Close
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Multielectrode arrays as neuron -silicon interfaces
(USC Thesis Other)
Multielectrode arrays as neuron -silicon interfaces
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
MULTIELECTRODE ARRAYS AS NEURON-SILICON INTERFACES
by
Walid Victor Soussou
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(NEUROSCIENCE)
December 2005
Copyright 2005 Walid Victor Soussou
UMI Number: 3219796
3219796
2006
Copyright 2005 by
Soussou, Walid Victor
UMI Microform
Copyright
All rights reserved. This microform edition is protected against
unauthorized copying under Title 17, United States Code.
ProQuest Information and Learning Company
300 North Zeeb Road
P.O. Box 1346
Ann Arbor, MI 48106-1346
All rights reserved.
by ProQuest Information and Learning Company.
ii
DEDICATION
To my friends and family who supported me lovingly while I completed this work.
iii
ACKNOWLEDGEMENTS
I have been very fortunate in my time at USC to have met and interacted with
many outstanding students, scientists, and faculty. When I first joined the
Neuroscience Graduate Program, John Fitzpatrick and Christian Perron were
excellent mentors who taught me both electrophysiology and lab operations. Drs.
John Walsh, Lou Byerly, and Garnik Akopian were most kind and helpful resources
as well, from my first laboratory rotation and throughout the years in the
neurophysiology journal club. Drs. Michel Baudry and Mark Thompson I thank for
inspirational leadership and treating me like a peer. Dr. Roberta Diaz Brinton is
certainly the kindest and most supportive person I have been fortunate to interact
with. I sincerely thank her for her guidance on my project, long discussions on a
vast range of topics, and for her genuine caring for my career and well-being. I was
also fortunate to collaborate with Dr. Ghassan Gholmieh whose drive and motivation
I greatly admire. I could not have wished for better colleagues than the ones I had in
Dr Berger’s lab: Zhuo Wang, Dong Song, Angelika Dimoka, Sageev George,
Ludwig Hamo, Jean-Marie Bouteiller, Hassan Heidarin, Ali Dibazar, and Dr.
Xiaping Xie, as well as Dr. Shuhua Chen, Michael Kim, and Rose Yamazaki from
Dr. Brinton’s lab. I also thank all my work-study students and other graduate
students for their helping hands. Last but not least, I wish to thank Dr. Theodore
Berger for taking a chance on me, and giving me my freedom to explore on my own,
while guiding me to remain focused in my research and challenging me to think
further and beyond.
iv
TABLE OF CONTENTS
DEDICATION............................................................................................................ii
ACKNOWLEDGEMENTS......................................................................................iii
TABLE OF CONTENTS..........................................................................................iv
LIST OF FIGURES ..................................................................................................vi
ABBREVIATIONS .................................................................................................viii
ABSTRACT...............................................................................................................ix
1. INTRODUCTION..............................................................................................1
1.1. The Neuron-Silicon Interface....................................................................1
1.2. Neuron-Silicon Interface Applications.....................................................3
1.2.1. Biosensors ............................................................................................3
1.2.2. Neural Prostheses.................................................................................6
1.2.3. Artificial Neuronal Networks.............................................................10
1.3. Multielectrode Arrays as the Interface ..................................................14
1.3.1. Implementations of Planar Arrays .....................................................14
1.3.2. High Density Conformal pMEAs.......................................................19
1.3.3. Biocompatibility Issues......................................................................21
1.4. Research Goals .........................................................................................24
2. TISSUE-BASED BIOSENSORS ....................................................................25
2.1. Hippocampal slices on pMEAs ...............................................................25
2.2. Materials and Methods............................................................................28
2.2.1. Acute Slice Preparation......................................................................28
2.2.2. Electrophysiology ..............................................................................29
2.2.2.1. Recording System..............................................................................29
2.2.2.2. Experimental Protocols.....................................................................29
2.2.2.3. Data Analysis ....................................................................................30
2.3. Speed and Reliability ...............................................................................30
2.3.1. Multiple Independent Recording Sites...............................................31
2.3.2. Multiple Spatially Distributed Stimulations.......................................33
2.3.3. System Stability .................................................................................35
2.4. Specificity, Sensitivity, and Breadth.......................................................36
2.4.1. Spatial Mapping of Electrical Activity ..............................................37
2.4.2. Spatio-Temporal Map of Electrical Activity......................................47
2.5. Durability: Cultured Tissue ....................................................................51
2.6. Future opportunities and obstacles ........................................................55
2.6.1. Portability and Ruggedness................................................................56
v
2.6.2. Confounding Factors and False Positives ..........................................56
2.6.3. Classification and Simplicity .............................................................58
2.6.4. Cost and Reusability ..........................................................................59
2.6.5. Conclusion .........................................................................................59
3. HIGH DENSITY CUSTOM MEAS...............................................................61
3.1. Trade-off Between Coverage Area and Density....................................61
3.2. Material and Methods .............................................................................64
3.2.1. Conformal pMEA design and fabrication..........................................64
3.2.2. Electrophysiology ..............................................................................65
3.2.3. Current-Source Density Analysis.......................................................65
3.3. Results .......................................................................................................66
3.3.1. Fabrication .........................................................................................66
3.3.2. CSD Mapping of Electrical Activity..................................................67
3.3.2.1. FP vs CSD.........................................................................................67
3.3.2.2. Higher spatial density yields better mapping of sinks and sources ..71
3.3.2.3. Multiple laminar profiles ..................................................................72
3.3.2.4. Multi-points stimulation....................................................................75
3.3.3. Specific Stimulation Site Selection....................................................78
3.4. Discussion..................................................................................................81
3.5. Conclusion.................................................................................................84
4. SUBSTRATE COATINGS CONTROL NEURON MORPHOLOGY AND
PHYSIOLOGY ON MEAS .....................................................................................86
4.1. Introduction..............................................................................................86
4.2. Materials and Methods............................................................................89
4.2.1. Substrate Preparation .........................................................................89
4.2.2. Cell Culture ........................................................................................89
4.2.3. Image Analysis...................................................................................90
4.2.4. Immunohistochemistry.......................................................................92
4.2.5. Electrophysiological Analysis............................................................93
4.3. Results .......................................................................................................95
4.4. Discussion................................................................................................109
4.5. Conclusion...............................................................................................118
5. CONCLUSION...............................................................................................119
ALPHABETIZED BIBLIOGRAPHY .................................................................137
APPENDICE A. Publications resulting from this work....................................156
vi
LIST OF FIGURES
Figure 1.1: Electrode positions matched to a hippocampal slice drawn with
representative cell bodies, axons, and dendrites traced from Nissl
stains (Ishizuka et al., 1995) ..............................................................20
Figure 1.2: Neuroprosthetic MEAs..........................................................................21
Figure 2.1: Hippocampal slice circuitry in relation to pMEAs................................27
Figure 2.2: Hippocampal slice paired-pulse (pp) responses recorded with a
pMEA.................................................................................................32
Figure 2.3: Effects of TEA on acute rat hippocampal slices. ..................................34
Figure 2.4: Long-Term Potentiation (LTP) of hippocampal slice on pMEA. .........36
Figure 2.5: Extent of LTP is different than extent of fEPSP response. ...................39
Figure 2.6: Timecourses of NMDA-induced LTP in hippocampal slice
stimulated at four locations on pMEA. ..............................................41
Figure 2.7: Topographic color maps showing NMDA-induced abolition of
responses at four stimulated sites in a hippocampal slice on pMEA. 43
Figure 2.8: NMDA induced potentiation and depression at different sub-regions
of hippocampal slice. .........................................................................44
Figure 2.9: NMDA-induced potentiation in DG in response to stimulation at
lateral PP. ...........................................................................................46
Figure 2.10: NMDA-induced depression in DG in response to stimulation at
medial PP. ..........................................................................................46
Figure 2.11: Three point stimulation of hippocampal slice on pMEA. ...................48
Figure 2.12: Effects of 1mM TMPP application on hippocampal slice...................49
Figure 2.13: Effects of TMPP on dissociated embryonic hippocampal cells
cultured on PDL coated pMEAs. .......................................................53
Figure 2.14: Effects of NMDA on dissociated hippocampal cells. .........................54
Figure 3.1: Three custom MEA (cMEA) designs with hippocampal slices
positioned on top................................................................................67
vii
Figure 3.2: Field Potential and Current Source Density in CA1 pyramidal cell
activity................................................................................................69
Figure 3.3: Effect of inter-electrode spacing on FP recording and CSD analysis. ..72
Figure 3.4: CSD laminar profile of CA1..................................................................75
Figure 3.5: Stimulation site independence along CA1’s pyramidal axo-dendritic
axis. ....................................................................................................77
Figure 3.6: Paired-pulse facilitation and depression in medial and lateral PP.........80
Figure 4.1: Chemical structure of polycationic substrates.......................................87
Figure 4.2: Phase micrographs of neurons cultured on MEAs coated with the
eight different substrate conditions. ...................................................96
Figure 4.3: Morphological feature scores for the various substrates and their
grouping. ............................................................................................98
Figure 4.4: Image analysis of surface coverage of cells and branches. .................100
Figure 4.5: Immunohistological staining of axons and dendrites and
measurement of their surface areas..................................................103
Figure 4.6: Analysis of electrophysiological activity parameters from two
sample PO+BM and PEI cultures. ...................................................107
Figure 4.7: Distribution of electrophysiological parameter values for PO+BM
and PEI.............................................................................................108
Figure 4.8: Correlation between morphological features and electrophysiological
parameters. .......................................................................................109
Figure 5.1: Photomicrographs of a culture patterned on a pMEA after 11 days in
vitro (DIV). ......................................................................................121
Figure 5.2: Neuronal network circuit design through adhesive biomolecules.......125
viii
ABBREVIATIONS
aCSF: Artificial Cerebro-Spinal Fluid
BCI: Brain-Computer Interface
BSA: Bovine Serum Albumin
CBW: Chemical and Biological Weapons
CSD: Current Source Density
DIV: Days In Vitro
DG: Dentate Gyrus
EPSP: Excitatory Post-Synaptic Potential
fEPSP: field Excitatory Post-Synaptic Potential
FP: Field Potential
FPGA/VLSI: Field Programmable Gate Array/Very Large Scale Integration
HFS: High-Frequency Stimulation
HSS: Hank’s Salt Solution
ISI: InterSpike Interval
LTP: Long-Term Potentiation
MEA: MultiElectrode Array
MEAs: MultiElectrode Arrays
cMEA: custom MultiElectrode Array
pMEA: planar MultiElectrode Array
mf: mossy fibers
NBM: Neurobasal Medium
NMDA: N-Methyl-D-Aspartate
pp: paired pulse
PP: Perforant Path
SchC: Schaffer Collaterals
STP: Short Term Plasticity
TEA: TriEthylAmine
TMPP: TriMethylolPropane Phosphate
ix
ABSTRACT
In order to probe the depth of the brain, explore the computational
capabilities of neuronal networks, and investigate the effects of drugs on such
networks, a tool is needed to extract this physiological information from neural
tissue. This tool, which will interface between neurons and silicon-based computers,
will have diverse applications including neuroprosthetics, and artificial neuronal
networks, and biosensors. In the evolution of such neuron-silicon interfaces,
multielectrode arrays (MEAs) have emerged as the most promising candidate.
This dissertation describes methodologies to harness the advantages of planar
MEAs (pMEAs) as neuron-silicon interfaces. pMEA technology offers several
advantages over traditional in vitro single electrode recording in the areas of high-
throughput testing, spatial mapping of electrical activity, temporal information
processing, spatio-temporal activity monitoring, long-term physiological
investigations, and neuroprosthetic-driven network dynamics elucidation. Chapter 2
presents experimental and analytical protocols for using pMEAs with hippocampal
slices as tissue-based biosensors for neurotoxin screening and pharmacological
investigations. The methods take advantage of the slice’s intrinsic circuitry and the
ability of pMEAs to stimulate and record at many locations in a slice simultaneously
to increase sensor detection speed, reliability, stability, specificity, sensitivity,
breadth and durability. Chapter 3 introduces high-density custom-designed
conformal pMEAs (cMEAs) as a solution to current limitations on electrode number.
These new pMEAs can selectively stimulate afferent fibers and record responses in
x
target regions. In addition, they enable Current Source Density (CSD) analysis
which allows physiological interpretation of recorded field potentials.
Towards the development of biocompatible surfaces for implantable
electrode arrays and the creation of designer neuronal networks, the impact of select
biochemical substrates is investigated in Chapter 4. It also explores the relation
between physiology and morphology in these neuronal networks and suggests that
electrophysiological activity of neurons is related to their network morphological
structure, and not solely intrinsically determined. This relation provides guiding
principles to consider in developing coating materials for implantable electrodes and
for creating designer neuronal networks on MEAs. In conclusion, the work
presented methods that augment the value of pMEAs as neuron-silicon interfaces for
neuroprosthetic, biosensor and neuronal network applications.
1
1. INTRODUCTION
1.1. The Neuron-Silicon Interface
This century will see the development of novel neural prosthetics and
neurobiological computers. These devices, which could significantly enhance brain
capacity or reparability, and provide new computing frontiers, will be based on an in-
depth understanding of the computational capabilities of the brain. The current
theoretical models that attempt to decipher the brain’s information processing and
learning rely on principles of neural networks. There is an abundance of computer
simulated neural networks based on various connectionist models and learning
algorithms; these execute or emulate many brain and non-brain tasks. The
experiments needed to demonstrate the validity of these models in biological systems
have been limited by yesterday’s technology. In order to probe the depth of the brain
and explore the computational capabilities of neuronal networks, and to experiment
with the effects of drugs on such networks, a tool is needed to extract this
physiological information from neural tissue. This tool, which will interface
between neurons and silicon-based computers, will have diverse applications
including neuroprosthetics, and artificial neuronal networks, and biosensors. In the
evolution of such neuron-silicon interface, multielectrode arrays (MEAs) have
emerged as the most promising candidate. MEAs are groups of electrodes that are
connected through hardware to a single computer. They can therefore record many
neurons’ intrinsic activity or stimulate them with electrical pulses, thereby enabling
two-way communication between neuronal cells and computers.
2
The brain encodes information in the spatio-temporal activity of neurons.
