gradient generation

PlnD1-PEG-Ac gradients and block-source hydrogels were generated in HA-SH/PEGDA hydrogels. The gradient hydrogel was generated using the same methods as described in Section 2.3, except that 60 μg PlnD1-PEG-Ac/mL final hydrogel volume was incorporated into the hydrogels of the outer syringes instead of microparticles and the PlnD1-PEG-Ac was allowed to react with HA-SH at 37 °C for one hour prior to the addition of the PEGDA crosslinker solution. Both the gradient hydrogels and the block-source hydro-gels were generated using PDMS molds. Once PlnD1-PEG-Ac gradient (PlnD1-gradient) hydrogels and PlnD1-PEG-Ac block-source (PlnD1 block-source) hydrogels were formed, they were incubated in a phosphate buffered saline (PBS) solution containing the FGF2–488 at 250 ng FGF2–488/mL PBS overnight for roughly 12 h. On Day 1, the FGF2–488-containing medium was removed and replaced with fresh PBS. Samples were imaged at Day 2 to visualize the initial formation of the HBGF gradients such that we could follow them over time and determine stability of each formed gradient. The samples were imaged using a Nikon A1 confocal microscope (10X objective). The confocal images were input into MATLAB and relative fluorescence intensity in the green channel graphed across the width of the hydrogel.

2.8. Primary human salivary gland stem/progenitor cell viability after flow through gradient maker device

Because high viability of cells is needed for tissue engineering applications, these delicate primary cells were chosen for this study using the MGMD. Primary human salivary parotid gland stem/progenitor cells (hS/PCs) were used as a primary human cell source to demonstrate that the gradient maker did not subject cells to shear stresses that caused cell death. Human salivary parotid gland tissues were procured from consented head and neck cancer patients under protocols approved by our donor site Institutional Review Boards (Christiana Care Health Systems and the University of Delaware). The gland tissue was dissociated and cultured from the gland according to previously reported protocols [52,53]. Primary hS/PCs were maintained in Corning Hepatocyte culture media kit (Fisher Scientific, CB-30055M) for up to 15 passages [52]. HA-SH and PEGDA precursor solutions were rehydrated according to protocol as described above to 1% and 0.5% (w/v), respectively. Once HA-SH fully dissolved, 3 × 10⁶ hS/PCs per milliliter final hydrogel volume were added to the solution and mixed gently. The pH of the solution was adjusted to 7.8 using NaOH to ensure rapid crosslinking of thiol groups on HA-SH with acrylate groups on PEGDA via Michael Addition. After adjusting pH, PEGDA was added to the cell/HA-SH solution and mixed thoroughly for about one minute. Prior to gelation, hS/PC and hydrogel precursor solutions were either seeded into cylindrical molds or pumped through the MGMD at 0.9 mL/min until the solution reached the end of the 10 mm wide channel and then the pump was turned off. The hydrogel solutions were allowed to crosslink for 1 h at 37°C prior to careful removal of the PDMS from the hydrogel. At this stage, the MGMD-generated hydrogel dimensions were 10 mm wide and 150 mm long and the cylindrical mold hydrogels were 6 mm wide pucks and 1 mm in height. The MGMD-generated hydrogel next was sectioned into 10 mm by 6 mm rectangular prisms using a scalpel and placed directly into medium in wells of two 12-well plates using sterile scalpels and spatulas. Cells were cultured in hydrogels for 1 to 7 days. Cell viability was qualitatively assessed using live/dead viability assay solution of Calcein AM (2 μM working solution in PBS, Biotium), ethidium homodimer-III (4 μM working solution in PBS, Biotium, 40050) and Hoechst 33342 nuclear stain (1.8 μM working solution in PBS, Enzo, ENZ-52401) after one and seven days of culture. Samples were rinsed with PBS then incubated at 37°C in live/dead viability assay solution for 30 min prior to imaging. Samples were imaged using the A1 Nikon Confocal microscope across the width of the hydrogels.

