| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Innovative Methodology
1 Center for Molecular and Behavioral Neuroscience, Rutgers, The State University of New Jersey, Newark, New Jersey 07102; 2 Department of Electrical Engineering and Computer Science, The University of Michigan, Ann Arbor, Michigan 48109-2122
Submitted 8 February 2003; accepted in final form 17 April 2003
ABSTRACT
Parallel recording of neuronal activity in the behaving animal is a prerequisite for our understanding of neuronal representation and storage of information. Here we describe the development of micro-machined silicon microelectrode arrays for unit and local field recordings. The two-dimensional probes with 96 or 64 recording sites provided high-density recording of unit and field activity with minimal tissue displacement or damage. The on-chip active circuit eliminated movement and other artifacts and greatly reduced the weight of the headgear. The precise geometry of the recording tips allowed for the estimation of the spatial location of the recorded neurons and for high-resolution estimation of extracellular current source density. Action potentials could be simultaneously recorded from the soma and dendrites of the same neurons. Silicon technology is a promising approach for high-density, high-resolution sampling of neuronal activity in both basic research and prosthetic devices.
INTRODUCTION
In the brain, there are both large- and small-scale levels of organization.
It has been suggested that the cortex is constructed from building blocks
(e.g., columns, barrels, lamellae) of similar intrinsic anatomical connections
and functional properties (Andersen et al.
1969
; Mountcastle
1997; Rice and Van der Loos
1977; Szentágothai
1975). This sub-millimeter level of organization is where
computational rules of neuronal interactions are expected to reside
(Churchland and Sejnowski
1992; Petsche et al.
1984). Several approaches have been introduced for achieving fast,
parallel recording of neuronal elements. Optical imaging of multiple single
neurons is a recent promising development in this field
(Mao et al. 2001). However,
this approach is not practical for monitoring deep cortical or subcortical
activity in behaving animals. An alternative approach is to use multiple wire
electrodes to record from large number of neurons simultaneously. Dozens of
putative single cells have been recorded in rodents and hundreds in a primate
using multiple wire electrodes (Csicsvari
et al. 1999; Hoffman and
McNaughton 2002; Wilson and McNaughton
1993,
1994). Several recent reviews
have documented progress in this field
(Chicurel 2001;
Deadwyler and Hampson 1995;
Eichenbaum and Davis 1988;
Hampson et al. 2001;
Nádasdy et al. 1998;
Nicolelis et al. 1997).
Sharp wire or glass electrodes have the advantage of isolating single cells
reliably (Evarts 1968;
Lang et al. 1999). However,
because of their high impedance, artifacts generated by the movement of the
behaving animal are frequent. Furthermore, the proximity of the electrode tip
to the neuron can mechanically stimulate it and change its firing properties.
Importantly, multiple-unit recordings require multiple movable drives and
wires, considerably increasing tissue damage. Monitoring the same neuronal
pool with closely spaced multiple-site electrodes ("stereotrodes,"
"tetrodes") has greatly facilitated the unit yield/tissue damage
ratio by introducing novel spike sorting ("unit clustering")
methods on the basis of the temporal coherence and waveform variance of spikes
across multisite electrodes (Drake et al.
1988; Gray et al.
1995; McNaughton et al.
1983; Recce and O'Keefe
1989; Wilson and McNaughton
1993). Wire tetrodes have numerous advantages, including easy and
cheap fabrication, low-impedance recording tips, and mechanical stability.
However, the exact configuration of the recording tips relative to the neurons
is not known (Jog et al.
2002). The spacing of multiple tetrodes is limited by the
mechanical drives and the tissue damage they inflict. Importantly, each
tetrode requires a separate drive for proper tip positioning and in situ
buffering with external preamplifiers that require extensive connector wiring.
These latter factors add substantially to the mass to be carried by the animal
and severely limit the number of recording sites in freely moving small
animals.
Micromachined silicon-based electrodes can eliminate or reduce some of the
technical limitations inherent in multiple wire electrodes
(Buzsáki et al. 1981;
Drake et al. 1988;
Kuperstein and Eichenbaum
1985; Petsche et al.
1984). In silicon devices, the thin-film recording sites are
defined lithographically. Any two-dimensional configuration can be achieved
with proper design and can be adapted to particular brain structures and
neuron types. Silicon probes can be combined with on-chip integrated circuitry
(Bai and Wise 2001;
Olsson et al. 2002) to reduce
cross-talk and spurious noise and perform anti-aliasing and impedance
transformations. We have designed and tested two-dimensional probes with 64
and 96 recording sites and examined their field- and unit-recording abilities
in the neocortex and hippocampus of the rat.
METHODS
Electrode configurations
Micro-electro-mechanical-system (MEMS)-based devices are particularly
useful for simultaneous recording local field and unit activity from large
numbers of sites with minimal damage to the nervous tissue. MEMS devices can
combine silicon integrated-circuit processing with thin-film microelectrode
sensing (see APPENDIX)
(Gingerich et al. 2001;
Najafi and Wise 1986;
Wise and Najafi 1991).
In the present experiments, three different types of two-dimensional micromachined probes were examined. The first probe was designed primarily to record extracellular currents in the hippocampus and neocortex of the rat. It contained six shanks (300-µm shank separation) and each shank had 16 iridium recording sites (108 µm2) with 100-µm vertical spacing (Fig. 1A). The interconnects (2 µm wide and 2 µm spacing) were passivated by a dielectric layer formed by stress-balanced stack of silicon dioxide and silicon nitride chemical vapor deposited films. The 12-µm-thick shanks tapered from a maximum width of 85 µm at the base to a sharp tip. The two-dimensional extent of the probe sampled the CA1, CA3 regions and the dorsal part of the dentate gyrus (Fig. 2). The second array consisted of eight shanks (200-µm shank separation) and each shank had eight recording sites (160 µm2) with 50-µm vertical spacing (Fig. 1B1). This probe was designed primarily for high-density recording of unit activity and extracellular current activity in the neocortex and hippocampus. The configuration of the third probe was similar to the second but the recordings sites were staggered to provide a two-dimensional arrangement of the tips (20 µm vertical separation; Fig. 1B2). Both "passive" and "active" configurations were tested. Active probes had on-chip preamplifiers (Fig. 1, C and D) to reduce movement-associated artifacts. The probes were attached to a stereotaxic drive (acute experiments) or a microdrive (chronic experiments), which allowed us to position the probes to the desired target depth during the experiment.