MEAs are used to study these patterns of activity in behaving animals or tissue
slices, or dissociated cells. The arrays enable extraction of network information that
would not be possible with single electrodes. The main advantage of MEAs is their
ability to record simultaneous activity at different points in the neuronal network,
thus creating a spatio-temporal map. These maps enable the location of a target
response, the selective stimulation of fibers, and tracking the propagation of activity.
The spatio-temporal preservation of information also allows advanced analyses, such
as correlations in neural firing to study connectivity, as well as current source density
(CSD) analysis to accurately localize the activity of the cells. Additionally, MEAs
can be used for electrical stimulation, which in combination with drug application,
can reveal the specific areas of a tissue which the drugs are effecting. The richness
and complexity of such network signal are useful for pharmacological investigations
or biosensor applications. Another developing application involves the creation of
designer neuronal networks to study the computational properties of neuronal
circuits. This information is useful for decoding the brain and developing
neuroprosthetic devices that would replace functionality lost in illness or injury.
For all of these applications, MEAs must be biocompatible and conducive to
neuronal attachment and growth. One of the methods of controlling the interaction
of neurons with the abiotic MEA materials involves treating their surface with
biologically active peptides. These peptides can limit tissue inflammatory responses
or specify where specific cell types or their processes attach. Selective peptide
3
application can be used to pattern the growth of regenerating neurons. Micro-scale
patterning methods are used with dissociated cell cultures for creating designer
neuronal networks or reproducible biosensors.
This work presents experimental and analytical methodologies for utilizing
MEAs as neuron-silicon interfaces in the context of their future real-world
applications in biosensors, neuroprosthetics, or neuronal network analysis.
1.2. Neuron-Silicon Interface Applications
1.2.1. Biosensors
Chemical and Biological Weapons (CBW) is considered the most serious
threat facing the United States and western society, in the twenty-first century. The
material and technical accessibility of these weapons to terrorist organizations along
with their portability and devastating effects make them a real danger that has been
demonstrated in several terrorist attacks, including the infamous Tokyo subway
attack by Japanese doomsday cult Aum Shinrikyo. (Falkenrath et al., 1998; Tucker,
2000)
There are a several types of sensors designed to detect biologically active
agents, whether for military defense or environmental monitoring. These hardware
sensors identify agents based on their physical and chemical properties, and include
mass spectrometry detection systems (military’s MM1 (Rostker, 1997)), ion mobility
spectrometry and semiconductor technology (M22 ACADA, Graseby Ionics, UK,
M90-D1-A, Environics, Milliken, Finland).
4
Though these sensors can accurately detect several nerve agents (e.g. Tabun,
Sarin and Soman, which are all acetycholinesterase inhibitors), blood agents
(cyanide), blister agents (Nitrogen Mustard), or biological agents (Anthrax), they are
limited, in that they can only detect known, and characterized agents. Terrorist
groups on the other hand, could produce modified chemicals or hybrid viruses that
would pass undetected by these types of sensors. There is therefore a need to
develop sensors that can accurately predict the physiological effects of novel
uncharacterized agents on a biological subject.
Biosensors consist of a hybrid between a biological sensing element that
detects changes in its environment and an interface with transducers that convert the
physiological signal to a quantifiable read-out. The biological sensing elements
provide for highly sensitive detection that is physiologically relevant. The choice of
biological moiety has ranged a wide gamete with different advantages and
disadvantages (Guilbault and Schmid, 1991; Roe, 1992; Byfield and Abuknesha,
1994; Higson and Vadgama, 1994; Paddle, 1996; Pancrazio et al., 1999). Antibodies
offer sensitive and precise detection of known antigens as the expense of being too
specific, irreversible and sometimes difficult to transduce into detectable signal
(Morgan et al., 1996). Receptors have similar sensitivity to antibodies with the
added advantage of providing some insight into potential physiological effects of the
agent, and the disadvantage of being difficult to collect in large quantities, stabilize
and transduce. Enzymes are sensitive, fast and can be made to amplify signal to
detectable levels, but are difficult to produce and maintain active (Kharitonov et al.,
5
2000). Lipid membranes can be attached to solid surfaces to provide voltametric
readouts of specific agonists (Tien et al., 1991). Organelles can be used as
biosensors to provide information about toxicity to specific cellular functions (da
Silva et al., 1998). Various cell types can be cultured and provide physiological
responses that can readily classify agents’ activities, but are delicate and sensitive to
handle (Gross et al., 1995; Pancrazio et al., 1999). Tissue preparations can be used
acutely or in culture to obtain rich information on the effects of a drug on a systemic
level and at physiologically meaningful concentrations, but also require care in their
production and maintainance (Wijesuriya and Rechnitz, 1993; Huotari, 2000;
Gholmieh et al., 2001). Finally, there is always the canary in the cage approach
where the vitals of an animal can be monitored for in vivo drug testing, and these
results are the most meaningful as they provide realistic toxicity results that include
entry, absorption, transport, etc.; however, they may sometimes not be sensitive or
rapid enough for certain agents. Of course, each of these biological sensors requires
appropriate hardware for transduction of their signal into meaningful data. These
interface devices therefore also span a wide range from optical sensors (Myszka,
1999) of chemilumenescence, fluorescence, or absorbance, to electrochemical
(Narayanaswamy, 1991; Boutelle et al., 1996; Sirkar et al., 2000) or piezo-electric
probes, to mechanical or acoustic actuators.
Beyond CBW defense, biosensor can be useful in pharmacological
investigations. There the effect of medicinal drugs on biological tissue can be
rapidly explored and quantified with similar biosensors. Chapter 2 describes the
6
advantages of hippocampal slices on MEAs as biosensors, and describes
experimental protocols applicable for pharmacological research.
1.2.2. Neural Prostheses
Brain diseases such as epilepsy, Parkinson’s, or Alzheimer’s, and damage
sustained through stroke or concussion are getting more treatment options in modern
medicine, ranging from preventative drugs to invasive surgeries to implantable
devices or stem cells (Mashour et al., 2005). The challenge of recovery stems from
the lack of regeneration of neurons and their processes (Gentleman, 1994). Though
some neurons have recently been shown to divide in adult animals, this is not in
response to neurological damage, but rather a continuous regenerative process
(Eriksson et al., 1998). And whereas receptive fields retain a certain degree of
competitive remapping after loss of their afferents, and there is evidence for
regaining function after brain lesions or surgeries, this is not due to axonal
regeneration, but rather to the continuous growth of undamaged axons. The glial
cells in the central nervous system inhibit nerve regeneration (Gentleman, 1994).
Recently, however, it has been shown that this inhibition can be blocked by nerve
growth factors such as NGF, BDNF, NT3 or NT4/5 (Tuszynski and Gage, 1995).
Nonetheless, the guiding cues available during development are often lost in the
adult brain, and hence regenerating axons may not find their proper targets. Whereas
some treatments such as stem cell transplantation or deep brain stimulation can
relieve symptoms, such as Bradykinesia or seizures, by substituting for deficient
neurotransmitter release or direct stimulation, lost cognitive functionality is hard to
7
restore in adults (Mashour et al., 2005). Brain prosthetics might allow replacement
of such complex brain functionality not only by substituting for the processing of
damaged neurons, but also by interacting appropriately with their afferents and
efferents (Berger et al., 2001). In order to replace the networks of neurons that
transduce the multitude of inputs to the brain into behavioral outputs, the neuronal
code must be elucidated.
The brain is a hierarchy of distributed neural networks that encode
information in the correlated activity of neurons (Deadwyler and Hampson, 1995;
Eichenbaum and Davis, 1998). The input information of this network is spread over
multiple processing neuronal elements. These neurons process multiple types of
information, and produce an ensemble output that is encoded in their pattern of
activation (Chapin, 1999). Therefore, examining one neuron’s activity to understand
how the brain encodes information is analogous to observing a single pixel and
trying to read a message on a screen. The more pixels are visible, the clearer the
image. In the same analogy, recording the activity of several neurons sequentially,
and trying to reconstitute their code is comparable to sampling several pixels one at a
time and trying to puzzle them back to form an image. While this could work for
still images, brain activity is more like a movie: it has spatial and temporal elements.
Thus in order to piece together a movie from individual pixels, the latter have to be
sampled at the same time. This pixelation analogy is similar to deciphering the
brain’s code using multi-neuronal recordings.
8
Population coding rests on the assumption that the firing patterns of the
constituent neurons are independent from one another (Hatsopoulos et al., 1999).
According to this assumption, the probability of a population firing in response to an
event is a product of the individual probabilities of each of its neurons. This premise
allows neurons recorded serially by single unit recordings to be combined into a
population. Whereas a basic premise behind ensemble coding is that the statistical
interaction between the neuronal activities provides additional information
(Hampson and Deadwyler, 1999). Thus ensemble coding mandates the use of
simultaneous multielectrode recording techniques, in order to capture the temporal
relations between neurons (Hatsopoulos et al., 1999). Neuroprosthetic applications
can draw different information from neuronal populations and ensembles.
Chapin et al. implanted multielectrodes into the motor cortices of rats trained
to control a lever that moves a robotic arm to dispense a water reward (Chapin et al.,
1999). The neuronal activity recorded during the task was analyzed by principle
component analysis and fed into an artificial neural network. After training, the
network could control the robotic arm in real-time based on its interpretation of the
rat motor cortex’s ensemble activity. After a few trials, the rat learned to control the
robotic arm without using the lever. This demonstrated that ensembles code enough
information that can be extracted by multivariate analysis, and used to control of
simple devices (Chapin, 1999). In an effort to predict the volition of rats during
specific tasks, Deadwyler and Hampson recorded simultaneously from the several
locations in hippocampus, using a custom designed MEA (Deadwyler and Hampson,
9
1997). The array consisted of 2 parallel rows of 8 electrodes that span 1.6 mm of
hippocampus longitudinally, and terminate in CA1 and CA3’s pyramidal layers. The
hippocampus is involved in spatial and working memory, and as such its neurons
encode task dependent information not easily amenable to population decoding.
Nonetheless, Deadwyler and Hampson were able to predict rats’ decisions at various
task phases based on the spatio-temporal firing relations recorded between neurons.
Similar algorithms are being used to decode ensemble activity in monkeys to
control robot arms in real-time and 3-dimentional space (Wessberg and Nicolelis,
2004). Human experiments are a step behind, as electrodes implanted in cortices of
paralyzed patients translated single unit firing rates to control a cursor on a computer
screen (Kennedy and Bakay, 1998). While these early human neuroprosthetics
allowed locked-in patients to communicate by typing on the screen, firing rate
analysis does not provide the necessary degrees of freedom to control prosthetic
limbs. Nonetheless, these experiments suggest the ability to use ensemble coding to
control machines in future Brain-Computer Interfaces (BCI) (Wolpaw et al., 2000).
In order to develop neural prosthetics that communicate back to the brain, an
understanding of how the ensembles will respond to electrical stimulation is also
required. Whereas this stimulus response has been well characterized and
successfully implemented for cochlear prostheses, where the cochlea is stimulated to
restore hearing (Clark and Hallworth, 1976; Black et al., 1983; McDermott, 1989;
Clark, 1998), it is still being investigated for other applications such as retinal
10
prostheses (Chichilnisky and Kalmar, 2003), and spinal chord stimulation (Waltz,
1997).
These multielectrode array experiments revealed that ensemble activity
encodes behavior with greater accuracy than multiple single-unit recordings. The
development of the interface of brain prosthetics is therefore directed towards MEA
configurations which can record ensemble activity as well as stimulate it, thereby
enabling efficient two-way communication (Clark and Hallworth, 1976; Lambert et
al., 1991; Deadwyler and Hampson, 1995; Branner and Normann, 2000).
1.2.3. Artificial Neuronal Networks
In order to study ensemble properties, many of the above papers utilized
simulated data sets to control some statistical variables. The reason for this, is that in
vivo recordings have several complicating factors: First, most brain areas have too
many neurons to model, and while one may record from a hundred, there are
thousands of neurons that are not detected. Second, connections between neurons
are not experimentally traceable. Third, there are often afferents not correlated to the
examined behavior that can complicate the signal. Fourth, behavior is not always
easy to quantify. These complexities with in vivo data led to another approach for
the study of the network activity.
In culture, dissociated embryonic neurons grow and synapse onto each other.
In the absence of anatomical or patterning guidance cues, the neurons form
“random” connections to each other. This connectivity can be observed by
microscopy in low density cultures, and complex networks are seen to form. These
11
are closed systems, as they do not receive afferent inputs, yet after several days in
cultures, the neurons become spontaneously active. Without external stimulation,
they demonstrate different activity patterns. Unfortunately, the analysis has often
been restricted to classification of activity states such as no activity, random spiking,
spiking and slow bursting, patterned bursting, periodic bursting, or intense tonic
spiking (Gross, 1994). Electrical stimulation has been used to produce controlled
changes, and allow further analysis of ensemble responses. Still, thus far, only
changes in firing patterns or increases in synchrony have been reported (Jimbo et al.,
1998; Jimbo et al., 1999). Alternatively, the effects of various drugs on culture
activity have been investigated, but their analysis has also been limited in to
descriptions of transitions between activity states (Gross et al., 1995; Canepari et al.,
1997). Ensemble activity in cultured networks has been difficult to investigate or
interpret due to the lack of behavioral correlates. Correlation of activity can lead to
confirmation of connectivity between neurons, and to elucidation of their driving
direction. The immediate response of networks can be analyzed using peri-event
cross-correlation analysis. Artificial neural networks can be used to analyze
temporal patterns on a longer time scale than other multivariate methods. And even
longer term changes in the network’s state can be assessed by state vector functions
such as the one used by Skaggs and McNaughton to describe the status of networks
during waking and sleeping behaviors (Skaggs and McNaughton, 1999).
