当前位置：文档库 › Multi-electrode array technologies for neuroscience and cardiology

Multi-electrode array technologies for neuroscience and cardiology

central goal of contemporary neuroscience is to understand the relationships between the functional connectivity-map of neuronal circuits and their physiological or pathologi-cal functions. At present, this goal seems impossible to achieve as available electrophysiological technologies only allow for extracel-lular recordings of large populations of neurons and are not suitable for providing simultaneous intracellular recordings of neural activ-ity from hundreds of individual neurons. Furthermore, because of their mechanical instability, intracellular glass electrodes cannot be used to monitor long-term electrophysiological correlates of plastic-ity and learning. An ideal multiunit readout system should provide information that covers the entire repertoire of electrophysiological parameters from the individually recorded neurons. These include: action potentials (APs), subthreshold excitatory- and inhibitory-postsynaptic potentials (EPSPs and IPSPs, respectively), and sub-threshold membrane oscillations. Moreover, it should be possible to modulate the activity of individual neurons within the network by current application.

The available methodologies for the recording of neural activ-ity include: (a) intracellular recordings and stimulation by sharp or patch electrodes, (b) extracellular recordings and stimulation by substrate-integrated microelectrode arrays (MEAs), (c) optical imag-ing and stimulation technologies of extrinsic fluorescent indicators or genetically encoded molecular probes, and (d) other methods such as functional magnetic resonance imaging, electroencephalography, electrocorticography and magnetoencephalography, designed to record activity from very-large-scale neural populations, which are not suitable for single-neuron resolution.

The invention of intracellular recording and stimulation technolo-gies were hallmark developments that enabled the biophysical ‘lan-guage’ by which individual neurons transmit electrical information, communicate and ‘compute’ subthreshold synaptic information to be deciphered 1–4. The power of intracellular recording systems is that they exhibit very good electrical coupling with the cell and provide accurate readout of the entire dynamic range of voltages generated by cells without distorting the readout over time. Y et, the use of sharp or patch microelectrodes is limited to individual neurons as steering Multi-electrode array technologies for neuroscience and cardiology

Micha E. Spira* and Aviad Hai

At present, the prime methodology for studying neuronal circuit-connectivity, physiology and pathology under in v itro or in v ivo conditions is by using substrate-integrated microelectrode arrays. Although this methodology permits simultaneous, cell-non-invasive, long-term recordings of extracellular field potentials generated by action potentials, it is ‘blind’ to subthreshold synaptic potentials generated by single cells. On the other hand, intracellular recordings of the full electrophysiological repertoire (subthreshold synaptic potentials, membrane oscillations and action potentials) are, at present, obtained only by sharp or patch microelectrodes. These, however, are limited to single cells at a time and for short durations. Recently a number of laboratories began to merge the advantages of extracellular microelectrode arrays and intracellular microelectrodes. This Review describes the novel approaches, identifying their strengths and limitations from the point of view of the end users — with the intention to help steer the bioengineering efforts towards the needs of brain-circuit research.

of the electrode tips into target cells requires the use of bulky micro-manipulators and the duration of intracellular recording sessions is limited by mechanical and biophysical instabilities.

In contrast, whereas the use of cell-non-invasive extracellular MEAs for in vitro recordings and polytrodes for in vivo recordings largely attenuate and temporally filter the electrical signals, it ena-bles the simultaneous recording and stimulation of large populations of excitable cells for days and months without inflicting mechani-cal damage to the neuron’s plasma membrane 5–10. Extracellular field potential recordings (ambiguously referred to as local field potentials (LFPs) or field potentials (FPs)) reflect the spike activity of individual neurons or the superposition of fast APs, synaptic potentials and slow glial potentials in both time and space. As the physical processes that underlie the generation of FPs are understood it is theoretically possible to reconstruct their sources. Even though a great deal of information can be gained by using polytrodes, the information har-boured in spike-pattern fingerprints is limited 11. For example, exten-sive spike sorting cannot provide information as to whether the firing of an individual neuron is triggered by endogenous mechanisms, a barrage of incoming excitatory inputs or the cessation of inhibition (Box 1). What terminates the firing of a given neuron? Is it a barrage of inhibitory synaptic inputs, cessation of excitatory inputs or hyper-polarization of the membrane potential by endogenous mechanisms? Neurons that do not fire APs during a recording session are not ‘vis-ible’ to extracellular electrodes (referred to as ‘dark neurons’; Box 1). In some brain areas, 90% of the neurons are not spiking or are firing occasionally at very low rates of <0.16 spikes per second (for review see ref. 12). Intracellular recordings of synaptic potentials from such neurons would disclose a great deal of information as to the role of this ‘silent majority’ in information processing and the importance of individual neurons to the circuit behaviour. A great deal of neu-roplasticity is associated with changes in the amplitude of synaptic potentials 13. Unless these changes reach firing threshold, extracellular recording systems are ‘blind’ to these critical events. It is conceivable that significant signalling between neurons is mediated by subthresh-old potentials (chemical or electrical synapses) and is thus undetect-able by conventional extracellular electrodes.

The Alexander Silberman Life Sciences Institute, and the Harvey M. Kruger Family Center for Nanoscience, The Hebrew University of Jerusalem, Jerusalem 91904, Israel. *e-mail: spira@cc.huji.ac.il

Since the development of the first MEA14–17, technological efforts improved the quality of information gained by extracellular recordings mainly by increasing the density and the number of the electrodes that can be constructed and addressed over a single MEA. In vitro MEAs may contain over 10,000 electrodes5,7,8,18,19. In vivo polytrodes may have over a hundred9,20. Nevertheless, the recording and stimulation qualities of these platforms (reflected by the electrical coupling coefficient between single neurons and the device, and the signal-to-noise ratio) remained poor. Typically the amplitudes of FPs range between 10 μV to 1 mV and a great deal of computational power is required to extract data and sort out the recorded signals21,22.

In parallel to the development of the extracellular MEA, efforts to develop optical imaging approaches began. These included imag-ing of membrane potentials by the use of voltage-sensitive dyes23–27, imaging of neuronal activity by monitoring the changes in the free intracellular calcium concentration28–30 and the monitoring of intrin-sic signals31. This was followed by very powerful methods to optically excite or inhibit individual neurons or neuronal ensembles32,33. The powerful optical imaging methods suffer from a number of limita-tions that at present prevent them from replacing electrophysiological approaches for studying neuronal circuits (for discussion see ref. 34).

Using nano- and micro-technologies, a number of laboratories began to merge the advantages of substrate-integrated extracellu-lar MEA technologies with the critical advantages of intracellular electrodes. Namely, the construction of nano- or microdevices that enable simultaneous, long-term, multisite, intracellular recording and stimulation from many neurons under in vitro conditions. Further development and implementations of these technologies are expected to revolutionize basic and applied neuroscience.

In this Review we describe recent developments, expected ben-efits from their use and the foreseen limitations of the different approaches. We review these developments from the end-user point of view, rather than from the technological point of view. By consid-ering the needs of contemporary ‘circuit neuroscience’ on the one hand and the recent technological developments in MEA fabrica-tion on the other, we hope to shift the focus of upcoming technical developments from the habitual engineering optimizations to the pressing needs of neuroscience.

To objectively evaluate the different approaches, we examine the principal achievements in relation to a list of biophysical parameters that are needed to decipher the functional connectivity map of a neuronal network. An ideal imaginary device would allow the user to: (a) simultaneously record and stimulate hundreds of individual

REVIEW ARTICLE NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2012.265

neurons intracellularly, (b) maintain a stable contact with the neu-rons for recording and stimulation for days and months, (c) moni-tor the transmembrane potential in the relevant cell-physiological range of ?80 to +30 mV, (d) detect subthreshold potentials such as excitatory and inhibitory synaptic potentials with amplitudes in the range of ±0.5–10 mV with a rise time of <1 ms and a slow decay time of 100–1,000 ms, and to record membrane oscillations in the range of ±5 mV at frequencies of 1–50 Hz, (e) record APs with amplitudes of ~100 mV and duration of 1–500 ms (long APs for recording from cardiomyocytes).

