Using High-Density MEAs For High-Throughput Retinal Ganglion Cell Type Classification
-
1
ETH Zurich, Department of Biosystems Science and Engineering, Switzerland
-
2
Friedrich Miescher Institute, Switzerland
-
3
Friedrich Miescher Institute, Switzerland
-
4
ETH Zurich, Department of Biosystems Science and Engineering, Switzerland
MOTIVATION
We are interested in developing a method for high-throughput screening of retinal ganglion cells (RGCs) by means of detecting and categorizing their functional properties determined by their light responses. These responses are extracellular action potentials that we record with high-density microelectrode arrays (HD-MEA).
It is estimated, that there are over 30 different RGC types in the mammalian retina (Baden et al., 2016), many of them have not yet been functionally characterized, but each cell type is thought to have a unique role in transmitting visual information to the brain. HD-MEAs can be used to measure extracellular action potentials of RGCs at a high spatiotemporal resolution. In mammalian retinas, RGCs are usually arranged in an almost 2D planar configuration. Therefore, we could in principle record from all ganglion cells in a given area of murine retina simultaneously.
Our goal is to find a set of visual stimuli, which allows us to distinguish different RGC types solely on the basis of their electrical responses, without the need for molecular or genetic markers. Potentially, a dedicated set of stimuli would enable us to get insight into the functional roles of RGC types in visual coding.
MATERIAL AND METHODS
Our HD-MEA system has an electrode density of 3460 electrodes/mm2 and 1024 readout channels at a sampling rate of 20kHz (Müller et al., 2015).
We recorded extracellular action potentials (spikes) from the ganglion cell layer of mouse retina while projecting light stimuli from the other side onto the photoreceptors. A fully automated and highly parallelized spike-sorting algorithm detected and classified spikes using shape and spatial distribution of their waveforms.
This process - when applied to the large amount of data produced by our device - is computationally expensive and was performed on a computer cluster with up to 500 parallel processors. The resulting spike-trains for each neuron in conjunction with the light stimuli were the basis for extracting features such as receptive fields, on-off-responses, spiking rates to pseudorandom stimuli, etc.
Clustering was performed based on these features with the goal of finding groups of functionally similar cells.
RESULTS
In one experiment with a single piece of retina, where we repeated the same protocol twice at different locations, spike-sorting yielded more than 700 output units. More than 200 showed clear receptive fields based on spike-triggered averages to a white noise stimulus. We then clustered those units based on features extracted from a pseudo-random fingerprinting stimulus. The clustering revealed several functionally distinct groups of cells with similar responses to the light stimulus. Cells spiked at very precise times upon repetition to the same stimulus, and also compared to other cells in the same cluster.
DISCUSSION
Our method allows us to record from a large number of RGCs at the same time with a spatiotemporal resolution that is high enough to detect and assign single action potentials to each cell. Currently, we record from roughly 2000 RGC/mm2 in an area of the mouse retina where we expect 5000-10000 RGC/mm2 (Jeon et al., 1998). We hope that we can increase the number of recorded cells even further by improving the tissue preparation, developing a better technique for pressing the retina to the electrodes and refining our spike-sorting algorithm.
We show that clusters correspond to putative cell types, within which the precise timing of spikes in relation to light stimuli is preserved. Our stimuli allow us to separate some clusters from the bulk of cells, but further stimuli have to be developed in order to obtain more clusters and to optimize cluster separability.
In the future, the method will be verified by correlating the clusters with genetically marked cell types.
CONCLUSION
We present a method for high-throughput RGC recordings with the goal of classifying them based on their response to light stimuli. Our high-density MEA produces a large quantity of data that necessitated the implementation of a fast, parallelized spike-sorting algorithm. Due to the high temporal precision of detected spikes, we are able to cluster cells based on the their spike times.
ACKNOWLEDGEMENTS
Financial support through the ERC Advanced Grant 267351 "NeuroCMOS" and the Swiss National Science Foundation Sinergia Project CRSII3_141801, as well as individual support for R. Diggelmann through a Swiss SystemsX IPhD grant are acknowledged.
REFERENCES
Baden, T., Berens, P., Franke, K., Román Rosón, M., Bethge, M., and Euler, T. (2016). The functional diversity of retinal ganglion cells in the mouse. Nature.
Jeon, C.J., Strettoi, E., and Masland, R.H. (1998). The major cell populations of the mouse retina. J. Neurosci. Off. J. Soc. Neurosci. 18, 8936-8946.
Müller, J., Ballini, M., Livi, P., Chen, Y., Radivojevic, M., Shadmani, A., Viswam, V., Jones, I.L., Fiscella, M., Diggelmann, R., et al. (2015). High-resolution CMOS MEA platform to study neurons at subcellular, cellular, and network levels. Lab Chip 15, 2767-2780.
Keywords:
spike sorting,
clustering,
retinal ganglion cell,
light stimulation,
cell type classification
Conference:
MEA Meeting 2016 |
10th International Meeting on Substrate-Integrated Electrode Arrays, Reutlingen, Germany, 28 Jun - 1 Jul, 2016.
Presentation Type:
oral
Topic:
MEA Meeting 2016
Citation:
Diggelmann
R,
Fiscella
M,
Drinnenberg
A,
Franke
F,
Roska
B and
Hierlemann
A
(2016). Using High-Density MEAs For High-Throughput Retinal Ganglion Cell Type Classification.
Front. Neurosci.
Conference Abstract:
MEA Meeting 2016 |
10th International Meeting on Substrate-Integrated Electrode Arrays.
doi: 10.3389/conf.fnins.2016.93.00015
Copyright:
The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers.
They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.
The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.
Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.
For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.
Received:
22 Jun 2016;
Published Online:
24 Jun 2016.
*
Correspondence:
Dr. Roland Diggelmann, ETH Zurich, Department of Biosystems Science and Engineering, Basel, Switzerland, roland.diggelmann@bsse.ethz.ch