The first question that wants to be asked in this simplified system is: Can
these simple neuronal networks learn and process information like in vivo
12
ensembles? If so, can this system be used to learn about neuronal network
processing? The answer requires first the ability to address the cultures, and then an
understanding of their language. Cultured neurons have been shown to respond to
electrical stimulation paradigms, which can thus be used as input events. The ease of
patching on dissociated neurons has led several laboratories to investigate synaptic
plasticity in culture (Tong et al., 1996; Vogt et al., 1997). By patching several
neurons simultaneously, Mu-ming Poo’s group has explored the propagation of
depression and potentiation along simple neural networks (Fitzsimonds et al., 1997;
Tao et al., 2000). They report that the plasticity observed depends on the correlation
between pre- and post-synaptic spikes, where potentiation occurs when a presynaptic
neuron fires before the postsynaptic one, while depression is observed in the reverse
case (Bi and Poo, 1998). In order to examine the effect of this plasticity on
networks, they record from neurons distal to the stimulation site, and attribute the
observed changes in firing to polysynaptic plasticity and delay lines (Bi and Poo,
1999). The patch recording approach however restricts the number of recordable
neurons in a network.
The tool to overcome this challenge of multi-cell recording in culture is the
planar MEA (pMEA). pMEAs are arrays of electrodes manufactured by depositing
thin metal lines on a glass substrate, covering them with an insulating material, and
exposing a the tip of the metal line. Dissociated neurons can be cultured on pMEAs,
and their activity recorded over prolonged periods, since the electrodes do not
penetrate the cells. Extracellular recording of neuronal networks has revealed some
13
of their spontaneous firing properties. Networks often tend to burst in synchrony
(Maeda et al., 1995; Kamioka et al., 1996), but can transition to other firing patterns,
such as tonic firing or sporadic spiking (Gross, 1994; Canepari et al., 1997). The
influence of culture medium components on network firing patterns has been
harnessed for the design of biologically based biosensors (Gross et al., 1995).
Networks are also responsive to stimulation from pMEA electrodes, and tetanic
stimulation can strengthen their synchronized activity (Jimbo et al., 1998).
Recording from up to 72 neurons at a time, Jimbo et al. (1999) observed that some
neurons in networks are potentiated while others are depressed in response to a same
tetanus (Jimbo et al., 1999). Cross-correlation analysis revealed that the observed
plasticity followed a similar rule to the one described by Poo (Bi and Poo, 1998),
namely that a neuron’s activity was enhanced if it had been correlated to spikes on
the tetanizing pathway prior to the tetanus. Correlations outside a 40 ms time
window were depressed. The connections between the 72 recorded neurons are
however not easily mapped, and this restricts the analysis of each neuron’s
contribution to the network’s activity.
A greater degree of control over network interconnections is necessary in
order to examine how neuronal networks process information (Stenger et al., 1998;
Maher et al., 1999b). The goal is to specifically connect neurons one to each other
and create designer neuronal networks that are amenable to electrophysiological
analysis. This would serve as a tool to compare the results of theoretical artificial
neural network computations with those of biological ensembles. These biological
14
neuronal networks often require uni-directional connections, where the axon of one
neuron extends to a second neuron but not vice-versa. The tools that would serve
this patterning purpose might also be used to guide axons on implants. One of the
influential factors controlling the behavior of neurons on electrode surfaces is
coating substrate. Understanding the effects of such substrates on the physiology of
the neurons will be essential for controlling the patterns of neuronal growth and
development at the neuron-silicon interface.
1.3. Multielectrode Arrays as the Interface
The next section introduces the utility and advantages of pMEAs as the
neuron-silicon interface in the three electrophysiological applications discussed
above. Section 1.3.2 then introduces a new generation of conformally designed
higher-density pMEAs as enhancement to MEA-based research.
1.3.1. Implementations of Planar Arrays
Over the last two decades, technological advances in the fields of microchip
and electronics manufacturing have enabled an increase in the production and use of
silicon-based MEAs (Singer, 2000). MEAs have come in a variety of shapes and
materials, but fall into two broad classes: thin and sharp (implantable) or dish-based
(planar). While many investigations are currently undertaking research in vivo with
implantable versions, this dissertation focuses on applications of pMEAs, which are
very well suited for in vitro experiments with slice or dissociated cells preparations.
Depending on the physiological preparation, there are two categories of
electrophysiological data that can be collected with current MEA technology: single
15
neuron action potentials (spike events in cell culture), and field potentials
(population synaptic and action potential activity in brain slices).
Currently, the research being undertaken on pMEAs ranges from studying
processes of neuronal plasticity underlying learning and memory, to tracking activity
development in networks, and also pharmacological drug screening and testing.
These diverse applications can be classified, based on the intricacy of their
methodology, into the following non-mutually exclusive categories: (1) MEAs can
be used as a multitude of single independent electrodes for rapid high-throughput
experiments; (2) The spatial relations between the electrode tips can be used
synergistically to map electrical activity to tissue location; (3) Recording
simultaneously from multiple electrodes allows correlation of temporal information,
which is not possible with many recordings from single electrodes; (4) The
combination of spatial and temporal monitoring reveals the spatio-temporal
dynamics of the neuronal network; (5) The ability to maintain cultured preparations
on pMEAs allows long-term physiological investigations; and (6) Recording and
stimulating through the pMEA creates two-way communication with the tissue that
is indispensable for investigating and developing neuroprosthetic applications.
High-throughput applications involve sampling several electrodes out of the
total number on an MEA and selecting a representative one, or treating subgroups
statistically as multiple samples from a homogeneous population. The electrodes
within a particular cytoarchitectural region of a slice usually record similar neural
responses. This redundancy of the observed signals can be used to enhance the
16
statistical significance of the results by grouping the responses into larger sample
sizes. Similar time savings are achieved in cell cultures, where the multitude of
electrodes records the activity of numerous cells at the same time, thereby decreasing
the number of individual experiments needed to reach a significant population
sample. Such high-throughput use of MEAs as biosensors has been applied to drug
screening using cell culture (Pine, 1980; Gross et al., 1995), and hippocampal slice
rhythmic activity (Shimono et al., 2000). In the first case, drugs are classified
according to changes in the firing activity of neuronal cells cultured on MEA (Gross
et al., 1995; Gross et al., 1999). In the second case, changes in the frequency of
carbachol-induced theta rhythmic oscillations in hippocampal slices are correlated
with specific drug properties (Shimono et al., 2000). In both cases, MEAs provided
multiple sample points in different regions of the network, which enabled either a
quick selection of an optimal site or averaging several channels for greater statistical
accuracy.
In contrast to using array electrodes as individual and independent streams of
data, the spatial arrangement of the electrodes can be used to generate spatial maps
of the activity in a slice. Any parameter of the recorded potentials can be plotted in a
color-coded matrix according to the relative spatial positions of the electrodes in
order to generate topographic activity maps. Such spatial activity maps can be
matched to a picture of the slice showing the actual electrode positions in order to
visualize activity in relation to sub-regions of a slice (Shimono et al., 2000) or map
the spatial extent of a response along a network (Jimbo and Robinson, 2000). In
17
addition, if the electrodes are close enough to each other, they enable CSD analysis,
which can elucidate the origins and meaning of complex field potentials (Wheeler
and Novak, 1986).
The ability to simultaneously record from all the electrodes over time enables
correlation of activity between different parts of a network to study its patterns and
plasticity in cell and tissue preparations. The temporal sequence of firing of
ensembles of cells can provide information on network states. Beggs et al. analyzed
cell bursting avalanches to describe the stability of networks (Beggs and Plenz,
2003). Jimbo et al. reported time-dependent synaptic plasticity in networks of
cultured cells by observing that connections between cells that fired within 40 ms
before the other were potentiated after tetanus, whereas connections between cells
negatively correlated within 40 ms were depressed (Jimbo et al., 1999).
pMEAs combine spatial and temporal information and enable the conversion
of static spatial activity maps into dynamic spatio-temporal map sequences. These
series of maps can be joined as frames of a movie to visually trace propagation of
spontaneous, evoked, or rhythmic activity across tissue slices. For example, Novak
and Wheeler studied the temporal propagation of seizure activity (Novak and
Wheeler, 1989), and Shimono et al. localized and spatio-temporally followed the
origin of theta rhythms generated by carbachol in hippocampal slices (Shimono et
al., 2000), both using CSD analysis of signals recorded from pMEAs.
The surface of pMEAs is ideal for long-term tissue cultures of both slices and
dissociated cells, as it provides a flat, biocompatible and sterilizable support with
18
embedded electrodes that can continuously monitor culture activity without
disrupting the closed system. Long-term experiments can track changes in activity
and plasticity of developing cultures and networks (Gross and Schwalm, 1994;
Stoppini et al., 1997; Thiebaud et al., 1997; Jahnsen et al., 1999) under different
chronic pharmacological treatments (Shimono et al., 2002).
While several neuroprosthetics, such as cochlear, cortical (Chapin et al.,
1999), and retinal (Humayun et al., 2003) devices rely on implantable in vivo MEA
technology, pMEAs still play a major role in understanding network connectivity
and dynamics (Meister et al., 1994; Warland et al., 1997). pMEAs are being used as
in vitro testing platforms to first characterize information processing in target
neuronal networks, before undertaking in vivo experiments. Berger et al.’s
hippocampal prosthesis goal is to replace hippocampal CA3 with a Field
Programmable Gate Array/Very Large Scale Integration microchip (FPGA/VLSI)
implementation of a non-linear model of the area (Berger et al., 2001). They are
using pMEAs to provide a functional proof-of-principle. Acute slice experiments on
pMEAs allow generation non-linear models and testing hardware implementations
parameters and conditions, rapidly, cost effectively and with fewer animals.
Similarly, the retinal prosthesis project relies on understanding underlying retinal
dynamics and plasticity in order to transform incident light into electrical stimulation
patterns that will produce correct visual percepts (Humayun et al., 2003). Retinal
stimulation and recording experiments are thus currently being undertaken on
pMEAs to develop a non-linear mathematical model of the retinal network that will
19
be implemented in the next generation of retinal prostheses (Chichilnisky and
Kalmar, 2003; Frechette et al., 2004).
1.3.2. High Density Conformal pMEAs
The above-mentioned applications of pMEAs illustrate their versatility and
advantages of having arrays multiple electrodes instead of single electrodes. These
advantages could be augmented by increasing the number of electrodes, i.e., by
increasing the number of sampling points and spatial resolution. Unfortunately,
physical as well as technological constraints limit the possible number of electrodes
on an MEA. The first obstacle is overcrowding of electrode leads, which can
produce unwanted noise and cross-talk between closely spaced lines, and requires
more complex manufacturing solutions, such as lead stacking or electrode
addressing. The second impediment involves connectors, which are difficult to keep
small and reliable. The third barrier is size and cost of signal modulation,
digitization, and storage hardware. That is of course without mentioning the
increased complexity and time consumption of data analysis. This limit on electrode
number creates a trade-off in array design: electrodes can either sparsely cover a
large surface area or be closely spaced (high-density) over a smaller one. Figure 1.1
illustrates the trade-off due to the constraint on electrode number with two
arrangements of electrodes overlaid on a hippocampal slice drawing.
Figure 1.1: Electrode positions matched to a hippocampal slice drawn with representative cell
bodies, axons, and dendrites traced from Nissl stains (Ishizuka et al., 1995). A) A square 8x8 array of
electrodes is overlaid on the slice. B) Two high-density 3x10 rectangular arrays whose angles and
separation conform to the slice cytoarchitecture are overlaid on CA2 and CA1 pyramidal cells.
(Electrode size and spacing not to scale)
The low-density square matrix 8x8 array in Figure 1.1A can record field
potentials from several different areas of slice while providing only two and rarely
three electrodes in a particular layer along the orientation of a cell group. The two
more densely packed 3x10 sub-arrays (Figure 1.1B) cover only two pyramidal cell
populations (one in CA1 and another in CA2), but match the correct orientation of
the cells’ dendritic axis, which allows 2D-CSD analysis and multiple points of
bipolar stimulation.
This trade-off has inspired the creation of several new high-density
conformal pMEAs that are custom-designed for specific experimental purposes.
These pMEAs have several high-density clusters of electrodes whose orientation and
location conform to the cytoarchitecture of rat hippocampal slices. The hippocampal
slice is an ideal tissue preparation for pMEAs, as its intrinsic 2-dimensional
trisynaptic excitatory cascade network from dentate gyrus (DG) through CA3, to
CA1 sub-regions (DG Æ CA3 Æ CA1) is preserved when hippocampus is sliced
20
along its longitudinal axis (Andersen et al., 1971). Chapter 3 describes the design of
new conformal high-density pMEAs emphasizing their advantages and applications.
1.3.3. Biocompatibility Issues
The neuron-silicon interface has many challenges inherent to the integration
of biological tissue with abiotic devices. These biocompatibility issues stem from
interaction of the biological systems immune response and abrasive saline
environment, and the devices’ material and physical properties. Requirements of the
interface vary depending on the specific brain area and function desired from a
device.
Figure 1.2: Neuroprosthetic MEAs. Aa) Cochlear stimulation array and Ab) Cochlear implant; B)
Retinal stimulation array (Humayun et al., 2003); C) Deep Brain Stimulation probes (Seletra probe,
Medtronics, Minneapolis, MN, and CenSeT electrode, U. of Kentucky, Lexington, KY); D) Cortical
stimulation and recording array (Rousche and Normann, 1998); E) Slanted nerve stimulation and
recording array (Branner and Normann, 2000).
Tissue cytoarchitecture determines the electrodes’ spatial layout (Figure 1.2).
Cochlear prosthetics spiral to fit in the cochlea (Clark, 1998), whereas retinal
stimulation MEAs must be flat and curved (Humayun et al., 2003), cortical implants
21
22
must be dense and penetrate to reach the cell layers of interest (Rousche and
Normann, 1998), nerve stimulators need to reach individual fascicles at various
depths (Branner and Normann, 2000), and deep brain stimulation requires long and
hard shafts to reach their target sites (Motta and Judy, 2005). Some implants aim to
record, others to stimulate and some do both. In order to record meaningful
electrical signals from MEAs, good coupling between tissue and electrodes is
required. Low electrode impedance allows field potential recording, whereas high
impedance electrodes are needed to record extracellular spikes. Material
electrochemical properties affect electrode impedance and charge capacity, as well as
lifespan of the device and toxicity on the tissue. Gold, Titanium Nitride, and
Platinum black are often used at the electrode tips to lower impedance and increase
charge capacity (Gross et al., 1982; Egert et al., 1998; Oka et al., 1999). Electrode
size also affects recording and stimulation, whereby smaller electrodes have higher
impedance and lower charge capacity, but can sample smaller areas with less
interference. Silicon micromanufacturing methods enable the construction of small
silicon-based MEAs with electrode densities that could not be achieved using regular
wire electrodes. The small size of the electrodes enables them to be embedded
permanently into cortical tissue (Rousche and Normann, 1999).
