Integration of Self-Assembled Microvascular Networks with Microfabricated PEG-Based Hydrogels

2.1. Device Microfabrication

To permit self-assembled microvascular network integration with perfusable microfabricated channels, we developed a dual mode, multilayer, microfabrication strategy to build a microfluidic poly(ethylene glycol) (PEG) hydrogel within a polydimethylsiloxane (PDMS) housing as shown in Figure 1. First a transparent PDMS housing with fluid access ports was fabricated using standard soft lithographic techniques. Next, an aqueous solution containing the provasculogenic PEG hydrogel precursors along with a 4:1 mixture of HUVEC and 10T1/2 cells was injected into the PDMS housing. A photomask was used to photolithographically crosslink the cell laden pro-vasculogenic hydrogel in a confined geometry between the perfusion channels. Finally, the PDMS-PEG device was sealed to cover glass and channel perfusion was initiated at flow rates derived from normal physiologic microvascular shear stresses (15 dyne/cm²) using two syringes pumps.

Hydrogel incorporation within a robust PDMS housing addresses several logistical challenges associated with microfluidic hydrogels without the need for complicated supporting apparatuses.^[11] Furthermore, this multilayer approach provides independent control over PDMS housing and PEG hydrogel geometry allowing for freeform user-defined fabrication of tissue-mimetic hydrogel structures within the PDMS housing. A particularly advantageous characteristic of this system is the fabrication of a solution-hydrogel interface between the perfused microchannel and the self-assembled vessel networks within the hydrogel bulk. This interface promotes the connection of fabricated channels with self-assembled vessel structures and facilitates mass transport across large length scales from fabricated channels 1 mm in diameter, a size amenable to surgical anastomosis during implantation, to microvascular networks 30 μm in diameter.

2.2. Evaluation of Self-Assembled Microvascular Networks

The self-assembly of encapsulated cells into vessel-like structures is confirmed by the confocal image in Figure 2. Here we note that HUVECs (green, fluorescent anti-PECAM) arrange in tubule networks that are stabilized on their external surface by pericyte-functioning 10T1/2 cells (red, fluorescent anti-smooth muscle α-actin). In order to promote spatially homogeneous microvascular network formation within the device, it is advantageous to maximize this network self-assembly as a function of distance from the perfused channel. Though the inclusion of microfluidic channels within engineered tissue constructs has been shown to improve mass transport and increase cell viability,^[12] the fraction of viable cells decreases significantly with increasing distance from the media-perfused conduit.^[6,7] Formation and perfusion of 3D, self-assembled microvascular networks between fabricated microchannels shifts the characteristic diffusion length from the process limited channel spacing^[13] to the shorter distance arising between self-assembled vessels formed from the co-culture.

We set out to determine the maximum distance from a perfused media channel that supports microvascular network self-assembly and cell survival in the pro-vasculogenic PEG hydrogel. For this study, dually-perfused microchannel devices were operated in either a Media-Buffer configuration, roughly analogous to capillary-lymphatic mass exchange, or a Media-Media configuration, more similar to inter-capillary transport. Spatiotemporal vessel formation was evaluated by morphological assessment of the number of tubules, as well as total tubule lengths and immunohistochemical analysis with an antibody to the endothelial mem brane marker PECAM. An apoptotic assay was conducted concurrently to determine the relationship between microchannel configurations and cell viability.

2.3. Discerning Transport Modes in Vascularized and Non-Vascularized Hydrogels

The vasculature provides mass transport over large distances in the body via rapid, lumen-supported convection followed by diffusion out of the microvascular networks and into the tissues.^[15,16] In vitro integration of self-assembled microvascular networks and perfusion with fabricated channels recapitulates in vivo network hierarchies and promotes efficient, biomimetic mass transport regimes with the potential to yield more clinically relevant engineered tissues. In validating the effectiveness of our microfluidic system, we compared the mass transport characteristics between vascularized and non-vascularized hydrogels. Pro-vasculogenic PEG hydrogels containing the co-culture of HUVEC and 10T1/2 cells were perfused in the Media-Media configuration for 96 h to drive microvascular network self-assembly and integration with fabricated microchannels. After 96 h of culture, growth medium was supplemented with high molecular weight fluorescent dextran (2 × 10⁶ Da, lysine fixable) and time-lapse confocal microscopy was used to image transport through the vascularized construct for 180 minutes. Non-vascularized controls were fabricated identically but without the inclusion of cells. Dextran accumulation and the spatiotemporal fluorescence intensity landscape in the region adjacent to the perfused microchannels were compared between the two systems.