|
|
Surgery and recording
For acute recordings, Sprague-Dawley rats (Hilltop laboratories or Zivic Miller laboratories) 300–500 g were anesthetized with urethan (1.5 g/kg) and placed in a stereotaxic frame. After preparing a 1 x 3-mm window in the skull, the dura was removed using a dissection microscope and the probe was positioned so that the tips avoided blood vessels. The skull cavity was filled with a mixture of wax and paraffin which decreased brain pulsation as well as provided lateral support for the probe shanks. The probe was attached to a micromanipulator and moved gradually to its desired depth position. The middle shanks were centered at anterior-posterior, AP = –3.5 mm and mediolateral, ML = 2.5 mm position. Extracellular units and local field were recorded by the silicon probes from either the somatosensory cortex or the hippocampus. After each acute experiment, the probe was rinsed with deionized water. Probes were used repeatedly for several months without noticeable deterioration in performance. A pair of stainless steel wires (60 µm in diameter) was placed into the fimbria-fornix/hippocampal commissure (AP = –0.8, L = 0.5, V = –4.2 mm) to stimulate the commissural afferents to the hippocampus. Extracellular signals were high-pass filtered (1Hz) and amplified (1,000 times) using a 64-channel amplifier (Sensorium, Charlotte, VT). All data were digitized at 20 kHz (DataMax System, RC Electronics, Santa Barbara, CA) and stored on computer disk for later analysis.
The general surgical procedures for chronic recordings have been described
(Csicsvari et al. 2003). In
short, rats of the Sprague-Dawley strain (400–900 g) were anesthetized
with a mixture (4 ml/kg) of ketamine (25 mg/ml), xylazine (1.3 mg/ml), and
acepromazine (0.25 mg/ml) and placed in the stereotaxic apparatus. The probes
were attached to a custom-prepared microdrive. The probes were implanted the
same way as in the acute experiments. During implantation, the tips of the
probe shanks were lowered 
400 µm below the brain surface. After
recovery from surgery, the probes were moved gradually and recordings were
made at several depth locations. A pair of stainless steel wires (60 µm in
diameter) were placed into the fimbria-fornix/hippocampal commissure (AP =
–0.8, L = 0.5, V = –4.2 mm) to stimulate the commissural afferents
to the hippocampus and another pair in the angular bundle (AP = –7.0, L
= 4.5, V = –3.5 mm) to stimulate the entorhinal input. The evoked
responses by these pathways helped the on-line identification of the recording
electrodes (Bragin et al.
1995). All procedures conformed to the National Institutes of
Health Guide for the Care and Use of Laboratory Animals.
The physiological data were collected during spontaneous exploration on
open field and during sleep episodes in the home cage. With passive probes,
instrumentation amplifiers, built into the female connector
(Szabó et al. 2001),
were used to reduce cable movement artifacts. Two synchronized 64-channel
DataMax systems (16-bit resolution; RC Electronics, Santa Barbara, CA)
continuously recorded electrical activity at 20 kHz to computer hard disk.
SPIKE SORTING. The continuously recorded wide-band signals were
high-pass filtered (0.8–5 kHz) digitally. Units were initially isolated
by an automatic algorithm ("KlustaKwik;" available at:
http://osiris.rutgers.edu/Buzsaki/software)
followed by manual clustering (Csicsvari
et al. 1998). Auto- and cross-correlations were calculated to
verify the clustering procedure.
CURRENT SOURCE DENSITY (CSD) ANALYSIS. One-dimensional CSD maps
were calculated in one direction (depth) as the second spatial derivative of
the local field potentials (Nicholson and
Freeman 1975). This approach assumes that the resistivity of the
extracellular medium is homogeneous and isotropic. Although some resistivity
differences are present in the different hippocampal layers, in practice these
are not large enough to significantly modify the spatial distribution of sinks
and sources (Holsheimer
1987).
Histological procedures
After completion of the experiments, the rats were deeply anesthetized and perfused through the heart first with 0.9% saline solution followed by a 10% buffered formalin phosphate solution. The brains were sectioned by a Vibroslice at 100 µm in the coronal plane. The sections were mounted on slides, Nissl-stained, and coverslipped. The tracks of the silicon probe shanks were reconstructed from multiple sections. DiI labeling around the shanks were viewed with a fluorescence microscope. The exact recording sites in the vertical direction for each session were determined by the evoked responses. The derived electrical patterns then were superimposed on the anatomical substrate that emanated them.
RESULTS
On-line estimation of the recording site positions
The 96-site probe provided a two-dimensional image of electrical activity
(1.6 mm deep, 1.8 mm wide area) across several subregions of the hippocampus
in the behaving rat. Electric stimulation of the entorhinal (perforant path)
and commissural/associational afferents, with known spatial distribution and
targets, provided a tool for a reliable calibration of the recording sites
(Fig. 2) (Ylinen et al. 1995). The
evoked sink and source currents (estimated from the field potentials; see
METHODS) precisely identified the activated layers. In response to
perforant path stimulation, monosynaptic excitatory responses were evoked in
the granule cell dendrites and the distal apical dendrites of CA3 pyramidal
cells (sinks indicated by arrowheads in
Fig. 2, left)
(Andersen et al. 1971; Do et al.
2002). Stronger stimulation evoked population responses of the
granule cells, followed by di- and trisynaptic activation of CA3 and CA1
pyramidal cells, respectively. The sink-associated sources in the cell body
layers reliably identified the CA3–CA1 pyramidal cell layers and the
dentate granule cell layer. Stimulation of commissural fibers evoked
antidromic discharge of CA3 pyramidal cells. Sinks at monosynaptic latency
were present in CA3 and CA1 dendritic layers. Stronger stimulation induced
orthodromic population spikes and, occasionally, a reverberating, entorhinal
cortex-mediated late response (Fig.
