Design, Fabrication and Implementation of a Novel Multi Parameter Control Microfluidic Platform for Three-Dimensional Cell Culture and Real-Time Imaging

Microfluidic Device Design

In order to satisfy these stated design criteria, a μFD was created (figure 1a) consisting of two parallel microfluidic channels for delivering cell culture medium and a central “gel cage” with a micro-pillar array for housing cell-seeded scaffolds. One primary objective was to facilitate the incorporation of most commonly used injectable cell culture scaffolds (hydrogels), which may be biologically derived (collagen gels or matrigel) or synthetic (self-assembling peptide gels)¹⁹. Gel cage geometry was optimized to permit the use of a range of hydrogel concentrations, with the posts providing physical support to prevent the fracture of brittle gels and provide stability for soft gels. The microfluidic network and gel cage were cast in PDMS which was bonded to glass after loading the scaffold (see details below). With the scaffold in place the two media channels are isolated from each other, communicating only through the gel containing region. This permits the establishment of gradients in both pressure and chemotactic factors across the gel. Either static or dynamic cultures can be established by integration with macro-scale plumbing. In dynamic cultures, the microfluidic chip is connected to an external flow circuit which allows the control of both surface shear stress and interstitial flow perpendicular to a cultured monolayer or past cells suspended in the gel. The external platform is equipped with two isolated media reservoirs having adjustable hydrostatic heads (producing a pressure gradient across the gel) at the inlet, two-way valves (to isolate the system when necessary) and outlet connectors to a programmable push/pull syringe pump (PHD 2000, Harvard Apparatus). Pressure gradients can also be generated by connecting reservoirs directly to the PDMS wafer. The entire μFD platform was designed to be small enough to fit on conventional microscope stage to allow real-time monitoring of cells over the course of an experiment.

Device Fabrication and Surface Modification

The design of the microfluidic network was created in AutoCAD (Autodesk, San Rafael, CA) with the dimensions of the microfluidic channels, gel “cage” and micro-pillars as provided in supplementary table 1. A transparency mask was created from the CAD file with a minimum geometric feature size of approximately 100 μm and printed by a high-resolution printer (PageWorks, MA). This transparency mask was used in photolithography of SU-8 photoresist to create a silicon wafer master. Microfluidic devices were made by replica molding²⁰ polydimethylsiloxane (PDMS) (Dow Corning, USA) and curing the degassed elastomer mix (10:1, base: curing agent) against the silicon master in an 80 °C oven for 2 hours. Polymerized PDMS wafers were peeled off the silicon master, individual microfluidic wafers (35 mm diameter, 0.8–1cm height) cut out and inlets and outlets cored down to microfluidic channels using a sharpened flat-ended 16 gauge needle. Prior to cell culture PDMS wafers were cleaned and sterilized at 120 °C for 20 minutes in a wet cycle followed by a dry cycle at 120 °C for 35 minutes (20 min sterilization/15min dry). Next, the PDMS surface was rendered hydrophilic by exposure to air plasma to facilitate scaffold loading (detailed below). Briefly, sterilized wafers were placed on trays in plasma cleaner (Harrick, CA) chamber (pattern side up). A pump down cycle (~ 2 minutes) was initiated followed by irradiation for 2 minutes with pink plasma. Surface treated wafers were stored in sterile container and used within 0.5 – 2 hrs following plasma treatment.

Scaffold Loading Microinjection System and Device Assembly

A microinjection station was created to load the cell culture scaffold (sub μL volumes) into the device under aseptic conditions (Fig. 1c). The system components included a manual micromanipulator (MN-151 Joystick Micromanipulator with H-7 Pipette Holder, NARISHIGE, NY), microliter syringe (Hamilton, 62RNR, 2.5 μL SYR, 22s/2″/3, VWR), digital microscope (Big Blue QX5, COMPUVISOR.COM, TX) (all housed in a laminar flow hood) and a monitor for visual guidance. The MN-151 joystick feature provided control of micro-scale adjustment in the XY plane with additional coarse adjustments along the X, Y and Z axes.