To acquaint the reader with the basic terminology used in the field and as a technical introduction we begin the discussion by describing the structural and electrical relationships formed between neurons and substrate-integrated MEAs and briefly explain the contribution of the various parameters to the electrical coupling between excitable cells (neurons and cardiomyocytes) and MEAs. We then review the micro- and nanoengineering approaches recently used in attempts to merge the benefits of extracellular MEAs with those of the sharp glass electrodes, as well as the critical problems that would have to be addressed in order for the new approaches to be incorporated in basic research and clinical applications. On the basis of this discussion we propose a tentative recipe that we believe might provide an optimal approach to construct an MEA that can provide simultaneous, long-term, multisite, non-destructive intracellular recording and stimula-tion of neurons.

Electrical circuit analogue of the neuron/electrode interface The structural relationships between a neuron and a substrate-inte-grated planar electrode along with the analogue electrical circuit are schematically depicted in Fig. 1. The neuroelectronic hybrid is com-posed of three components (a) a neuron, (b) a cleft formed between the neuron and the substrate surface, and (c) the electrode. Typically, differentiated neurons are non-isopotential structures with a cell body from which neurites in the form of a dendritic tree and a single long cylindrical axon emerge. These neuronal compart-ments adhere to MEA substrates by electrostatic or chemical inter-actions between adhesion molecules that protrude from the lipid membrane of the neurons and molecules deposited on the MEA plat-forms by the experimentalist18,35. The cleft formed between the cell membrane and the MEA substrate is filled with the ionic solution. For the simplified model depicted in F ig. 1, the neuron surface area is subdivided into: a junctional membrane that faces the sensing

pad(s) (with R j as the junctional resistance), and the non-junctional membrane (with R nj as the non-junctional resistance) that faces the bathing solution and the substrate. Propagating APs or synaptic potentials produce complex extracellular current flow between acti-vated compartments and other parts of the neuron. Fractions of these extracellular currents flow between the non-junctional and the junc-tional membranes. The cleft, formed between the neuron and the sensing element, generates a resistance that is referred to as the seal resistance (R seal). The voltage formed over R seal directly modulates the gate voltage of a field-effect transistor (FET), or the charge dispersal across a passive metal electrode36–38.

Neuron–device electrical coupling

The electrical coupling between a neuron or a cardiomyocyte and a sensing pad is defined here as the ratio between the maximal voltages recorded by the device in response to the maximal voltage generated by an excitable cell. Figure 2 illustrates how individual parameters of a passive analogue electrical circuit contribute to the coupling coef-ficient of slow membrane oscillations, medium-frequency synaptic potentials, and fast APs (Fig. 2a). The parameters used for the simula-tion approximate the passive physical properties of an excitable cell cultured on a substrate-integrated electrode (Fig. 2a). Simulation of voltaic events of various frequencies illustrates that the coupling coefficient for higher-frequency APs is attenuated more strongly compared with postsynaptic potentials and membrane oscil-lations (Fig. 2b and Fig. 3).

Theoretical and experimental considerations revealed that reduc-tively, the amplitude and shape of the FPs are determined by the mul-tiplication of the R seal value by the current that flows across it (Fig. 2c). For these reasons, intensive efforts were devoted to increase the value of R seal (refs 35,37,39). Most of these research efforts yielded only lim-ited improvements in the signal-to-noise ratio of the various devices. Studies of cell/electrode interfaces showed typical cleft thicknesses between 40–100 nm (refs 40–44). F or most cell types this would correspond to R seal in the range of 1–2 M? and FP recordings in the range of a few tens to few hundreds of microvolts.

Another factor that largely affects the electrical coupling coef-ficient between cells and an MEA is the input impedance of the sensing pad (F ig. 1 and F ig. 2d). The currents of living cells and electronic devices are fundamentally different: the former are formed by ions in solution whereas the latter, by electrons in mostly solid-state metals and semiconductors37,45. The ramification of this difference is the effect of the impedance of the device on the electri-cal coupling, reflected by the sensing pad geometry and the mate-rial it is composed of. Typically the impedance of the sensing pad, either constructed from noble metals or insulated semiconduc-tors, is attributed to the ‘blocking’ ion bilayer formed at the device

Figure 1 | Schematic layout depicting the spatial relationships between

a neuron and a substrate-integrated electrode and the analogue passive electrical circuit. The cell body of a neuron (light blue) resides on a sensing electrode (orange) integrated in the culture substrate (yellow). The electrode is coupled to an amplifier (yellow). A cleft filled by the culturing media (ionic solution) interposes between the cell membrane and the electrode–substrate. The neurons plasma membrane is subdivided into two: the part that faces the electrode (blue) is defined as the junctional membrane and is represented by the junctional membrane resistance (R j) and the junctional membrane conductance (C j). The rest of the membrane, defined as the non-junctional membrane (red), faces the bathing solution and the culture substrate. This part of the membrane is represented by the non-junctional resistance (R nj) and the non-junctional capacitance (C nj). The physiological solution within the cleft generates the seal resistance (R seal) to ground. The electrode (orange) impedance is represented by

the electrode resistance and capacitance (R e and C e, respectively). The electrode can be a passive element or a transistor. For simulation purposes of APs or intracellular current injections, current can be injected into the analogue cell-circuit in-between R nj and R j. Under physiological conditions current is generated by transient changes in the membrane conductances. The colour coding shown here is used in Fig. 4 to depict the different components of the analogue electrical circuit.

REVIEW ARTICLE

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2012.265

active region and the ionic solution in which the neurons reside. For example, standard planar gold electrodes with radius of 30 μm have impedances of 50 K? at 1 KHz in electrolyte solution. Reducing the surface area of individual sensing pads to match the dimen-sions of individual neurons enables the density of the MEA and its spatial resolution to be increased 5,7,8,46,47. This however, is reflected by the reduction of the FP amplitudes as a result of the increase in impedance and consequent reduction of the signal-to-noise ratio. Thus, the electrodes geometry and the ensuing impedance both place constraints on reducing the electrode size. Increasing the sur-face area by using nanostructures such as spongy platinum black or Ti 3N 4 (refs 48–50), gold nanoflakes and nanopillars 51,52, or car-bon nanotubes 53,54 is used to compensate for the dimensions of the electrode surface ‘visible’ to the cell. Although effective in reducing the impedance values up to 95% at approximately 1 KHz (ref. 54), in practice the recorded F Ps are still in the range of hundreds of microvolts. This is most likely due to averaging of the complex posi-tive and negative currents concomitantly generated by a number of sources, over fractions of the large surface area of the electrode. This ‘averaging’ usually results in reduced amplitude of the electrical

readout. It should be noted, however, that reduction of the electrode impedance may be very effective in improving the readout signals when applied under conditions in which a single cell ‘covers’, engulfs or internalizes a single electrode such as those described in this Review (Fig. 4).

The junctional membrane is defined as the resistance and capaci-tance of a membrane patch that faces the sensing pad or the gate of an FET (Fig. 1). The surface area of the junctional membrane can be anywhere between a very small fraction of the cell surface area, up to approximately 50% in cells that flatten while adhering strongly to substrate-integrated sensing pads. This variable depends on the geometry of the sensing pad and the morphology and adhesion char-acteristics of the specific cell. The junctional membrane can thus be of very high resistance and low capacitance. This implies that only a small fraction of the current generated across the neuron’s membrane, flows through the junctional membrane. Reduction of the junctional membrane resistance would be very effective in improving the elec-trical coupling coefficient between a neuron and an electrode (Fig. 2e and Fig. 3). This is in fact the approach used by the classical meth-ods of sharp electrodes, whole-cell patch electrodes or the perforated

501001502000

102030Non-junctional membrane

resistance (M ?)