Though encapsulated cells in vascularized hydrogels have the capacity to alter the diffusive properties of the material via polymer degradation or extracellular matrix (ECM) deposition during long term culture, our observations suggest that the vascularized system supports enhanced mass transport characterized by combined convection and diffusion. Non-vascularized hydrogels exhibited a uniformly advancing dextran intensity front characteristic of 1D Non-Steady State Fickian diffusion of a solute through a hydrogel matrix (Figure 6a and Figure 7a),^[12,17,18] while vascularized hydrogels displayed rapid, convective-like dextran transport in regions spatially coinciding with vessel structures (Figure 6b, c, and d). At each time point, fluorescent dextran can be seen at greater distances from the perfused microchannel in the vascularized construct when compared to the non-vascularized construct (Figure 6a and b, and 7a). Accordingly, at a fixed distance from the microchannel, the rate of dextran accumulation in the vascularized system was significantly greater than the non-vascularized system during the first 30 minutes of perfusion (p < 0.05, n = 9, Figure 7b). Interestingly, this difference in movement in the vascularized and nonvascularized hydrogels (Figure 7b,% Convective Transport) is most extreme at early time points when convection is dominant and decreases over time as system transport becomes diffusion limited.

Spatially, we observed rapid intensity increases in dextran distribution in the vascularized system occurring ahead of the diffusion front, indicative of convective, vessel-supported transport. Examples of this occurrence are highlighted by the arrows in Figure 6c, and are also visible in Figure 7a. This observation suggests a shift in transport regimes from Fickian diffusion to rapid lumen-supported convection through the vessel networks. To further validate that microvascular lumen supported convective transport, dextran perfused vascularized constructs were fixed and immunohistochemically evaluated for the endothelial marker PECAM to determine the proximity of perfused dextran to vessel lumen. Confocal imaging showed perfused dextran predominantly co-localized with the lumen of microvascular structures, further demonstrating transport via the newly formed vessels (Figure 6d).

In addition to supporting convective transport, vascularized constructs displayed a greatly reduced characteristic diffusion length (λ_V = 67 μm) when compared to the acellular system (λ_D = 3 mm). Diffusion lengths in the vascularized hydrogel are defined by vessel density and spacing, not overall construct geometry (Figure 7a). This significant reduction in required solute travel highlights an important shift towards a biomimetic transport regime wherein characteristic diffusion lengths are less than the critical necrotic distances. Time lapse dextran intensity profiles also show broadening of profile intensities and a decrease in peak intensity aspect ratio over time (Figure 6b and Figure 7c and d). Together these observations suggest vascularized constructs support initial convective-like transport and extra-vessel diffusion into the interstitial space, recapitulating the functional properties of the microvasculature.

Cell Maintenance

Human umbilical vein endothelial cells (HUVECs) (Lonza, Walkersville, MD) were cultured in endothelial growth medium (EGM-2, Lonza) supplemented with 2 mm L-glutamine, 1000 U mL^–1 penicillin, and 100 mg L^–1 streptomycin (Sigma-Aldrich, St. Louis, MO) and used from passages 3–8. 10T1/2 cells (American Type Culture Collection, Rockville, MD) were maintained in Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 2 mm L-glutamine, 1000 U mL^–1 penicillin, and 100 mg L^–1 streptomycin (Sigma-Aldrich) and used from passages 16-20. All cells were incubated at 37 °C in a 5% CO₂ environment.

Proteolytically Degradable PEG Pre-Polymer Synthesis

Extracellular matrix (ECM) derived peptide sequences have been shown to be specifically cleaved by matrix metalloproteases (MMPs) secreted by tissue cells.^[19] Incorporation of the collagen mimetic peptide sequence (GGPQGIWGQGK) into materials based on non-proteolytically degradable synthetic polymers has been shown to result in complete degradation of the polymer network by specific MMPs as well as other broad acting proteases.^[10] Furthermore, the conjugation of the GGPQGIWGQGK sequence to monoacrylate, photocrosslinkable PEG macromers in an ABA block co-polymer configuration has demonstrated the ability to promote hydrogel remodeling and support microvascular network formation by HUVECs and 10T1/2 co-cultures.^[10]