Demonstration of Concentration Gradient across Gel in OFD with and without Channel Flow

Collagen gels were formed as previously described. Following polymerization, microfluidic channels were filled with cell culture media (without supplements). Gradient studies were performed under static (no flow) (Fig. 2c) and dynamic (flow through the channels) conditions (detailed in supplementary material) with the media in one microchannel replaced by a dilute solution of fluorescent dextran (40 kDa, Invitrogen) at an initial concentration of 20 μg ml⁻¹. Fluorescent intensity was visualized with a Nikon TE300 microscope (Nikon Instruments Inc., NY). A series of fluorescent images (4× magnification) of the gel region were acquired with a Hamamatsu camera (Hamamatsu, Japan) using Openlab (Improvision, MA) data acquisition software and stored for further analysis. Images were processed to obtain the changes in fluorescent intensity across the gel at each time point. Image processing of time-lapse fluorescent images was performed using a custom written ImageJ (NIH public-domain image processing software) macro and MATLAB (MathWorks, MA) code. Briefly, parallel diagonal lines were drawn across the gel region from the sink to the source channel (Fig. 2a) in each fluorescent image. A clear path was defined which avoided the PDMS micro-pillars. Pixel intensities and corresponding location from the “source” channel were recorded for these line sections. Average fluorescent intensities were calculated for pixels at the same distances from the dextran channel for all pixels across the length of the gel. At each time point a plot of the normalized average intensity profile across the gel was generated. Variations in fluorescent intensity as a function of time were plotted at fixed locations within the gel region and compared to theoretical predictions (detailed in supplementary material) obtained from a finite element model generated in FEMLAB (Comsol, USA).

Cell Culture and in vitro capillary morphogenesis models

All cell cultures were maintained in a humidified incubator at 5% CO₂ and 37 °C. Human adult dermal microvascular endothelial cells (HMVEC-ad, LONZA, USA) were propagated in EGM-2MV media system with 5 % fetal bovine serum (LONZA formerly CAMBREX, USA). Cells were expanded on collagen-coated flasks and used at passages 6–8.

Characterization of capillary morphogenesis and tube-like structures

Phase-contrast, epifluoresence and confocal microscopy were used to characterize capillary morphogenesis and the 3D morphology of EC structures. Fluorescent and phase contrast images were acquired on Nikon TE300 microscope equipped with a Hamamatsu camera and Openlab image acquisition software. Time-lapse images were taken of live samples every 12 – 24 hours with phase contrast microscopy. At the end of an experiment, samples were fixed with 4.0 % paraformaldehyde (PFA) and tagged with a fluorescent marker for actin cytoskeleton and cell nuclei. Confocal images were collected using spinning disk confocal microscope (Zeiss Axiovert 200M) furnished with Imaging Suite (PerkinsElmer Life Science) acquisition software. A series of 100 optical serial sections (1 μm thick) were obtained. The aligned images were stacked and rendered for 3D visualization using Imaris (Bitplane, MN). In addition to confocal microscopy, in order to demonstrate the presence of open lumen-like structures (in fixed samples) the cell seeding channel was perfused with fluorescent micro-beads (supplementary movie 3).

Microvascular Endothelial Cells Sprouting Movies

To demonstrate the capability to monitor cells in real-time, time-lapse movies were recorded of endothelial cells during sprouting angiogenesis. An endothelial monolayer was formed on collagen gel as previously described. The μFD was kept in a custom built environmental control chamber at 37 ºC and 5% CO₂ and cells visualized with a Zeiss inverted microscope. To minimize evaporation during the course of the experiment, media reservoirs (with zero height differentials) were connected directly at each inlet and outlet ports. The device was then placed in a secondary container with a humidified local environment and cut-out glass window in the bottom for visualization. Bright-field images were taken with the AxioCam MRm (Carl Zeiss) (at single optical plane) at 2 minutes interval with AxioVision image acquisition software.

