The microfluidic platform comprises two 600 × 900 × 70 μm chambers, each with a volume of 36 nL, linked by three 250 μm long, 125 μm wide channels (Figure 1). Three inlets are connected to a common channel that is connected to both chambers and ends in a common flush outlet. One inlet is dedicated to medium and the other two to different cell populations, thus avoiding cross-contamination. Each chamber also has a dedicated, independent outlet.

Integrated pneumatic valves composed of 150 μm wide channels allow the routing of medium and cells to each chamber independently, the isolation of the microenvironment of each chamber, and the blocking of each inlet to the common channel (Figure 2). The channels between the chambers are also equipped with an integrated valve.

Before inserting cells, each device was sterilized and primed. Ethanol 70% was inserted into the microfluidic chip with a syringe pump (New Era Pump Systems, NY, USA) after which all inlets except one were plugged with a steel plug. Pressured air was then applied in the unplugged inlet at 80 kPa for 5 min to release all existing air bubbles (Kim et al., 2007). The channels were then rinsed with phosphate buffered saline (PBS) and coated with fibronectin (FN) to improve cell adhesion and proliferation (Klebe et al., 1981; Lee et al., 2004). The coating was performed by flowing a solution of 50 μg/mL of FN in PBS and incubating for at least 1 h at 37°C. Finally, before seeding the cells in the cell chambers, the device was rinsed with cell culture medium.

The pneumatic channels were also filled with DI water by filling the capillary tubing and applying about 200 kPa of pressure. By filling the channels with water one is able to effectively actuate the valves without running the risk of generating air bubbles in the medium by diffusion of gas through the PDMS membrane (Studer et al., 2004). The effectiveness of this type of valves was shown previously (Unger et al., 2000; Thorsen et al., 2002; Araci and Quake, 2012), and its leak-proof character was further verified with experiments using air bubbles and fluorescein isothiocyanate (data not shown).

Activation of the valves was controlled by switches connected to a simple custom-made printed circuit board (PCB; Supplementary Figure 1). The PCB was connected to a 24 V power source and the switches control small solenoid valves (SY144-5MZ-Q, SMC Corporation, Tokyo, Japan) mounted on a manifold, itself connected to a regulator and a pressured-air line. Each solenoid valve was linked to an integrated valve by polyethylene tubing (Instech Laboratories Inc., PA, USA).

Human H4 neuroglioma cells (HTB-148, LGC Standards, Barcelona, Spain) were cultured in Opti-MEM (Gibco, Thermo Fisher Scientific Inc, Waltham, MA, USA), supplemented with 10% heat inactivated fetal bovine serum (FBS, reference 10500-064, lot 07Q2330K, Gibco, Thermo Fisher Scientific Inc, Waltham, MA, USA) and 1% penicillin-streptomycin (pen-strep) and maintained at 37°C and 5% CO₂. Before transfection, cells were plated in T-25 flasks and allowed to reach 70% confluence. Cells were transfected with wild type aSyn tagged with GFP (αSyn-WT-EGFP, WTSynEGFP plasmid previously described; McLean et al., 2001) or GFP (EGFP, pEGFP-N3 plasmid; Clontech, Palo Alto, CA, USA) using FuGENE (Promega Corporation, Wisconsin, USA). To select stably transfected populations, the medium was replaced with Opti-MEM complete medium supplemented with 400 μg/mL of G418 (Thermo Fisher Scientific Inc, Waltham, MA, USA) 24 h after transfection. A control flask containing non-transfected H4 cells was used to assess the efficacy of the selection. After the selection was completed, transfected cells were maintained in Opti-MEM complete medium supplemented with 200 μg/mL of G418.