The peptide sequence GGPQGIWGQGK was generated by standard solid phase FMOC chemistry using an Apex396 peptide synthesizer (Aapptec, Louisville, KY). Following synthesis and purification, the peptide product was verified by matrix-assisted laser desorption ionization time of flight mass spectrometry (MALDI-ToF; Bruker Daltonics, Billerica, MA). A diacrylate ABA block co-polymer of the peptide sequence was synthesized by reacting heterobifunctional acrylate-PEG-succinimidyl carboxymethyl (Laysan Bio, Arab, AL; acrylate-PEG-SCM, 3400 Da) in a 1:2.2 (peptide:PEG) molar ratio in anhydrous dimethylsulfoxide (DMSO, Sigma-Aldrich) in the presence of 15 mm N,N-Diisopropylethylamine (DIPEA, Sigma-Aldrich) overnight. This reaction conjugates PEG monoacrylates to both the amine terminus of the peptide sequence and to the lysine amino side-chain at the peptide carboxy terminus using standard succinimideamine chemistry. The resulting reaction volume was slowly diluted 2:1 in an ice bath with ultrapure H₂O and dialyzed (5000 MWCO regenerated cellulose, Spectrum Laboratories, Rancho Dominguez, CA) against ultrapure H₂O for 48 h to remove DIPEA, DMSO, and un-reacted PEG and peptide moieties. The product was then frozen at –80 °C, lyophilized for 72 h and stored under argon at –80 °C until further use. The cell adhesive ligand Arg-Gly-Asp-Ser, (RGDS, American Peptide Sunnyvale, CA) was conjugated to acrylate-PEG-SCM in a 1:1.1 (PEG:RGDS) molar ratio under the same conditions as the PEG-Peptide-PEG block co-polymer. Successful polymer conjugations were confirmed via gel permeation chromatography with UV-vis and evaporative light scattering detectors (Polymer Laboratories, Amherst, MA).

Multilayer Photolithographic Fabrication of Perfused PEG Scaffold

Poly(dimethylsiloxane) (PDMS, Sylgard 184) housings were prepared using standard soft lithographic techniques as previously described.^[12] Briefly, PDMS (15:1, silicone elastomer: curing agent) was molded to photoresist masters (SU-8 2100, 350 μm in height, Microchem, Newton, MA) to create a robust housing with perfusion access ports and transparent optical qualities for ease of imaging (Figure 1). All PDMS surfaces were treated with the photoinitiator acetophenone (2,2-dimethoxy-2-phenyl acetophenone in n-vinylpyrrolidone, 300 mg mL^–1, Sigma) for 5 min to promote free radical induced interfacial polymerization of the acrylate terminated PEG pre-polymer solution to the housing. Excess photoinitiator was removed via sequential rinsing with 100% ethanol and ultrapure H₂O, which was followed by drying with filtered air. Acetophenone-coated PDMS housings were sealed to glass slides treated with SigmaCote (Sigma-Aldrich, St. Louis, MO) and overlaid with a transparency photomask (20,000 dpi, CAD/ART Services, Bandon, OR) aligned to control spatial photopolymerization and hydrogel geometry within the PDMS (Figure 1).

The MMP-degradable PEG block copolymer and PEG-RGDS cell-adhesive derivative were dissolved in EGM-2 to final concentrations of 2.5% (w/v) and 2 mm, respectively. The photoinitiator Irgracure-2959 (Ciba, 100 mg mL^–1 in 200 proof ethanol) was added to the pre-polymer solution at 30 μL mL^–1. HUVECs and 10T1/2 cells were suspended in the pre-polymer solution in a 4:1 ratio (HUVEC:10T1/2) to a final concentration of 30 × 10⁶ cell mL^–1. The resulting cell/pre-polymer solution was injected into the PDMS housing via the perfusion access ports and exposed through the photomask to long wavelength UV light (365 nm, 10 mW cm^–2) for 5 min to fabricate the PEG hydrogel structures. The PDMS housing with the embedded photocrosslinked PEG hydrogel was gently rinsed with sterile phosphate buffered saline (PBS, pH 7.4), dried with canned air and sealed to #1 cover glass. The cell/hydrogel construct was perfused at 600 μL h^–1 with EGM-2 at one interface and sterile PBS on the opposite interface to establish a gradient of growth factors, nutrients, and waste species across the construct (Media-Buffer configuration, Figure 1). Some experiments were conducted in a Media–Media configuration in which both hydrogel construct channel interfaces were perfused with EGM-2 in order to determine the maximum distance microvascular networks would develop from the perfused nutrient source.

Immunohistochemistry

Immunofluorescence analysis was performed to determine spatiotemporal microvascular network characteristics and the cell-cell relationships within the co-culture. At 0, 48 and 96 h perfusion was terminated and hydrogel microdevices were fixed, permeabilized, and reacted with primary and secondary antibodies (Figure 2). The primary antibodies used were mouse anti-smooth muscle α-actin (Sigma-Aldrich) to label the 10T1/2 cells and goat anti-PECAM (Santa Cruz Biotechnology, Santa Cruz, CA) to label membranes of HUVECs. The secondary antibodies were anti-mouse and anti-goat conjugated with either Alexa fluor 488, Alexa fluor 532, or Alexa fluor 647. DAPI (300 nm, Invitrogen) was added as a nuclei counter stain for all immunofluorescence images.