2, right; arrowhead). In addition to the evoked
responses, the presence of multiple-unit discharges at several sites helped
identify the pyramidal and granule cell layers. Large-amplitude field ripples
(140–200 Hz) during immobility provided additional information for the
identification of CA1 pyramidal layer
(Buzsáki et al. 1992).
At the termination of the experiments, the brain was perfused with the probe
left in situ. Histological localization of the electrode tracks was used to
verify the prediction of the recording sites. In all six rats, there was a
perfect match between the physiology-predicted and histologically identified
sites of the probe. Once the two-dimensional positions of the recording sites
were determined, spontaneously occurring patterns (e. g., theta, gamma, sharp
wave-ripples) were recorded and their sink-source distributions were related
to the known anatomical afferents
(Csicsvari et al. 2003). The
approach illustrated in the preceding text can be used for on-line
"calibration" in the neocortex as well, using thalamic and
callosal inputs (Castro-Alamancos and
Connors 1996; Kandel and
Buzsáki 1997).
Although the spacing of the recording sites (100 µm) on the 96-site
probes is not ideal for hippocampal unit separation, typically two neighboring
sites on an individual shank recorded from the same units in the CA1–CA3
pyramidal cell layers. Unit clustering with such "stereotrode"
configuration (McNaughton et al.
1983) yielded 
60 clear unit clusters in the hippocampus of the
behaving rat. For clustering neocortical pyramidal cells with long apical
shafts, the 100-µm recording-site separation appeared adequate. In fact,
many single neurons could be recorded from four to six sites
(Buzsáki and Kandel
1998). In some shanks, every single site recorded some unit
activity and as many as 20 unit clusters could be separated by a single shank
in the neocortex. Because the two-dimensional spatial position of the recorded
units can be precisely determined with the 96-site probes, these measurements
will allow for quantitative studies of the intra- and inter-columnar
interactions of cortical neurons.
Field and unit recordings by 64-site probes
Figure 3 illustrates
recordings with the 64-site probe (staggered recording sites;
Fig. 1B2) from deep
layers of the somatosensory cortex of the urethan-anesthetized rat. The
alternation of active and silent periods at 0.5 Hz reflects slow
oscillations characteristic of the anesthetized brain
(Fig. 3A) (Steriade
2001). All 64 recording sites yielded unit activity, and the same units were
recorded by several recording sites of the same shank
(Fig. 3B). The spike
amplitude and waveform of the neuron varied across recording sites. The
voltage gradient of spikes across the recording sites provided an approximate
location of the cell body of the recorded neuron. The spike amplitude from
presumed perisomatic locations often exceeded 0.3 mV.
|
Using probes with linearly arranged recording sites
(Fig. 1B1), the speed
of soma-dendritic back-propagation of the action potentials could be measured
for numerous neurons (Fig. 4A)
(Buzsáki and Kandel
1998). In future experiments, this will allow for the examination
of behavior modulation of back-propagating action potentials
(Buzsáki et al. 1996;
Harris et al. 2001;
Quirk et al. 2001). Nine to 18
neurons per shank could be routinely discriminated. Due to the high spatial
density of the recording sites, the relative vertical position of the recorded
units could be reliably determined (Fig.
4B). This feature will be useful in determining
spatiotemporal activation on neuronal populations under physiological
conditions. The sufficiently large number of recording sites per shank also
yielded a high-resolution CSD of various local field potentials (not shown).
This important feature of the 64-site probe should be emphasized because local
inhomogeneities in field potential generation and the contribution of unit
discharge patterns to the generation of local field can now be rigorously
studied (Pesaran et al.
2002).
|
Probes with on-chip active circuitry had two major advantages in
chronically implanted animals. First, the preamplifier head-stage was
eliminated, and therefore the rat had to carry substantially less bulk.
Second, the interconnect between the high-impedance recording sites (1–3
M
at 1,000 Hz) and the preamplifier input was reduced to <6 mm. A
large part of this interconnect was shielded by the brain. The length of the
interconnect is critical because it serves as an "antenna" and
introduces large slow-frequency artifacts, originating from movement of the
electrically charged whiskers, cable movements, fast head movement-associated
"microphonia," and other environmental noise sources. Although
these artifactual sources pose less of a problem for unit recordings, they can
effectively prevent proper detection and interpretation of field potentials at
low frequencies. Using the active probes, we obtained artifact-free recordings
even during vigorous movements or jumps
(Fig. 5). Because the physical
location of the recorded neurons can be reliably predicted, this device will
be useful to address issues like spatial extent of inhibition and organization
of hippocampal representations (Hampson et
al. 1999; Hirase et al.
2001; Redish et al.
2001).
|
Unit isolation
Silicon probes offer a unique opportunity for massively parallel unit
recordings. Given the relatively large vertical span of the eight-site shanks
(350 or 140 µm), routinely five to seven shanks could be positioned in the
CA1 pyramidal layer. Shanks spanning the pyramidal layer typically recorded
8–15 high-quality units, i.e., the same number or more than from an
average wire tetrode. The quality of the isolated units was quantitatively
determined using the "isolation distance" measure of unit quality
(Fig. 6, inset)
(Harris et al. 2001). This
measure indicated that unit-isolation quality was similar to those obtained by
wire tetrodes (Fig. 6).
|
Full utilization of the power of silicon probes, however, requires advanced
methods of spike detection and cluster isolation. First, when using such a
large-dimensional feature space, automatic cluster separation is a necessity.
Software for this approach is available
(Harris et al. 2000). However,
silicon probes pose new challenges for which spike-sorting methods, developed
for wire tetrodes, may not be optimal. In particular, tetrode spike detection
typically cannot identify spikes of different cells that occur within 1
ms of each other due to the problem of resolving overlapping spikes. With the
silicon probes this problem is more serious because larger numbers of recorded
cells are associated with more overlapping spikes. However, this problem is
potentially solvable because the different cells may be measured on different
sites. Furthermore because each cell is detectable, only on some of the
channels, including channels that do not record from a particular cell add
unnecessary noise. Spike-sorting methods that take advantage of the additional
spatial and temporal information afforded by the silicon arrays are needed to
exploit the full potentials of silicon probes.