C o u p l i n g (%)

Membrane oscillations (1–50 Hz)

PSPs (100 Hz)

APs

(500–1,000 Hz)

Solution

Electrode

0102030405060

1020304050Stray capacitance (pF)

C o u p l i n g (%)

106

10710810910100102030

Junctional membrane

resistance (?)

C o u p l i n g (%)

106

107108109101010111020304050Electrode impedance (?)

C o u p l i n g (%)106

107108109

0102030Seal resistance (?)

C o u p l i n g (%)

Frequency (Hz)

C o u p l i n g (%)

Figure 2 | Simulation of the contribution of individual electrical components to the neuron-electrode electrical coupling coefficient. a , Left: the analogue passive electrical circuit and default parameters used for the simulation parameters of a representative passive analogue electrical circuit. Right: the simulations were conducted for low-frequency signals depicting membrane oscillations, medium frequencies depicting postsynaptic potentials (PSPs) and fast frequencies depicting APs. C amp , input capacitance of the amplifier; Z stray , stray impedence. For the simulation, all default parameters were kept constant as shown in (a ) while the tested parameter was varied in the range indicated by the horizontal axis. b –g , The coupling coefficient as a function of: voltage pulse frequency (b ), the seal resistance (c ), the sensing pad impedance (d ), the junctional membrane resistance (e ), the non-juctional membrane resistance (f ) and the stray capacitance (g ). Simulations based on refs 64,65.

REVIEW ARTICLE

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2012.265

patch configuration3,55 (F ig. 4a,b). Attempts to improve the electri-cal coupling coefficient between cultured cells and planar MEAs by expression of ion channels in the plasma membrane provided experi-mental demonstrations of the contribution of the conductance of the junctional membrane37. Nevertheless, experimental manipulation to improve the electrical coupling by expression of ion channels in neu-rons should not be used in studies of neuronal networks as such a manipulation changes the electroanatomy of the cells and their excit-ability and thus alters the functioning of the network being studied. In some recent studies, local increases in the junctional membrane con-ductance by localized electroporation have been used to transiently increase the coupling coefficient56–60 (Fig. 4e).

Unavoidable stray capacitance along the conducting lines, together with the input impedance of the amplifying circuitry, further attenu-ates the recorded signals (Fig. 2g).

Recently, a number of laboratories began to merge the advantages of extracellular MEAs and intracellular microelectrodes. In the fol-lowing paragraphs we describe the novel approaches, identifying their strengths and limitations from the point of view of the end users. Neurons actively engulf protruding electrodes

The first series of studies reporting on successful multisite, non-invasive, intracellular recording and stimulation by MEAs were pub-lished by our laboratory between 2007 and 201061–65. In these studies we increased the neuron–microelectrode electrical coupling coeffi-cient from approximately 0.1% as recorded by a planar extracellular MEA to approximately 50% by the use of a chemically functionalized micrometre-size mushroom-shaped gold protrusion as the sensing electrode (F ig. 4d and F ig. 5a). The increased coupling coefficient was associated with an intracellular recording of a monophasic posi-tive attenuated intracellular AP instead of a typical biphasic FP. The unique neuron–electrode configuration used in this work, made it possible, for the first time, to record with a MEA action potentials as well as synaptic potentials (Fig. 5c). The key to the multi-electrode-array ‘in-cell recording’ approach developed by us is the outcome of three converging cell biological principals: (a) the activation of endocytotic-like mechanisms in which cultured Aplysia neurons are induced to actively engulf gold mushroom-shaped microelectrodes (gMμE) that protrude from a flat substrate, (b) the generation of high R seal between the cell’s membrane and the engulfed gMμE, and (c) the increased junctional membrane conductance.

The neuron/gMμE interface was generated by chemically ‘luring’ the neurons to engulf the protruding gMμE by a highly conserved cell biological mechanism — endocytosis (which is a cell biological mechanism that underlies the internalization of particles into the cells66). The shape and the dimension of the gMμE were selected to mimic the geometry and dimensions of dendritic spines67. To facilitate the engulfment, the gMμEs were chemically functionalized by an RGD-based peptide35. This is one of a group of well-known molecular recognition motifs that trigger adhesion and engulf-

ment mechanisms68. The localized presentation of the peptide by the gMμE (and possibly the electrode geometry itself) led to active engulfment of the electrodes by Aplysia neurons and a number of cell lines61–65. The engulfment of the microelectrodes is generated by molecular cascades that include the restructuring of the cytoskel-eton to form an actin ring around the stalk of the ‘gold mushroom’ (Fig. 4d and Fig. 5a, left panel).

Using biophysical parameters obtained by direct measure-ments of the non-junctional resistance, non-junctional capacitance, electrode resistance, electrode capacitance and R seal, as well as the estimated value of the junctional membrane properties calculated according to the geometry of the gMμE (Table 1), we simulated the expected recordings of APs and subthreshold potentials using an equivalent electrical circuit. We found that the calculated val-ues used for the junctional resistance were insufficient to gener-ate the neuron–electrode coupling coefficient as that obtained experimentally. To reach the coupling level observed in the experi-ments we had to increase the junctional membrane conductance by at least an order of magnitude64,65. The mechanism underlying the increased junctional membrane conductance is not known. It is conceivable that in association with the cytoskeleton restructuring around the gMμE, voltage-independent ion channels are recruited to the junctional membrane to sufficiently increase the junctional conductance without leading to noticeable effects on the passive and active membrane properties. An alternative explanation could be that the mechanical tension generated at the curvature of the gMμE generates non-specific membrane nanopores58,69. An important question that was brought up in this relation is whether the growth of Aplysia neurons on a matrix of gMμEs is altering the neuron’s physiological properties. We found that the growth of the neurons

on gMμEs did not alter the biophysical properties of the neurons

5 mV

1 nA Current

injection

Device

High-frequency APs Low frequency

(hyperpolarization)

R j = 100 M?

R j = 1 G?

R j = 100 G?

20 mV a

Sharp glass

micropipette

100%

Figure 3 | Dependency of the electrical coupling on the junctional membrane resistance and pulse duration. Shown is a simulation of

the cell–device coupling of APs and a long hyperpolarizing pulse at

three different values of R j (100 MΩ, 1 GΩ and 100 GΩ) a, Schematic illustrations of the depolarizing (left) and hyperpolarizing current pulse (right) delivered to generate two APs and membrane hyperpolarization, respectively. b, Simulation of the ensuing intracellular potentials recorded by an intracellular electrode (red). c–e, The recorded potentials by an extracellular-located electrode (as shown in Figs 1 and 2) under different junctional membrane values (blue). Note that the amplitude of the extracellularly recorded APs (short pulses) is reduced faster than the voltage generated by the long pulse. That is, the coupling coefficient of

the AP is more sensitive to the value of R j than that of the long pulse.

The values used for the simulation are identical to those shown in Fig. 2 although R j is altered as indicated.

REVIEW ARTICLE

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2012.265

or their synaptic communication. It does however alter the typical growth patterns of the neurites63.

Stimulation of cells by substrate-integrated electrodes (either passive metal electrodes or F ETs) often involves undesired elec-trochemical reaction products at the electrode interface and consequently damage to the cells that goes beyond transient elec-troporation. Measurements of the neurons input resistance before and after stimulation by gMμEs showed no change, affirming that high enough charge transfer can be applied to evoke APs without damage to the cell64.

