Multielectrode arrays
From Scholarpedia
| This article has not been peer-reviewed or accepted for publication yet; It may be unfinished, contain inaccuracies, or unapproved changes. | ||||||||||||||||||||
Author: Dr. Guenter W. Gross, Dept. of Biological Sciences, University of North Texas, Denton, TX
The large number of MEA designs that have been generated and used cannot be summarized in a short article. Therefore this section is limited to planar, passive arrays with substrate-integrated microelectrodes for use in neuronal cell culture.
History
The concept of using groups of microelectrodes for recording from or stimulating neural tissue is as old as the realization that the monitoring of dynamic neural systems requires the simultaneous capture of multiple electrophysiological events. However, the lack of appropriate microelectronic techniques and the large data streams prevented systematic fabrications and application until the second half of the 20th century. By the end of this century, technology had caught up with the demands of neuroscientist resulting in an explosion of designs, applications, and concomitant novel neurophysiological data. The domain of microelectrode arrays is broad and encompasses sets of movable microelectrodes (Crain, 1976), bundles of wire electrodes and fixed array needle electrodes (Schmidt, 1999), microelectronic multisite recording “daggers” (Wise and Angell, 1975). These categories are further complicated by devices with on-board electronics (Wise), or arrays with FETs that are often classified as “active arrays” (Fromherz, et al.,1991). Such arrays must be discussed separately from this article. The review article by Pine provides more detail of early developments (Pine, 2006). Theory, design, and modeling of thin film electrodes s also well represented in a review article by Kovacs (1994).
The first functional planar array using photolithographed thin film conductors was conceived and fabricated by Thomas et al (1972) with gold/nickel conductors on glass and used successfully for recording of action potentials from cardiac myocytes. The first recordings of action potentials from individual neurons was reported in 1977 using spontaneously active Helix pomatia (snail) ganglia (Gross et al., 1977, Gross, 1979). These experiments used a glass plate decorated with 36 thin film gold conductors on titanium,12 um wide, insulated with a polysiloxane and de-insulated with single laser shots (pulsed nitrogen laser, 337 nm). The next major step forward was achieved by J. Pine (1980) when the first signals from mammalian central nervous system cultures were obtained with a 16-electrode array. This was quickly followed by recording from murine spinal neuronal networks (Gross et al., 1982a,b; Gross and Lucas, 1982, Droge et al., 1986) and the introduction of transparent indium tin oxide as an electrode material (Gross et al., 1985) to minimize the optical loss induced by opaque electrode materials.
In addition to the early developments of passive planar arrays and subsequently by commercial enterprises such as Multichannel Systems in Reutlingen, Germany, Panasonic USA, and Ayanda in Switzerland, the simplicity of photolithography has allowed various research groups to fabricate custom arrays suited to their research purposes. This is especially true for slice recording where specific array geometries were introduced to optimize monitoring of activity from different brain regions with specific geometries: retina (Grumet et al., 2000; Meister et al., 1994), spinal cord (Borkholder et al., 1997; Streit et al., 2006), and hippocampus (Boppart et al., 1992; Egert et al., 1998; Novak and Wheeler, 1988; Oka et al., 1999; Thiebaud et al., 1999, Gholmieh et al. 2006). 3-D arrays with tip-shaped electrodes to penetrate the dead cell layer of slices have also been introduced (Heuschkel et al., 2006).
A hippocampal conformal MEA (cMEA) fabricated by the Berger research group at USC is shown in Fig. 3 (Gholmieh et al, 2006).
Cell-electrode coupling in primary cell cultures
Dissociate embryonic tissue forms intimate contact with a hydrophilic insulation surface. However, adhesion-enhancing molecules are generally applied to improve adhesion and long-term stability. A large number of adhesion-promoting materials have been tried among which polylysine or polyornithine ( McKeehan and Ham, 1976), and laminin (Hunter et al., 1991) are the most common. Hydrophobic surfaces can be “activated” by plasma etching or flaming before they accept decoration with polylysine. The latter procedure involves a one second exposure to a butane flame and is simple and economic (Lucas et al., 1986). It has the additional advantage of direct generation of adhesion islands by flaming through masks.
The adhered ”monolayer” is actually a shallow 3-dimensional volume with neurons residing on top of a glial carpet and neurites both on top and below the carpet. Regardless of the covalent surface decoration selected, the use of polylysine or polyornithine, as well as the unavoidable cell debris introduced during cell seeding, masks the chemical characteristics of the original surface, and generates a rather complex non-specific adhesion milieu. Very high signal-to-noise ratios are obtained when an active process crosses a recording crater capped by glia. In the early stages of network development, it has been observed that glia climb over adhered neurites, that neurites climb onto and over glia, but that adhered neuronal cell bodies are lifted by approaching glia that grow between the substrate and the membrane of the soma. Control over this stratefication and the glial capping has not yet been achieved, but would yield very high (several mV) signals from neurites.
It is important to recognize that the ratio of cell mass to adhesion area cannot be too large, as it leads to regional or global retraction of network components. Large fascicles, if allowed or even encouraged to form in networks, develop tension and usually lift off the substrate. The best stability is obtained from quasi monolayers of neurons and glia with neuronal densities not exceeding 500 per mm2 and glia growth controlled by antimitotics such as fluoro-2’-deoxyuridine (Ransom et al., 1977).