Tissue displacement/damage by silicon probes
Histological examination of the electrodes revealed very little effect on
brain tissue. In contrast to wire tetrodes with blunt cut ends, it was
possible to record from the same cell layer numerous times after moving the
probe up and subsequently back to the previous recording locations. Recordings
from isolated single units showed that the same cell could be recorded from by
several successive sites while the probe was moved by 50-µm steps. Focal
bleeding from capillaries around the tracks was very rare. In fact, after
acute recordings, the electrode track could not be recovered in most cases
because in the absence of gliosis no other landmarks identified the tracks. In
subsequent experiments, the shanks were dipped into concentrated DiI solution
and dried prior to implantation. This procedure did not interfere with the
recording qualities of the probes, and the tracks could be positively
identified by the vertical bands of fluorescent processes. In chronic
experiments, the tracks could be recognized by the tissue indent left by the
shanks (Csicsvari et al.
2003). No obvious tissue reaction was visible around the tracks,
in support of a recent detailed report on the biocompatibility of
silicon-based devices (Kristensen et al.
2001).
Although moving the eight-site shanks (62 µm wide at the top recording site) up and down in the tissue several times yielded apparently the same units, in experiments with the 16-site probes, we noticed that when recordings sites 12–16 penetrated through the CA1 pyramidal cell layer, the quality of units deteriorated. Similar deterioration of unit quality was obvious also in cortical layers II and III. The width of the 16-site probe at the 16th recording site is 82 µm (12-µm thick). Thus we believe that the main cause of the reduction of unit quality is due to the tissue damage caused by the widening of the shank.
DISCUSSION
Parallel recording of neuronal activity in the intact brain is a prerequisite for our understanding of neuronal representation and storage of information. Micro-machined silicon microelectrode arrays are particularly promising for achieving this goal. The two-dimensional probes described here provided exceptionally high-density recording of unit and field activity with minimal tissue displacement or damage. The on-chip active circuit eliminated movement and other artifacts and greatly reduced the mass of the headgear. The precise geometry of the recording tips allowed for the estimation of the spatial location of the recorded neurons.
Massively parallel recording of single units with silicon arrays
To date, methods for efficient spatial and temporal resolution of neuronal
activity in the depth of the brain have been limited to recording from
multiple single cells with closely spaced sensors. With these methods, the
number of recording sites and the amount of tissue damage, isolation quality,
and long-term stability compete with each other. Recording from large numbers
of neurons distributed broadly over the cortex and other cortical region
regions is possible with wire electrodes
(Deadwyler and Hampson 1995;
Hoffman and McNaughton 2002;
Kralik et al. 2001;
Nicolelis et al. 1997).
However, recording from a physiologically representative portion of cells from
a small piece of tissue to obtain insight into the operations of local
networks requires a different approach. In addition to the large number of
parallel-recorded units, it is important to acquire information about their
spatial layout and, preferably, their cell type identity. The two-dimensional
silicon probes described here fulfill some of these requirements. First, the
tissue displacement and damage inflicted by a silicon probe is significantly
less than that of a wire tetrode. The diameter of a twisted tetrode is
35–50 µm (i.e., 1,800 µm2 tip surface area), and
the tips of tetrodes often splay apart after chronic implantation
(Jog et al. 2002). The blunt
area of the tetrode tip is twice as large as that of the silicon probe at its
uppermost site (for 8-site shanks) and three to five times larger than the tip
of the silicon probe. The tapered feature of the silicon shank further reduces
damage on depth advancement of the probe. Because all shanks are moved
simultaneously in the brain, the effects of differential drifts at various
recording sites, inherent in multiple tetrode recordings, are reduced, thus
improving recording stability.
The recording sites of the probe can directly line up parallel with the
somata and dendrites. Because the voltage is inversely proportional to
distance from the soma, the voltage gradient across the recording sites
provides an approximate location of the cell body of the recorded neuron
(Henze et al. 2000;
Holt and Koch 1999). The known
geometry of the recording sites in multiple shank recordings provides precise
information about the lateral distances of the recorded units, i.e., whether
they are in the same or different micro-architectonical areas, such as columns
or barrels. In addition to spike occurrence, several intracellular features,
such as the maximum rate of depolarization and repolarization of the action
potential, relative polarization state of the membrane from rest, and the
speed and extent of somadendritic back-propagation of the action potential can
be estimated from the features of the extracellularly recorded spikes
(Buzsáki et al. 1996;
Henze et al. 2000). The
co-linear arrangement of the recording sites allows for the estimation of high
spatial resolution CSD. For unit separation, this is an added advantage
because it eliminates electomyographic noise, which can be substantial in rats
and mice due to the close proximity of the large masseter muscles to the
recording sites, and other high-frequency electronic noise. Importantly, high
spatial resolution of local currents is important not only for investigation
of network rules but can be exploited for practical purposes as well, e.g.,
controlling prosthetic devices (Pesaran et
al. 2002).
Spikes of pyramidal shaped neurons generate an extracellular dipole
parallel with the somadendritic axon of the neuron
(Drake et al. 1988;
Nádasdy et al. 1998).
The extracellular spike can be monitored several hundred micrometers away in
the vertical direction. In the horizontal direction, pyramidal cells can be
detected 
100 µm from the recording site
(Henze et al. 2000;
Jog et al. 2002). In support
of these previous observations, we rarely recorded from the same putative
pyramidal cell with neighboring shanks of the 64-site probe. Therefore
200-µm shank separation appears adequate for maximum sampling of the
pyramidal cell population in the cortex of the rat. In the hippocampal CA1
region a 64-site probe can record, in principle, from all neurons in a 200
x 1,400 µm strip of tissue, provided that all eight shanks are
properly positioned in the pyramidal layer. A remaining fundamental problem is
adequate separation of single cells
(Harris et al. 2000) that can
be improved by increasing spatial sampling further (see following text) and by
developing novel spike-sorting algorithms. The high dimensionality of
multiple-site monitoring of single-unit activity by silicon probes requires
spike sorting methods that are fundamentally different from the currently used
ones for tetrode architectures.