Spatiotemporal Microvascular Network Morphology Analysis

Spatiotemporal PECAM immunofluorescence images were analyzed at 0, 48 and 96 h post device fabrication to determine the effects of perfusion time and distance from the EGM-2 perfused microchannel on microvascular network morphology (Figure 3a and 3b). The PECAM immunofluorescence was imaged using a confocal microscope (Zeiss LSM 5 Live, 20x/NA 0.8 objective, FOV 318 × 318 μm, vertical z-stack: 1 μm slices, 50 slices per region, Carl Zeiss Inc., Thornwood, NY). Image regions were tiled at 300 μm intervals radially outward from the perfused media microchannel wall to the PBS microchannel wall for a total of 10 regions spanning the 3 mm thickness of the hydrogel construct.

Pre-processing of the LSM images was accomplished by first applying a maximum intensity projection to the z-stack using the image browser function in the LSM Image Examiner software (Carl Zeiss Inc., Thornwood, NY). After maximum intensity projection, channels were split and the PECAM immunofluorescence channel was thresholded to preserve the tubule elements, while decreasing the background signal. The resulting thresholded image was then used to measure the lengths of the tubule elements within the construct. Each tubule was measured manually using the image browser measurement tool, where single tubules were determined as starting at one branch and ending at another. Both the tubule lengths and the total tubule number were recorded for each image at different locations within the constructs.

Spatiotemporal TUNEL Apoptosis Assay

Spatiotemporal apoptotic profiles at 48 h and 96 h of culture were analyzed using the Click-It TUNEL assay for DNA fragmentation (Invitrogen, Carlsbad, CA). Samples were fixed, permeabilized, and labeled according to manufacturer recommended protocols and counterstained with DAPI (300 nm, Invitrogen). Fragmented DNA fluorescence was imaged using a confocal microscope (Zeiss LSM 5 Live, 20x/NA 0.8 objective, FOV 318 × 318 μm, vertical z-stack: 1 μm slices, 50 slices per region). Again image regions were tiled at 300 μm intervals radially outward from the perfused media microchannel wall to the PBS microchannel wall for a total of 10 regions spanning the 3 mm thickness of the hydrogel construct. At each region and time point, the ratio of apoptotic cells (TUNEL positive) to total cells (DAPI positive) was analyzed and plotted to indicate percentage of apoptotic cells. Control samples treated with 70% ethanol were used to set the TUNEL positive imaging parameter values and to avoid false positive TUNEL signal.

Analysis of Acellular and Vascularized Mass Transport

System mass transport properties were analyzed in vascularized and acellular PEG hydrogel constructs after 96 h of culture. High molecular weight fluorescent dextran (dextran- FITC, lysine fixable, 2 ×10⁶ Da; Invitrogen, Carlsbad, CA) was added to the perfusion media and flow was set at 600 μL h^–1. Images of the region immediately adjacent to the microchannel wall were collected for 120 min via time lapse live-cell fluorescent confocal microscopy (Figure 4a, 4b, and 4c) (Zeiss LSM 5 Live, 20x/NA 0.8 objective, FOV 318 × 318 μm, vertical z-stack: 10 μm slices, 6 slices per region). Following time lapse imaging, a subset of vascularized hydrogels was fixed using 4% paraformaldehyde and then reacted with the PECAM antibody and DAPI nuclear counterstain to determine the spatial localization of dextran relative to vessel structures (Figure 4d).

Differences in the dextran fluorescent intensity landscape of the acellular and vascularized images were assessed using the Plot Profile function in Image J (NIH, Bethesda, MD). Raw intensity values were normalized to maximum intensity in order spatially overlay the data (Figure 5a). As shown in Figure 5b, the rates of dextran transport in the acellular and vascularized systems were assessed by measuring the total fluorescence intensity in a region 50 μm from the channel wall over time using the Image Processing Toolbox in MatLab (Mathworks, Natick, MA). The difference in vascularized and acellular dextran accumulation (Figure 5b,% Convective Transport, blue trace) was determined by taking the difference of vascularized and acellular values at a time point and plotting as a percentage of the vascularized value as in Equation 1. To evaluate extra-vessel diffusion, changes in vascularized intensity peak widths and peak aspect ratios (Equation 2) were measured over time at a constant intensity value with the results plotted in Figure 5c and 5d.

Equation 1 : Equation for calculation of % convection transport trace.

Equation 2 : Calculation of peak aspect ratio.