Out of the five criteria to evaluate of the benefits of the approaches, the gMμE-based MEA provided multisite, simultaneous, intracellular recording and stimulation for periods of days (which is for as long as we carried out the recordings). The filtering properties of the gold electrodes and the a.c.amplifier used do not enable the resting poten-tials of the neurons to be recorded. Nevertheless, the configuration b

a c e g

Micro?uidic channel

Figure 4 | Different forms of the electrode/neuron interface configuration. The colour code represented here for the different components is the same

as for Fig. 2. a, A sharp glass intracellular microelectrode. b, Whole-cell patch-electrode configuration. The ‘mixing’ of orange and blue schematically illustrates the perfusion of the cytosol by the electrodes content. c, A neuron cultured on a substrate-integrated planar extracellular electrode. Note the cleft (white) separating the junctional membrane and the electrode. d, A neuron engulfing a gold mushroom-shaped protruding microelectrode. Note actin rings surrounding the mushrooms stalk stabilizing the configuration. e, Nanopillar electrodes extending into a cultured cardiomyocyte but that do not penetrate the plasma membrane (i). After the application of an electroporating pulse (ii) the nanopillar gains access to the cytoplasm. The electroporation is transient and the junctional membrane resistance recovers to control level within minutes (iii). f, An array of nanopillars that penetrate the plasma membrane forming direct physical contact with the cytosol. g, A nanopillar that serves as the gate for a nano-FET penetrates the cell’s membrane. h, Patch clamping of cultured neurons. The mixing of the ionic solution of the microfluidic system with the cytosol is depicted. For more details see main text.

REVIEW ARTICLE NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2012.265

20 mV 1 nA

0.25 s

d 5 mV

V NW

g e

Figure 5 | Recently developed MEA devices record intracellular potential from excitable cells. a , Gold mushroom-shaped microelectrodes (gM μEs) functionalized with RGD-based peptide (top left, scanning electron microscopy (SEM) image) are engulfed by Aplysia neurons (top right, transmission electron microscopy image), which induces cytoskeletal reorganization around the structure (bottom, confocal microscopy images). The gM μE is 1.42 μm high. b ,c , Stimulation and recording of APs and subthreshold synaptic potentials is achieved by the gM μEs with a signal-to-noise ratio similar to sharp-glass and patch-clamp micropipettes. Green, current injection into the neuron; red, intracellular recording by a glass microelectrode; blue, in-cell recording by an extracellular gold mushroom microelectrode. d , SEM (left) and optical (right) images of vertical nanowire electrode arrays (VNEAs). Scale bars, 2.5 μm. e ,f , Vertical nanowire electrode arrays (VNEA) can stimulate (e ) and record (f ) from rat cortical neurons monophasic APs. V p , patch clamp voltage; V NW , nanowire recordings voltage. g –i , A phospholipid-functionalized silicon nanotube as the gate electrode of an FET device records APs from cardiomyocytes with a signal-to-noise ratio similar to glass electrodes. g , SEM images of the device (left) and of cardiomyocytes grown on it (bottom right). h , Recording from a nanotube before peneration (left) of the cell’s plasma membrane and after penetration (right). i , Left: FPs recorded by the nanotube before penetration of the cell. Right: a single intracellularly recorded potential. Figure reproduced with permission from: a , T op: ref. 62, ? 2009 RSC; Bottom: ref. 63, ? 2009 IOP; b ,c , ref. 64, ? 2010 Am. Physiological Soc.; d –f , ref. 71, ? 2012 NPG; g –i , ref. 78, ? 2012 NPG.

REVIEW ARTICLE

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2012.265

successfully monitored subthreshold synaptic potentials and APs (Fig. 5c, middle panel). The filtering nature of the recording system can be deconvoluted and thus unfiltered high-quality recordings of APs and synaptic potentials can be retrieved. A stable electrical cou-pling between gMμE and a neuron coincided with the formation of cytoskeletal actin rings surrounding the stalks of the mushroom-like structure59,63 (Fig. 5a, lower panel). Individual gMμE enables both volt-age recordings and application of current64. So far, attempts to obtain in-cell recordings and stimulation from rat hippocampal neurons and primary cardiomyocytes were unsuccessful. It should be noted never-theless that these attempts were limited to gMμE functionalized with poly-d-lysine rather than by the engulfment promoting peptide. Nanopillars for intracellular recordings and stimulation Sharp glass electrodes (Fig. 4a) and patch electrodes (Fig. 4b) pro-vide excellent intracellular recording by penetrating the plasma membrane and directly accessing the cytosol while generating effective R seal with the plasma membrane (hundreds of M? to a few G?, respectively, Table 1). A recent study from Park’s laboratory71 used vertical nanowire electrode arrays (VNEAs) constructed from a doped silicon core encapsulated by silicon dioxide and tipped by Ti/Au, to generate an identical configuration to sharp intracel-lular electrodes(Fig. 4f and Fig. 5d). In the study, 3 × 3 arrays of 9 nanopillars, 150 nm in diameter, 3 μm in height at 2 μm pitch were grown on 16 sensing pads. Embryonic rat cortical neurons or HEK293 cells were then cultured on the VNEAs for a number of days. About 50% of the VNEAs spontaneously penetrated through the plasma membrane of the HEK293 cells as demonstrated by the fact that current injection through the VNEAs generated a voltage drop across the plasma membrane. In cases where spontaneous penetration of the membrane was not evident, an electroporating pulse (approximately ±6 V, 100 ms) was applied to penetrate the membrane of the neuron. The effect of the electroporating pulse on the integrity of the membrane was not shown. The seal resistance formed between the VNEAs and the plasma membrane was esti-mated to be 100–500 M?. The VNEAs were used in two regimes: in the Faradic regime, when a bias of ~?1.5 V was applied to the nanowire and the access resistance was reduced to 300 M?, and in the capacitive regime, when no bias was delivered and the access resistance was infinite. Accordingly, in the Faradic regime the elec-trical coupling for APs between the cell and the VNEA was about 10% whereas in the capacitive regime the attenuation was larger, reaching a lower coupling of ~0.1–0.3% (Table 1).

Consistent with the intracellular positioning of the VNEA, all recorded APs were positive monophasic (as discussed in refs 18,70). However, the coupling coefficient and signal-to-noise

REVIEW ARTICLE NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2012.265

ratio were insufficient to enable recordings of subthreshold synap-tic potentials. In fact, in relation to the five examination criteria listed above, the recordings obtained by the VNEA do not provide significant advantages over extracellularly positioned gMμE in cultured hippocampal neurons70. The relatively low coupling coef-ficient of the VNEA-based sensor is most likely due to the high impedance of the VNEAs (Table 1). An advantage of recording by VNEAs over recordings by classical substrate-integrated planar electrodes (but not small electrodes7) is that a single pad records APs from an individual neuron. Although Robinson et al.71 emphasize the usefulness of VNEAs to address individual neurons by electrical stimulation and record the ensuing synaptic poten-tials by a patch electrode, it should be noted that different types of extracellular micrometre-size planar (Fig. 4c) and protruding microelectrodes have been shown to effectively record or stimu-late single cells19,56,72,73. For example, Hofmann et al.73 constructed a liquid-filled nanocavity (50 nm in height) that accesses a low impedance electrode by the fluid that fills the cavity. The functional contact of a cell with the electrode is through an aperture with a diameter smaller than the cell’s. This configuration increases the spatial resolution of a MEA and increases the R seal formed between the cell and the device. The cell–nanocavity configuration ena-bles negative almost monophasic FPs with amplitudes >1 mV to be recorded. Using an array of 600-nm-thick tungsten protruding micronail-like electrodes Huys et al.19 recorded biphasic or mono-phasic negative FPs of 50–100 μV. These micronails also enabled single cells to be stimulated effectively.