Increasing the number of recording sites inevitably increases the size and
weight of the headstage carried by the small animal. For wire electrodes,
independently movable drives are needed for each electrode for the exact
positioning of the recording tips near the cell bodies. In addition, each
high-impedance wire should be buffered to reduce low-frequency noise. With a
large number of electrodes, this not only increases the weight but also the
length of the interconnects, making the biological signal vulnerable to
various artifacts. The high-density, two-dimensional probes used in this study
require only a single microdrive for positioning. The vertical span of the
eight recording sites is sufficient so that most or all shanks can be placed
in the same layer even if the structure is curved, such as the hippocampal CA1
pyramidal layer. In the neocortex, all sites of all shanks recorded
extracellular units. A further considerable reduction in volume and weight is
due to the on-chip electronics that eliminates the need of preamplifiers. The
low-impedance outputs of the active silicon probes can be directly connected
to high-density connectors. Besides volume and weight reduction, the most
important advantage of on-chip circuitry is the reduction of movement and
environment-derived artifacts due to the substantially reduced interconnect
lengths between the recording sites and preamplifier inputs. Although in the
present experiments the amplifiers provided only unity gain, amplification by
10 or 100 is possible using essentially the same circuitry
(Olsson et al. 2002).
The triangulation method of unit clustering is based on the tacit
assumption that neurons are point sources of the extracellularly recorded
currents. Therefore best separation is expected from three-dimensional
monitoring devices (McNaughton et al.
1983; Recce and O'Keefe
1989). To obtain maximum unit separation, the recording sites
should be evenly separated in all three dimensions. However, this requirement
is not met by either the silicon probes tested here and or by wire tetrodes.
Wire tetrode tips typically form a diamond shape on two-dimensional surface
(Jog et al. 2002). The site
arrangements of the silicon probe form either a one- or two-dimensional array
(staggered recording sites). The staggered sites have both vertical and
horizontal (i.e., 2-dimensional) separation, mimicking wire tetrodes.
Importantly, amplitude distribution of extracellular spikes is strongly biased
by the morphology of neurons with vertically elongated fields for pyramidal
cells (Henze et al. 2000).
Therefore the efficacy of unit separation depends not only the recording site
configuration but also the cytoarchitecture of the recorded neurons and
whether the electrode approaches the cell body from the basal or apical
dendrites. In the hippocampal CA3 region, similar cluster qualities were
obtained by either wire tetrodes or one-dimensional (4-site) silicon probes
(Csicsvari et al. 2000). Each
neuron type may require different tip configurations for ideal unit
isolation.
Outlook
An ideal recording electrode has a very small volume, so that tissue damage
is minimized, but has very large number of monitoring sites so that the
recorded spikes, emanating from different neurons, can be reliably separated.
These competing demands limit the number of useful recording sites on a given
shank. The microfabrication technology used here employed a 2-µm-wide
interconnect lead (on 4-µm centers) per recording site (2 µm features).
Thus >10–12 recording sites require a shank wide enough to produce
noticeable tissue damage and less favorable conditions for unit recording.
However, present industrial production uses 0.18-µm features and multiple
levels of metal. Line widths as small as 22 nm are being targeted for the year
2010 (Geppert 2002). Clearly,
interconnect will not limit the eventual realization of very large numbers of
sites on probes displacing very small volumes of tissue. Indeed, the only
reason for having any silicon under the present shanks (and not just the very
thin dielectric layers encapsulating the interconnects) is the need for
sufficient strength that allows the probes to penetrate brain tissue. Thus
silicon probes are strength limited rather than technology limited in
shrinking overall shank dimensions, and there is no doubt that probes will
evolve to ever smaller feature sizes in coming years as they become more
widely used in neurophysiological research. Recently, we have tested a
32-site, single-shank probe (50 µm vertical site separation), manufactured
by double-sided deep reactive ion etching (DRIE) method with 1-µm
interconnect lines and spacings (Norlin et
al. 2002). Recordings from the neocortex and hippocampus yielded
results similar to those presented in the preceding text. To utilize feature
sizes consistent with present industrial practice, it is important to make the
probes commercially available with fabrication in commercial foundries to help
minimize the high tooling costs associated with manufacturing a limited number
of probes for research.
Further development of on-chip interface circuitry is another important
direction. Circuitry is needed to provide on-chip amplification, filtering,
time-division multiplexing (Olsson et al.
2002), programmed microstimulation through the recording sites,
and, potentially, in vivo, real-time signal processing. On-chip amplification
can eliminate the need for expensive multichannel amplifier systems so that
the probe output can be directly interfaced with a computer. These active
interfaces will not only enhance the performance of silicon probes for basic
research but are a prerequisite for fully implantable neural prosthetic
devices (Donoghue 2002;
Nicolelis 2001;
Rousche et al. 2001).
APPENDIX
Probe fabrication
The fabrication of probes with integrated CMOS circuitry was carried out
with a merged bulk-micromachined 2P/1M CMOS process that has been described in
detail elsewhere (Bai and Wise
2001; Gingerich et al.
2001). The probe body and shanks were defined with a deep boron
diffusion which acted as an etch stop during the final release of probes from
the wafer. A second boron diffusion of shorter duration allowed the definition
of sharp probe tips, resulting in a rounded edge profile for the shanks. A
stack of low-pressure chemical-vapor-deposited (LPCVD) silicon
oxide/nitride/oxide insulated the polysilicon shank interconnects from the
silicon substrate below and the tissue above. Through proper balancing of the
compressive and tensile stresses of the oxide and nitride films, respectively,
the residual stress in the shanks could be cancelled
(Cho et al. 1992). This
allowed the shanks to remain straight even when they are quite long (5 mm in
the current probes). Normal CMOS circuit processing then followed. The bottom
nitride layer in the dielectric stack acted as the masking nitride for field
oxidation in the circuit area. A second polysilicon layer allowed the
formation of poly/poly capacitors, which were used in the frequency
compensation of the probe buffers. The circuit metal was a stacked Ti/Al
layer, and TiN contact plugs were used to prevent junction spiking during
circuit passivation with 1 µm of low-temperature oxide (LTO). A layer
of sputtered or electro-plated gold was then deposited on top of the LTO. This
layer is helpful in reducing the optical sensitivity of the probe circuitry
and was also used as a ground shield against (chemical and electrical)
interference. Following patterning of the gold circuit shield, iridium
recording sites were patterned by lift-off, and gold bonding pads were formed.
