pumping for the parallel trapping of single neurons onto a microsieve electrode array

for the parallel trapping of neurons onto the l SEA chip with the goal to improve its biological performance. This method uses the capillary pumping between two droplets (a “pumping droplet” on one side of the chip and a “reservoir droplet” on the other side) to create a stable and controllable ﬂow. Due to simpliﬁcation of the handling procedure, omitting the use of a syringe and additional connections to the l SEA chip, the set-up is compatible with real time microscopy techniques. Hence, the authors could use optical particle tracking to study the trapping process and record particle velocities by video imaging. Analyzing the particle velocities in the passive pumping regime, the authors can conﬁrm a gentle uniform particle ﬂow through the 3D micropores. The authors show that passive pumping particle velocity can be tightly controlled (from 5 to 7.5 to 10.4 l m/s) simply by changing the droplet volume of the pumping droplets from 20, 40, and 60 l l and keeping the reservoir drop constant (10 l l). The authors demonstrate that neuron capturing efﬁciency and reproducibility as well as neuronal network formation are greatly improved when using this passive pumping approach. V C 2017 Author(s). All article content, except where otherwise noted, is licensed under a Creative Commons Attribution (CC BY) license (http://creativecommons.org/ licenses/by/4.0/) .


I. INTRODUCTION
Microsystems for biology typically use microfabricated platforms (micrometer range) with MEMS and microfluidic components to study concepts ranging from electrical engineering to chemistry and biology. 2 Lab-on-a-chip technologies, for which the idea is to miniaturize a fully operational laboratory on-a-chip of a few centimeters in size, 3 have led to the development of organ-on-a-chip concepts where researchers are able to reconstruct part of human organs or recapitulate specific organ functions on-a-chip. 4 Some examples include liver, 5 kidneys, 6 and lungs 7 on-a-chip. In this research, our end goal is to create a platform technology for brain-on-a-chip 8 to study and better understand epileptic seizures. This in vitro model can also potentially be extended to other brain disorders such as Alzheimer's and Parkinson's. The development of in vitro brainlike tissue constructs is also important for the understanding of healthy brain physiology. In vivo studies are slow, low throughput, complex, costly, and arguably immoral. 9 In addition, animal use could be dramatically reduced for drug screening studies if relevant human in vitro organ-on-chip models were engineered, as they have highthroughput, reproducibility, and robustness, as well as costeffectiveness considering today's demand in pharmacological developments. 4,9 Consequently, there is a need for improving current neuronal cell culture models to create in vivo-like brain tissue constructs on-a-chip. These miniaturized analytical displays can be used to study brain development and a) Electronic mail: R.Luttge@tue.nl complex brain cells' interactions, leading to diseases in the form of novel biological assays. 10 Engineering brain tissue constructs on-a-chip however is challenging and requires a multidisciplinary approach in the integration of a large variety of scientific skills. The brain is a complex yet highly organized network of cells communicating chemically and electrically with each other in a very specific manner and with an advanced hierarchical structure that enables its functionality. 11 Neurological disorders and diseases arise when the brain cellular network is disturbed (i.e., by structural, biochemical, or electrical damage), which can lead to Alzheimer's, Parkinson's, and epilepsy, to name a few. Growing neurons in a spatially standardized fashion (for example, in arrays) will ease the analysis of neurite connectivity (which is a hallmark neurodevelopmental end point indicator) and will lead to an easier method for relating changes in connectivity to electrophysiology and biological functions.
Our current concept on brain-on-a-chip has two main parts, a microbioreactor (MBR) enabling 3D cell/tissue cultures and a microsieve electrode array (lSEA) 1 for pairing single neurons to electrodes (Fig. 1). The MBR serves as a 3D neuron culture chamber and contains a porous interface that will permit the diffusion of nutrients, and therefore, the neuronal cells are continuously fed via microfluidic principles. The lSEA enables the parallel trapping and pairing of neurons onto an electrode array containing 3D micropores and will be the main focus of this article. Commercially available microelectrode arrays (MEAs) already permit the recording and stimulation of neuronal activity 12 but not to study events occurring for an individual neuron within a network over time. This is because cell cultures are dynamic and neurons migrate over time on a planar configured MEA, changing their locations on the array surface and making it impossible to follow the process over time. Accordingly, the MEA was modified into a lSEA to include 3D micropores at the electrodes in order to be able to trap hundreds of single neurons in a well-organized array conformation. The 3D micropores were also functionalized with an electrode matching the design of multichannel systems MEA readout electronics (Fig. 2). 13 This enables an organized positioning of neurons and the formation of a spatially controlled neuronal network inbetween the 3D micropores. In this contribution, we demonstrate a new cell trapping procedure to capture single neurons within the lSEA without the need to use syringes or pumps but by exploiting passive pumping and capillary phenomena. This allows the generation of reproducible flow rates that are compatible with cell capture and cell survival whilst being also compatible with microscopy and eliminating the need for additional equipment.