N9 cells (a kind gift from Dr. Paula Ricciardi-Castagnoli) were cultured in RPMI 1640 (Gibco, Thermo Fisher Scientific Inc, Waltham, MA, USA) supplemented with 10% heat-inactivated FBS and 1% pen-strep. Activation of N9 cells was achieved by replacing the medium with RPMI 1640 complete medium supplemented with 1 μg/mL of lipopolysaccharides (LPS, from Escherichia coli 055:B5, chromatographically purified by gel filtration, Sigma-Aldrich, St. Louis, MO, USA) and incubating for at least 12 h.

All images were acquired with a Nikon Eclipse Ti-E inverted microscope (Nikon Corporation, Tokyo, Japan) equipped with a motorized stage, a cage incubator for temperature control and a stage-top gas chamber for CO₂ control (Okolab S.R.L., Pozzuoli, NA, Italy). Transmitted light and fluorescent images were acquired with a dry 10 × Nikon CFI Plan Fluor objective with a numerical aperture of 0.30 and a high sensitivity Andor iXon Ultra 897 EMCCD camera (Oxford Instruments, Oxfordshire, UK).

Acquired images were analyzed using ImageJ software (NIH, Bethesda, MD, USA) and stitched using the pairwise stitching plugin (Preibisch et al., 2009).

Cells in suspension were loaded in a syringe and inserted into the device at one of the inlets and a controlled flow was maintained by a syringe pump; liquid was allowed to flow through the common channel to the flush outlet until cells were seen entering the inlet; valves routed the cells into the chamber and the valve at the entrance of the chamber was opened and closed until the appropriate number of cells was inside the chamber. Since inertia does not play a significant role in this flow regime, the fluid flow stops immediately once the valve is closed and the number of cells can be quickly assessed. Cells were then allowed to sediment and adhere to the substrate for at least 90 min: it was seen that this was the necessary time for cells to spread on the substrate and acquire their typical epithelial-like morphology. It should be noted that several seeding rounds are possible: since cells are quick to sediment and adhere to the substrate in a few minutes, new cells in suspension can be routed into the chamber without detaching the previously seeded ones. Throughout the seeding process the channels linking both chambers were kept closed to prevent cross-contamination, and after the seeding of each cell population the common channel was rinsed with medium at a high flow rate (above 20 μL/min). Since during seeding the suspended cells are in constant perfusion through the common channel, and we can seamlessly switch the incoming flow from the cell to the medium inlets, cells do not have the chance to sediment and adhere to this channel, further minimizing the chance for cross-contamination.

After the static adhesion time, medium inside each chamber was renewed by flowing medium at 5 μL/min for 10–20 s—which was also used to flush non-adhered or dead cells. Two hours after insertion cells were already expressing aSyn-GFP, as evidenced by a strong fluorescence signal. At this time point, a solution of Triton X-100 (Panreac Química S.L.U., Barcelona, Spain) 0.05% in PBS was perfused for about 5 s at 5 μL/min in the cell chamber containing the cells with aSyn-GFP, after which the chamber was isolated by closing the valves at the entrance and exit. The cells were kept isolated for 20 min, after which the central channels were opened. This concentration of Triton was optimized so that it is sufficient to permeabilize the membranes of the cells inside the chamber, releasing aSyn-GFP in the medium, without affecting cells in the neighboring chamber once the valve is opened (Figure 3).

After the central channel was opened, aSyn-GFP diffused to the neighboring chamber (Figure 4). Both cell chambers were kept isolated but with the central channels open for 10 h, after which they were rinsed with PBS, fluorescent images were taken and a viability staining was performed. For the staining 10 μg/mL of propidium iodide (PI) was introduced in the device and left to incubate for about 20 min.

For N9 and H4 co-culture, two sets of experiments were made: one with activated, and another with non-activated N9 cells. Seeding of both N9 and H4 was made as described above. In the assay with LPS-activated N9, previously activated cells were suspended in RPMI 1640 complete medium supplemented with 1 μg/mL LPS, while non-activated N9 were suspended in complete medium without LPS prior to seeding. Cells were left to settle and adhere for 2 h, after which medium was changed for both populations. In both sets of experiments, the fresh medium did not contain LPS. The central channel was opened 30 min after the medium change, and diffusion was allowed to occur for 10 h. At the end of the time allowed for diffusion, DHE at 30 μM in Opti-MEM complete medium was inserted in the chambers and left to incubate for 20 min at 37°C.