2.9. Preosteoblastic migratory response to HBGFs in PlnD1 block-source hydrogels

Cells from the migratory mesenchymal preosteoblastic murine cell line MC3T3-E1 were used for this study. Cells were maintained in alpha-MEM with 10% (v/v) FBS, 1% (v/v) penicillin/streptomycin, and 2 mM L-glutamine. Preliminary experiments with these cells demonstrated the need for the addition of cell attachment factors and enzymatically degradable crosslinkers to the hydrogel matrix to support cell binding and motility. The crosslinker was generated following a previously reported protocol in our lab [54]. Peptides KGGGPQG;IWGQGK (PQ peptide) with N-terminal acetylation and GRGDS with C-terminal amidation were purchased from Genscript USA Inc. (Piscataway, NJ). Briefly, the MMP-cleavable PQ peptide was reacted with acrylate-PEG-SVA (3400 g/mol, Laysan Bio Inc., Arab, AL) at a molar ratio of 1:2.1 in HEPBS buffer adjusted to pH 8.0 using 1 N NaOH (20 mM HEPBS [Santa Cruz Biotechnology Inc.], 100 mM NaCl, 2 mM CaCl₂, and 2 mM MgCl₂) [54,55]. The reaction tube was protected from light and put on a rocking table overnight at 4°C. The reaction then was dialyzed against ultrapure water for two days using 3500-Da MWCO dialysis membrane (Spectrum Laboratories Inc.). The dialyzed product was sterile-filtered using a 0.2 μm pore size filter, frozen at −80 °C, and lyophilized for two days. The final lyophilized product of AC-PEG-PQ-PEG-AC (PQ) was stored at −20 °C until use. Similarly, the cell-binding motif, GRGDS, was also pegylated following the previous protocol. Briefly, GRGDS was reacted with acrylate-PEG-SVA at a molar ratio of 1.2:1 in HEPBS buffer. The final lyophilized product of GRGDS-PEG-Ac also was stored at −20 °C until use. Upon use, PEGylated peptides were weighed and rehydrated to precise concentrations in PBS and used immediately. HA-SH was hydrated at 10 mg/mL and used as the backbone for these hydrogels, PQ was reconstituted to 22.5 mg/mL, and GRGDS was reconstituted to 73.7 mg/mL and used at 4:1:1 (v/v/v) ratios of HA-SH/PQ/GRGDS. The final concentration of PQ was 3.7 mg PQ/mL hydrogel that corresponds to a 6:1 SH:AC crosslinking ratio; 12.3 mg GRGDS/mL hydrogel was used for a 3 mM final concentration. In the hydrogels that contained PlnD1, PlnD1-PEG-Ac was added to the hydrogel at a concentration of 60 μg PlnD1-PEG-Ac/mL hydrogel. MC3T3-E1 cells were encapsulated at an initial seeding density of 1 × 10⁶ cells/mL. Cells were encapsulated in hydrogel precursor solutions, plated as a block hydrogel (1 mm in height, 2 mm wide, and 10 mm long; 20 μL), onto glass-bottomed wells for optical clarity, and crosslinked directly next to either a blank block hydrogel or a PlnD1-containing block hydrogel without cells of equal volume (20 μL). To study the effects of PlnD1 and growth factor binding on cell motility of encapsulated cells, the serum in the medium was reduced to 1% (v/v) FBS because FBS is a rich source of HBGFs. Two HBGFs were exogenously added to the culture medium, FGF2 (Biolegend, San Diego, CA) and PDGF-BB (Tonbo Biosciences, San Diego, CA), at 100 ng/mL, and supplied to the hydrogels immediately after the hydrogels crosslinked. Medium was changed every 2 days and no additional growth factor was added after the initial dose. An inverted light microscope Nikon Eclipse TE300 was used to image the hydrogels across the width of the hydrogels after three, five, and seven days of culture and the images stitched together using Nikon NIS Elements software.

2.10. Breast cancer cell migratory response to HBGF gradients

The highly migratory breast cancer cell line MDA-MB-231 (ATCC, Manassas, VA), able to migrate by both mesenchymal and amoeboid mechanisms [56], was used for these studies. Cells were cultured in DMEM/F12 media with 10% (v/v) FBS, 1% (v/v) penicillin/streptomycin, and 2 mM L-glutamine [57]. MDA-MB-231 cells were encapsulated into either (1) PlnD1-gradient hydrogels or (2) PlnD1-uniform hydrogels to determine effects of the gradient on the migration of these cells. Briefly, HA-SH was reconstituted according to protocol to 1% (w/v) solution using degassed deionized water and adjusted to pH 8.0 using 1 N NaOH. The PQ crosslinker was reconstituted in PBS to 15 mg PQ/mL PBS for a final 9:1 SH:AC crosslinking ratio. GRGDS was reconstituted in PBS to 73.7 mg GRGDS/mL PBS for a final 3 mM concentration. In PlnD1-gradient hydrogels, 60 μg PlnD1-PEG-Ac/mL hydrogel was used as the highest concentration and the concentration decreased across the hydrogel. In PlnD1-uniform hydrogels, 60 mg PlnD1-PEG-Ac/mL hydrogel concentration was used throughout the hydrogel. MDA-MB-231 cells were encapsulated as a cell gradient opposing the gradient of PlnD1. Fluorescently-labeled 10 μm microparticles (Polysciences, #18140–2) were encapsulated uniformly throughout the hydrogel at a concentration of 3.25 × 10⁵ microparticles/mL hydrogel to serve as stationary fiduciary markers to track cell migration. These particles also permitted us to clearly visualize hydrogel borders during imaging, given that the side of the hydrogel with low cell seeding density is optically clear. The particles also allowed us to accurately measure the hydrogel size throughout culture. To form a gradient of cells opposing a gradient of PlnD1 or uniform PlnD1, three one-mL syringes were used simultaneously to pump three solutions of hydrogel precursor solutions through the MGMD for either the gradient or uniform condition. The three syringes were filled with the same hydrogel precursor solutions with added microparticles except for the addition of PlnD1 and cells to different syringes. In the gradient PlnD1 group, PlnD1-PEG-Ac was added into solution for syringe 2 at 30 μg/mL and into solution for syringe 3 at 60 μg/mL, while cells were added into solution for syringe 1 at 1 × 10⁶ cells/mL and into solution for syringe 2 at 5 × 10⁵ cells/mL. In the uniform PlnD1 group, PlnD1-PEG-Ac was added into solutions for syringes 1, 2, and 3 at 60 μg/mL (the high concentration in the gradient condition), while cells were added into solution for syringe 1 at 1 × 10⁶ cells/mL and into solution for syringe 2 at 5 × 10⁵ cells/mL. The highest cell seeding density was 1 × 10⁶ cells/mL hydrogel and cell density decreased across the hydrogel. The GRGDS solution and PQ crosslinker solution were mixed with HA-SH solution at a 4:1:1 ratio (v/v/v) of HA-SH:PQ:GRGDS [54]. After PQ was added to the solutions, the mixture was allowed to crosslink for three minutes prior to loading into syringes. The 45-degree angled syringe pump was loaded with the three syringes and the blunt end needles were inserted into the MGMD. A constant flow rate of 0.9 mL/min was used to form the gradient hydrogel. The hydro-gel solutions were allowed to crosslink for 1 h at 37°C prior to careful removal of the PDMS from the hydrogel. The hydrogel was sectioned and placed into separate wells of a 24-well plate using an autoclaved coverglass as a cutting tool (60 mm × 25 mm) into respective media conditions. Both the PlnD1-gradient and PlnD1-uniform hydrogels were cultured in three different media formulations: 10% (v/v) FBS, 1% (v/v) FBS with 50 ng/mL transforming growth factor-β1 (TGF-β1; #100–21, PeproTech Inc., Rock Hill, NJ) and 50 ng/mL heparin binding-epidermal growth factor (HB-EGF; #100–47, PeproTech Inc.), or 1% (v/v) FBS in base medium of DMEM/F12 with 1% (v/v) penicillin/streptomycin, and 2 mM L-glutamine. The hydrogels were cultured for time periods up to seven days, with samples fixed using 4% (v/v) formalin at Day 1 and Day 7. Samples were washed in PBS before fixation, then rinsed in PBS after fixation and stored at 4 °C until use in immunocytochemistry (ICC). For ICC, the primary antibodies used were anti-RhoA (1:50, mouse monoclonal, Santa Cruz Biotechnology SC418) and anti-vimentin (1:100, rabbit monoclonal, Abcam ab92547, Cambridge, MA). The secondary antibodies used were Alexa Fluor 488 goat anti-mouse IgG (1:1000, Life Technologies, A11029, Carlsbad, CA) and Alexa Fluor 568 goat anti-rabbit (1:1000, Life Technologies, A11011). The samples were washed thoroughly with PBS, then permeabilized using PBS with 0.2% (v/v) Triton X-100 for 20 min. Samples were blocked in 3% (v/v) goat serum in PBS with 0.2% (v/v) Triton X-100 (blocking solution) for 2 h at room temperature. Then, samples were incubated overnight at 4°C in primary antibody-containing blocking solution along with plain blocking solution for secondary control samples. Then, samples were washed using PBS and incubated at room temperature for 2 h in secondary antibody solution: PBS with 0.2% (v/v) Triton-X100 with 1:1000 dilution of both secondaries, 1:40 dilution of phalloidin (200 U/mL stock, CF647 Phalloidin #00041, Biotium, Fremont, CA) and 1:500 dilution of DAPI (5 mg/mL stock, D1306, Invitrogen). Samples were thoroughly washed after staining in PBS. The groups were imaged using the A1 Nikon Confocal microscope across the width and through the depth of the hydrogel creating z-stack images using 4× and 20× objectives. Using z-stacks generated from 4× objective, Imaris (Bitplane, Zurich, Switzerland) software was employed to quantify the volume of vimentin- and F-actin(phalloidin)-positive cells across the end of the hydrogel with low cell seeding density (30% of the hydrogel on the right side) in all groups. Vimentin and F-actin were used to detect the volumetric location of the cells. The low cell-seeded side of the hydrogel was a region of interest (ROI) from Day 1 to Day 7 as it contained the cells that responded to the PlnD1-HBGFs by moving into that region. The volumes of vimentin- and F-actin-positive cells within the most distant 30% (30% in the x-axis) of the hydrogel were quantified for each group (n = 3 per group) as the volumetric sum within a given volume and divided by the volume within each hydrogel that was quantified to normalize the value before comparing to other groups. When staining for F-actin, small clusters and large clusters were visible, but not single cells. In contrast, with vimentin, single cells were visible but not all large clusters stained positive. Therefore, F-actin was used to quantify the volumes of clusters of cells and vimentin was used to quantify volumes for single cells. For volumes of single cells, vimentin in Day 1 samples was used to determine the maximum volume value in Imaris to use for all Day 7 values as the maximum value thresholded for quantification of vimentin-positive single cells at Day 7. The mean of the samples for each group was quantified and the error bar represents the standard deviation for a given group.

