At first, a SU-8 master was fabricated using standard lithography. Briefly, in order to produce channels with a thickness of 40 µm, SU-8 3025 (MicroChem Corp., Newton, USA) was spin-coated at 500 rpm for 10 s and spread at 1000 rpm for 30 s. A prebaking step was carried out at 95 ∘C for 30 min. The SU-8 initiator was activated by exposure to UV for 1 min. Subsequently, a post-exposure bake was performed at 65 ∘C for 1 min and 95 ∘C for 5 min. Finally, the samples were developed with the SU-8 mr-Dev 600 developer (MicroChem Corp., Newton, USA), which is mainly composed of propylene glycol monomethyl ether acetate (PGMEA) and rinsed with isopropanol. Ultimately, for the purpose of prolonging the masters' lifetime and facilitating PDMS removal from the mould, the master was exposed to a (tridecafluoro-1,1,2,2-tetrahydrooctyl)-1-trichlorosilane (United Chemical Technologies, USA) atmosphere for a period of 2 h. The silanized silicon master was then used to pattern the microfluidic channels by soft lithography. Polydimethylsiloxane (PDMS) (Sylgard 184 Silicone elastomer, Dow Corning, Midland, USA) silicone base and curing agent were mixed in a 10:1 ratio by weight, respectively, and degassed. The mixture was subsequently injected into the MicCell platform (GesiM, Großerkmannsdorf, Germany) over the PC base plate. The curing process was done at 100 ∘C for 60 min. Once cured and cooled down, the MicCell platform was disassembled, releasing the base plate with the fluidic channel built in the PDMS polymer.

The membrane is responsible for sealing the microfluidic channels and enabling the enclosed electrolyte solution to flow through the openings up to the cell culture chamber. Once the device demands for an optically transparent and biocompatible membrane were met (Nemani et al., 2013), a lift-off technique using SU-8 and standard photolithography was selected. Additionally, this method facilitates obtaining of high-resolution geometric patterns. In order to facilitate the SU-8 patterned membrane release from the substrate, an OmniCoat^™ sacrificial layer (MicroChem Corp., Westborough, USA) was spin-coated and soft-baked prior to resist deposition. The following photolithographic steps were processed as previously described, only with minor adjustments to obtain an 80 µm thick photoresist layer. During the development step, the solution was gently stirred so as not to introduce additional mechanical stresses until the lift-off process was complete.

Lastly, the channel plate, the SU-8 free-standing membrane and the cell chamber were assembled. A covalent bonding of the PDMS fluidic channels to the SU-8 membrane is important for a leakage-free device. With that concern, a surface functionalization of the SU-8 free-standing membrane was performed. This was a two-step process: first the SU-8 surface was activated by O2 plasma (100 W, 150 s, 100 W). Subsequently, the plasma-treated layer was soaked in 5 % APTES (Sigma-Aldrich, Munich, Germany) solution and heated at 40 ∘C on a hotplate for 20 min. The APTES solution is responsible for introducing a silanized layer on the substrate, forming amine groups on the SU-8 surface. Meanwhile, the PDMS channel plate was activated by O2 plasma (same parameters). At this point, the functionalized SU-8 membrane was aligned over the fluidic channels using an optical microscope and subsequently placed over a hotplate for bonding. The optimized bonding temperature and time for our experiments were 70 ∘C and 10 min, respectively. During the bonding process, the activated amine and silanol groups condense to reduce their surface free energy. Consequently, Si–O–Si covalent bonds are formed between the SU-8 and PDMS pieces, leading to strong, irreversible bonding of the two materials (Li et al., 2012). Finally, the cell chamber was prepared by cutting a slab of PDMS (22 mm2) and punching it in the middle with an 8 mm diameter Harris, Uni-Core^™ biopuncher (Ted Pella, Redding, USA). The PDMS provides a watertight reversible sealing mediated by van der Waal forces, enabling a functionalization free bonding.

The current in the DC microfluidic biochip makes its path from the Ag/AgCl electrodes via the electrolyte medium which fills the channels to the central well chamber.

Cell viability was measured using the LIVE/DEAD assay kit (Invitrogen, California, USA). After 2 h of EF stimulation, cells were washed with phosphate-buffered saline (PBS) containing 0.5 µm of calcein AM and 6 µm of ethidium homodimer-1 (EthD-1) and were incubated at 37 ∘C with 5 % CO2 for 15 min. The staining solution was removed and the samples were then viewed under an inverted microscope (Olympus IX81; Olympus, Hamburg, Germany) with 494 nm (green, calcein) and 528 nm (red, EthD-1) excitation filters. Images were captured using Xcellence software (Olympus, Hamburg, Germany).