The prion-like spreading of aSyn pathology in the brains of PD patients is intriguing, and the precise molecular mechanisms involved are still elusive (Li et al., 2008; Olanow and Prusiner, 2009). Here, we took advantage of the microfluidic platform we developed to study and monitor the transfer of this protein between two cell populations in real time. By co-culturing cell populations expressing or not expressing aSyn, we studied conditions that affected the release and uptake of the protein. For this purpose, we used a GFP-tagged version of aSyn (aSyn-GFP; previously described in McLean et al., 2001), to enable the imaging of the process in living cells and in real-time. In this particular proof-of-concept study we focused on the uptake of aSyn and, therefore, promoted the release of aSyn-GFP by permeabilizing the cell membrane.

To study the spreading of aSyn between cells (Lee et al., 2008; Lashuel et al., 2012) naïve H4 cells were cultured with H4 cells expressing aSyn-GFP in the two chambers of the device as described above. This cell line chosen since H4 cells and other types of neuroblastoma have been extensively used in the study of the release of aSyn and other proteins (El-Agnaf et al., 2003; Alvarez-Erviti et al., 2011; Herrera et al., 2011; Chutna et al., 2014).

To assess if the uptake of aSyn-GFP occurred, the fluorescence signal of naïve H4 cells was measured before the opening of the central valve and after a 10 h exposure to aSyn-GFP in solution followed by washing with PBS. Regions of interest (ROIs) corresponding to the cells' perimeter were defined using a transmitted light microscopy image and these were used to measure the signal intensity in the fluorescence microscopy image. To ensure that the ROIs captured the fluorescence signal of the entire cells, transmitted light and fluorescence images were captured at the same time point. Cells that stained positive for PI were discarded from this selection. Since other studies reported spreading of aSyn between cells in as little as 5 min (Lee et al., 2008), we surmised that 10 h was an appropriate time frame to observe uptake events.

Since cells change shape and divide throughout the duration of the experiment, it was not practical to track each single cell and evaluate the change in fluorescence signal. The method used to determine if aSyn-GFP entered the cells, either through diffusion or endocytosis, was to compare the average signal inside the cell to the background fluorescent signal inside the cell culture chamber.

Interestingly, in the experimental conditions used, we were not able to detect the uptake of aSyn-GFP (Figure 5A). In fact, although these cells have a slightly higher average fluorescence than the background (about 5%) this relation does change significantly after a 10 h exposure to aSyn-GFP. Although these observations seem to go against the general consensus of the literature, several possible explanations might justify these results: (i) the concentration of aSyn-GFP resulting from cell release is too low; (ii) aSyn molecules that find their way inside the cell might be rapidly degraded or re-released—in which case experiments would need to be carried out in a shorter time frame; (iii) molecules enter the cells in a very low number, and consequently have no visible influence on the fluorescence signal; (iv) GFP affects the uptake of or diffusion of aSyn, since it is a larger protein than aSyn itself. Control experiments in which H4 were co-cultured with H4 producing GFP seem to support the last explanation, since no uptake of GFP was observed (Figure 5B), however, other studies reported diffusion of GFP into cells in similar conditions (Reyes et al., 2015). It should be noted that in other studies using extracellular aSyn, internalization was observed after as little as 5 min for monomeric aSyn and 1 h for aSyn fibrils (Lee et al., 2008), which suggests that 10 h is an appropriate time scale for such an experiment. It is important to point out that a visible increase in fluorescence signal was found in cells that later showed positive PI staining, which would indicate that aSyn-GFP was able to enter cells with a compromised membrane (Figure 6). The same effect was verified with GFP alone. This could either indicate that active transport is necessary to constantly remove the molecule from inside the cell or that diffusion is less efficient across a non-compromised membrane. In any case, additional experiments will be required to clarify the mechanisms involved.