2.11. Statistical analysis

All data were presented as mean ± standard deviation. Statistical analysis was performed using GraphPad Prism (GraphPad, San Diego, CA). Statistical analysis was performed using Student’s t-tests. A threshold of p < 0.05 was used to determine statistical significance.

3. Results

3.1. 3D printing and PDMS casting of multichannel gradient maker device

Solidworks CAD modeling software facilitated the generation of several prototypes of the MGMD. To form gradients of HBGFs in hydrogels, a mixing device was designed first to be large enough in size in both the cross-sectional area of each channel to flow hydrogel precursor solutions with cells suspended in solution through the device without damaging the cells, and second of sufficient size to allow the final hydrogel construct to be tested for regenerative medicine applications. The design parameters for the device included the: (1) final hydrogel dimensions as output for the device, (2) ability to form a gradient profile after mixing through the device, (3) ability to maintain cell viability after flow through device, and (4) the ability to be autoclaved. We used the unitless Reynolds Number (Re) to assess the fluid flow patterns in channels as being either turbulent (above 4000) or laminar (below 2300). Turbulent flow patterns can effectively mix two solutions together, whereas laminar flow patterns cause little to no mixing due to characteristically steady flow. The Reynolds Number for a rectangular duct, as seen in the cross section of the MGMD channels, was rectangular, not tubular, and was defined as Re = ρVD_H/μ, where ρ = density of the fluid (kg/m³), V = velocity (m/s), D_H = hydraulic diameter of channel cross section (m), and μ = dynamic viscosity (m²/s). In the MGMD, the cross-sectional hydraulic diameter ranged from 6.7 × 10⁻⁴ to 1 × 10⁻³ m, so the velocity needed to be large and the viscosity small to achieve turbulent flow, which was found not to be achievable in our MGMD. Additionally, turbulent flow of precursor solutions with encapsulated cells could cause cell lysis and death and thus would be an undesirable design parameter. To facilitate passive mixing, a redesigned split-and-recombine channel flow with chaotic mixing within channels was designed and tested. First, straight channels that diverged and converged resembling a honeycomb structure were generated as a “split-and-recombine” flow design, but distinct solutions barely mixed together. Next, S-shaped channels were used, but the mixing did not improve greatly. Finally, S-shaped channels with herringbone features in a split-and-recombine design were generated and the herringbone features facilitated mixing at various flow rates. A CAD model was generated (Fig. 1A) to facilitate the printing of a negative mold of the MGMD (Fig. 1B). Finally, the MGMD was formed using PDMS and the 3D printed mold. Lastly, 70% glycerol with food coloring was utilized as a viscous solution to visualize the gradient formation prior to the use of hydrogel precursor solutions (Fig. 1C). This design was adopted for the MGMD in all future studies.