Intracellular recordings by membrane electroporation

As pointed out by Robinson et al.71, only a fraction of the nanow-ires spontaneously penetrate the plasma membrane. Nevertheless, application of current through the pillars readily leads to their func-tional penetration.

Four recent studies demonstrated for the first time that localized membrane electroporation may lead to transient intracellular record-ings of attenuated APs. Xie et al. demonstrated electroporation of cultured cardiomyocytes by vertical nanopillar electrodes57, Hai and Spira58, and F endyur and Spira59 demonstrated electroporation of cultured Aplysia neurons and primary cultured rat cardiomyocytes, respectively, by the gMμEs, and Breaken et al. demonstrated single-cell cardiomyocytes electroporation using micrometre-sized TiN protruding electrodes60. In all four studies the intracellular access was transient suggesting that electroporation activates repair mechanisms that seal off the electroporated nanopores58, leaving the protruding nano- or microelectrodes out of the cell (Fig. 6). In these studies, the electroporating pulses reduced the junctional membrane resistance sufficiently to transform a typical biphasic cardiomyocyte F P to a 1–11 mV positive monophasic AP. The shape of the attenuated AP was typical of cardiomyocyte potentials.

The transient nature of the electrical coupling and the attenua-tion of the APs indicate that electroporation cannot be the method of choice to improve the quality of the interface formed between micro- or nano-based MEAs and the cell’s membrane. Nevertheless, as it was demonstrated by the Melosh laboratory74–76 and already applied by the Lieber laboratory77,78, it is conceivable that membrane–electrode fusion and G? seal formation can be facilitated by proper surface functionalization of the electrodes using lipid-based agents.

It is of interest to note that the nanopillar approach does not pro-vide a significant advantage over planar extracellular electrodes as the recorded potentials are attenuated by at least an order of magni-tude by the inherent high electrode impedance and the insufficient R seal. It should also be noted that even if a single sensing pad carries

?102649113125148

ii iii iv

Time (s)

300607

24 h

2 mV

200 ms

0.2 mV

10 ms

b c

i iii iv

Figure 6 | From extracellular field-potential recordings to intracellular recordings and the recovery process. a, Before electroporation, a gMμE recorded a rat cardiomyocyte extracellular FP (a, (i); enlarged in b, (i)). After the delivery of an electroporating pulse (100 ms, 1 V) (at 0 s) the biphasic FP transformed into a monophasic 5 mV positive potential with a shape similar to that reported by intracellular recordings. The amplitude of the AP diminished over time 125–148 s, gradually resuming the shape of the extracellular FP after electroporation. Thereafter the shape of the FP gradually recovered (between 148–607 s, enlarged in b, i–iv) regaining the typical biphasic shape (a, (iv) and b, (iv)). b, Enlargements of the FPs before electroporation (i) and 148 s (ii), 607 s (iii) and 24 h (iv) after electroporation. c, Schematic drawing of the presumed reversible effects of an electroporating pulse on the plasma membrane facing a gMμE. Figure reproduced with permission from ref. 59, ? 2012 Frontiers Media.

REVIEW ARTICLE NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2012.265

multiple nanopillars, and a number of them penetrate the plasma membrane, the electrodes impedance is too high to enable recording of subthreshold potentials57,71 (Table 1). Theoretically, the impedance problem could be solved by increasing the density of the nanopillars over the sensing pad. Nevertheless, when the density of protruding nanostructures exceeds a certain bound, the pillars do not penetrate the cell membrane, analogous to a dense ‘bed of nails’. F or exam-ple, Bruggermann et al. fabricated densely packed vertical (60 nm diameter, 300–400-nm-high) gold pillars on 15-μm-diameter pads52. In spite of the nanometric dimensions of the pillars, spontaneously firing HL-1 cells cultured on the MEA generated large ~1 mV neg-ative monophasic FPs. This clearly pointed out that the pillar nano-electrodes maintained an extracellular position. In a similar manner, densely packed (2 μm pitch) micronail-shaped structures record low amplitude (~100 μV) biphasic or negative-monophasic extracellu-lar FPs from cultured rat cardiomyocytes79. In conclusion, although physical contact between a nanoparticle and the plasma membrane may be sufficient to induce particle penetration through the plasma membrane80, higher-density nanocontacts (aiming at lowering the electrode impedance) may prevent membrane penetration from occurring. Thus, optimization of the pillar number, to reduce the impedance and pillar densities to promote internalization is critical. Overcoming the constraints of nanopillar impedence Using advanced semiconductor-based nanotechnology and the clas-sical concepts of mechanically penetrating the cell plasma membrane by sharp glass microelectrodes, Lieber’s laboratory demonstrated intracellular recordings of 80–100 mV APs from beating cultured car-diomyocytes (Fig. 5g). This was done either by the so-called kinked nanowires77 or pillar-shaped protruding silicon nanowires78, nanofab-ricated as the sensing gate electrode of an FET.

In the work described by Tian and Cohen-Karni et al., an FET was generated at the tip of an ~80 nm kinked silicon nanowire by way of in situ doping77. In the work described by Duan et al. a pro-truding silicon nanowire was integrated onto the gate of the FET78 (Fig. 5g). To facilitate the penetration of the electrodes into the cells the device’s surface was modified by phospholipids77. Spontaneous fusion of the applied phospholipids with the lipid membrane of the cells seems to underlie the formation of the G? seal. Using both types of nanosensors, full blown cardiac APs of 75–100 mV, ~200 ms were recorded (Fig. 5h,i). The high-quality measurements of the APs were made possible by three factors: (a) the nanoscale size of the sensors that enabled its insertion into the cytosol through the plasma mem-brane of the cells, (b) the formation of G? resistance between the plasma membrane and the nanostructures, (c) the fact that the size of the sensing area does not affect its sensitivity78,81,82. It should be noted that whereas the aspects of nano-dimensions and G? seal formation are essential components to enable high-quality recordings, the key to the success is the use of the gate electrode of an FET as the sensing electrode rather than passive metal or silicon-based micro- or nano-electrodes. Thus, in contrast with passive conducting lines, where the signal is significantly attenuated due to stray capacitance, an FET effectively amplifies the signal in situ. It should be noted that FETs are more susceptible to failure due to leakage currents, whereas passive electrodes are not affected as dramatically by device imperfections. The approach used by Lieber’s laboratory to insert the nanopillars into the cells77,78,83,84 was to transfer a layer of ‘mature’ cardiomyocytes grown on a thin piece of polydimethylsiloxane, upside down, onto the device surface and apply gentle downward pressure onto the substrate. This manipulation led to the insertion of the nanoelectrodes into the cells within approximately 45 s. Whereas at the ‘proof-of-concept’ level the approach taken by Lieber’s laboratory is sufficient, it should be further developed to enable experiments in which cardiomyocytes and neurons can be cultured and grown in continuous contact with the MEA substrate and maintain the cells–electrodes contact for long periods rather than be acutely manipulated.

When considering the branched intracellular nanotube-F ET or kinked nanoelectrode devices as tools to map functional synaptic connectivity, a major hurdle is the signal-to-noise level of the device. Examination of some recordings (for example, see Fig. 4 in ref. 78) reveals noise levels of more than 20 mV. Whereas the noise is attrib-uted to the nano-dimensions of the FETs and thus can be reduced by adjusting the FET size, the present device does not provide the resolution to enable the recording of miniature potentials, synaptic potentials and small membrane oscillations. Another unsolved prob-lem that would need further study is to enable the accurate recordings of the resting potentials.