For the successful implementation of such artificial parts into the brain,
neurons must attach to the MEA surface, and extend appropriate branches to the
inputs and outputs of the MEA. This specific interconnection must occur after
invasive and damaging implantation procedures. Even with sterile procedures, the
23
brain can reject inorganic materials. The brain’s initial response to implants involves
astrocyte activation, secretion of cytokines that further kill neurons, and the
formation of a glial sheath around the foreign body. This reactive gliosis distances
the neurons from the device’s surface, therefore reducing its ability to interact with
them. This can create unstable recording or sometimes completely abolish them
(Rousche and Normann, 1998). Various surface modifications are being investigated
to reduce glial scarring and increase neuronal interaction with implanted devices.
For example, Twain et al. have coated probe surfaces with dexamethasone, an anti-
inflammatory compound, before implantation, and showed reduced gliosis for up to
six weeks (Shain et al., 2003). Other biological molecules have been selectively
patterned on surfaces to spatially confine cellular attachment in vitro (Kleinfeld et
al., 1988; Stenger et al., 1998; James et al., 2000). Another challenge in brain
prosthetic design is targeting of axons toward recording surfaces, dendrites toward
stimulating electrodes. Coating MEA surfaces with growth factors that enhance cell
attachment and promote axonal elongation can provide a solution to this obstacle
(Ignatius et al., 1998). Kennedy and Bakay pretreated their conical electrodes with a
proprietary mixture of growth factors, which promoted axonal extension into the
cones, and enabled high quality and stable long-term recording from the motor
cortices of paraplegic subjects (Kennedy and Bakay, 1998). The preliminary success
of this approach commands a methodological study of substrates and surface features
that can be used to coat or shape electrodes, and their effects on neuronal attachment
and physiology. Chapter 4 examines such effect of substrates on neuronal network
24
morphology and physiology, and correlates parameters of the two. Its reported
findings will influence the selection of biological modifications for the surface
treatment of neuroprosthetic devices and the application of patterning methods for
the creation of designer circuits for network analysis.
1.4. Research Goals
This dissertation describes methodologies to harness the advantages of
pMEAs as a neuron-silicon interface. Chapter 2 presents experimental and analytical
protocols for using pMEAs with hippocampal slices as tissue-based biosensors for
neurotoxin screening and pharmacological investigations. The methods take
advantage of the slice’s intrinsic circuitry and the ability of pMEAs to stimulate and
record at many locations in a slice simultaneously to increase sensor detection speed,
reliability, stability, specificity, sensitivity, breadth and durability. Chapter 3
introduces high-density custom designed pMEAs (cMEAs) as a solution to current
limitations on electrode number. These new pMEAs can selectively stimulate
afferent fibers and record responses in target regions. In addition, they enable CSD
analysis which allows physiological interpretation of recorded field potentials.
Chapter 4 investigates the effect of coating pMEAs with biological substrates on
dissociated cell cultures. It also explores the relation between physiology and
morphology in these neuronal networks. In conclusion, the work presented here
describes methods that extend the value of pMEAs as neuron silicon interfaces for
neuroprosthetic, biosensor and neuronal network applications.
25
2. TISSUE-BASED BIOSENSORS
2.1. Hippocampal slices on pMEAs
One way to characterize an unknown agent’s effects on the nervous system is
to use live brain tissue as part of the detection system. By pre-screening the tissue’s
responses to known classes of drugs, new agents can be rapidly classified based on
their physiological effect. This is the premise for a biological tissue-based sensor, or
biosensor. Several functional characteristics have been suggested for biosensors:
Selectivity, sensitivity, detection limit, reversibility, response time, size, ruggedness,
reliability, cost, and signal recovery (Roe, 1992). Neurons are very sensitive to their
environments. The changes they detect affect their physiology and can often be
rapidly analyzed by electrophysiological means. Several neuron-based biosensors
have been developed for drug detection and analysis (Gross et al., 1995; Gholmieh et
al., 2001) (Pancrazio et al., 2003). These biosensors use MEAs to record
extracellular potentials from neurons or slices. The large number of electrodes
provides redundancy necessary for error control, and maximizes the likelihood that
viable neurons are located on top of electrodes. The biosensor implementation
described in this chapter consists primarily of acute hippocampal slices on MEAs,
with some examples utilizing dissociated neurons.
Hippocampus is an associational cortical area involved in learning and
memory of the declarative type as well as spatial mapping and face recognition
(Brown and Zador, 1990). Hippocampus receives several inputs originating in
entorhinal cortex, septum, diagonal band, contralateral hippocampus, brainstem’s
26
raphe nuclei and locus coeruleus, as well as hypothalamic and thalamic projections
(Brown and Zador, 1990). These along with a multitude of intrinsic inhibitory cell
types, such as basket and hilar cells, modulate activity through several
neurotransmitters, including acetylcholine, norepinephrine, serotonin, GABA, and
several neuropeptides (Brown and Zador, 1990). In addition, estrogenic receptors on
pyramidal cells are also reported to affect plasticity in hippocampus (Kim et al., in
submission). This receptor richness make the hippocampal slice an ideal biosensor
tissue, which can predict neuro-cognitive effects of various drugs and CBW agents.
The hippocampal slice offers an additional benefit in the form of preserved circuitry
along a single plane. Although hippocampus has significant longitudinal
connections that are disrupted during transverse sectioning, resulting slices still
maintain electrophysiological functionality of its trisynaptic laminar circuitry for
several hours in acute experimental conditions and up to weeks in culture (Andersen
et al., 1971; Amaral and Witter, 1989; Stoppini et al., 1997; Andersen et al., 2000).
This trisynaptic circuit receives its main inputs from entorhinal cortex, through the
perforant path (PP), which terminates mostly on DG’s granule cell, and less
extensively in CA3 and CA1 molecular layers. The granule cells extend mossy
fibers (mf) that synapse on CA3’s pyramidal cells, which extend their axons, known
as the Schaffer Collaterals (SchC), to CA1’s pyramidal cells (Figure 2.1 A).
Figure 2.1: Hippocampal
slice circuitry in relation
to pMEAs. A) Schematic
diagram of the
hippocampal slice’s
preserved tri-synaptic
pathway: Entorhinal
cortex (Ento) provides
the input to the Dentate
Gyrus (DG) through the
Perforant Pathway (PP)
(synapse 1). Dentate’s
granule cells send mossy
fibers (mf) to CA3’s
pyramidal cells (synapse
2), which extend their
axons (Schaffer
Collaterals, SchC) to
CA1’s pyramidal cells
(synapse 3). The blue
trace is a field Excitatory
Post-Synaptic Potential
(fEPSP) that is evoked at
the CA1 area marked by
the blue dot, in response
to SchC stimulation. The
red trace shows a
population spike induced
below the molecular
layer of DG (red dot), in
response to PP
stimulation. B)
Photomicrograph of an
acute rat hippocampal
slice positioned on a
planar MEA (pMEA).
27
The hippocampal slice’s trisynaptic pathway is therefore ideal for planar
array investigations, because its intrinsic single plane circuitry allows investigation
of drugs’ effects on circuit transmission and plasticity. Elucidation of interactions
between different anatomical areas, or observation of signal propagation requires the
ability to stimulate and record from multiple regions of the slice simultaneously
(Figure 2.1B). pMEAs readily meet this requirement and additionally enable CSD
analysis. This analysis, described further in the next chapter, is used to localize sinks
28
and sources of current flow in biological tissue, which provides more accurate
information on the signal flow in circuits.
The goal of a biosensor is to detect, characterize, and classify a new agent
based on its physiological activity. It needs to do so quickly and simply yet reliably,
and with a good balance between sensitivity, specificity, and breadth. This chapter
describes a tissue-based biosensor that uses rat hippocampal slices on pMEAs to
meet these criteria.
2.2. Materials and Methods
2.2.1. Acute Slice Preparation
Young adult male Sprague Dawley rats (1-3 months old or 150-250g) were
anaesthetized with halothane prior to decapitation. The hippocampi were dissected
from the brains and cut into blocks under cold cutting solution consisting of: (in
mM) Sucrose, 220, NaCl, 20; KCl, 2.5; NaH
2
PO
4
, 1.25; NaHCO
3
, 26; Glucose, 10;
MgSO
4
, 2; ascorbic acid, 2; CaCl
2
, 2, and oxygenated with a mixture of 95% O
2
and
5% CO
2
to maintain physiological pH of 7.2. The blocks were mounted onto a
vibratome (VT1000S, Leica, Germany) with SuperGlue and sliced transversely with
respect to the longitudinal axis of hippocampus. 400-500 µm thick slices were then
incubated at 32 °C in artificial cerebrospinal fluid (aCSF, (in mM) NaCl, 128; KCl,
2.5; NaH
2
PO
4
, 1.25; NaHCO
3
, 26; Glucose, 10; MgSO
4
, 2; ascorbic acid, 2; CaCl
2
,
2) for at least an hour before being carefully positioned on an MEA over an inverted
microscope (DMIRB, Leica, Germany). Their position on the MEAs was held by a
nylon mesh strung across a platinum ring, and documented with a digital camera
29
(Spot RT, Diagnostic Instruments, MI, USA). Slices were left to equilibrate for at
least 15 minutes at 32 °C before electrophysiological experiments. During the entire
experimental duration on the MEAs, slices were submerged and perfused at a
constant flow rate of 2 ml/min with aCSF reduced in its MgSO
4
concentration to
1mM, and in some experiments supplemented with 100µM picrotoxin.
2.2.2. Electrophysiology
2.2.2.1. Recording System
The MEA electrophysiological recording system is a commercially available
turnkey system (MEA60, Multi Channel Systems, Reutlingen, Germany). It consists
of amplifiers (1200x gain with cut-off filters at 0.1Hz and 5kHz), a data acquisition
card (10-25 kHz sampling frequency per channel), an eight-channel stimulation box,
and collection and analysis software. Its MEAs consist of 60 electrodes of 10 or
30µm diameter at 100, 200, or 500µm spacing, and use an external Ag/AgCl
reference electrode.
2.2.2.2. Experimental Protocols
15-30mins after positioning slices on MEAs, afferent fibers were stimulated
by current injection through pairs of electrodes, while the rest of the MEA electrodes
measured evoked field potentials at target areas. The bipolar stimulations were
biphasic with each phase lasting 100µs, and their intensity was in the range of 10-
100µA. Experimental stimulation intensity was chosen from an Input/Output
stimulation curve (IO curve) generated by plotting the amplitude of responses to
stimulations of increasing intensity. The stimulation that yielded a response equal to
30
a third of the maximal response was used. Experimental protocols consisted of trains
of single or paired pulses (pp) delivered at various intervals, with drugs delivered in
aCSF 20-30mins after the start of the experiment. pMEAs recorded field Excitatory
Post-Synaptic Potentials (fEPSP) as well as field population spikes.
2.2.2.3. Data Analysis
Recorded potentials were displayed in McRack (Multi Channel Systems,
Reutlingen, Germany), which also enabled extraction of waveform amplitude and
timecourse display. Further analysis was conducted in Matlab (The Mathworks,
Natick, MA, USA), using the MEAtools toolbox from U. Egert
(Univeristy of Freiburg, Germany, http://www.brainworks.uni-
freiburg.de/projects/mea/meatools/overview.htm), and custom written scripts for
CSD Analysis (see section 3.2.3). Spatial activity maps were generated by bicubic
interpolation between electrode data values in Matlab. Data overlays on slice images
were accomplished in Adobe Photoshop (Adobe, San Jose, CA, USA).
2.3. Speed and Reliability
The first element of a biosensor is that it should be reliable. That means that
it should have a low failure and error rate, yield meaningful and reproducible results
that are stable over the course of the experiment. Current commercial pMEAs
consist of planar arrays of 60 or more electrodes etched on a glass surface through
silicon micromanufacturing technology. The electrodes can electrically stimulate
biological tissue or record its electrophysiological activity. Such pMEAs have been
used for recording from acute (Novak and Wheeler, 1988; Oka et al., 1999) and
31
cultured hippocampal slices (Stoppini et al., 1997; Egert et al., 1998), as well as
cultured dissociated cells (Gross et al., 1993; Gross et al., 1995; Jimbo et al., 1998).
The large number of electrodes provides for redundancy of the recorded signal. If
one or two electrodes fail through oxidation or wear-and-tear, there are 58 more to
use. The large number of channels also provides means for internally cross-checking
responses. A single change of signal would not trigger an alarm if its adjacent
signals remain unaffected. This comparison of the signals from the multiple
electrodes would decrease false-positive and false-negative signals. The redundancy
of the recorded signals also enables high throughput pharmacological investigations
that rapidly select an optimal recording site or group results from different sites to
increase statistical significance.
2.3.1. Multiple Independent Recording Sites
pMEAs allowed rapid determination of both optimal stimulation and
recording sites without damaging the tissue by the traditional method of repeated
electrode insertions. Figure 2.2 shows a pp experiment with a hippocampal slice
positioned over a pMEA and extracellular potentials recorded at all the electrodes.
The slice was stimulated with a pp at SchC. The electrodes with the largest
responses can be easily selected for further analysis.
Figure 2.2: Hippocampal slice paired-pulse (pp) responses recorded with a pMEA. A. Hippocampal
slice positioned over an 8x8 array. Recorded potentials at each electrode are displayed in the square
encompassing it, at a time scale of 150 ms per square, and a y-axis of ± 250 mV. The stimulation, at
the red dots, consisted of two biphasic pulses of ± 75 mA for 100 ms each at an interstimulus interval
of 100 ms. Perfusion aCSF contained 100 µM picrotoxin. pp stimulation induced significant
facilitation in CA1. The second fEPSP is much larger than the first one, with some even showing a
population spike. B. Expanded view of the optimal recording, the one with the largest amplitude
fEPSP highlighted in red in A. pp facilitation is easily discernible.