A. Microsieve electrode arrays
A lSEA chip is developed 1 enabling hydrodynamic trapping of single neurons within highly uniform 3D micropores (Fig. 2). This lSEA chip has a surface area of several square millimeters. The 3D micropores are fabricated by means of corner-lithography and wet chemical etching in {100}-silicon. In brief, the fabrication process consists of a silicon sieving structure obtained by corner lithography with a patterned boron doped poly-silicon, connecting the contact electrodes within the 3D micropores. A LPCVD silicon-rich silicon nitride layer was used as insulation, and this new technology platform for multisite electrophysiology recordings was termed lSEA. A more detailed fabrication protocol can be found in the study by Schurink et al.  Polystyrene microparticles of 1 lm in size (Micromer V R , micromod Partikeltechnologie GmbH) were diluted in phosphate buffer saline (PBS) to reach a final concentration of 500 000 particles/ml and were used to characterize the passive pumping flows. The neuroblastoma cell line SH-SY5Y (ATCC V R , CRL-2266 TM ) was used to characterize the cell trapping procedures. The original cell line was isolated from bone marrow taken from a young human female with neuroblastoma. 15 SH-SY5Y neuroblastomas were cultured in Dulbecco's Modified Eagle's Medium (DMEM)/F-12 media (1:1) supplemented with 10% fetal bovine serum and 1% pen/strep and grown in an incubator at 37 C, 5% CO 2 . When cell confluency was reached, trypsin (Â1) was used to harvest the cells and centrifuged at 900 rpm for 5 min. A concentration of 200 000 cells/ml was used throughout the trapping experiments. Following the trapping, the cells were exposed to retinoic acid for 3 days at 10 lM in DMEM/F-12 media to differentiate the cells into neurons.

C. Trapping microparticles and SH-SY5Y neurons
Here, a passive pumping method was exploited to observe the capturing flow pattern of particles and consequently allow us to optimize capture conditions for trapping single neurons on the lSEA. First, a 10 ll empty (free of particles) droplet is placed under the lSEA (called reservoir drop, bottom side), and the chip is positioned into a clean Petri dish. A 20, 40, or 60 ll droplet (called pumping drop) containing the microparticles is immediately placed on top of the lSEA chip (top side), and the platform is positioned under the microscope for video recording. For the cell trapping procedure, the same approach was applied, but only 20 ll pumping drops were used.

D. Tracking microparticles
In order to characterize particles' and cells' flowing speeds and trapping velocity, particle tracking software (IMAGEJ, Mosaicsuite plugin) was used. 16 Using these imageprocessing algorithms, videos of particles flowing inside the 3D micropores were analyzed for three different volumes of the pumping droplet (20, 40 and 60 ll). The particle tracking detection parameters had to be tuned for every video. Particle detection was carried out where the approximate radius of the particles was optimized (1 lm); the cut-off score for nonparticle discrimination was constant and the percentile which determines which bright pixels are considered as particles was the most variable. In addition, particle linking parameters such as displacement (maximum pixels a particle can travel between frames) and link range (to match the optimal correspondence matching) were also adapted for every video. To estimate the speed of the particles, first, the total pixel changes in both x and y axis were recorded. The total pixel changes were converted to micrometer. Then, the total distance traveled in both x and y directions was calculated, and the full trajectory over time was deduced by calculating the hypotenuse value (x 2 þ y 2 ¼ square root of the distance traveled). Then, the total amount of frames was converted to seconds, and the total distance traveled by a particle was converted to micrometer per second.