Planar patch-clamp MEA technology

Another line of investigation to overcome the problems of high junctional membrane resistance, electrode impedance and of low R seal revolves around microfluidic-based MEAs able to patch clamp neurons under in vitro conditions. The approach is extensively used to acutely patch cells in suspension relying on suction to draw individual cells to the aperture and to form a giga-seal resist-ance. Thus far the approach was not suitable for studies of long-term adhering cultured neuronal networks85,86. Recently Martina et al.87 began demonstrating that the approach has a potential to be adapted to neuronal networks. In this study a silicon oxide sub-strate or silicon oxide laminated by polyimide film containing 2 to 4-μm-sized apertures were used. Each aperture was connected to a microfluidic channel. Two isolated cell bodies of identifiable Lymnaea neurons were co-cultured over the apertures for 8–12 h. Within this time the cell bodies formed chemical synapses and adhered to the substrate to spontaneously form an unusually high G? seal resistance. Negative pressure pulse through the microflu-idic system broke the junctional membrane establishing a classical whole-cell patch-clamp configuration. In a fraction of the experi-ments the patch configuration was stable for a number of hours and the properties of the synapses formed between the two cells could be investigated. The signal-to-noise ratio obtained by the planar patch-clamp device matches that of conventional patch-clamp recording. Although promising, it should be noted that the somata of the isolated neurons adhered to each other but did not extend neurites on the culture substrate. Thus the neuron–device configuration did not simulate the complex growth pattern of cul-tured mammalian neurons but rather is closer to the cell suspen-sion mode of patch-clamp recordings.

When considering the potential use of planar patch-clamp MEAs as tools to map functional synaptic connectivity among cultured neurons, two major problems have to be dealt with: (a) the record-ing duration is expected to be limited by the perfusion of the neuron by the microfluidic solutions (Fig. 4h), (b) the sensor’s density (aper-tures) is expected to be limited by the backside fluidic system. Conclusions

On the basis of the results reviewed above and theoretical considera-tions, we estimate that nano- and micro-electrophysiological tech-nologies enabling simultaneous, long-term, multisite, intracellular recording and stimulation from many neurons under in vitro and in vivo conditions will become available to the neuroscientist com-munity within a number of years.

The approaches that at present reveal the best potential are: the bioinspired use of protruding electrodes that are engulfed by neurons, and the use of nanostructures that penetrate the plasma membrane in a similar way to classical sharp microelec-trodes. Experimentations with passive nanopillar-based protrud-ing structures functionalized by lipid layers revealed that whereas single or multiple nanopillars can penetrate the cell’s plasma membrane forming relatively high R seal, the high nano-electrode impedance/R seal ratio value attenuates the recorded potential to a level that makes it impossible to record synaptic potentials or

REVIEW ARTICLE NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2012.265

subthreshold membrane oscillations. Attempts to reduce the elec-trode impedance by fabricating multiple nanopillars on a single sensing pad did not solve the problem71. It seems that increasing the number and density of the nanopillars to reach low enough impedance (and improve the electrode impedance/R seal ratio) is limited by cell biological properties as cells cultured on high-den-sity nanopillars do not extend the plasma membrane into narrow spaces between the pillars52,79 and that physically inserting a dense population of nanopillars would be damaging to the cell. It is for these reasons that current passive nanopillar electrodes cannot be used as substitutes for traditional glass micropipette electrodes for intracellular recordings from neurons. Another severe limita-tion with the use of nanopillars is the instability of the intracellu-lar configuration. So far, mechanisms to stabilize the intracellular positioning of the nanopillars have not been addressed.

On the other hand, the use of either kinked nanowires77 or pillar-shaped protruding silicon nanowires78, as the gate electrodes of an FET device, bypasses the problem of high electrode impedance (as FET recordings are independent of gate impedance78,81,82). Because of their nanoscale dimensions, the use of FET-based nanosensors can also enable the simultaneous recordings from subcellular compart-ments (dendrite, somata, axon, varicosities and others). The current limitations of the nanopillars–F ET devices for intracellular record-ings are: the inherent noise level of the nano-F ETs, and the need to mechanically manipulate the cultured cells and the substrate on which they grow into physical contact with the electrodes. The prob-lem of noise may be dealt with by modulating the transistor size and geometry88 while taking into consideration the limitations of ambi-ent thermal noise in electrolytic solutions89. The issue of having to mechanically press the cells and device to acutely form physical con-tact may be solved by merging some of the concepts developed by us and the Lieber laboratory.

So far, the gMμE-based MEA is the only device that enabled the recordings of both APs and subthreshold synaptic potentials, and that can also be used for effective intracellular stimulation64. Nevertheless, this coupling was demonstrated using large Aplysia neurons but as of yet has not successfully been applied to rat hippocampal neurons and primary rat cardiomyocytes.

It is conceivable that merging the cell-biological principles of evoking engulfment of the electrodes on the one hand and the use of FETs on the other may provide both a stable neuron–electrode con-figuration and intracellular access. Once achieved, such a device may be applied in arrays that make use of the well-established multiplex-ing capabilities of ultra-large-scale integrated transistor arrays. Received 2 August 2012; accepted 18 December 2012; published online 5 February 2013

References

1. Hodgkin, A. L. & Huxley, A. F. Action potentials recorded from inside a nerve

fibre. Nature144, 710–711 (1939).

2. Grundfest, H. The mechanisms of discharge of the electric organs in relation to

general and comparative electrophysiology. Prog. Biophys. Biophys. Chem.

7, 1–85 (1957).

3. Sakmann, B. & Neher, E. Patch clamp techniques for studying ionic channels in

excitable membranes. Annu. Rev. Physiol.46, 455–472 (1984).

4. Verkhratsky, A., Krishtal, O. A. & Petersen, O. H. From Galvani to patch clamp:

the development of electrophysiology. Pflugers Arch.453, 233–247 (2006).

5. Hutzler, M. et al. High-resolution multitransistor array recording of electrical

field potentials in cultured brain slices. J. Neurophysiol.96, 1638–1645 (2006). 6. Eytan, D. & Marom, S. Dynamics and effective topology underlying

synchronization in networks of cortical neurons. J. Neurosci.

26, 8465–8476 (2006).

7. Berdondini, L. et al. Active pixel sensor array for high spatio-temporal

resolution electrophysiological recordings from single cell to large scale

neuronal networks. Lab Chip9, 2644–2651 (2009).

8. Frey, U., Egert, U., Heer, F., Hafizovic, S. & Hierlemann, A. Microelectronic

system for high-resolution mapping of extracellular electric fields applied to

brain slices. Biosens. Bioelectron.24, 2191–2198 (2009).9. Hochberg, L. R. et al. Neuronal ensemble control of prosthetic devices by a

human with tetraplegia. Nature442, 164–171 (2006).

10. Blanche, T. J., Spacek, M. A., Hetke, J. F. & Swindale, N. V. Polytrodes:

high-density silicon electrode arrays for large-scale multiunit recording.

J. Neurophysiol.93, 2987–3000 (2005).

11. Buzsaki, G., Anastassiou, C. A. & Koch, C. The origin of extracellular fields and

currents — EEG, ECoG, LFP and spikes. Nature Rev. Neurosci.

13, 407–420 (2012).

12. Shoham, S., O’Connor, D. H. & Segev, R. How silent is the brain: is there a “dark

matter” problem in neuroscience? J. Comp. Physiol.192, 777–784 (2006).

13. Mayford, M., Siegelbaum, S. A. & Kandel, E. R. Synapses and memory storage.

Cold Spring Harb. Perspect. Biol.4, a005751 (2012).

14. Thomas, C. A. Jr, Springer, P. A., Loeb, G. E., Berwald-Netter, Y. & Okun, L. M.

A miniature microelectrode array to monitor the bioelectric activity of cultured

cells. Exp. Cell Res.74, 61–66 (1972).