Finally, a photoresist mask was applied to the completed probes, and the
dielectrics between the probes were etched with reactive ion etching. The
wafer was then thinned from the backside with an isotropic silicon etchant (a
hydrofluoric acid and nitric acid mixture) and then immersed in a
water/ethylene diamine pyrocatechol (EDP) mixture at 80°C. All areas of
the wafer, not defined by the deep or shallow boron diffusions, were etched
away, releasing the probes from the wafer. Circuit areas were protected from
being attacked by the EDP with large guard rings of deep boron diffusion.
Additional protection was provided at the corners of the circuit area using
corner compensation (Puers and Sansen
1990) and dielectric bridges. When released, the probes were
removed from EDP, rinsed thoroughly in de-ionized water, acetone, and
isopropyl alcohol and then soaked overnight in methanol. The probes were then
wire-bonded to a custom-designed printed circuit board and the bonds were
potted in medical grade silicone to prevent body fluids from short-circuiting
the contacts. Soldering pads along the edges of the printed circuit board
served to connect the board to high-density (0.6 mm) connectors (Omnetics,
Minneapolis, MN).
Noise reduction by on-chip buffering
The two major sources of noise in extracellular recording are externally coupled noise and the thermal (Johnson) noise intrinsic to the electrode itself. Successful extracellular neural recording requires buffering the neural signal at a location close to the recording site to minimize the effect of externally coupled noise. With passive probes and wire electrodes, the "antenna" effect of the interconnect (e.g., wire bonds, printed circuit board traces, connectors and in some cases cabling) between the recording site and buffer input is considerable, providing a means for capacitive and/or inductive noise coupling. On-chip buffering shortens the length of the high-impedance input node and decrease the magnitude of charge coupled into the input. A headstage buffer is itself a source of noise, and it is desirable that this electronic noise be insignificant compared with the thermal electrode noise in the bandwidth of interest. The on-chip buffers in the current design exhibit a baseline noise of 6.4 µV in the 10-kHz bandwidth of interest. While some commercially available operation amplifiers have better noise characteristics, it is important to note that the intrinsic electrode noise (12.7 µV for the iridium sites used in the present experiments) is significantly higher than the electronic noise of the preamplifier. Because independent noise sources add as the sum of squares, electronic noise contributes only 13% to the total baseline noise level.
DISCLOSURES
This work was supported by National Institutes of Health Grants NS-34994, NS-43157, MH-54671, NO1-NS-0-2329, and 1P41RR-09754.
ACKNOWLEDGMENTS
We thank P. Norlin for the gift of 32-site linear probes.
FOOTNOTES
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Present addresses: J. Csicsvari, MRC Anatomical Neuropharmacology Unit, Dept. of Pharmacology, Oxford University, Mansfield Road, Oxford OX1 3TH, UK; D. A. Henze, Merck Research Laboratories, West Point, PA, 19486; B. Jamieson, Detector Systems Branch, NASA/Goddard Space Flight Center. Greenbelt, MD 20771.
* These authors contributed equally to this work. 
G. Buzsáki, Center for Molecular and Behavioral, Neuroscience, Rutgers University, 197 University Ave., Newark, NJ 07102 (E-mail: Buzsaki{at}axon.rutgers.edu).
REFERENCES
Andersen P, Bliss TV, Lomo T, Olsen LI, and Skrede KK. Lamellar organization of hippocampal excitatory pathways. Acta Physiol Scand. 76: 4A–5A, 1969.[Medline]
Bai Q and Wise KD. Single-unit neural recording with active microelectrode arrays. IEEE Trans Biomed Eng 48: 911–920, 2001.[ISI][Medline]
Bragin A, Jandó G, Nádasdy Z, Hetke J, Wise K, and Buzsáki G. Gamma (40–100 Hz) oscillation in the hippocampus of the behaving rat. J Neurosci 15: 47–60, 1995.[Abstract]
Buzsáki G, Grastyán E, Czopf J, Kellényi L, and Prohaska O. Changes in neuronal transmission in the rat hippocampus during behavior. Brain Res 225: 235–247, 1981.[ISI][Medline]
Buzsáki G, Horvath Z, Urioste R, Hetke J, and Wise K.
High-frequency network oscillation in the hippocampus.
Science 256:
1025–1027, 1992.
Buzsáki G and Kandel A. Somadendritic backpropagation of
action potentials in cortical pyramidal cells of the awake rat. J
Neurophysiol 79:
1587–1591, 1998.
Buzsáki G, Penttonen M, Nadasdy Z, and Bragin A. Pattern
and inhibition-dependent invasion of pyramidal cell dendrites by fast spikes
in the hippocampus in vivo. Proc Natl Acad Sci USA
93: 9921–9925,
1996.
Castro-Alamancos MA and Connors BW. Cellular mechanisms of the
augmenting response: short-term plasticity in a thalamocortical pathway.
J Neurosci 16:
7742–7756, 1996.
Chicurel M. Windows on the brain. Nature 412: 266–268, 2001.[Medline]
Cho ST, Najafi K, and Wise K. Internal stress compensation and scaling in ultrasensitive silicon pressure sensors. IEEE Trans Electron Devices 39: 836–842, 1992.
Churchland PS and Sejnowski TJ. The Computational Brain. Cambridge, MA: MIT Press, 1992.
Csicsvari J, Hirase H, Czurkó A, and Buzsáki G. Reliability and state dependence of pyramidal cell-interneuron synapses in the hippocampus: an ensemble approach in the behaving rat. Neuron 21: 179–189, 1998.[ISI][Medline]
Csicsvari J,
Hirase H, Czurkó A, Mamiya A, and Buzsáki G. Oscillatory
coupling of hippocampal pyramidal cells and interneurons in the behaving rat.
J Neurosci 19:
274–287, 1999.
Csicsvari J, Hirase H, Mamiya A, and Buzsáki G. Ensemble patterns of hippocampal CA3–CA1 neurons during sharp wave-associated population events. Neuron 28: 585–594, 2000.[ISI][Medline]
Csicsvari J, Jamieson B, Wise K, and Buzsáki G. Mechanisms of gamma oscillations in the hippocampus in vivo. Neuron 2003.