This advantage of pMEAs was used in an implementation of a hippocampal
slice-based biosensor which exploited variations in Short-Term Plasticity (STP) for
rapid neurotoxins detection (Gholmieh et al., 2001). In this sensor, STP state of a
slice was described by first and second order Volterra kernels based on population
spike amplitudes triggered by random interval electrical impulse sequences. The
kernels changed depending on the properties of perfused chemical compounds,
which enabled their rapid classification. Selecting the largest CA1 population spike
signal was a critical parameter for this approach.
32
33
2.3.2. Multiple Spatially Distributed Stimulations
The description of the hippocampal tri-synaptic pathway above alluded to
more intricate anatomy and complex synaptic connections. Each synapse location
has a different distribution of receptors and channels that corresponds to the various
cell types, and which gives the network more complex dynamics (Brown and Zador,
1990). By stimulating afferents to the three main sub-regions before and after drug
application, the effect on all three synaptic types and their interactions can be
explored in a single slice. The stimulation protocol consists of stimulating PP, mf,
and SchC sequentially, with ten seconds between stimulations.
This paradigm is useful since the effect of a drug may not be uniform across
the entire slice, and recorded responses may therefore miss an effect. An example of
such an overlooked effect is presented in Figure 2.3, where 25mM triethylamine
(TEA) is applied to an acute rat hippocampal slice. TEA is a potassium channel
blocker that has been reported to induce synaptic modification in hippocampal CA1
(Song et al., 2001). These experiments missed a large effect if TEA because they
recorded only from the sites of maximal response to stimulation before TEA
application in order to analyze the effect of drug application. Recording DG
responses to PP stimulation showed a small increase in fEPSP amplitude and a latent
small positive deflection in the field potential followed by a small negativity (Figure
2.3A). However, electrodes at hilus indicate a very large population spike followed
by a long hyperpolarization that match the small deflections observed close to DG’s
granule cells (blue traces in Figure 2.3B). This large latent response spreads through
the entire slice, but appears only as a very small response in DG. Similarly,
examining only the sites of larges response to SchC stimulation prior to TEA
application would have indicated response potentiation and CA1 population spike
but missed the much larger hilar population spike visible on the rest of the electrodes
(red traces in Figure 2.3B). On the other hand, slice responses to mf stimulation
were not affected by TEA application, neither at CA3 nor elsewhere (green traces in
Figure 2.3B).
Figure 2.3: Effects of TEA on acute rat hippocampal slices. 25mM TEA was perfused for 20
minutes and evoked responses to sequential stimulations of PP, CA3, and CA1 were recorded before
(green) and during (blue) perfusion. A. Sample voltage responses (Rec.) from pairs of electrodes in
DG, CA3, and CA1, to stimulations (Stim.) at the PP and SchC. Note the appearance of a new
component, 100ms after PP stimulation that propagates through CA1. In response to SchC
stimulation, multiple population spikes appear in CA1, followed by hyperpolarization. CA3 remains
inactive, with the activity observed probably due to current propagation through the tissue. B. Image
of the slice from which the recordings were taken. Recorded responses to stimulation at PP (blue
electrodes and waveforms), mf (green), and SchC (red), with the stimulation sites marked with the
corresponding colors. Yellow boxes delimit the sites sampled in A.
The multitude of recording elements therefore can not only provide
redundancy for enhanced accuracy and rapid optimal site selection, but they further
enable the sampling of several different locations of heterogeneous tissue, thereby
providing a more complete assessment of a drug’s effects. These 60 electrodes
34
35
pMEAs, with inter-electrode spacing of 200 µm, provides several electrodes in each
sub-region of adult rat hippocampal slices thereby providing redundancy and
breadth.
2.3.3. System Stability
Investigations into long-term synaptic plasticity often require the system to
produce stable recordings over extended periods. Likewise, for the reliable detection
of long-term or chronic drug effects, biosensors must be able to maintain a steady
baseline for prolonged periods. An example of a Long-Term Potentiation (LTP)
experiment is shown in Figure 2.4. Potentiation is induced in CA1 by 100 Hz titanic
stimulation, and the response assessed by pp sent every 10 seconds. The pMEA
graphs over the electrodes plot the fEPSP amplitude of both first (blue) and second
(black) responses over an hour long experiment (Figure 2.4A). After tetanus, both
first and second pulse responses were potentiated. Since not all sites showed similar
potentiation, the electrode with the greatest response was easily selected (Figure 2.4
B and C). In section 2.4.1, further information is extracted from this distribution of
responses.
Figure 2.4: Long-Term Potentiation (LTP) of hippocampal slice on pMEA. A. Each square plots
the timecourse of trough magnitude of fEPSPs recorded in its included electrode over the hour and 15
min long experiment. Blue lines are for the first fEPSP of test pp, and black for the second. Test pp
stimulation, at the red electrodes, consisted of 2 biphasic pulses of ±75 mA for 100 ms each and with
an interstimulus interval of 100 ms. After 15 minutes of test stimulation, a High Frequency
Stimulation (HFS) tetanus consisting of four trains of ten biphasic pulses of ±75 mA at 100 Hz was
delivered, followed by more test pulses for over an hour. Perfusion aCSF contained 100 µM
picrotoxin. Y-axis is 800mV. B. Pre- (blue) and post (red) tetanic stimulation fEPSPs. C.
Expanded view of optimal site timecourse to the first pulse. X-axis 1h10min. Y-axis 500mV.
2.4. Specificity, Sensitivity, and Breadth
Specificity in a biosensor refers to the ability to discriminate different drugs
from each other. Sensitivity, on the other hand, is a function of the lowest detectable
concentration of a particular drug. And breadth is the number of distinct classes that
can be distinguished. Specificity and sensitivity in MEA based biosensors can both
be enhanced by a similar general approach: increasing spatial and temporal
complexity of input stimuli. In order to increase the odds that an individual drug will
have a discernible signal, test dimensions need to be increased. One approach is to
stimulate different locations on sensing slice preparations, in order to take advantage
36
37
of regional differences in anatomical structure, and of network interconnectivity.
Another approach relies on increasing the temporal input richness by stimulating
slices with pulses at random intervals (Gholmieh et al., 2003). A biosensor’s
breadth is limited by the responsiveness of selected biological tissues to presented
drugs. This is a function of receptor types expressed and susceptibility of other
intrinsic biological mechanisms that would affect tissue electrophysiology.
Hippocampal brain slices have many receptor types including glutamatergic,
acetylcholine, cholinergic, noradrenergic, and estrogenic, and peptide specific ones
(Brown and Zador, 1990), thus making them receptive to many classes of drugs
including common chemical nerve agents which are often anti-cholinergic drugs.
While it may be possible to make a sensor tissue responsive to more drugs, by
genetic insertion of additional receptors for example; this would make its response
less indicative of the agent’s effect on a natural human brain. However, breadth
would be enlarged indirectly by increasing a biosensor’s specificity and sensitivity,
as drugs could be classified into a larger number of categories. The following
sections present methods to increase the specificity and sensitivity of pMEA-based
biosensors. These methods draw on benefits and advantages of pMEAs, and are
illustrated with examples of drug responses in hippocampal slices.
2.4.1. Spatial Mapping of Electrical Activity
In a heterogeneous tissue like the hippocampal slice, which has several
regions populated by different cell types with different interconnectivities, all the
electrodes of a pMEA do not record the same signals. The spatial distribution of
38
electrodes determines the sub-regions that are sampled, which may respond
differently to different stimuli. The uniqueness of information carried by each
electrode therefore reflects underlying slice cytoarchitecture, and provides
information on the functional physiology of various sub-regions. By arranging all
recording electrode responses according to their pMEA topography, spatial maps of
responses are generated. The waveforms in figures Figure 2.2 and Figure 2.4 are
arranged according to their respective positions on the pMEA, and illustrate how
responses vary over the slice. In Figure 2.2, fEPSP amplitude decreases in relation
to distance from the stimulation site, while extending further orthodromically along
SchC than antidromically. In Figure 2.4, the degree of potentiation also varies
depending on the proximity to the stimulation site. These spatial maps show the
entire response at every electrode, however their formats may be hard to visually
interpret. Another approach to display spatial maps involves extracting a parameter
of interest, transforming its value into a color code, and then plotting it according to
its electrode position. By interpolating between adjacent electrode values, a color
map is generated, which enables simpler visualization of the parameter’s topographic
distribution.
Figure 2.5: Extent of LTP is different than extent of fEPSP response: A. fEPSP topographic map
showing CA1 response to SchC stimulation (color map reflects interpolation between fEPSP
amplitudes at recording electrodes, where red indicates 50mV, dark blue –50mV, and green 0mV). B.
Same as in A only mapping response after HFS potentiation. C. Timecourse of LTP experiment
plotting amplitude over time at each electrode location. Top inset shows overlay of fEPSP waveforms
pre and post-HFS for the electrode boxed in red. Bottom inset show that electrode’s timecourse.
(Box sizes 500µV and 80ms). D. Spatial map showing extent of spread of potentiation (color map
represents interpolation between ratios of post over pre-HFS fEPSP amplitude at recording electrodes,
where red indicates 300%, and dark blue 0%).
Figure 2.5 illustrates the usefulness of color spatial maps for visual
presentation of LTP experiments run on a 60 electrode pMEA. In LTP experiments,
the amplitude of fEPSP is compared before and after high-frequency stimulation
(HFS). In MEA experiments, fEPSP amplitude can be extracted and converted into
color maps that are overlaid on top of pictures of experimental slices. fEPSP spread
is hence quickly visualized, red areas in Figure 2.5A; and post-HFS response (Figure
2.5B) is notably darker than pre-HFS’, thereby indicating amplitude potentiation,
with little spread of response area. The ratio of post-HFS amplitude over pre-HFS
39
40
amplitude is calculated to determine the percentage of potentiation at each electrode.
This percentage is color-coded and mapped in Figure 2.5D, which reveals in red the
extent of CA1 that exhibited potentiation. This spatial map of LTP is a lot simpler to
read than Figure 2.5C, which shows the time course of the LTP experiment at each
electrode. It is interesting to observe that not all areas that showed fEPSP responses
in Figure 2.5A and B were actually potentiated in D. The potentiation at hilus is due
to an increased response over a very small or no response prior to HFS. Later
sections will discuss an analysis method that could eliminate such artifacts of field
potential spread by examining actual current sinks and sources (Section 3.3.2).
LTP can also be induced chemically, with the application of a neuronal
excitatory compound, N-Methyl-D-Aspartate (NMDA). The NMDA receptor is
known to be essential for the induction and maintenance of LTP. NMDA binding to
its receptor opens its associated calcium channel, if there is sufficient cellular
depolarization to remove its Magnesium block. The channel opening allows calcium
ions influx which triggers cascades that produce changes in synaptic efficacy.
In the experiment shown in Figure 2.6, 5µM NMDA was applied in aCSF
perfusion to an acute rat hippocampal slice for a duration of five minutes then
removed. The slice was being stimulated at four different locations with 5sec
between each stimulation location: PP, mf, SchC, and more distally along SchC.
Figure 2.6 shows timecourse plots of fEPSP amplitude over the duration of the
experiment. After a stable 30min baseline, NMDA application completely abolishes
responses to all stimulation sites. Responses return gradually during drug wash-out,
however they do not all return to the same levels. Responses to PP stimulation
remain depressed, especially at DG. mf stimulation only excited CA3, and those
responses returned to their original level. CA2 responses to SchC stimulation were
however potentiated for 30 mins before gradually returning to their original levels.
Responses to distal SchC stimulation were weakly potentiated around CA2, and
depressed at CA1.
Figure 2.6: Timecourses of NMDA-induced LTP in hippocampal slice stimulated at four locations
on pMEA. A. Timecourses at three electrodes in a column at each sub-region in response to
stimulation at PP, mf, SchC, and SchC distal. B. Timecourses of LTP in response to SchC
stimulation (red electrodes) recorded on entire pMEA and overlaid on slice. Yellow boxes mark
extracted electrode location in A, and green electrodes mark other stimulation sites. X-axis 10
4
sec (2h
45min); Y-axis 500µV.
In order to visualize these different responses topographic color maps were
generated for each stimulation site showing percentage of amplitude ratio compared
to baseline. Figure 2.7 presents the ratio of depression at each electrode for the 5min
41
42
NMDA administration and 15min wash-out during which no fEPSP were observed at
any location. Ratios were calculated by averaging amplitudes of responses recorded
during 20min periods and dividing by the average baseline amplitude. Dark blue
represents a ratio of 0, or abolition of signal, green is for a ratio of 1 or no change,
and red is set ratios of 2 or double the baseline. During the 5 min NMDA
application and for 15min thereafter, the response was abolished, and therefore
appears as blue areas in Figure 2.7. This depression of the signals is attributed to the
continual cell depolarization which inactivates sodium channels. The slight oranges
in Figure 2.7D are due to small fluctuations in a very small signal, and are not
significant.
After NMDA washed out from the slice, DG and CA3 responses to PP and
mf stimulation respectively, came back but remained depressed (Figure 2.8C and D).
The area around CA2 however, was potentiated 200% for 30mins, in response to
SchC stimulation and somewhat less potentiated in response to antidromic
stimulation (Figure 2.8B). Returning response at distal CA1 on the other hand, was
depressed for both SchC stimulation location (Figure 2.8A).
Figure 2.7: Topographic color maps showing NMDA-induced abolition of responses at four
stimulated sites in a hippocampal slice on pMEA. A. Response to distal SchC stimulation is
completely depressed. B. Response to SchC stimulation is depressed where it was present, and
unchanged where there was no response to begin with. C. Stimulation of PP yields no responses. D.
CA3 responses to mf stimulation are depressed, other areas did not have any response in the first
place, and thus the yellows are due to fluctuations of background noise. Color maps are of ratios of
average response amplitude during NMDA application and wash-out divided by average baseline
fEPSP amplitude. Color scale: Red represents a ratio of 2, green 1, and blue 0. Red dots mark
stimulating electrodes.