15. Gross, G. W., Williams, A. N. & Lucas, J. H. Recording of spontaneous activity

with photoetched microelectrode surfaces from mouse spinal neurons in

culture. J. Neurosci. Methods5, 13–22 (1982).

16. Regehr, W. G., Pine, J., Cohan, C. S., Mischke, M. D. & Tank, D. W. Sealing

cultured invertebrate neurons to embedded dish electrodes facilitates long-term stimulation and recording. J. Neurosci. Methods30, 91–106 (1989).

17. Connolly, P., Clark, P., Curtis, A. S., Dow, J. A. & Wilkinson, C. D. An

extracellular microelectrode array for monitoring electrogenic cells in culture.

Biosens. Bioelectron.5, 223–234 (1990).

18. Nam, Y. & Wheeler, B. C. In vitro microelectrode array technology and neural

recordings. Crit. Rev. Biomed. Eng.39, 45–61 (2011).

19. Huys, R. et al. Single-cell recording and stimulation with a 16k micro-nail

electrode array integrated on a 0.18 μm CMOS chip. Lab Chip

12, 1274–1280 (2012).

20. Schwartz, A. B. Cortical neural prosthetics. Annu. Rev. Neurosci.

27, 487–507 (2004).

21. Fee, M. S., Mitra, P. P. & Kleinfeld, D. Automatic sorting of multiple unit

neuronal signals in the presence of anisotropic and non-Gaussian variability.

J. Neurosci. Methods69, 175–188 (1996).

22. Brown, E. N., Kass, R. E. & Mitra, P. P. Multiple neural spike train data analysis:

state-of-the-art and future challenges. Nature Neurosci.7, 456–461 (2004). 23. Kauer, J. S., Senseman, D. M. & Cohen, L. B. Odor-elicited activity monitored

simultaneously from 124 regions of the salamander olfactory bulb using a

voltage-sensitive dye. Brain Res.418, 255–261 (1987).

24. Siegel, M. S. & Isacoff, E. Y. A genetically encoded optical probe of membrane

voltage. Neuron19, 735–741 (1997).

25. Shoham, D. et al. Imaging cortical dynamics at high spatial and temporal

resolution with novel blue voltage-sensitive dyes. Neuron24, 791–802 (1999).

26. Kralj, J. M., Douglass, A. D., Hochbaum, D. R., Maclaurin, D. & Cohen, A. E.

Optical recording of action potentials in mammalian neurons using a microbial rhodopsin. Nature Methods9, 90–95 (2012).

27. Loew, L. M., Cohen, L. B., Salzberg, B. M., Obaid, A. L. & Bezanilla, F. Charge-

shift probes of membrane potential. Characterization of aminostyrylpyridinium dyes on the squid giant axon. Biophys. J.47, 71–77 (1985).

28. Stosiek, C., Garaschuk, O., Holthoff, K. & Konnerth, A. In vivo two-photon

calcium imaging of neuronal networks. Proc. Natl Acad. Sci. USA

100, 7319–7324 (2003).

29. Higley, M. J. & Sabatini, B. L. Calcium signaling in dendrites and spines:

practical and functional considerations. Neuron59, 902–913 (2008).

30. Rothschild, G., Nelken, I. & Mizrahi, A. Functional organization and

population dynamics in the mouse primary auditory cortex. Nature Neurosci.

13, 353–360 (2010).

31. Grinvald, A., Lieke, E., Frostig, R. D., Gilbert, C. D. & Wiesel, T. N. Functional

architecture of cortex revealed by optical imaging of intrinsic signals. Nature 324, 361–364 (1986).

32. Nagel, G. et al. Channelrhodopsin-1: a light-gated proton channel in green

algae. Science296, 2395–2398 (2002).

33. Bernstein, J. G. & Boyden, E. S. Optogenetic tools for analyzing the neural

circuits of behavior. Trends Cogn. Sci.15, 592–600 (2011).

34. Homma, R. et al. Wide-field and two-photon imaging of brain activity with

voltage- and calcium-sensitive dyes. Phil. Trans. R. Soc. Lond. B

364, 2453–2467 (2009).

35. Spatz, J. P. & Geiger, B. Molecular engineering of cellular environments: cell

adhesion to nano-digital surfaces. Methods Cell Biol.83, 89–111 (2007).

36. Rutten, W. L. Selective electrical interfaces with the nervous system. Annu. Rev.

Biomed. Eng.4, 407–452 (2002).

37. Fromherz, P. Three levels of neuroelectronic interfacing: silicon chips with ion

channels, nerve cells, and brain tissue. Ann. NY Acad. Sci.

1093, 143–160 (2006).

38. Jones, I. L. et al. The potential of microelectrode arrays and microelectronics

for biomedical research and diagnostics. Anal. Bioanal. Chem.

399, 2313–2329 (2011).

REVIEW ARTICLE

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2012.265

39. Bershadsky, A. D., Balaban, N. Q. & Geiger, B. Adhesion-dependent cell

mechanosensitivity. Annu. Rev. Cell Dev. Biol.19, 677–695 (2003).

40. Wrobel, G. et al. Transmission electron microscopy study of the cell-sensor

interface. J. R. Soc. Interface5, 213–222 (2008).

41. Braun, D. & Fromherz, P. Fluorescence interference-contrast microscopy of cell

adhesion on oxidized silicon. Appl. Phys. A65, 341–348 (1997).

42. Iwanaga, Y., Braun, D. & Fromherz, P. No correlation of focal contacts and close

adhesion by comparing GFP-vinculin and fluorescence interference of Dil. Eur.

Biophysics J. Biophysics Lett.30, 17–26 (2001).

43. Lambacher, A. & Fromherz, P. Luminescence of dye molecules on oxidized

silicon and fluorescence interference contrast microscopy of biomembranes.

J. Opt. Soc. Am. B 19, 1435–1453 (2002).

44. Gleixner, R. & Fromherz, P. The extracellular electrical resistivity in cell

adhesion. Biophys. J.90, 2600–2611 (2006).

45. Fromherz, P. in Neuroelectronic Interfacing: Semiconductor Chips with Ion

Channels, Nerve Cells, and Brain (ed. Waser, P.) Ch. 32, 781–810 (Wiley, 2003).

46. Maccione, A. et al. Experimental investigation on spontaneously active

hippocampal cultures recorded by means of high-density MEAs: analysis of the spatial resolution effects. Front. Neuroeng.3, 1–12(2010).

47. Buitenweg, J. R., Rutten, W. L. & Marani, E. Geometry-based finite-

element modeling of the electrical contact between a cultured neuron and a microelectrode. IEEE Trans. Biomed. Eng.50, 501–509 (2003).

48. Pine, J. Recording action potentials from cultured neurons with extracellular

microcircuit electrodes. J. Neurosci. Methods2, 19–31 (1980).

49. Oka, H., Shimono, K., Ogawa, R., Sugihara, H. & Taketani, M. A new planar

multielectrode array for extracellular recording: application to hippocampal acute slice. J. Neurosci. Methods93, 61–67 (1999).

50. Grumet, A. E., Wyatt, J. L. Jr & Rizzo, J. F. 3rd Multi-electrode stimulation and

recording in the isolated retina. J. Neurosci. Methods101, 31–42 (2000).

51. Kim, J. H., Kang, G., Nam, Y. & Choi, Y. K. Surface-modified microelectrode

array with flake nanostructure for neural recording and stimulation.

Nanotechnology21, 85303 (2010).

52. Bruggemann, D. et al. Nanostructured gold microelectrodes for extracellular

recording from electrogenic cells. Nanotechnology22, 265104 (2011).