3.2. Flow rate variation on microparticle gradient formation

Fluorescent microparticles were used to determine the effects of hydrogel precursor flow rate through the multichannel gradient maker device on the microparticle gradient profiles. A schematic of the hydrogel precursor solutions with or without microparticles used to generate gradients of microparticles across the hydrogel is displayed in Fig. 2A. Three flow rates were tested, 0.1, 0.3, and 0.9 mL/min. The lowest flow rate value was chosen to ensure the hydrogel precursor solutions did not crosslink within the device during the mixing process; the highest flow rate was chosen as the highest achievable flow rate in this device. After generation of the microparticle gradients at various flow rates, the hydrogels were imaged across the width of the hydrogel as described in Methods (Fig. 2B). The traces displayed in Fig. 2C were generated using MATLAB to quantify a running average of the fluorescence intensity values across the hydrogel. The fluorescent microparticles used in this work nearly saturated the detector of the confocal microscope even at the lowest laser power level. Because of this near-saturation, each microparticle appeared as a very high intensity value. The average intensity was measured in each column of pixels across the entire image. This produced a very noisy trace because the space immediately surrounding the microparticles had very low intensity. A running average calculation in MATLAB was utilized to smooth the curve. The highest flow rate generated a gradient of microparticles across the hydrogel with the steepest profile (Fig. 2C, red line), the profile of the middle flow rate was less steep (Fig. 2C, green line), and the lowest flow rate generated the least steep profile of the three flow rates tested (Fig. 2C, blue line). The profiles are reflected in the images in Fig. 2B, with the most mixing into the center of the hydrogel at the lowest flow rate and the least mixing at the center of the hydrogel with the highest flow rate. The highest flow rate of 0.9 mL/min was chosen for future studies with PlnD1 and HBGFs as it provided a reasonably steep, linear gradient to which cells were judged to be likely to respond. The flow rate that achieved the steepest microparticle gradient profile was chosen for future studies using PlnD1 and HBGFs because HBGFs are small enough to diffuse through the hydrogel pores and could have formed longer range gradients than the microparticles because of diffusion and the varying binding affinities of HBGFs to HS. We chose 0.9 mL/min to account for HBGF diffusion from HS/PlnD1.

Fig. 2.

Flow Rate Effects on Microparticle Gradient Profile. Fluorescent microparticle gradients in hydrogels were generated using an MGMD and three different flow rates. (A) The hydrogel composition, thiolated hyaluronic acid (HA-SH) with poly(ethylene glycol) diacrylate (PEGDA) crosslinker, across the width of the hydrogel was constant except for the incorporation of microparticles shown in 1, 2, 3 which indicate precursor solutions in syringes 1, 2, 3 respectively prior to mixing in MGMD. (B) Confocal-generated fluorescent images of microparticle gradients across the width of the hydrogel at three specified flow rates. (C) The relative fluorescence intensity was measured across the width of the hydrogel demonstrating that microparticle gradients were affected by the flow rate of precursor solutions prior to gelation. The steepest gradient profile was the red curve, 0.9 mL/min. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

3.3. Primary human salivary gland stem/progenitor cell viability after flow through gradient maker

Primary hS/PCs from salivary glands were used as a delicate primary human cell source that could undergo cell damage or death in response to excessive shear stresses caused by flow through the MGMD. In Fig. 3A, confocal images displayed hS/PCs within hydrogels formed using the MGMD and stained after one or seven days in culture with live cell marker (Calcein AM, green), dead cell nuclei marker (ethidium homodimer-III, red), and all cell nuclei marker (Hoechst, blue). Fig. 3B displayed confocal images of hS/PCs within hydrogels formed by pipetting precursor solutions into cylindrical molds and stained after one or seven days in culture with the same live cell, dead nuclei, and all nuclei identifying reagents listed above. Qualitative comparison after one day in culture revealed similar levels of live cell staining and dead cell staining as indicated by the green and red cell staining, respectively. We routinely check the viability of our hS/PCs and viability is generally around 80% in this hydrogel formulation (Martinez et al., under review). The trend continued through seven days of culture (Fig. 3A/B, bottom images) with similar staining of live and dead cells present in both cultures. Additionally, hS/PCs formed spheroids after seven days in culture in both conditions, as displayed in phase images (Fig. 3A/B, bottom images). To visualize the viability of single cells encapsulated in the hydrogels across the entire hydrogel construct after one day of culture, we imaged the whole hydrogel at a low magnification at Day 1. After one day of culture, hS/PCs remained as single cells without obvious multicellular structures. After seven days in culture, the hydrogel constructs were imaged at higher magnification to visualize the features of newly assembled multicellular structures forming in both types of hydrogels. While hS/PCs were used as a delicate cell type, both MC3T3-E1 and MDA-MB-231 cells maintained high viability after flow through the MGMD (data not shown).

Fig. 3.