The above examples of spatial topographic response maps illustrate the
advantage of pMEAs in mapping a response’s location and spread. In the case of
NMDA application, the various hippocampal sub-regions reacted very differently to
the drug, and this difference could be used as a marker to classify this drug. PP’s
sub-regions also reacts differentially to NMDA application. NMDA receptor
antagonists produce differential effects on medial and lateral PP evoked EPSPs
43
recorded in DG’s molecular layer (Dahl et al., 1990). In fact, it had been reported
that bath perfusion of NMDA (10µM, 5min) results in induction of LTP in medial
PP, and induction of long-term depression (LTD) in the lateral perforant path (Rush
et al., 2001). The difference in NMDA-induced plasticity in the two sub-pathways is
attributed to a differential increase in intracellular Ca
2+
at the synapses. A
mechanism for such difference in Ca
2+
influx could arise as a result of a difference
in NMDA receptor properties or distribution.
44
Figure 2.8: NMDA induced potentiation and depression at different sub-regions of hippocampal
slice. Topographic color maps showing NMDA-induced plasticity of responses at four stimulated
sites in hippocampal slice on pMEA. A. Response to distal SchC stimulation is depressed in CA1,
but weakly potentiated closer to CA2. B. Response to SchC stimulation is potentiated close to CA2
and hilus and depressed more distally. C. Stimulation of PP yields depressed responses. D. CA3
responses to mf stimulation are depressed. Color maps are averaged response amplitudes over 20 min
starting 15 min after NMDA washout divided by average baseline fEPSP amplitude. Color scale: Red
represents a ratio of 2, green 1, and blue 0. Red dots mark stimulating electrodes.
45
PP is the major excitatory projection into DG and consists of two
anatomically and functionally discrete subdivisions, the medial and lateral PP. The
lateral PP originates in ventrolateral entorhinal cortex, and terminates in the outer
third of DG molecular layer, while the medial PP originates in dorsomedial
entorhinal cortex, and synapses in the middle third of DG molecular layer (Steward,
1976). Experimentally, the two pathways are distinguished by their anatomical
location, and their responses to pp stimulation, whereby lateral PP shows pp
facilitation while medial PP responses exhibit pp depression (McNaughton and
Miller, 1984; Dahl et al., 1990). Experiments in which NMDA was applied to
hippocampal slices were therefore separated into two groups based on their pp
responses to PP stimulation. Figure 2.9 and Figure 2.10 show examples of slices in
which medial and lateral PP were stimulated, respectively. pp stimulation at 40 ms
interstimulus intervals yielded either slight depression or slight facilitation of fEPSP
amplitude (2
nd
pulse / 1
st
pulse x 100%). Color mapping this plasticity reveals pp
depression at DG in light blue (Figure 2.9B) and thus reflects medial PP stimulation,
whereas in Figure 2.10B yellowish-green demarks DG pp facilitation indicating
lateral PP stimulation. 50µM NMDA was perfused for 5 mins in these experiments,
and then the response was allowed stabilize for 20 mins. The change in fEPSP
amplitude (post-NMDA / pre-NMDA x 100%) is then color mapped. Figure 2.9C
shows potentiated responses especially in DG’s molecular layer, whereas Figure
2.10C illustrates DG’s depressed response.
Figure 2.9: NMDA-induced depression in DG
in response to stimulation at lateral PP. A.
Hippocampal slice on pMEA with red
electrodes marking stimulation site. B. Long-
Term Depression indicated by blue area below
the stimulating electrodes, which corresponds
to DG. C. pp facilitation indicated by the red
and yellow in upper blade of DG.
Figure 2.10: NMDA-induced potentiation in
DG in response to stimulation at medial PP. A.
Hippocampal slice on pMEA with red electrodes
marking stimulation site. B. Long-Term
Potentiation indicated by yellowish green below
the stimulating electrodes, which corresponds to
DG. C. pp depression indicated by the light
blue in DG.
46
In five experiments, selected electrode pairs thus stimulated either one of two
PP sub-pathways according to their pp responses and produced corresponding
changes in responses after NMDA application. These results are consistent with
reported findings (Rush et al., 2001), and thus demonstrated the capabilities of
pMEAs to produce reliable electrophysiological results that can be used to classify
drugs based on their regional effects. The determination of stimulated sub-pathway
was however always post-hoc as it was not possible to select the sub-pathway to
stimulate using these sparsely arranged pMEAs. Section 3.3.3 illustrates how this
differential responsiveness to NMDA in adjacent pathways could be explored in a
same slice with higher density pMEAs.
47
2.4.2. Spatio-Temporal Map of Electrical Activity
The examples above illustrated the power of pMEAs in generating static
topographical activity maps by interpolating values between electrodes. The ability
to simultaneously record from all electrodes over time, transforms these static maps
into dynamic ones that allow visualization of spatio-temporal propagation of activity
across a slice.
The ability of this hippocampal slice based pMEA biosensor to detect and
characterize drugs was assed with trimethylolpropane phosphate (TMPP), a
byproduct of the partial combustion of synthetic lubricants used in jet fuel
combustion and a reported GABA-A inhibitor and epileptogenic agent (Lin et al.,
1998; Keefer et al., 2001; Lin et al., 2001). In this experiment, three locations were
stimulated at regular intervals in order to establish a stable baseline of responses in
the slice sub-regions (Figure 2.11). All three stimulation point produced localized
responses with PP stimulation triggering population spikes in DG only (blue
waveforms), mf stimulation yielding small fEPSPs in CA3 (red waveforms), and
SchC stimulation producing large fEPSPs in CA1 (green waveforms).
Figure 2.11: Three point stimulation of hippocampal slice on pMEA. PP stimulation at blue
electrodes triggered population spikes in DG (blue waveforms). mf stimulation at red electrodes
produced small fEPSPs in CA3 (red waveforms). SchC stimulation at green electrodes yielded large
fEPSPs in CA1 (green waveforms).
After establishing a stable baseline, TMPP was perfused onto the slice in
aCSF. At 1mM (Figure 2.12), TMPP disinhibited the slice sufficiently that
stimulation at PP produced monosynaptic population spikes in dentate granule cells
followed by a disynaptic burst of population spikes in CA3. This epileptiform
bursting in CA3 spread antidromically into DG, and propagated slow orthodromic
fEPSPs into CA1. mf stimulation produced large monosynaptic population spikes in
CA3 immediately followed by bursting similar to that produced by PP stimulation
and followed by a disynaptic fEPSP in CA1. Finally, stimulation of SchC produced
monosynaptic fEPSPs in CA1 identical to the pre-drug ones and no activity
48
anywhere else in the slice. Overlaying the responses to the three stimulations clearly
visualized the activity’s propagation across the slice by revealing response delays.
The time difference between the first population spike’s peaks in CA3 triggered by
PP and mf stimulations was 12-13ms. These first spikes in CA3 where followed by
bursts of spikes 3-5 ms apart, and in the case of PP stimulation, the second spike was
the largest. The response delay between these two stimulations corresponded to the
fEPSP lag observed in CA1 for the same two stimulations (14 ms delay at the
troughs). In addition, disynaptic fEPSPs evoked by mf stimulation was delayed 12
ms from the monosynaptic potential elicited by direct stimulation of SchC.
Figure 2.12: Effects of 1mM TMPP application on hippocampal slice. A. Sample responses
recorded at DG, CA3 and CA1, evoked by stimulation at PP (blue dots and waveforms), mf (red dots
and waveforms), and SchC (green dots and waveforms). B. Photomicrograph of slice on pMEA
overlaid with responses to the 3 stimulations. Yellow rectangles mark electrodes that are expanded in
A. x-axis 100 ms, y-axis 500 µV
These lags reveal that propagation of activity across the slice was facilitated
by TMPP. The observed epileptiform response is consistent with TMPP being a
GABA-A inhibitor, which would reduce the inhibition of the slice. A disinhibited
CA3 is likely to seize due to its large number of recurrent axons that are usually
49
50
inhibited by gabaergic interneurons. More specifically, TMPP disinhibited the
inhibitory neurons in CA3, thus when a sufficient stimulation arrived through mf to
CA3, its disinhibited pyramidal cells all fired in synchrony (population spike). This
population spike spread uninhibited through the recurrent collaterals thereby
triggering the observed bursting. All CA3 spikes also traveled along SchC and
produced the observed trisynaptic fEPSP in CA1. The delays in CA1 fEPSP
reflected the number of synapses traversed between the initial stimulation site and
the CA1 pyramidal cells. Thus SchC stimulation still produces monosynaptic
responses in CA1, while mf stimulation makes a disynaptic responses in CA1, and
the most latent fEPSPs are trisynaptic responses to PP stimulation. The pMEAs thus
enabled not only the localization of the major site of action of TMPP at CA3
inhibitory interneurons, but also unmasked its facilitatory effect on propagation of
activity throughout the entire trisynaptic pathway of the hippocampal slice. TMPP’s
effect can thus be described as inducing CA3 seizures, and facilitating multi-synaptic
transmission.
This example of drug testing shows how the multi-site stimulation paradigm
increased sensitivity to TMPP. Stimulating only SchC and monitoring CA1 response
would have missed the drug’s effect completely, and stimulating only PP and
recording in DG would have observed antidromic propagation of the seizures, but
not deciphered its source of origin. This multipoint stimulation also increased
sensitivity to 1µM TMPP (data not shown), whereby only CA3 showed a very small
and transient epileptogenic response to mf stimulation, while CA1 and DG responses
51
showed no change to any stimulation site. This paradigm thus enabled localization
of the area of the drug’s effect. Furthermore, interactions of different elements of the
circuit before and after drug application provide complexity that can be used to sort
different drugs into more discrete classes, and thus enhance biosensor specificity.
These approaches enrich the information obtained from pMEAs and therefore
increase specificity, sensitivity and breadth at the same time. Sensitivity is not
increased beyond the tissue’s limitations, however, responses are more likely to be
observed due to the larger sampling area and regional differences in sensitivity and
responsiveness to drugs. Specificity is improved by comparative examination of the
drug effects on different areas of the slice. Breadth is enhanced through the
combination of sensitivity and specificity, which would allow more non-overlapping
drug effects. In summary, pMEAs allow sampling multiple tissue locations which
provides a richer response profile to base drug classification on, and thereby imparts
more specificity and breadth to a biosensor.
2.5. Durability: Cultured Tissue
Beyond its abilities to detect and classify unknown agents, a biosensor must
last and function for some duration of time after its manufacture and deployment into
the field. Biological tissue can be cultured for prolonged periods, and would provide
a self-regenerating system. Slices and dissociated cells can be cultured on pMEAs
for several weeks or even years (Stoppini et al., 1997; Thiebaud et al., 1997; Potter
and DeMarse, 2001) (see Section 4.2.2 for methods). These cultures enable longer
term experiments to investigate effects of chronic drug exposure (Gross et al., 1995;
52
Duport et al., 1998). Dissociated culture systems on pMEAs can function as reliable
long-term biosensors, as the large number of electrodes still allows selection of
optimal recording sites, thereby yielding low error rates and fault tolerance. Figure
2.13 shows rate histograms of spontaneous activity recorded at different electrodes in
a hippocampal cell culture before and after application of 10µM TMPP, a byproduct
of jet fuel combustion and a GABA-A inhibitor. Some electrodes showed a short
lasting increase in activity after TMPP application, some became more synchronized
and burst faster and for longer durations, while electrodes with little or no
spontaneous activity were not affected. These results are consistent with reported
findings on the effects of TMPP on cell cultures (Keefer et al., 2001), and can
therefore be used to characterize TMPP as an inhibitor of inhibitory pathways based
on its synchronizing effects on network electrophysiology.
Figure 2.13: Effects of TMPP on
dissociated embryonic hippocampal cells
cultured on PDL coated pMEAs. A)
Electrode and cell photomicrographs are
overlaid with spontaneous activity rate
histograms in green, where spike number
in 1sec bins are plotted over time. B)
Same photomicrographs overlaid in red
with rate histograms of spontaneous
activity recorded in the presence of 10µM
TMPP. X-axis is a period of 10min (63,
28, and 32) or 1min (56). Y-axis indicates
the number of spontaneous spikes per bin
(range 0-10).
Another experiment was conducted with the neuronal excitatory compound
NMDA. In the experiment shown in Figure 2.14, 5µm NMDA was added to
spontaneously firing cells cultured on pMEAs. Cell firing rate was measured in 1
sec bins and plotted over time before and after NMDA application.
53
Figure 2.14:
Effects of NMDA
on dissociated
hippocampal cells.
Photomicrographs
of electrodes show
cell coverage on
particular recording
electrodes. The top
electrode (13) with a
large clump of cells
was spontaneously
firing continuously,
with almost regular
short bursts. These
bursts were
eliminated by
NMDA application,
as the cells began
firing continuously
at a higher firing
rate. Electrodes 64
and 65 both had
smaller clumps
which fired
sporadically at a low
firing rate (64) or
not at all (65). The
firing rate on both of these electrodes increased after NMDA application. The sporadic activity
recorded from the small clump on top of electrode 51, was not affected by NMDA application.
Longer-time course shows an initial burst in activity, followed by brief quiescence, then a return to
heightened activity levels.
Photomicrographs of the culture show the position and size of cell clusters on
top of recording electrodes (Figure 2.14). Rate histograms comparing 600sec of
activity before and after NMDA application reveal a clear increase in spontaneous
firing. Regular bursting patterns are eliminated and replaced by continuous high
frequency firing. Some low or no activity small clumps increased in firing
frequency, whereas others were visibly not affected. Immediately after NMDA
application, there is a rapid increase in activity, which is followed by period of 5-
10min in which firing is reduced, before a return to heightened activity. The
different response properties observed are attributed to the population heterogeneity
54
55
of cultured hippocampal cells and to their local interconnectivity. Sampling multiple
electrodes therefore provides a more accurate image of the effects of NMDA on the
culture.
In addition to dissociated cells, slices can also be cultured on pMEAs
(Stoppini et al., 1997; Thiebaud et al., 1997; Egert et al., 1998; Duport et al., 1999;
Jahnsen et al., 1999). Such organotypic slice cultures can also be used to investigate
long-term drug effects on the tissue preparation with the advantages of preserved
cytoarchitecture (Shimono et al., 2002). Both cells and slices require a week to grow
on pMEAs and become electrophysiologically active after 10 days in vitro (DIV),
but can then last up to months or years under proper conditions (Potter and DeMarse,
2001).