The use of aSyn tagged with fluorescent molecules has been used in various studies to monitor the uptake and release of this protein. However, these studies rely either on the use of previously prepared conditioned medium (Lee et al., 2008, 2010), i.e., medium in which cells released aSyn, or on culturing a populations expressing tagged aSyn together with another population expressing another fluorescent molecule (Reyes et al., 2015). This platform allows us to monitor the spreading of aSyn without the need for using recombinantly produced protein. Instead, we can directly monitor the release of cell-produced aSyn into the medium. The fact that we can perform all these steps inside the device increases the speed of the assay and improves reproducibility due to the simplification of the experimental variables. Furthermore, by using the platform we can also keep cells separated, thereby ensuring that any spreading happens via mechanisms that do not require cell-cell contact.

Since one of the main mechanisms through which microglia damages neurons is by the production of ROS, we used this microfluidic platform to monitor the effects of N9 microglia on naïve H4 cells. N9 microglial cells share many properties with primary microglia, namely activation by LPS and production of interleukin (IL)-6, tumor necrosis factor (TNF), IL-1, and ROS (Stansley et al., 2012) and are widely used in the research of neurodegenerative diseases. There are two main ways by which we can assess the impact of N9 on H4 cells: either by studying indirect effects such as changes in cell viability—which can be difficult to unequivocally relate to the presence of ROS or might not even occur if the concentration of ROS is small—or by directly evaluating the amount of ROS to which the cells are subjected using fluorescent dyes.

To demonstrate how microglia-neuron interactions can be studied in this platform, two sets of experiments were designed: on the first, H4 cells were co-cultured with non-activated N9 cells, as a control condition; on the second, H4 cells were co-cultured with LPS-activated N9 cells. The production of ROS and their effect on the H4 cells was evaluated by staining these cells with dihydroethidium (DHE), a fluorescent dye that becomes red and intercalates with DNA when oxidized by superoxide (Carter et al., 1994; Bindokas et al., 1996; Owusu-Ansah et al., 2008; Kalyanaraman et al., 2012).

Cell seeding was performed as described above and the fluorescence signal was evaluated by selecting ROIs corresponding to the stained nuclei of the cells in the fluorescence microscopy images.

The results of the DHE assay showed that H4 cells co-cultured with activated N9 cells exhibit almost twice the fluorescence signal of cells co-cultured with non-activated N9 cells, which suggests that they are in the presence of higher levels of ROS (Figure 7). This difference cannot be explained merely by the difference in the number of N9 cells, since in the assay with non-activated cells the culture chamber contained 297 cells whereas in the assay with LPS-activated cells, there were 451 cells, which amounts to an increase of 52%. Furthermore, a two-tailed Student's t-test with unpaired data showed statistical significant difference from the mean signal of the two populations with p < 0.001. Although our observations produced the expected results, i.e., that LPS-activated N9 cells increased the overall level of ROS in H4 cells, more thorough analyses can be added to this type of preliminary study. Since cells are confined to microfluidic chambers, it is possible to know the exact number of both H4 and N9 cells—and not just an estimate as in traditional cell culture platforms. As such, it would be possible to find the relation between the number of N9 cells and the increase in DHE fluorescence signal. Furthermore, other ROS dyes could be used, such as 2′,7′-dichlorfluorescein-diacetate (DCFHDA; Macedo et al., 2014), and effects such as early apoptosis could be screened (Vermes et al., 1995; Datta et al., 2002) to see what other effects activated N9 have over the population. Although N9 cells pre-activated with LPS were used in this experiment, this platform allows the studying of the interplay of aSyn and microglia by stimulating resting microglia with aSyn, either released by cells or in supplemented medium. This will in turn allow replicate and monitor the mechanisms of sustained, chronic inflammation that occur in vivo.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.