Primary Human Salivary Stem/Progenitor Cell Viability After Flow Through MGMD. Salivary cells (P12) encapsulated in HA-SH/PEGDA demonstrated comparable viability at Days 1 and 7 after (A) flow through MGMD with subsequent crosslinking and culture or (B) crosslinking directly in cylindrical molds (control) followed by culture. hS/PCs form characteristic spheroids after 7 days in culture (bottom two images from A and B). Cells were stained with live-cell cytoplasmic marker Calcein AM (green), all cell nuclei stain Hoechst (blue), and dead cell nuclei marker Ethidium Homodimer-III (red). Phase images were displayed at Day 7 as well to show cell cluster architecture. Day 1 samples were imaged at 10× (scale bars 1 mm) and Day 7 samples imaged at 20X (scale bars 50 μm). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

3.4. Recombinant perlecan domain I generation and PEGylation

To covalently bind PlnD1 to HA-SH hydrogels, PlnD1 was PEGylated using acrylate-PEG-succinimidyl valerate (Ac-PEG-SVA, 3.4 kDa) as described in Methods, where SVA reacts with free amines on surface lysines of the core protein of PlnD1 to form PlnD1-PEG-Ac [58]. Once formed, a western blot was used to confirm an appropriate increase in molecular weight between PlnD1 and PlnD1-PEG-Ac. A schematic of the PlnD1 and PlnD1-PEG-Ac samples with or without digestion using endoglycosidases was displayed in Fig. 4A. Because of the presence of HS chains with variable lengths attached to the core protein of PlnD1, intact PlnD1 appeared as a smear during western blotting as is typical for this molecule (Fig. 4B, Lane 1; Supplemental S2 displayed blots at varying exposure times). The undigested sample of PlnD1, in lane one, and PlnD1-PEG-Ac, in lane 3, both appeared as smears in the western blot, but PlnD1 in lane one was only barely visible with 10 s exposure of the film between 37 and 50 kDa (Fig. 4B Lane 1). The smeared appearance of undigested samples was expected as these samples contain both HS/CS chains and/or PEG chains of varying lengths and charge. The digested samples of PlnD1, in the second lane, and PlnD1-PEG-Ac, in the fourth lane, demonstrated increased electrophoretic mobility along with a reduction in the apparent molecular weight range when compared to their undigested counterparts. This behavior was expected because of the removal of most of their GAG chains after endoglycosidase treatment. However, digested PlnD1-PEG-Ac resolved to several bands, which corresponded to the addition of 3.4 kDa AC-PEG-SVA groups to PlnD1. Additionally, the increase in MW of digested PlnD1-PEGAc and digested PlnD1 was approximately 12.3 kDa corresponding to an average of three to four PEG groups per PlnD1 molecule. Digested PlnD1 displayed a MW of approximately 29.5 kDa in lane 2. Lane 4 contained digested PlnD1-PEG-Ac, with the top band of highest molecular weight being first, the middle band being second, and the bottom band of lowest molecular weight being third. Densitometry analysis was performed comparing the first band and second band relative to the third (lowest MW) band in Lane 4. The first band with MW of approximately 53.7 kDa (MW of band minus MW of digested PlnD1 divided by the MW of PEG = 7.1 PEG chains) with a density of 1.5 relative to the lowest band. The second band (the densest band) with MW of approximately 41.8 kDa (~3.6 PEG chains) had a density of 5 relative to the lowest band. The third band had the lowest MW of 32.2 kDa (~0.8 PEG chains) and acted as the reference band for comparison to the other bands. Because the digested and undigested PlnD1 samples differentially bound to nitrocellulose, exposure times were varied to best visualize all lines and these are shown side by side in Fig. 4. Western blots with constant exposure times are included in the supplementary materials (Supplemental S2) for comparison. The densitometry analysis of the three bands in lane 4 revealed that the PlnD1-PEG-Ac solution contained PlnD1 with varying degrees of PEGylation: 67% PlnD1 was modified with 3–4 PEG chains, 20% PlnD1 with 7 PEG chains, and 13% with 1 PEG chain.

Fig. 4.

PEGylation of PlnD1. (A) Schematic of modified and unmodified PlnD1 with and without enzymatic digestion removing GAG chains: (1) PlnD1, (2) PlnD1 after GAGase digestion, (3) PlnD1-PEGAc, and (4) PlnD1-PEG-Ac after GAGase digestion. (B) Western analysis of enzymatically digested modified (4) and unmodified (2) PlnD1 demonstrated that PlnD1 was successfully conjugated to PEG (MW 3400). This conjugation was indicated by increase in MW of PlnD1-PEG-Ac Digested (4) over PlnD1 Digested (2) corresponding to the attachment of PEG chains. The film exposure time and protein content varied per lane and were as follows: Lane 1 contained 2 μg PlnD1 with film exposure for 10 s; Lane 2 contained 2 μg digested PlnD1 with 5 s film exposure; Lane 3 contained 1 μg PlnD1-PEG-Ac with 3 s film exposure; and Lane 4 contained 0.5 μg digested PlnD1-PEG-Ac with 3 s film exposure. Detection in panel B using primary antibody against perlecan (N-20; sc-27448 Santa Cruz Biotechnology; 1:2000). A bracket identifies the location of the polydisperse PlnD1 sample in Lane 1.