53. Shein, M. et al. Engineered neuronal circuits shaped and interfaced with carbon

nanotube microelectrode arrays. Biomed. Microdevices11, 495–501 (2009). 54. Keefer, E. W., Botterman, B. R., Romero, M. I., Rossi, A. F. & Gross, G. W.

Carbon nanotube coating improves neuronal recordings. Nature Nanotech.

3, 434–439 (2008).

55. Akaike, N. & Harata, N. Nystatin perforated patch recording and its

applications to analyses of intracellular mechanisms. Jpn. J. Physiol.

44, 433–473 (1994).

56. Braeken, D. et al. Single-cell stimulation and electroporation using a novel

0.18 μ CMOS chip with subcellular-sized electrodes. Conf. Proc. IEEE Eng. Med.

Biol. Soc.2010, 6473–6476 (2010).

57. Xie, C., Lin, Z., Hanson, L., Cui, Y. & Cui, B. Intracellular recording of action

potentials by nanopillar electroporation. Nature Nanotech.7, 185–190 (2012).

58. Hai, A. & Spira, M. E. On-chip electroporation, membrane repair dynamics

and transient in-cell recordings by arrays of gold mushroom-shaped

microelectrodes. Lab Chip12, 2865–2873 (2012).

59. Fendyur, A. & Spira, M. E. Toward on-chip, in-cell recordings from cultured

cardiomyocytes by arrays of gold mushroom-shaped microelectrodes. Front.

Neuroeng.5, 21(2012).

60. Braeken, D. et al. Open-cell recording of action potentials using active

electrode arrays. Lab Chip12, 4397–4402 (2012).

61. Spira, M. E. et al. Improved neuronal adhesion to the surface of electronic

device by engulfment of protruding micro-nails fabricated on the chip surface.

Transducers ‘07 and Eurosensors XXI 1247–1250 (IEEE, 2007).

62. Hai, A. et al. Spine-shaped gold protrusions improve the adherence and

electrical coupling of neurons with the surface of micro-electronic devices.

J. R. Soc. Interface6, 1153–1165 (2009).

63. Hai, A. et al. Changing gears from chemical adhesion of cells to flat substrata

toward engulfment of micro-protrusions by active mechanisms. J. Neural Eng.

6, 066009 (2009).

64. Hai, A., Shappir, J. & Spira, M. E. Long-term, multisite, parallel, in-cell

recording and stimulation by an array of extracellular microelectrodes.

J. Neurophysiol.104, 559–568 (2010).

65. Hai, A., Shappir, J. & Spira, M. E. In-cell recordings by extracellular

microelectrodes. Nature Methods7, 200–202 (2010).

66. Aderem, A. & Underhill, D. M. Mechanisms of phagocytosis in macrophages.

Annu. Rev. Immunol.17, 593–623 (1999).67. Hayashi, Y. & Majewska, A. K. Dendritic spine geometry: functional

implication and regulation. Neuron46, 529–532 (2005).

68. May, R. C. & Machesky, L. M. Phagocytosis and the actin cytoskeleton. J. Cell

Sci.114, 1061–1077 (2001).

69. Cohen, A., Shappir, J., Yitzchaik, S. & Spira, M. E. Reversible transition of

extracellular field potential recordings to intracellular recordings of action

potentials generated by neurons grown on transistors. Biosensors Bioelectronics 23, 811–819 (2008).

70. Fendyur, A., Mazurski, N., Shappir, J. & Spira, M. E. Formation of essential

ultrastructural interface between cultured hippocampal cells and gold

mushroom-shaped MEA — toward “IN-CELL” recordings from vertebrate

neurons. Front. Neuroeng.4, 14(2011).

71. Robinson, J. T. et al. Vertical nanowire electrode arrays as a scalable platform

for intracellular interfacing to neuronal circuits. Nature Nanotech.

7, 180–184 (2012).

72. Schoen, I. & Fromherz, P. Extracellular stimulation of mammalian neurons

through repetitive activation of Na+ channels by weak capacitive currents on a silicon chip. J. Neurophysiol.100, 346–357 (2008).

73. Hofmann, B., Katelhon, E., Schottdorf, M., Offenhausser, A. & Wolfrum, B.

Nanocavity electrode array for recording from electrogenic cells. Lab Chip

11, 1054–1058 (2011).

74. Almquist, B. D. & Melosh, N. A. Fusion of biomimetic stealth probes into lipid

bilayer cores. Proc. Natl Acad. Sci. USA107, 5815–5820 (2010).

75. Almquist, B. D. & Melosh, N. A. Molecular structure influences the stability of

membrane penetrating biointerfaces. Nano Lett.11, 2066–2070 (2011).

76. Almquist, B. D., Verma, P., Cai, W. & Melosh, N. A. Nanoscale patterning

controls inorganic-membrane interface structure. Nanoscale3, 391–400 (2011).

77. Tian, B. et al. Three-dimensional, flexible nanoscale field-effect transistors as

localized bioprobes. Science329, 830–834 (2010).

78. Duan, X. et al. Intracellular recordings of action potentials by an extracellular

nanoscale field-effect transistor. Nature Nanotech.7, 174–179 (2012).

79. Huys, R. et al. A novel 16k micro-nail CMOS-chip for in-vitro single-cell

recording, stimulation and impedance measurements. Conf. Proc. IEEE Eng.

Med. Biol. Soc.2010, 2726–2729 (2010).

80. Doherty, G. J. & McMahon, H. T. Mechanisms of endocytosis. Annu. Rev.

Biochem.78, 857–902 (2009).

81. Cohen-Karni, T., Timko, B. P., Weiss, L. E. & Lieber, C. M. Flexible electrical

recording from cells using nanowire transistor arrays. Proc. Natl Acad. Sci. USA 106, 7309–7313 (2009).

82. Sadiku, M. N. O. Elements of Electromagnetics (Oxford Univ. Press, 2000).

83. Tian, B. & Lieber, C. M. Design, synthesis, and characterization of novel

nanowire structures for photovoltaics and intracellular probes. Pure Appl.

Chem.83, 2153–2169 (2011).

84. Gao, R. et al. Outside looking in: nanotube transistor intracellular sensors.

Nano Lett.12, 3329–3333 (2012).

85. Ionescu-Zanetti, C. et al. Mammalian electrophysiology on a microfluidic

platform. Proc. Natl Acad. Sci. USA102, 9112–9117 (2005).

86. Lau, A. Y., Hung, P. J., Wu, A. R. & Lee, L. P. Open-access microfluidic patch-

clamp array with raised lateral cell trapping sites. Lab Chip

6, 1510–1515 (2006).

87. Martina, M. et al. Recordings of cultured neurons and synaptic activity using

patch-clamp chips. J. Neural Eng.8, 034002 (2011).

88. Abidi, A. A. High-frequency noise measurements on FET’s with small

dimensions. Electron. Dev., IEEE Trans. on33, 1801–1805 (1986).

89. Voelker, M. & Fromherz, P. Nyquist noise of cell adhesion detected in a neuron-

silicon transistor. Phys. Rev. Lett.96, 228102 (2006).

Acknowledgements

Spira’s laboratory is currently supported by EU FP7 MERIDIAN Grant agreement no. 280778., EU FP7 Marie Curie ITG, Grant agreement no. 264872., and the Charles E. Smith and Prof. Elkes Laboratory for Collaborative Research in Psychobiology. A. Hai was supported by a scholarship from The Israel Council for Higher Education.

Additional information

Reprints and permissions information is available online at https://www.wendangku.net/doc/bb7121320.html,/reprints. Correspondence should be addressed to M.E.S.

Competing financial interests

The authors declare no competing financial interests.

REVIEW ARTICLE NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2012.265