Cytoskeleton and Nuclei Staining

F-actin distribution and number of cells involved in “capillary – like” network or tube structures were assessed after 2–7 days culture in the μFD. F-actin and nuclei staining were performed after fixation with 4.0 % PFA (30 minutes). The fixed samples were rinsed twice with 1X phosphate buffered saline (PBS), treated with 0.1 % Triton-X (1–2 minutes), rinsed with 1X PBS followed by the infusion of a mixture of DAPI and rhodamine phalloidin (30 minutes) and a final wash step with 1X PBS.

Microfluidic with 3D cell culture capabilities (OFD)

A new μFD was designed and fabricated from PDMS using standard soft lithography and replica molding techniques. To facilitate precise control of the fluidic microenvironment (shear levels and soluble factors), the device consists of two parallel flow channels and a central gel cage which was designed to immobilize injectable scaffolds (soft hydrogels) for cell culture. A staggered array of micro-pillars is incorporated in gel cage to provide mechanical stabilization for the scaffolds which can biologically-derived or synthetic. We have found that this specific design provides optimal mechanical support (supplementary Fig. 1) for gels and can sustain pressure differentials in excess of several cm of water. With the gel in place the two microfluidic channels are essentially isolated from each except for the diffusion or convection of soluble factors through the porous scaffold from one channel to the other.

Concentration Gradient maintained in μFD

A typical time course of the concentration profile in the gel cage following the introduction of a fluorescent dextran to one channel is shown in Fig. 2a. Normalized florescence intensity (C−C_min)/(C_max−C_min) in the gel is plotted as a function of normalized distance (x/x_max) from the dextran (40 kDa) “source” channel. Depending on the width of the gel region, a steady state concentration profile was reached on the order on an hour. For gradients established under static (no flow) conditions, we have demonstrated the ability to sustain a stable gradient for up to 40 hours (without replenishing either the source or sink channels). While a linear gradient can be maintained, there is depletion and accumulation of dextran in the source and sink channels respectively In order to maintain a constant concentration, it is necessary to continuously replenish the sink and source channels. This was achieved by establishing a slow flow (10 μl min⁻¹) in the two microfluidic channels, one with control fluid and the other containing fluorescent dextran, to maintain a gradient across the gel (supplementary Fig. 4) for several hours. These results demonstrate our ability to simultaneously control surface shear stress (on the surface of the endothelial monolayer) and gradients in our device. In this study, 40 kDa dextran (non-reactive solute) was chosen because it is similar in size to several growth factors of interest including VEGF, bFGF and IGF²⁴.

The experimental results for gradients under static conditions were compared to results from a finite element (supplementary Fig. 3) model assuming a diffusion coefficient of 4 × 10⁻⁷ cm² s⁻¹ with the gel region. Figure 2d shows the comparison between the theoretical results (solid lines) and experimental results for two “gel cage” geometries. This value for diffusivity agrees well with the range of values reported in the literature^{25, 26}. The ability to generate gradients of soluble factors across a 3-dimensional matrix provides the potential to simulate physiologically relevant mechanisms during directional migration including sprouting angiogenesis, tumor metastasis and immune response. The dynamic motility of migratory cells can be probed in a controlled microenvironment and monitored in real-time. In addition, the spatial and temporal presentation of such factors provides another level of control which would be physiologically relevant but not possible in most current systems. Saadi et al. however, has recently demonstrated the ability to generate concentration gradients in their microfluidic ladder chamber²⁷.

Microfluidic cultures, model system: capillary morphogenesis

As a case study we used our μFD system to study microvascular endothelial morphogenesis. The primary mechanism by which new blood vessels or capillaries are formed in vivo, angiogenesis²⁸, involves a series of well delineated steps including matrix degradation, cell migration, proliferation and lumen formation. This is a tightly regulated process that is affected by metabolic stress^{29, 30}, mechanical stresses^31–33, soluble factors³⁴ and ECM matrix components^{35, 36}. With this wide range of stimuli and morphological responses they elicit this serves as a model system to demonstrate the unique capabilities of our μFD.