3.5. HBGF gradient formation

FGF2 was fluorescently tagged to allow for visualization of HBGF gradients in PlnD1-gradient hydrogels. The resultant FGF2–488 molecule was analyzed using the NanoDrop spectrometer and determined to have approximately 0.62 fluorescent molecules per FGF2 molecule (labeling scheme and data shown in Supplemental S3). Fig. 5A displays the schematic of the PlnD1 block-source hydrogel containing no PlnD1 gradient, and the PlnD1-gradient hydrogel containing a high-low-high PlnD1 concentration gradient across the width of the hydrogel. To visualize the gradient formation, PlnD1-PEG-Ac gradient hydrogels were submerged in a solution of FGF2–488 and imaged after two days, as displayed in Fig. 5B. The gradient of FGF2–488 was formed (Day 2, Fig. 5B) and maintained for at least a month (Supplemental S4). The profile of the FGF2 bound to PlnD1 block-source hydrogel (Fig. 5B, top) showed high binding of FGF2 to the PlnD1-containing block hydrogel followed by a steep drop off in intensity or binding of FGF2 to the hydrogel adjacent to the block that did not contain PlnD1, demonstrating that FGF2 bound to the PlnD1 in the block hydrogel and that it did not form visible long-range gradients of FGF2 into the adjacent HA-SH/PEGDA hydrogel. In the PlnD1-gradient hydrogel (Fig. 5B, bottom), FGF2–488 bound highly to both edges of the hydrogel and decreased gradually towards the center of the hydro-gel in a high-low-high concentration gradient of FGF2–488. As seen in the images, the highest irregularity is seen in the portion of the hydrogels at the periphery. This may occur during the removal of the PDMS mold from the crosslinked hydrogel. The relative fluorescence intensity plotted in Fig. 5C further displayed the high-low-high gradient of FGF2–488 in the PlnD1 gradient hydrogel and the steep drop of the FGF2–488 intensity in the block-source hydrogel. The gradient of FGF2–488 displayed in Fig. 5B appears smoother than the gradient of microparticles in Fig. 2B because of the size difference between the microparticles and the FGF2–488 molecules. The confocal microscope used to image these gradients can readily resolve micron-sized particles, but not nanometer-scale molecules, as individual objects. The microparticles used were 10 μm in diameter and the FGF2–488 molecules were on the scale of nanometers, roughly 1000-fold smaller than the microparticles. The microparticles were encapsulated as static “cell substitutes” whereas the unbound FGF2–488 could diffuse freely through the hydrogel.

Fig. 5.

Gradient and Block-Source PlnD1-Conjugated Hydrogels Sequester FGF-2. (A) Schematic of “block-source” presentation of PlnD1 on left side of hydrogel in 1 mm wide strip along side of hydrogel and of PlnD1 as a gradient from left and right sides of hydrogel. (B) After 2 days of incubation and washes every two days using plain PBS, AlexaFluor488-labeled FGF2 (FGF2–488) in hydrogels was imaged using confocal microscopy. Gradient of FGF2 established only after incubation in PlnD1 gradient, not upon incubation with “block-source” PlnD1 as is displayed in (C) relative fluorescence intensity measured across hydrogel. The plot in 5C was generated using MATLAB and quantified as average fluorescence intensity in columns of one pixel width across the hydrogel and begins at the edge of the hydrogel, indicated using arrowheads, ▲, where the fluorescence intensity is greatest.

3.6. Preosteoblastic cell response to HBGFs in PlnD1 block-source hydrogels

MC3T3-E1 cells were used to test the ability of HBGFs to elicit a directional migratory response in a cell line capable of rapid motility. The hydrogel schematics for both the PlnD1 block-source hydrogel and the treatment control hydrogel are displayed in Fig. 6A. MC3T3-E1 cells were encapsulated in hydrogels directly next to PlnD1 block-source hydrogels (Fig. 6A) or blank hydrogels (Fig. 6B) and were treated with FGF2- and PDGF-containing medium. The phase images to the right of the schematics (Fig. 6A & B, right) displayed cells in hydrogels after seven days of culture. In the PlnD1 block-source hydrogel condition, cells are displayed as growing in a branched or honeycomb-like pattern into the left hydrogel containing PlnD1 (Fig. 6A). These cells migrated towards the PlnD1/HBGF hydrogel. In the treatment control hydrogel condition, the cells encapsulated in the right hydrogel remained in the right hydrogel (Fig. 6B). MC3T3-E1 cells did not migrate toward the blank hydrogel as they did towards the PlnD1-hydrogel, even though HBGFs were supplied exogenously to both groups.

Fig. 6.

MC3T3-E1 Cell Response to HBGFs bound to PlnD1 Block-Source Hydrogel. (A) Schematic of “block-source” hydrogel on left, with cells in the right block hydrogel and PlnD1 in the left block hydrogel. With the addition of exogenous FGF2 and PDGF to the culture medium, MC3T3-E1 cells migrated to the PlnD1-containing hydrogel over seven days of culture. (B) Schematic of control blank hydrogel in the left block and cell-encapsulated hydrogel in the right block. After seven days of culture with exogenous FGF2 and PDGF, MC3T3-E1 cells do not migrate into the left, blank hydrogel.

3.7. Breast cancer cell migratory response to HBGF gradients

To visualize the migration of highly invasive MDA-MB-231 breast cancer cells over time, these cells were encapsulated in a high-medium-low cell density gradient across the hydrogel width (Fig. 7A), and cell density was quantified in the low seeding density side (target region of interest [ROI], outlined with the white dotted line in Fig. 7C) of each of the hydrogels. Because MDA-MB-231 cells formed clusters in these hydrogels, it was not possible to separate fluorescence intensity of individual cells stained for F-actin and vimentin using Imaris software algorithms, instead, cell volumes were quantified as a surrogate value of cell density. As seen in Fig. 7B and C, the cell density in the target ROI was greatest in the PlnD1-gradient hydrogel in 10% FBS-medium. The cells in PlnD1-gradient hydrogels in 1% FBS with HBGF migrated to the same extent as those in the PlnD1-uniform hydrogels in 10% FBS as quantified in the cell density graph. The cells in PlnD1-uniform hydrogels with 1% FBS with HBGFs migrated to a lesser extent than did the cells in the PlnD1-gradient hydrogel in the same medium, but the same as the gradient hydrogel in 1% FBS. Finally, the cells migrated the least in PlnD1-uniform hydrogels in 1% FBS. During live imaging of the cells in 12-hour intervals, the cells moved in a characteristically amoeboid-like fashion with rounded cells sending blebs through gaps in the matrix almost as if oozing through the hydrogel, instead of mesenchymal movement during which cells would appear stretched out and elongated during migration. The MDA-MB-231 cells stained positive for RhoA (Supplemental S5), a marker for amoeboid movement and further evidence for amoeboid movement in these hydrogel matrices.

Fig. 7.