Maturation and survival of primary cultures
With present culture maintenance protocols, survival of primary cultures for 6 to 12 months is possible (Gross, 1994; Potter and DeMarse, 2005). For practical reasons, most experiments use 4 to 8 week old cultures. Morphological and electrophysiological stability is generally attained by 4 weeks of growth in vitro. After this time, only minimal cell movement occurs and unique tissue-specific activity patterns have a high probability of being expressed (see below).
Self-organization
Networks in culture are not random systems, but represent self-organized dynamic ensembles (Bettencourt et al., 2007). Connection specificity has been reported (A. Ghosh) and native, spontaneous activity patterns are unique for networks formed from tissue of different brain regions. Whereas the cortical networks show highly coordinated bursting, midbrain tissue expresses high spike rates with embedded, minimally coordinated bursts, and spinal cord networks display multiple, simultaneous patterns with strong bursts of relatively long duration (Fig. 5 )
Applications
In addition to the obvious quest for understanding how neuronal ensembles function and how they generate or process spatio-temporal action potential patterns (cf Taketani and Baudry, 2006), primary networks in culture have shown a surprising histiotypic (like the parent tissue) pharmacologic behavior, if prepared to contain ratios of neuronal and glial components similar to those found in the parent tissue (Gross and Gopal, 2006). Consequently, it has been proposed that such networks on MEAs can be used as effective and rapid screening platforms in pharmacology, toxicology, for drug testing, and even tissue-based biosensors (Gross and Pancrazio, 2007). The MEA approach is crucial to obtain average system information, fault tolerant read-out based on many cells, individual neuron response profiles, and action potential wave shape data .
Given the histiotypic nature of pharmacological and toxicological network responses, the application of these platforms to rapid screening of compounds represents a natural next step. Such platforms provide the following specific advantages:
- Long-term (days to weeks) multisite action potential readout with cell identification based on wave shape templates.
- Control over the biochemical environment.
- Determination of functional changes (in the absence of cell death).
- Quantification of cytotoxicity using activity decay as the primary measure.
- Close correlation between electrophysiological and morphological changes.
- Combined electrophysiological and fluorescence readout.
In addition, one mouse with 12 embryos can provide enough central nervous system tissue to seed up to 1000 networks This is a very high tissue utilization efficiency that, however, cannot be put to practice until the technology for massively parallel multi-array platforms is developed.
Pharmacology on MEAs
Figure 7 shows the response of a spinal cord network in terms of mean burst duration plotted against mean burst rate to a sequential addition of different compounds designed to block inhibitory and, partially excitatory, synapses culminating in a network driven almost exclusively by NMDA synapses (Keefer et al, 2001). Starting with native activity, the culture received 40 uM bicuculline to block GABAA inhibition, followed by 1 uM strychnine to remove glycinergic inhibitory influences, 50 uM SCH50119, a GABAB antagonist, and 20 uM NBQX to block AMPA/kainate receptors. Charybdotoxin (15 nM) and apamin (1 uM) antagonize, respectively, the large and small conductance Ca++ activated K+ channels and were used to explore the effect of the Ca++ activated BK channel on activity patterns. A surprising result was the rapid development of a highly regular burst pattern after addition of NBQX at burst periods of 2.9+/-0.3 sec (18 – 23 bursts per min). Subsequent additions of cholinergic and dopaminergic blockers had no measurable effect on the final “NMDA-only” state (Keefer et al., 2001).
Prototype multi-array platforms
Applications in research benefit from parallel recording arrays that allow simultaneous experimental and control studies and provide statistical results in a short period of time. For drug efficacy and toxicity screening these approaches to high throughput are essential and platforms must be scaled up to a maximum possible number of networks. These efforts present a significant technical and programming challenge in the areas of life support, pipetting accuracy and repeatability, and automated data analysis.
A prototype 8-network MEA has been developed by the CNNS and has shown promise in numerous experiments. Life support is provided by a 10% CO2 atmosphere and 70% humidity with robotic water additions, determined empirically, to maintain osmolarities. Both the robot (Biotek Precision 2000) and the preamplifiers function well in a 70% humidity environment.
Modification of passive MEAs with carbon nanotubes
A remarkable reduction of electrode impedances (up to a 23 –fold reduction) can be achieved by decorating the exposed metal recording sites with carbon nanotubes (CNTs) or organic conductive polymers (Keefer et al. 2008). This simple electrolytic modification allows further down sizing of conductor dimensions, recording site diameters, and a substantial increases in electrode densities without compromising optical access and signal-to-noise ratios. Especially the potential size reduction of the optically opaque recording site, which normally interferes with morphological determinations of cell-electrode coupling, can make important contributions to circuit tracing.
Application to studies of pattern generation and information processing
Whereas quantitative pharmacology and the determination of substance toxicity have proven to be relatively simple if stable life support is assured and appropriate application protocols are used, investigations of plasticity, learning, pattern processing, and fault tolerance have been found to be more difficult. It is assumed that the former domain is governed primarily by the pharmacological sensitivities of the cell types, receptors, and synapses found in the culture rather than by the detailed circuitry. However, information processing appears to be influenced at least to an equal degree by the circuitry that is established in vitro. Such circuitry, even in these highly simplified systems, is difficult to determine morphologically. However, progress is now being made in establishing functional connectivity (Eytan and Marom, 2006; Bettencourt et al., 2007; Eckmann et al., 2008; Ham et al. 2008). Subtle plasticity phenomena are also yielding to investigation (Jimbo et al, 1999; Shahaf and Marom, 2002) and dynamic attractors in activity patterns are being identified (Beggs and Plenz, 2004; Wagenaar et al., 2006).