Deadwyler SA and Hampson RE. Ensemble activity and behavior: what's the code?
Science 270:
1316–1318, 1995.
Do VM, Martinez
CO, Martinez JL Jr, and Derrick BE. Long-term potentiation in direct
perforant path projections to hippocampal CA3 regions in vivo. J
Neurophysiol 87:
669–678, 2002.
Donoghue JP. Connecting cortex to machines: recent advances in brain interfaces. Nat Neurosci 5, Suppl: 1085–1088, 2002.
Drake KL, Wise KD, Farraye J, Anderson DJ, and BeMent SL. Performance of planar multisite microprobes in recording extracellular single-unit intracortical activity. IEEE Trans Biomed Eng 35: 719–732, 1988.[ISI][Medline]
Eichenbaum H and Davis JL. Neuronal Ensembles: Strategies for Recording and Decoding. New York: Wiley-Liss, 1988.
Evarts EV. A technique for recording activity of subcortical neurons in moving animals. Electroencephalogr Clin Neurophysiol 24: 83–86, 1968.[ISI][Medline]
Geppert L. The amazing vanishing transistor act. IEEE Spectrum 39: 28–33, 2002.
Gingerich MD, Hetke JF, Anderson DJ, and Wise KD. A 256-Site 3D CMOS microelectrode array for multipoint stimulation and recording in the central nervous system. International Conference on Solid-State Sensors and Actuators (Transducers'01), Munich, Germany, 2001.
Gray CM, Maldonado PE, Wilson M, and McNaughton B. Tetrodes markedly improve the reliability and yield of multiple single-unit isolation from multi-unit recordings in cat striate cortex. J Neurosci Methods 63: 43–54, 1995.[ISI][Medline]
Hampson RE, Simeral JD, and Deadwyler SA. Distribution of spatial and nonspatial information in dorsal hippocampus. Nature 402: 610–614, 1999.[Medline]
Hampson RE, Simeral JD, and Deadwyler SA. What ensemble recordings reveal about functional hippocampal cell encoding. Prog Brain Res 130: 345–357, 2001.[Medline]
Harris KD,
Henze DA, Csicsvari J, Hirase H, and Buzsáki G.. Accuracy of
tetrode spike separation as determined by simultaneous intracellular and
extracellular measurements. J Neurophysiol
84: 401–414,
2000.
Harris KD, Hirase H, Leinekugel X, Henze DA, and Buzsáki G. Temporal interaction between single spikes and complex spike bursts in hippocampal pyramidal cells. Neuron 32: 141–149, 2001.[ISI][Medline]
Harris KD, Henze DA, Hirase H, Leinekugel X, Dragoi G, Czurko A, and Buzsáki G. Spike train dynamics predicts theta-related phase precession in hippocampal pyramidal cells. Nature 417: 738–741, 2002.[Medline]
Henze DA, Borhegyi Z, Csicsvari J, Mamiya A, Harris KD, and Buzsáki G. Intracellular features predicted by extracellular recordings in the hippocampus in vivo. J Neurophysiol 390–400, 2000.
Hirase H,
Leinekugel X, Csicsvari J, Czurkó A, and Buzsáki G.
Behavior-dependent states of the hippocampal network affect functional
clustering of neurons. J Neurosci
21: RC145,
2001.
Hoffman KL and
McNaughton BL. Coordinated reactivation of distributed memory traces in
primate neocortex. Science 297:
2070–2073, 2002.
Holsheimer J. Electrical conductivity of the hippocampal CA1 layers and application to current-source-density analysis. Exp Brain Res 67: 402–410, 1987.[ISI][Medline]
Holt GR and Koch C. Electrical interactions via the extracellular potential near cell bodies. J Comput Neurosci 6: 169–184, 1999.[ISI][Medline]
Jog MS, Connolly CI, Kubota Y, Iyengar DR, Garrido L, Harlan R, and Graybiel AM. Tetrode technology: advances in implantable hardware, neuroimaging, and data analysis techniques. J Neurosci Methods 117: 141–152, 2002.[ISI][Medline]
Kandel A and
Buzsáki G. Cellular-synaptic generation of sleep spindles,
spike-and-wave discharges, and evoked thalamocortical responses in the
neocortex of the rat. J Neurosci
17: 6783–6797,
1997.
Kralik JD, Dimitrov DF, Krupa DJ, Katz DB, Cohen D, and Nicolelis MA. Techniques for long-term multisite neuronal ensemble recordings in behaving animals. Methods 25: 121–150, 2001.[ISI][Medline]
Kristensen BW, Noraberg J, Thiebaud P, Koudelka-Hep M, and Zimmer J. Biocompatibility of silicon-based arrays of electrodes coupled to organotypic hippocampal brain slice cultures. Brain Res 896: 1–17, 2001.[ISI][Medline]
Kuperstein M and Eichenbaum H. Unit activity, evoked potentials and slow waves in the rat hippocampus and olfactory bulb recorded with a 24-channel microelectrode. Neuroscience 15: 703–712, 1985.[ISI][Medline]
Lang EJ,
Sugihara I, Welsh JP, and Llinas R. Patterns of spontaneous purkinje cell
complex spike activity in the awake rat. J Neurosci
19: 2728–2739,
1999.
Mao BQ, Hamzei-Sichani F, Aronov D, Froemke RC, and Yuste R. Dynamics of spontaneous activity in neocortical slices. Neuron 32: 883–898, 2001.[ISI][Medline]
McNaughton BL, O'Keefe J, and Barnes CA. The stereotrode: a new technique for simultaneous isolation of several single units in the central nervous system from multiple unit records. J Neurosci Methods 8: 391–397, 1983.[ISI][Medline]
Mountcastle VB. The columnar organization of the neocortex.
Brain 120:
701–722, 1997.
Nádasdy Z, Csicsvari J, Penttonen M, Hetke J, Wise K, and Buzsáki G. Extracellular recording and analysis of neuronal activity: from single cells to ensembles. In: Large Scale Recording of Neuronal Activity. edited by Eichenbaum H and Davis JL. New York: Wiley, 1998, p. 17–55.