MDA-MB-231 Cell Response to HBGFs associated with PlnD1 Conjugated as a Gradient or Uniform Concentration. (A) Schematic of PlnD1 Gradient hydrogel (left) and PlnD1 Uniform hydrogel (right) with PlnD1 (blue), cells (red), and microparticles (green). Hydrogels were then separated into groups treated with media containing different FBS concentrations or the addition of exogenous HBGFs. (B) After 7 days in culture, hydrogels were fixed and imaged across the hydrogel width and migrated clusters of cells were quantified as volume of cells staining positive for vimentin and F-actin in the side of the hydrogel with the low cell seeding density and normalized to volume. Statistical significance was determined using Student’s t-tests between groups Symbols *, ** denote statistically significant differences between the means, p < 0.05. (C) Representative maximum projection images generated using confocal microscopy for the corresponding groups were displayed, scale bar = 1 mm. Dotted white line represents ROI of 30% of hydrogel on the low cell density side of the hydrogel. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

4. Discussion

HSPGs aid in the establishment of HBGF gradients in vivo that are integral to tissue formation in embryogenesis, tumor growth in cancer biology, and tissue repair in a wound response. While heparin has been used extensively to sequester growth factors, proteoglycans have not been used as widely in tissue engineering applications [59]. This work utilized PlnD1 to ionically bind HBGFs to form physiologically-relevant gradients for cell migration.

Commercially available 3D printers were used to produce the negative mold for the MGMD in this work, demonstrating the broad utility of 3D printers in biomedical applications. With 3D printing, the user is limited to the resolution of the printer itself, and as technology improves over time the possibility of using 3D printed parts to generate templates for microfluidic devices becomes a possibility. In our MGMD, the herringbone features were the smallest features and were considered large compared to microfluidic devices, as 0.5 mm was the smallest dimension. So, while our MGMD was largely unaffected by the resolution limitations of 3D printing, microfluidic devices would be greatly affected by these limitations, making photolithography still the gold standard and most common practice for generating microfluidics systems. This work, however, demonstrated the use of 3D printing to generate negatives of the intended device design and for use in larger scale cell biology and tissue engineering applications.

Microparticles acted as a surrogate for cells during the development of the MGMD. Although the microparticles used in these studies were fluorescent and not functionalized to bind biological molecules, the generation of microparticle gradients using the MGMD demonstrated the utility of the MGMD for gradient formation of micron-sized particles and cells alike. One future application of the MGMD could include the generation of microparticle gradients in hydrogels, in which the microparticles contain a drug or other biological molecule, to determine the effects of the drugs on co-encapsulated cancer cells. Gradients of chemotherapeutic drugs at therapeutic doses through a matrix dense with co-encapsulated cancer cells could mimic the gradient of drugs through tumor tissues and results be used to determine if the drug can kill cancer cells on the lose-dose side of the hydrogel. Additionally, varying the flow rate of the hydrogel precursor solutions yielded microparticle gradient profiles varying in steepness which could be modulated to achieve the gradient profile desired for a specific application. For instance, to mimic a cancer tumor with leaky vasculature, perhaps a more gradual or less steep gradient profile is desired, so the flow rate of precursor hydrogel solutions through the MGMD can be decreased to achieve such a profile.

The high viability of primary hS/PCs after flow through the MGMD that we observed suggested that the MGMD did not exert stresses upon the cells in excess of what was exerted on the cells from pipetting into the cylindrical molds. Flow through the MGMD at even the highest flow rate can maintain cell viability. The formation of hS/PC spheroids after seven days in culture after flow through the MGMD and after standard hydrogel formation methods suggested that these cells were largely unaffected by the MGMD. Primary hS/PCs continued to proliferate and form multicellular structures after seven days in culture. In addition to the hS/PCs, cell lines such as the MC3T3-E1 and MDA-MB-231 have demonstrated high viability after flow through the MGMD demonstrating the broad applicability of the MGMD for use with multiple cell types.

Although PlnD1 has been conjugated to microparticles [58,60] and electrospun scaffolds [61], this work marks the first time that the PlnD1 has been PEGylated for conjugation as a concentration gradient to a hydrogel backbone. Instead of directly conjugating the HBGFs to the hydrogel scaffold, conjugating PlnD1 allowed for diffusion of HBGFs through the matrix and bioaccumulation on HS chains. HS chains allow for diffusion and bioavailability of these HBGFs due to patterned sulfation at the 2- and 6-O-sulfation sites [12]. The regions of low sulfation of HS leaves segments of the GAG chain accessible for heparanase to cleave the HS chains [62]. This allows for the ternary complexation of HBGF-HS-receptor binding that is key to cell signaling [18,63,64]. Without the complexation of HS with HBGF and receptor, little or no signal would be transduced. Thus, PlnD1 is an ideal physiological way to provide HS for HBGF gradient formation and delivery. PEGylated PlnD1 proved to be an excellent way to mimic how HBGF gradients are established in vivo.

The visualization of PlnD1-PEG-Ac gradients using antibodies against N-terminal perlecan and alcian blue staining to display the HS chains were attempted, but were not successful. The antibody staining of the hydrogels was undertaken with an antibody optimized for western blotting and did not work well in 3D hydrogels; alcian blue stained the entire hydrogel even when the pH was modulated to stain sulfated regions preferentially. However, gradients of HBGFs was the ultimate goal for the use of PlnD1, so utilizing fluorescently tagged FGF2 was optimal to demonstrate both the PlnD1-gradient formation as well as HBGF-gradient formation. The distribution of the FGF2–488 in the PlnD1-gradient hydrogel (Fig. 5) demonstrated a pattern of FGF2 binding similar to the profile generated with microparticles (Fig. 2). These results suggested that gradients of PlnD1 were necessary to generate long-range gradients of HBGFs and not just block-source PlnD1 that would rely solely on diffusion to produce gradients of HBGFs.

