3 Results & Discussion
Typical CNNs experiments in Petri dishes usually concentrate the cell population in an approximately circular area (radius ~5 mm), which had been previously coated for cell adhesion. However, this procedure does not allow an accurate selection of the area to be functionalized, which finally results in a certain degree of cell dispersion and the reduction of the cell concentration in the area of interest. The custom design of our chips permits different shapes and sizes to develop the desired areas to grow cells. The work area contention is achieved by designing and fabricating just a circular area connected to the inlet and outlet entries.
We conducted two culture on chips experiments along this study. In our experiments it was successfully observed that the microflux drag force does not affect the cells adhesion and their uniform distribution. As expected, neurons moved slightly to form neuronal aggregates, but they were heavily attached to the surface, and no cell loss or detachment due to a failing in the coating or the flow were observed.
Continuous flow can be regulated to supplement the amount of fresh medium wanted. Although the flow was able to drag small particles and debris, larger dissociation particles as void ganglia capsules were not removed. The flow also washed off particles and toxic metabolites products of cellular activity, as melanin, particularly important in cell cultures from insects when hemolymph is added to the medium and whose accumulation can be cytotoxic due to phenol oxidase activity and can damage the cells. Consequently, the CNNs developed in the chip survived longer, at least 22 days, than the ones grown on Petri dish (up to 14-18 days) , where the medium is not changed. In fact the CNN was still connected in a good conditions until that day.
Another point to analyse was the homogeneity of the flow distribution. It has not been observed greater process of detachment of the cells in the most peripheral areas of the chip, and in addition, network processes have also been formed in these areas of the chip, so it seems that the flow has been homogeneous in all the culturing volume.
In experiments where the culture status evaluation and measurement are done by means of optical microscopy, the microfluidic device design must fulfil several requirements. In addition to be transparent enough, the chip dimensions must fit the microscope holding plate. Also, the overall thickness must allow to focus the culturing surface, and therefore must fit the range imposed by the microscope focal.
Once both requirements are met, the new neuron-on-a-chip culture is connected to a microfluidic system with hemolymph enriched medium, as explained above. The chip is monitored for at least 18 days following the described observation protocol.
The longitudinal observation of the development of the CNN, as reflected in Fig. 3, seems to follow a similar path as reported in our previous works. Isolated neurons in the first day in vitro (Fig.2 A1) started to grow neurites and being linked to their neighbouring neurons around 3-6 days of culture, getting closer among to form clusters. At DIV 7 (Fig.2 A2), connections among farther neurons are observed and junctions previously formed are more entrenched. In this step we observed an increase in the number of connections and a larger network is settled. After this stage the networks grow, forming bigger clusters, until the CNNs settled as a consolidated and mature network at day 15 of culture (Fig.2 A3). No significant changes in the network were observed beyond this point under our observation until day 22.
In order to properly quantify the different network states with the tools of graph theory, the resulting graphs from the segmented mosaics (Fig.2 B1-3) were mapped into adjacency matrices (Fig.2 C1-3) whose elements aij are equal to 1 if nodes i and nodej are connected and 0 otherwise. Results of the evolution and development of the graph of the two CNNs grown on chips have similar trends to those grown on Petri dishes with similar initial density. As shown in Fig.3A, the number of connected nodes, nodes having at least 1 connection, increases until the 6-7 days in vitro, as a result of the neurite growth and connections being made. After this stage, the network development is steady, and in fact, there is a slightly decrease of the number of connected nodes because when connected neurons form clusters, the segmentation algorithm identifies them as a single node.
Other network properties we focused on were the longitudinal progression of the averaged clustering coefficient (C) and of the shortest path length L, normalised by the size of the largest connected component SGC1, L / SGC (Fig. 3B). Regarding the normalized shortest path, high values were observed at the start of the experiment. Around 4-6 days in vitro, this value decreases, as links between nodes are formed. Beyond that, L/SGC keeps a low constant value in the more mature stage. Clustering coefficient shows near 0 values at the start, as expected as there are no connections between neurons. In the critical stage of network formation, around 3-6 days in vitro, C increases revealing the new connections between neurons. When the mature network is settled, the C value keeps constant at a relatively high value indicating the presence of loop circuits in the network.
Both values, low shortest path values, and the high clustering coefficient are studied frequently as characteristics of a small world structure, where the high tendency to form clusters of nodes in highly interconnected subgroups and short distance between them contribute to an optimal functionality in the network based in high efficiency. These results are very similar to the ones obtained in Petri dishes mentioned above, with no significant alteration, so we can confirm that the use of our chips to grow optimal functional CNNs with the desired parameters not only does not affect their development but they promote a longer network survival.