For the immunofluorescence staining, after 2 h of EF stimulation, cells were washed with PBS (pH 7.2–7.4), fixed in 4 % paraformaldehyde for 5 min at room temperature (RT), permeabilized with 0.5 % Triton X-100 for 6 min and subsequently blocked with 2 % bovine serum albumin (BSA) for 30 min. To detect focal contacts, cells were incubated with mouse anti-human vinculin (1:200, Serotec, Martisried, Germany) at 4 ∘C overnight. After the incubation time, a PBS washing step followed and was later incubated with a fluorescein–isothiocyanate (FITC)-coupled goat anti-mouse antibody (1:1500, Dianova, Hamburg, Germany) along with tetramethylrhodamineisothiocyanate (TRITC)-conjugated phalloidin (1:300, Sigma Aldrich, Munich, Germany) at RT for 1 h. TRITC was added to visualize actin. Finally, the nuclei were stained with 4'6-diamidino-2-phenylindole dihydrochloride, DAPI (1:5000, Sigma-Aldrich, Munich, Germany) at RT for 5 min. Cells mounted in DABCO were imaged under fluorescence microscope.

A real-time observation system (Fig. 2) consisting of an inverted microscope (Olympus IX81), a CCD camera (Olympus DP70), and the Xcellence imaging software together with an incubation system were used for observation of the cell migration at the microfluidic biochip. Images were recorded every 3 min for the duration of the experiment and a time-lapse video was created.

For data analysis, captured images were imported into ImageJ (ImageJ 1.37v by W. Rusband, National Institutes of Health, Baltimore, USA). Cell movement was quantified between the frames of a temporal stack by manually tracking the centre of mass of each cell. Subsequently, the data sets of XY coordinates for each point of time were imported into the Chemotaxis and Migration Tool (v. 1.01, distributed by ibidi, Munich, Germany). After the coordinate transformation, the ibidi Chemotaxis and Migration Tool automatically sets all initial points to (0.0). This tool computed the cell migration speed and y-forward-migration index (y-FMI) of cells and plotted the cell migration pathway. The y-FMI of the cell was defined as the straight-line distance along the y axis between the start position and the end position of a cell divided by accumulated distance.

The LIVE/DEAD assay enables the differentiation of metabolic active cells from damaged and dead cells (Fig. 2). Live and dead cells were checked for control as well as for stimulated samples using calcein and ethidium homodimer dyes. Cells were exposed to a non-homogenous field of the order of 1 V cm-1 for a period of 120 min. Membranes without electrical stimulation were taken as controls. The cells stained with green are viable cells. In contrast, cells stained with red are non-viable cells. The biocompatibility of the microfluidic device was confirmed as the cells exhibit no signs of apoptosis in both control and DC stimulations.

The performance of the microfluidic biochip for electrotaxis studies was validated by studying the 661 W cell line electrotactic response to non-homogenous DC EF stimulation (Fig. 3). Electric fields that produce electrotaxis are typically in the range of 0.1 to 10 V cm-1 (McCaig et al., 2009). The electric field strength at the region of interest was on average 1 V cm-1, resembling the fields observed in wound tissue (Zhao et al., 2006). This field induced a current of the order of 3 mA, which was stable for the duration of the experiment (120 min).

The migratory behaviour of cells was quantified between frames of a temporal stack by tracking the centre of mass of at least 25 cells per experiment. Images were taken every 3 min for the duration of the experiment and later a time-lapse video was created from the image stacks. The migration pathways were traced for both control and stimulated samples. In the absence of an applied electric field (control samples, Fig. 3a) cells had a random migration in all directions with a scattered distribution, whereas when a DC EF was applied, 661 W cells preferentially migrated towards the cathode, as observed in Fig. 3b. Electrotactic migration was verified to be highly directional once almost all the cells migrated cathodally. These results are in agreement with the results obtained for DC homogeneous EF stimulation (data not shown). The results confirm that the microfluidic biochip enables the application of physiologically relevant EFs and can serve as a platform for further electrotaxis experiments.

Immunofluorescence labelling enables the visualization of cell biocompatibility parameters such as cell attachment and cell spreading, but also cytoskeleton organization and focal adhesion formation. The actin cytoskeleton is a highly dynamic network composed of actin polymers and a panoply of associated proteins that mediate intra- and extra-cellular movement and structural support. This dynamic structure rapidly changes shape and organization in response to stimuli.

Indeed, the migration process encompasses a cascade of intracellular signalling events that coordinate actin polarization, protrusion, cellular and membrane polarization and adhesion mechanisms (Chung et al., 2001).

In Fig. 4 are depicted images of immunofluorescence labelling for control and electrically stimulated samples. In both cases the samples were placed on the device for 120 min before being labelled. However, stimulated samples were electrically stimulated for the above-mentioned time frame, whilst control samples without any stimulation were taken as controls. In Fig. 4b it is noticeable that, overall, cells spread and elongate in the same direction. One can see that the actin filaments from each cell elongate in a common direction, which is perpendicular to the electric field vector present in the area. In contrast, the orientation of the cytoskeleton (actin filaments) and the distribution of the focal contacts (vinculin) in Fig. 4a do not seem to follow any directional cue, as cells appear to spread arbitrarily in all directions: some have actin filaments in the x direction, with others in the y direction. There is no external cue guiding in a specific direction, and hence cells orient and migrate randomly.