The incorporation of PlnD1-PEG-Ac into the hydrogel precursor solutions, while necessary for HBGF binding, added an additional crosslinking molecule to the precursor solutions because PlnD1-PEG-Ac had an average of 3–4 PEG groups per PlnD1 molecule. However, while the crosslinker, PQ, at a SH:AC ratio of 9:1 had a concentration in the hydrogel of 3.1 × 10⁻⁴ M and the cell binding motif, GRGDS, concentration was 3 × 10⁻³ M, the concentration of PlnD1-PEG-Ac was roughly 8 × 10⁻⁷ M, so PlnD1 was incorporated at a concentration that was orders of magnitude less than the other components of the hydrogel. The PQ crosslinker must be used at concentrations corresponding to specific crosslinking ratios that can vary for different cell types that require unique hydrogel mechanical properties. For instance, MC3T3-E1 cells are preosteoblastic, requiring slightly more crosslinked, stiffer hydrogels than those preferred by MDA-MB-231 breast cancer cells. The optimal crosslinking ratio for each cell type must be determined empirically. However, the PQ crosslinker is very versatile because of the range of cell-secreted MMPs that can cleave the α(I) collagen-derived peptide sequence allowing for local matrix degradation [54,65–68].

The migration of MC3T3-E1 cells into hydrogels with PlnD1 suggested that the HBGF in the medium associated with the heparan sulfate chains in the PlnD1-containing hydrogel and acted as a growth factor reservoir and a short-range gradient for cell migration. In contrast, in the hydrogels without PlnD1, the HBGF did not get absorbed into the gel so the cells in the adjacent hydrogel did not sense a concentration gradient in the direction of the hydrogel without PlnD1. Similarly, MDA-MB-231 cells demonstrated a migratory response to the PlnD1-gradient hydrogels and to a much lesser extent in the PlnD1-uniform hydrogels. We propose that in PlnD1-uniform hydrogels the cells were surrounded by the highest concentration of PlnD1 used in the gradient hydrogels, so the cells were surrounded by HS/HBGFs. In contrast, in the PlnD1-gradient hydrogel the HS/HBGFs were on one side of the cells, so the cells sensed a concentration gradient and moved along the gradient towards the source. Additionally, MDA-MB-231 cells reportedly express high levels of heparanase to access HBGFs to stimulate proliferation, invasion, and metastasis [69]. MDA-MB-231 cells were previously shown to express high levels of MMPs [70] capable of cleaving the MMP-cleavable crosslinker used in this research. Interestingly, both latent and active forms of heparanase increased RhoA activity in brain metastatic melanoma cells [71]. Increased expression of both heparanase [69] and RhoA [72] by MDA-MB-231 cells could explain the amoeboid-like movement of the cells in the PlnD1-gradient hydro-gel systems detailed herein. Also interesting, the use of 10% FBS medium with the MDA-MB-231 cells showed the highest migratory response of the cells, so the addition of TGF-β1 and HB-EGF at 50 ng/mL was not enough to induce the response seen with 10% FBS. While the growth factor and protein contents of FBS vary by manufacturer and lot, the HBGFs FGF2 and TGF-β1 have been measured in FBS previously [73]. The presence of these HBGFs and other factors in the FBS provided a stimulus for cell migration.

In summary, typical 3D hydrogel systems employ uniformly distributed molecules that surround encapsulated cells but do not closely mimic the native tissue. Here, the incorporation of PlnD1 gradients throughout the matrix was used to simulate in vivo morphogen gradients and to stimulate a migratory response of encapsulated MC3T3-E1 and MDA-MB-231 cells. By changing the flow rate or by incorporating other cells or morphogenic factors, this MGMD platform could be leveraged for other applications in which HSPGs and HBGFs are key determinants of cell behavior. Both micro- and nano-scale objects such as cells, particles, and proteoglycans can be formed into gradients in hydrogels using this MGMD, providing a versatile platform on which to create a variety of specially configured gradient hydrogels for a host of applications in cell and cancer biology and tissue engineering.

5. Conclusions

A multichannel gradient maker device that facilitated the formation of gradients of microparticles, cells, and proteoglycans across the width of hydrogel was developed in this work. We demonstrated the fabrication of soluble HBGF gradient hydrogels in the millimeter to centimeter-scale size, making this construct relevant for future in vivo applications. No previous work has demonstrated the production of heparan sulfate proteoglycan gradients within 3D hydrogels in which cells can be co-encapsulated, marking this work as a first. While this platform has demonstrated usefulness to monitor migratory responses of cells to physiologically relevant short-range and long-range HBGF signaling, this platform was ultimately developed to provide gradients of HBGFs as morphogenic factors similar to those seen in organogenesis, wound healing, and cancer biology. This hydrogel platform will be used in combination with various HBGFs and cell types to engineer tissues that more closely mimic the functional state of the target tissue. One area of interest in our lab is the use of these PlnD1/HBGF-gradient hydrogels to induce branching morphogenesis of encapsulated hS/PCs by subjecting the cells to morphogen gradients. These gradient hydrogels represent a step towards replicating the in vivo morphogenic microenvironment, making the MGMD a versatile tool for healthy and cancerous tissue engineering applications.

Statement of Significance.

Gradients of heparin binding growth factors (HBGFs) direct cell behavior in living systems. HBGFs bind electrostatically to gradients of HS proteoglycans in the extracellular matrix creating HBGF gradients. We recreated HBGF gradients in physiological hyaluronate-based hydrogels using a 3D-printed multi-channel gradient maker device (MGMD) that created gradients of HS proteoglycan-derived perlecan/HSPG2 domain I. We demonstrated the ability of a variety of cells, including primary salivary stem/progenitor cells, pre-osteoblastic cells and an invasive breast cancer cell line, to be co-encapsulated in gradient hydrogels by flowing them together through the MGMD. The versatile device and the ability to create HBGF gradients in hydrogels for a variety of applications is innovative and of broad utility in both cancer biology and tissue engineering applications.