A. Modified microsieve electrode array (lSEA) setup
The lSEA chip was produced as previously described by Schurink et al. (Fig. 2). 1,13,14 The original protocol for trapping single neurons inside the 3D micropores involves the use of syringes and pumps and some additional polydimethylsiloxane (PDMS) parts to provide an active pumping mechanism. The original setup 13 and protocol are shown in Fig. 3. However, we noticed that following the trapping of the neurons, the capture efficiency was not reproducible (ranging from 21% to 90%) 13 and most of the neurons did not survive the trapping procedure and therefore were not able to network inbetween the 3D micropores. There were several reasons for the limited applicability of the procedure in the original test setup. One is that the trapping cannot be visualized under the microscope due to the syringe/PDMS construct blocking the light path. So, there is no mean to check if the neurons are being trapped during the procedure properly. Second, the trapping of neurons needs to be performed in sterile conditions to avoid culture infections, and so, the pump and the syringe have to be used under a cell culture hood, which is cumbersome and impractical for the end user. Third, the removal of the bottom PDMS construct connecting the lSEA to the syringe can cause negative back-flow. This means that neurons flow out of the 3D micropores. Fourth, the capture flow speed cannot be tightly controlled as tubing and plastic syringes have dead volumes which can impact the reproducibility  of the trapping procedure and requires time to adjust for every experiment. To summarize these points, a more user friendly and more reproducible trapping procedure was developed that is easy to be adopted by a state-of-the-art biologists' laboratory. The new trapping protocol only requires a standard micropipette for controlled dispensing of the reservoir and pumping drops in the passive pumping to start the seeding procedure. The details of the process are described in Secs. III B and III C below.

B. Microparticle trapping and tracking by passive pumping
To improve the performance of the single cell trapping protocol, a passive pumping principle was investigated [ Fig. 4(a)].
Passive pumping is a simple method for pumping fluids in a semiautonomous way, 17 which eliminates the need for expensive or cumbersome external equipment. Passive pumping is a principle that relies on surface tension and adhesion forces present in a small drop of liquid to create a flow through the microchannel. In our experiments, however, the reservoir drop (10 ll) is flattened between the lSEA and a Petri dish and the pumping drop (20, 40, or 60 ll) is placed on top [ Fig. 4(b)]. When this occurs, there is a weak capillary driven force that is generated inbetween the lSEA and the petri dish. This directs the flow from the pumping drop to the reservoir drop and consequently through the 3D micropores, enabling the particles to be directed inside the 3D micropores. For flow speed characterization experiments, the pumping drop consists of particles in PBS with a concentration of 500 000 particles/ml and the  reservoir drop consists of PBS only. The flow rate is determined from the volume of the pumping drop with respect to the reservoir drop. 17 We report that using this approach, passive flows are generated reproducibly. Next, we changed the pumping drop sizes from 20, 40 to 60 ll to investigate the effect of the drop size on flow rates. Using IMAGEJ MosaicSuite particle tracking package, we analyzed the videos [ Fig. 5(a)] for each droplet size and plotted the particle speeds as a measure of flow rates generated [ Fig. 5(b)]. We demonstrate that not only are the flow rates extremely gentle, from 8.6 to 13.3 lm/s, but they also appear to be dependent on pumping drop sizes. The average total pumping flow rates were determined to be between 2 and 2.5 ll/min. Although there is a large spread in the data, the average seems to reveal a trend in that the bigger the pumping droplet, the faster the flow rates.

C. Neuron cell pairing by passive pumping
Following the successful loading and characterization of the microparticle flow rates, we chose a droplet regime of 10 ll for the reservoir drop and 20 ll for the pumping drop as an optimum for neuron trapping. Using the same passive pumping approach and using the capillary effect inbetween the lSEA and the Petri dish, we report that single neurons can indeed be trapped in the 3D micropores and paired to the integrated electrodes using this approach [ Fig. 6(a)]. We also observe that cell survival (90%) and therefore the onset of the neuronal network [ Fig. 6(b)] are not issues anymore using this approach compared with the active pumping approach, indicating that the neurons are able to survive this gentle trapping procedure into the 3D micropores of the lSEA chip. This observation is also supported by the low cell capture flow rates which are compatible with cell survival. Furthermore, the flow rates are noticeably lower when using cells instead of microparticles (5.81 lm/s compared to 8.6 lm/s). We hypothesize that the reason for the lower trapping flow rates is due to 2 factors: first, a cell is more buoyant than a polystyrene particle, which means that it will move slower in liquid. Second, a cell is considerably larger than a particle (10 lm vs 1 lm) and therefore experiences more drag force. The observed lower cell capture speeds by passive pumping are an actual advantage considering that shear forces through the 3D micropores were previously impacting cell survival while using the syringes and pumps. In addition, this passive pumping procedure allows the user to reproducibly perform the trapping within minutes, in a sterile setting without the need for extra pumping equipment or loading parts whilst being compatible with microscopy.

IV. SUMMARY AND CONCLUSIONS
Passive pumping offers a fast (minutes) and simple route for the spontaneous and gentle trapping of hundreds of single neurons in parallel within the 3D micropores of our modified lSEA, eliminating the need for pumping equipment. This will facilitate the acceptance of this technology by the biology community. The passive pumping principle can also be extended to other cell types for single cell analysis 18 where there is a need to trap single cells in an arrayed format. Examples include cancer, stem cells, reproductive biology, and many more. 19