ECs invade 3D matrix to form capillary-like structures reminiscent of sprouting angiogenesis

HMVEC-ad were cultured until sub-confluence on a collagen coated flask in EGM-2MV complete media, harvested and subsequently cultured in the μFD. HMVEC-ad remain viable for a period of several days. Within a few hours following cell seeding EC form a monolayer on collagen gels. In order to demonstrate the capability to better characterize and study the cellular mechanism during sprout formation time-lapse movies were made. In traditional sprouting models this capability is limited since cells are viewed through the monolayer. Here the directional sprouting and migration occurs in the microscopic viewing plane. As a result we were able to capture the dynamics that occurs which have not been demonstrated in other systems. Figure 3, panels (a)-(i) show the sequence of morphological changes that occurs. Time–lapse imaging shows a “lead-cell” as it invades the underlying 3D collagen matrix. In case of single sprout formation (supplementary movie 1); the lead-cell extends filopodial projections into the underlying matrix while the neighboring ECs on the monolayer remain non-invasive. Cell invasion follows a period of dynamic projection and retraction of filopodia while maintaining contact with the monolayer and remain highly polarized. An initial root-like structure is formed in the direction of migration that persists for several minutes with more dynamic smaller extensions. Subsequent morphological changes included increased penetration depth, filopodial diameter and translocation of cell from monolayer (evident by movement of nucleus) (Fig. 3c & d) followed by conical structure (onset of lumen formation). The invading cells subsequently form sprouts with open lumen structures (Fig. 3(h) shows a particle trapped in the newly developing lumen). With this system, we were able to demonstrate all the sequential cellular mechanisms that occur during sprouting angiogenesis in vivo. This provides a very powerful tool for future studies with the capability to answer and clarify some of the controversy in this area of research.

Figure 3(j) shows a representative result for long term cultures. ECs maintained in cultured for several days form multi-cellular capillary-like structures. Endpoint F-actin and DAPI labeling shows the organization and the complexity of these structures. However, capillaries maintained under static conditions regress and lose their connection to the monolayer. One of the hallmarks of capillary formation is development lumen structures. To demonstrate the presence of open lumens, fluorescent microspheres were added to the channel on the apical surface of the monolayer. Figure 3, panels (k) and (m) provides confirmation of open lumen (microspheres collected in structure).

ECs suspended in 3D collagen gels form capillary-like structures

The culture of EC populated collagen gels have been previously studied in macro-scale systems³⁷ but not yet in a microscale device. Endpoint results are shown in Fig. 5. Isolated cells cultured in 3D formed multi-cellular chords and EC rings. To demonstrate the effect of biochemical stimuli, 3D encapsulated ECs were cultured in medium supplemented with bFGF, VEGF and PMA (Fig. 5b). As expected, there was a drastic difference in morphology compared to control samples (Fig. 5a). In control sample, cells migrate and organize to form isolated multi-cellular ring-like structure. Cells stimulated with pro-angiogenic factors remodel to form complex interconnected multi-cellular capillary – like structures. Figure 5e shows the variation of the pressure differential over time which is used to estimate the velocity through the gel. In the presence of interstitial flow (~ 90μm min⁻¹) ECs form multi-cellular structures within the gel and the monolayer at the gel/liquid interface (Fig. 5c).

2D surface migration promotes formation of tube-like structures

Microvascular ECs cultured as shown in Fig. 2b (bottom) experienced extensive morphogenesis. ECs on the fibronectin coated channel retain their characteristic cobblestone phenotype while a remarkable difference in morphology is evident at the gel surface (Fig. 4a). Prior to sheet or tube formation, the cells migrated as a contiguous structure into the gel region with a prominent increase in vacuoles and blebs (Fig. 4b & 4c). These structures were highly dynamic (supplementary movie 2) but eventually evolved into more stable sheets (Fig. 4d) and tubes (Figs. 4e & 4f). Serial sections of fixed samples obtained from confocal imaging and subsequent 3D reconstruction of EC networks confirm the presence of circular and flattened lumen-like structures which extend throughout the length of the vessel. The existence of continuous lumens is further demonstrated by flowing beads through the vessels under a small pressure drop (supplementary movie 3). Some beads can be observed flowing all the way across the gel cage and others collect at necked-down regions in the vascular structure.