Najafi K, Ji J, and Wise KD. Scaling limitations of silicon multichannel recording probes. IEEE Trans Biomed Eng 37: 1–11, 1990.[ISI][Medline]
Najafi K and Wise KD. An implantable multielectrode array with on-chip signal processing. IEEE J Solid-State Circuits 21: 1035–1045, 1986.
Nicholson C and
Freeman JA. Theory of current source-density analysis and determination of
conductivity tensor for anuran cerebellum. J
Neurophysiol 38:
356–368, 1975.
Nicolelis MA, Ghazanfar AA, Faggin BM, Votaw S, and Oliveira LM. Reconstructing the engram: simultaneous, multisite, many single neuron recordings. Neuron 18: 529–537, 1997.[ISI][Medline]
Nicolelis MA. Actions from thoughts. Nature 409, Suppl: 403–407, 2001.[Medline]
Norlin P, Kindlundh M, Mouroux A, Yoshida K, and Hofmann UG. A 32-site neuronal probe fabricated by DRIE of SOI substrates. J Micromech Microeng 12: 414–419, 2002.
Olsson RH III, Gulari MN, and Wise KD. Silicon neural recording arrays with on-chip signal processing for in-vivo data acquisition. Proceedings of the 2nd Annual IEEE-EMBS Conference on Microtechnology Medicine and Biology, Capri, Italy, p. 237–240, 2002.
Pesaran B, Pezaris JS, Sahani M, Mitra PP, and Andersen RA. Temporal structure in neuronal activity during working memory in macaque parietal cortex. Nat Neurosci 5: 805–811, 2002.[ISI][Medline]
Petsche H, Pockberger H, and Rappelsberger P. On the search for the sources of the electroencephalogram. Neuroscience 11: 1–27, 1984.[ISI][Medline]
Puers B and Sansen W. Compensation structures for convex corner micro-machining in silicon. Sensors Actuators A A 23: 1036–1041, 1990.
Quirk MC, Blum
KI, and Wilson MA. Experience-dependent changes in extracellular spike
amplitude may reflect regulation of dendritic action potential
back-propagation in rat hippocampal pyramidal cells. J
Neurosci 21:
240–248, 2001.
Recce M and O'Keefe J. The tetrode: a new technique for multiunit extracellular recording. Soc Neurosci Abstr 15: 1270, 1989.
Redish AD,
Battaglia FP, Chawla MK, Ekstrom AD, Gerrard JL, Lipa P, Rosenzweig ES,
Worley PF, Guzowski JF, McNaughton BL, and Barnes CA. Independence of
firing correlates of anatomically proximate hippocampal pyramidal cells.
J Neurosci 21:
RC134, 2001.
Rice FL and Van der Loos H. Development of the barrels and barrel field in the somatosensory cortex of the mouse. J Comp Neurol 171: 545–560, 1977.[ISI][Medline]
Rousche PJ, Pellinen DS, Pivin DP Jr, Williams JC, Vetter RJ, and Kipke DR. Flexible polyimide-based intracortical electrode arrays with bioactive capability. IEEE Trans Biomed Eng 48: 361–371, 2001.[ISI][Medline]
Steriade M. Corticothalamic resonance, states of vigilance and mentation. Neuroscience 101: 243–276, 2000.[ISI][Medline]
Szabó I, Czurkó A, Csicsvari J, Hirase H, Leinekugel X, and Buzsáki G. The application of printed circuit board technology for fabrication of multichannel micro-drives. J Neurosci Methods 105: 105–110, 2001.[ISI][Medline]
Szentágothai J. The "module-concept" in cerebral cortex architecture. Brain Res 95: 475–496, 1975.[ISI][Medline]
Wilson MA and
McNaughton BL. Dynamics of the hippocampal ensemble code for space.
Science 261:
1055–1058, 1993.
Wilson MA and
McNaughton BL. Reactivation of hippocampal ensemble memories during sleep.
Science 265:
676–679, 1994.
Wise KD and
Najafi K. Microfabrication techniques for integrated sensors and
microsystems. Science 254:
1335–1342, 1991.
Ylinen A, Bragin A, Nádasdy Z, Jandó G, Szabó I, Sík A, and Buzsáki G. Sharp wave-associated high-frequency oscillation (200 Hz) in the intact hippocampus: network and intracellular mechanisms. J Neurosci 15: 30–46, 1995.[Abstract]
This article has been cited by other articles:
|
|
 |
|
  S. Schrader, S. Grun, M. Diesmann, and G. L. Gerstein Detecting Synfire Chain Activity Using Massively Parallel Spike Train Recording J Neurophysiol, October 1, 2008; 100(4): 2165 - 2176. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. M. Montgomery, A. Sirota, and G. Buzsaki Theta and Gamma Coordination of Hippocampal Networks during Waking and Rapid Eye Movement Sleep J. Neurosci., June 25, 2008; 28(26): 6731 - 6741. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
V. Itskov, E. Pastalkova, K. Mizuseki, G. Buzsaki, and K. D. Harris Theta-Mediated Dynamics of Spatial Information in Hippocampus J. Neurosci., June 4, 2008; 28(23): 5959 - 5964. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. M. Montgomery and G. Buzsaki Gamma oscillations dynamically couple hippocampal CA3 and CA1 regions during memory task performance PNAS, September 4, 2007; 104(36): 14495 - 14500. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. Geisler, D. Robbe, M. Zugaro, A. Sirota, and G. Buzsaki Hippocampal place cell assemblies are speed-controlled oscillators PNAS, May 8, 2007; 104(19): 8149 - 8154. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. Gold, D. A. Henze, C. Koch, and G. Buzsaki On the Origin of the Extracellular Action Potential Waveform: A Modeling Study J Neurophysiol, May 1, 2006; 95(5): 3113 - 3128. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. Fujii and A. M. Graybiel Time-varying covariance of neural activities recorded in striatum and frontal cortex as monkeys perform sequential-saccade tasks PNAS, June 21, 2005; 102(25): 9032 - 9037. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. J. Blanche, M. A. Spacek, J. F. Hetke, and N. V. Swindale Polytrodes: High-Density Silicon Electrode Arrays for Large-Scale Multiunit Recording J Neurophysiol, May 1, 2005; 93(5): 2987 - 3000. [Abstract] [Full Text] |