2.6. Future opportunities and obstacles
With CBW proliferating among states and into the hands of sub-state
organizations, the risk of terrorists using chemical or biological attacks in major
metropolitan cities is increasing daily with current global political events (Tucker,
2001). There is therefore an urgent need to develop and deploy sensors that will
detect and classify attacks rapidly and accurately to emergency response teams.
Biosensors are unique in their ability to detect unknown agents and classify them
based on their functional activity. Despite this great opportunity for biosensors,
there are numerous challenges that need to be overcome before they become as
commonplace as security cameras.
56
2.6.1. Portability and Ruggedness
In order for this biosensor technology to become field deployable and serve
as early warning against chemical and biological weapons, several methodological
obstacles still need to be surmounted. First the preparation of tissue sample requires
a sterile lab environment and some specialized skill, and since biological tissue does
not last indefinitely, it needs to be prepared continuously. This can become
impractical if biosensors are to be placed in every train stations around the world.
(however, this could provide a lucrative self-sustaining business and a job for me!)
Furthermore, assembled sensors should be able to survive shipping and handling.
The culture system requires a warm and moist 5-10% CO
2
atmosphere. Currently,
pMEAs are small enough to carry by hand, but the majority of data processing
happens on a desktop computer with amplifiers and A/D boards. These would all
have to be miniaturized, along with perfusion systems. Air sampling devices would
also have to be included. These and other portability challenges have recently been
overcome at the Naval Research Laboratory, and a rugged briefcase size biosensor
chamber has been developed which utilizes dissociated cell cultures that can be
shipped by FedEx (Pancrazio et al., 2003). This rugged briefcase biosensor is the
first step to deployment in the field.
2.6.2. Confounding Factors and False Positives
There is, however, a major problem that caused the M8a1 CBW detector to
fail constantly during the first Gulf war: a dirty environment. The M8a1 was
plagued by false alarms because it was not designed to operate in a fuel combustion
57
products laden atmosphere, such as the one presented during the gulf war (Gulflink,
1997). Similarly, biologically based biosensors will undoubtedly report that the air
is contaminated with innumerable bacteria and toxic agents that smog-up urban
areas, but which humans have adapted to, or are surviving in. Most biosensors are
not yet suitable for testing of realistically complex samples, such as blood or soil
(Byfield and Abuknesha, 1994; O'Sullivan and Alcock, 1999), however, purification
systems such as capillary electrophoresis or gas chromatography could be attached
for serial sample testing (Bossi et al., 2000). The above-described biosensor
becomes a very sensitive system that is certainly more sensitive than a live human
brain which is protected by many biological systems. The increased sensitivity of
the biosensor is due to the absence of homeostatic body functions that are designed
to protect the brain from the environment: such as the skin, lungs, and the blood-
brain-barrier which shields brains from many blood-born drugs or toxins, and the
absence of kidneys, liver, and an immune system that clear out harmful agents.
While increased sensitivity is generally desirable, one should keep in mind that
agents applied to a biosensor slice, are not necessarily what would reach a human
brain. An in vitro blood-brain barrier culture system has been developed for
modeling and studying drug effects on the brain in a more realistic fashion (Duport
et al., 1998). Such a system could prove useful in the development of more
biologically relevant biosensors. However, a biosensor does not have to predict the
exact effect of an agent on the human body based on its tissue response, but merely
to characterize that compound’s effect in order to classify it.
58
2.6.3. Classification and Simplicity
A useful biosensor, will not only warn you of unknown agents that affect
your body and brain, but will be able to classify them based on a database of known
drugs. One must therefore construct a broad library with different pharmacological
agents before testing unknown agents. Gross et al. constructed a small database of
drug effects on dissociated cultured neurons (Gross et al., 1995). They later used
this information to correctly classify TMPP as an epileptogenic agent (Gross et al.,
1999; Keefer et al., 2001). However, this database was qualitative in its description
of drug effects, and therefore analysis of TMPP samples required subjective and
expert evaluation. New quantitative methods have been proposed to automate and
objectify this classification of unknown agents. These methods create fingerprints of
known chemicals based on their effects on short-term plasticity (Gholmieh et al.,
2003) or network activity (Gramowski et al., 2004; Selinger et al., 2004), and then
classify unknown compounds by matching their fingerprints in a library of known
compounds. The challenge for these approaches is the creation of a comprehensive
database which should sample every known compound at several concentrations and
various exposure times with a few replications of each for accuracy and error margin
estimations. However, the reference database does not need to be complete, but
should at least contain all known classes of compounds, and efficient search tools for
matching agents’ fingerprints extracted from biosensor data. Future biosensors will
be fully automatable and output an agent’s class and lists precautions, treatments,
and containment procedures directly to emergency response teams.
59
2.6.4. Cost and Reusability
Finally, the price the world is willing to pay for potentially savings millions
of lives is difficult to estimate, however, the cost of manufacturing, deploying, and
maintaining biosensors must be competitive with other thecnologies. Current
mechanical and electrochemical sensors are cheap and durable. The price of a
biosensor must be commensurate with its advantage of detecting unknown agents. It
may be cheaper to produce slice or cell cultures than to purify and stabilize enzymes
or antibodies. However, the month-long lifespan, and requirements for maintaining
tissue alive, may offset the savings. In addition, reusability of biosensors might not
be possible as baseline signal recovery depends on the dynamics of a drug and its
receptors on the tissue. Therefore, in order to compare an agent to a database, it
would be best to replace the tissue regularly and after any detection, as a second
agent’s response might be affected by a prior tissue exposure. As for the cost of
pMEAs, they are not cheap to manufacture, and the oxidative culture environment is
an endurance challenge to their manufacturing materials. It is, however, possible to
clean and reuse pMEAs repeatedly until the signal quality deteriorates to below an
acceptable threshold. The cost of maintaining biosensors alive around large cities is
therefore one of the major obstacles to the deployment of this technology.
2.6.5. Conclusion
With all the limitations listed above on some of the criteria for successful
sensors, namely, selectivity, sensitivity, detection limit, reversibility, response time,
size, ruggedness, reliability, cost, and signal recovery (Roe, 1992), there is still a
60
desperate need for a biosensor that would detect and classify unknown agents in the
field rapidly and accurately. Its uncontested advantage over conventional CBW
detectors is that it can characterize unknown compounds based on their physiological
effects in a very short amount of time. Since the threat of CBW includes novel
agents, it is imperative that a tool exists for when all the other tools fail to identify
the hazard. Biosensor certainly have potential for environmental inspection
agencies, army base labs, or pharmacological research and development companies
that wish to compare or classify unknown compounds based on their effects on a
biological tissue or the nervous system more specifically. The system described here
should be highly reliable, and presents error tolerance and self-checking modes
through the redundancy of its electrodes. The methods presented above enable the
sensor to probe brain slices with rich spatio-temporal inputs, which enhances its
sensitivity, specificity and breadth. Furthermore, multielectrode recordings lend to
biological interpretation and mechanistic understanding which could be useful for
pharmacological research industries. Such multielectrode-based biosensors are
already more commonly used in drug development research (Baudry et al., 2005;
MultiChannelSystems, 2005).
61
3. HIGH DENSITY CUSTOM MEAS
3.1. Trade-off Between Coverage Area and Density
The capability of simultaneously recording electrical activity at multiple sites
in vitro has enabled investigations of neuronal network dynamics previously not
possible with single electrode recordings (Droge et al., 1986; Warland et al., 1997;
Singer, 2000). pMEAs present one currently available technology to record from
multiple neurons simultaneously in vitro (Novak and Wheeler, 1988; Gross and
Schwalm, 1994; Stoppini et al., 1997; Egert et al., 1998; Duport et al., 1999; Jahnsen
et al., 1999; Oka et al., 1999; Jimbo and Robinson, 2000). The spatial distribution of
pMEA electrodes captures spatio-temporal relationships of neuronal activity, while
their transparent solid support permits microscopic visualization of the relative
position of tissue to electrodes. pMEA electrodes can be used for both recording and
stimulation, thereby providing self-contained systems with no need for external
electrodes (Novak and Wheeler, 1988; Gross et al., 1993). These advantages offered
by pMEAs over traditional extracellular pulled-glass or sharp-wire electrodes depend
on the number of electrodes, which is limited by current technological constraints,
such as electrode and lead overcrowding, cross-talk, connector design, and data
acquisition. This limitation creates a trade-off between spatial sampling resolution
and area, preventing stimulation and recording from every location of a tissue
preparation. The methodology described in this chapter resolves this compromise by
designing cMEAs that have a high-density of electrodes in tissue-conformal
configurations for specific experimental applications.
62
Low cost of photolithographic fabrication coupled with advances in signal
acquisition hardware, processor speed, and data storage, has led several groups to
independently develop their own pMEAs. These investigators developed thin-film
pMEAs in a variety of configurations to monitor extracellular electrophysiological
activity in acute and cultured slices from different brain areas such as retina (Meister
et al., 1994; Grumet et al., 2000), spinal cord (Borkholder et al., 1997), and
hippocampus (Novak and Wheeler, 1988; Boppart et al., 1992; Egert et al., 1998;
Oka et al., 1999; Thiebaud et al., 1999; Heuschkel et al., 2002). In addition, several
commercial planar multielectrode recording systems have recently become available
such as MEA60 from Multi Channel Systems, Reutlingen, Germany and MED64
from Panasonic, CA, USA. Most of these currently available pMEAs have
electrodes distributed in square matrix configurations with relatively large inter-
electrode spacing (greater or equal to 100µm). These regular arrangements provide a
versatile solution to the trade-off between coverage area and electrode density,
forced by the limited number of recordable electrodes on pMEAs. However, since
neural cytoarchitecture changes dramatically in the spatial domain in a non-
symmetrical fashion, these low-density geometrically regular electrode arrangements
do not provide the necessary resolution for selectively stimulating afferent pathways
or flexibility in recording from small subregions. Furthermore, the low electrode
density does not permit adequate CSD analysis (Freeman and Nicholson, 1975;
Nicholson and Freeman, 1975; Nicholson and Llinas, 1975; Wheeler and Novak,
63
1986). Therefore there is a need to create tissue-specific high-density cMEA that
conform to the cytoarchitecture of the nervous tissue of interest.
This study describes methodology for designing, fabricating and using such
conformal high-density cMEAs, and presents three examples of such pMEAs suited
for CSD analysis, all of which are conformally-mapped to hippocampal slice
cytoarchitecture. The conformality refers to electrode distributions that correspond
to the organization of intrinsic hippocampal circuitry. Stimulating electrodes are
thus concentrated under afferent fibers or presynaptic cells, while recording
electrodes are arranged under postsynaptic dendrites and somas. These layouts were
designed for in vitro stimulation and recording from different hippocampal
subregions (Figure 3.1): cMEA #1 is a 3x20 rectangular array created for CSD
analysis, and is well suited for electrophysiological investigations of pyramidal and
granular cells of the hippocampus, since these cells are densely packed into columns
of parallel dendrites. cMEA #2 is designed to stimulate SchC afferents to CA1 and
records their responses, whereas cMEA #3 is intented for stimulation of PP fibers to
excite DG and CA3. These designs with conformal topographical mapping and
high-density spatial placement of electrodes empowered cMEAs with fine control
and easy optimization of stimulation and recording sites.
Large signal-to-noise ratios (>10:1) and high spatial density of electrodes has
enabled one-dimensional CSD analysis of responses recorded from all three designs.
This analysis disentangles field potentials to accurately map sources and sinks of
synaptic currents. CSD was combined with sequential stimulations through a
64
column of electrodes to generate a laminar profile of CA1, and to demonstrate
independence of spatially distinct inputs. Selective stimulation of afferent fibers was
hence optimized with ease even with adjacent pathways. These experiments with
acute rat hippocampal slices established that conformal high-density MEAs can be
custom-designed for slice preparations to ideally suit experiments requiring selective
stimulation of afferent pathways and CSD analysis of responsive areas.
3.2. Material and Methods
3.2.1. Conformal pMEA design and fabrication
Cytoarchitectural measurements were taken from 5-10 photomicrographs of
adult rat hippocampal slices (see Section 2.3). Experiment-specific pMEA layouts
were drawn using Illustrator (Adobe, CA, USA), with electrode positions
conforming to several acute slice photomicrographs and according to known
hippocampal cytoarchitecture. Small groups of electrodes were arranged in sub-
arrays perpendicular to their target’s cell layer. These electrode arrangements were
transferred to VLSI layout design tool (L-edit software, Tanner Inc.) where leads and
contact arrangements matching one of the electrophysiology recording setups were
added to the design. Layouts were sent to Photosciences Inc. for mask manufacture,
where three masks were created for each pMEA design: one for fiducials, one for
leads and contacts, and one for electrode tips. pMEAs were fabricated using
standard photolithographic techniques similar to those described in (Gross et al.,
1993), briefly: (more details can be found in (Gholmieh et al., In Preparation))
1) Nickel fiducial marks were deposited by electron-beam evaporation onto
transparent Indium Tin Oxide (ITO)-coated glass substrates, for use as alignment
marks for subsequent patterned masks.
2) ITO was etched with an acid solution to pattern electrodes leads and tips.
3) Silicon Nitride (SiN
x
) was deposited as insulating material.
4) Electrode tips and bonding pads were exposed using RIE etching in CF4-plasma.
5) The tips were covered by deposition of titanium, platinum, and gold.
6) O
2
-plasma was used to clean the pMEA surfaces make them more hydrophilic.
3.2.2. Electrophysiology
See Section 2.2.2
3.2.3. Current-Source Density Analysis
CSD was calculated using only one dimension by assuming that the current’s
variations along the other two dimensions were negligible and that the medium
conductivity is pseudo-isotropic (Holsheimer, 1987). Equation (D3) from Freeman
and Nicholson (Freeman and Nicholson, 1975) was applied to recorded field
potentials (FP):
) 1 ( 7 / ) 2 2 2 (
2 2 h x h x x h x h x x m
I
+ + − −
− + + + − − = φ φ φ φ φ σ
Where I
m
is current density, Φ is FP, σ is an estimate of biological tissue
conductivity along the analyzed dimensions and h is the spatial sampling interval
along that dimension.
FP and CSD topographical activity maps were generated in Matlab by
assigning a color scheme to the measurement range and applying piecewise cubic
65
66
Hermite interpolation between vertical electrodes data values. The 10kHz sampling
rate eliminated the need for interpolation along the time domain. Data overlays on
slice images were accomplished in Photoshop (Adobe, San Jose, CA, USA).
3.3. Results
3.3.1. Fabrication
cMEAs for specific experiments on adult rat hippocampal slices were
fabricated to fit into MEA recording systems. These ITO-based cMEAs had up to 64
gold recording/stimulating microelectrodes consisting of either 28 µm-diameter disks
or 36 µm-squares; either 50 or 60µm center-to-center spacing was used (Figure 3.1).
Four configurations of high-density cMEA were designed to conform to
hippocampal slice cytoarchitecture in order stimulate and record electrical activity in
specific areas. Electrodes were grouped into arrays radial to the cell layer curvature,
to allow for variability in slice size and positioning. cMEA#1 was a rectangular
3x20 matrix suited for CSD analysis of CA1 pyramidal cells (Figure 3.1A).
cMEA#2 had one sub-array of 2x8 electrodes arranged for bipolar stimulation of
SchC and another 4x12 sub-array for recording corresponding CA1 FPs and
calculating their CSDs (Figure 3.1B). cMEA#3 was intended for investigating DG-
CA3 interactions by stimulating the PP from two 3x7 sub-arrays and recording CA3
FPs (Figure 3.1C).
Figure 3.1: Three custom MEA (cMEA) designs with hippocampal slices positioned on top. A.
cMEA#1: 20x3 rectangular array of circular electrode 50µm apart with tips of 28µm diameter and
176k Ω impedance at 1kHz. B. cMEA#2: two sub-arrays of square electrodes 60µm apart with tip
sides of 36µm and 85k Ω impedance at 1kHz. The 2x8 sub-array is designed to stimulate SchC fibers
and a 4x12 sub-array to record output responses from CA1 pyramidal cells C. cMEA#3: two 3x7 sub-
arrays to stimulate PP and record in DG and one 3x6 sub-array to record CA3 output. Electrode tips
similar to A.
These cMEAs were durable, and some have been used to record useful
signals in over 20 experiments consisting of hundreds of stimulations each.
Electrochemical characterization of these cMEAs including impedance
measurements, cyclic voltammetry testing and corresponding current injection limit
determination has been described in (Han et al., 2002).
3.3.2. CSD Mapping of Electrical Activity
3.3.2.1. FP vs CSD
The spatial maps in section 2.4 give an excellent picture of what is happening
over the entire slice during experiments; however, they do not provide information
about the exact location of activity. There are not enough sampling points to
produce accurate information about differences along the dendritic tree of pyramidal
cells. This electrode low spatial resolution is further complicated by the fact that
field potentials intermix in the tissue, and their sources and sinks may be missed
between electrodes. High-density pMEAs sample more points per unit area, and
enable CSD analysis. (Nicholson and Llinas, 1975; Miyakawa and Kato, 1986) The
67
68
latter is a spatial filter that reveals sources and sinks of currents, and allows more
meaningful interpretation of field potential recordings.
In order to assess the applicability of high-density pMEAs for CSD analysis,
hippocampal slices were oriented on cMEA#1 such that CA1’s pyramidal axo-
dendritic axis was parallel to the long side of the rectangular array (Figure 3.2A).
Monopolar biphasic stimulations were then delivered through one of the electrodes
in a peripheral columns, and evoked responses were recorded on the fifty-nine other
electrodes. The left panel in Figure 3.2B shows FP activity recorded at the middle
column of electrodes in response to a single stimulation in stratum radiatum (marked
by a red dot). After a 3-5ms delay, positive FPs were observed in stratum oriens,
and negative potentials in strata radiatum and moleculare. 5ms after stimulus,
population spikes, seen as a sharp deflection in the waveform, were detected on
many electrode traces around the cell layer. These currents gradually switched
polarity as cells repolarized themselves, and hyperpolarized 20ms after stimulation.
Combining multiple simultaneous recordings along the length of pyramidal cells
constituted a FP laminar profile of CA1 responses. By interpolating between
adjacent recordings, a topographical map was generated in which voltages were
assigned colors (Figure 3.2B). In these color maps, yellow/red marked the spread of
positive fEPSPs, while blue areas delineated negative fEPSPs, or the spread of
population spikes through positive fEPSPs. CSDs of the laminar profiles were
obtained by applying 1-Dimentional CSD analysis to recorded voltages. The spatial
spread of population spikes was markedly narrower in CSD topographical maps, and
69
yet could be more easily distinguished as fEPSP spread was also reduced (Figure
3.2B). CSD also more finely localized reversal regions of EPSPs and population
spikes, zones where their polarity inverts, to stratum pyramidale. Additionally, CSD
analysis unmasked a current source in stratum radiatum that was not discernible in
FPs.
Figure 3.2: Field Potential (FP) and Current Source Density (CSD) in CA1 pyramidal cell activity.
A. Photomicrograph of a hippocampal slice positioned on cMEA#1 with a pyramidal cell drawing
from the Neuron Morphology Database (http://neuron.duke.edu) to illustrate spatial extent of cells and
strata relative the electrodes. B. Blow-up of cMEA’s span of CA1 and laminar profile recorded in
response to 200 µA monopolar stimulation at the electrode marked in red. The red traces show
recorded FP (left) or calculated CSD (right) from the middle electrode column (red box), with
numbers corresponding to electrode position starting from the top. Topographical maps converted
voltage to a color scale with cubic interpolation between electrodes. The trace drawing of a cell
illustrates the relative position of the electrodes with respect to CA1 pyramidal cells. In addition,
strata are marked with dashed lines and labeled in white on their respective electrodes (o: oriens; p:
pyramidale; r: radiatum; m: moleculare (CA1 and DG)). Stimulation artifact is visible at time=0. x-
axis span 35 ms, y-axis and color range -400 to +400 µV or CSD units.
70
71
3.3.2.2. Higher spatial density yields better mapping of sinks and sources
After illustrating the classical advantage of CSD over FP, the effect of
different electrode densities on CSD resolution was investigated. FP recordings
from a 20 electrode column were separated into groups of 10 Odd and 10 Even
electrodes, and CSDs from these subgroups were compared to CSDs obtained from
the entire 20 electrode column. Figure 3.3 shows FPs and CSDs obtained from a
single stimulation at an electrode in stratum radiatum (same as in Figure 3.2).
Topographical maps indicated that for a same stimulation, there was minor
difference between data collected on Odd and Even subgroubs, mostly consisting of
slight signal size variations. The most significant observation was the effect of
electrode separation on CSDs, whereby CSD from Even and Odd subgroups, whose
electrodes were separated by 100µm, did not narrow sinks and sources as did CSD
from the entire array with 50µm interelectrode spacing.
Figure 3.3: Effect of inter-electrode spacing on FP recording and CSD analysis. FP and CSD
profiles of the responses shown in Figure 3.2A divided into electrode subsets: All (50 µm separation),
Odd and Even (100 µm). The slice, stimulation site and intensity, scales and strata markers are the
same as in Figure 3.2.
3.3.2.3. Multiple laminar profiles
CA1 pyramidal cells have multiple afferent fibers synapsing on their
dendrites. Activation of these fiber bundles produces different laminar profile
72
73
responses in the cells. cMEA#1 enables visualization of these different response
maps. CSD analyses of multiple laminar profiles of CA1 were generated by
sequentially stimulating electrodes in a column along the entire length of the
pyramidal cells. Figure 3.4 shows CSDs of laminar profiles recorded from the
middle column of the array depicted in Figure 3.2, with stimulus sites denoted by red
dots. The stimulation spanned the entire dendritic range of CA1 and into DG.
Starting with stimulation at alvear fibers, were the slice has less thickness due to
slicing along the hippocampal curvature, no response was generated (Profiles 1,2,3).
As the stimulation site moved closer to the cell body layer, fiber volleys were
observed followed by a current sink was observed in stratum oriens with a
corresponding source in stratum pyramidale (4,5,6,7). The fEPSP is interrupted by a
sharp population spike sink which moves from pyramidale to radiatum over a 5 ms
duration, while its source moves from oriens to pyramidale. After the fEPSP
attenuates, sinks and sources switch positions marking cell hyperpolarization.
Exciting the cell body layer directly produced a source in radiatum with
corresponding two small sink in oriens and distal radiatum (8), which were all
inverted in polarity for stimuli at stratum radiatum (9). As the stimulating site
moved more distally along the apical dendritic tree, sinks and sources first grew to
their largest values (10,11) and the population spike reappeared. The population
spike sink appears to start just apical to stratum pyramidale, partially masked by the
fEPSP, while its corresponding source starts basal to the cell layer and spreads into
oriens. fEPSP amplitude decreased and the population spike disappeared with
74
stimulations at more distal radiatum (12-13). fEPSP amplitude was larger and a
population spike reappeared with stimulation closer to stratum lanucosum
moleculare (14) with similar temporal dynamics and spatial spread to the
stimulations in radiatum closer to the soma. Stimulation at lanucosum moleculare
(15-17), however, yielded no evoked responses. Stimulating close to the fissure did
not produce responses in neither CA1 nor DG (18). Finally, stimulation in stratum
moleculare of DG (19, 20) produced large current sinks responses in the granule cell
dendrites.
Figure 3.4: CSD laminar profile of CA1. CSD topographical maps of laminar profiles of the slice in
Figure 3.2A in response to sequential stimulations along the peripheral column. The red dot marks
the stimulation electrode’s relative position for each panel with stimulation starting basal to the
pyramidal layer in stratum oriens and proceeding into DG’s stratum moleculare. The slice,
stimulation site and intensity, scales and strata markers are the same as in Figure 3.2.
3.3.2.4. Multi-points stimulation
The sequential laminar profile presented above exhibited reversal of sinks
and sources when the stimulation site switched from basal to apical dendrites. Multi-
point stimulations enabled examination of possible interactions between these
spatially distinct afferent pathways. The response to two simultaneous monopolar 50
75
76
µA stimulations, one at basal and another at apical dendrites, were recorded and a
CSD laminar profile generated (Figure 3.5A). Two sinks were present, one in apical
and another in basal dendritic areas suggesting simultaneous dendritic activation,
with a large and complex source in between. In order to demonstrate independence
of the two responses, paired pulses were delivered at 50 ms interstimulus intervals
through the two electrode sites, either on the same side (cis) or opposite sides (trans)
of the cell body layer, and the four corresponding CSD topographical maps were
generated (Figure 3.5B-E). When the two pulses were in cis configuration, the
second pulse resulted in facilitation, as usually observed with pp stimulation in CA1,
for both basal (Figure 3.5B) and apical (Figure 3.5C) stimulations. When the paired
pulses were in trans configuration, the second pulses resembled a single pulse from
that location, suggesting that the stimulation sites do not have any short-term
plasticity effects on each other in this stimulation protocol (Figure 3.5D and E).
This experiment illustrates the ease with which high-density cMEAs, with
their closely spaced electrodes are able to selectively activate adjacent pathways.
Such sub-pathway segregation is extremely difficult to achieve in the same slice with
more widely spaced MEAs or inserted twisted-wire stimulating electrodes. Section
3.3.3 illustrates this advantage further.
Figure 3.5: Stimulation site independence
along CA1’s pyramidal axo-dendritic axis.
A. CSD topographical map of the
response of the slice in Figure 3.2A, to
simultaneous stimulation at electrodes on
basal (#6) and apical (#10) dendrites. B
and C show CSD of the response to pairs
pulses with 50 ms interstimulus intervals
at these electrodes where both pulses are
on the same side (cis): B. basal. C.
apical. D and E. show CSD of pp
responses when the pulses are on poosite
sides of the cell bodies (trans): D. First
pulse basal and second apical, E. First
pulse apical and second basal. The slice,
stimulation intensity and strata markers
are the same as in Figure 3.2, however the
color scale spans from -200 to 200.
77
78
3.3.3. Specific Stimulation Site Selection
The primary advantage of cMEA#2 and #3’s conformal design was the ease
and convenience of optimizing stimulation and recording locations. These novel
pMEA designed to match hippocampal slice cytoarchitecture eliminate the need to
reposition the slice or stimulation electrode for optimizing stimulation and recording
locations. cMEA#2 was conceived to rapidly localize the optimal stimulation
location to target SchC and select the largest CA1 response. An optimization
protocol and example for that array are described in (Gholmieh et al., In
Preparation). cMEA#3 had three sub-arrays designed to study PP-DG-CA3 synapses
with an emphasis on differentiating between lateral and medial PP fibers. The sub-
arrays span outer- and inner-blades of DG (3x7 sub-arrays) and CA4/CA3 (3x6 sub-
array) in hippocampal dorsal slices (Figure 3.1C). Section 2.4.1 introduced PP and
its sub-pathways and the use of pp stimulation to distinguish between them
electrophysiologically. In order to demonstrate the ability to discriminate between
these two pathways in a same slice, high-density conformal cMEA are used to
stimulate lateral and medial PP. Each of these sub-pathways encompasses
approximately 100 µm, and the electrodes are spaced 50 µm apart, therefore,
according to the Nyquist theorem, each sub-pathway should be activated from at
least one row of electrodes.
Figure 3.6 illustrates cMEA#3’s ability to distinguishing between medial and
lateral PP in acute rat hippocampal slices. Slices can be positioned in different
orientations on the arrays allowing recording from different subregions. In this
79
example, one sub-array (marked by a yellow box) is positioned perpendicular to the
molecular layer of DG’s upper blade. This enables bipolar stimulation through
adjacent electrode pairs in rows, and recording DG responses on the remaining
electrodes in this sub-array. The electrodes highlighted in green were used to
stimulate the outer third of PP (Figure 3.6 B), which corresponds to lateral PP. A 40
ms interval pp produced an increase in the amplitude of the second fEPSP compared
to the first, hence exhibiting pp facilitation. The ratios of the 2
nd
pulse’s fEPSP
amplitude over that of the first is color-coded at each electrode, and these values are
interpolated into a spatial color map that reflects facilitation in yellow-red. In
contrast, stimulating the same slice through electrodes at medial PP (Figure 3.6 C)
revealed pp depression (blue) for the same 40 ms interval. The facilitation and
depression observed respectively at the anatomical lateral and medial pathways
confirm that the two sub-pathways were indeed selectively excited. Paired pulse
stimulations at the electrodes in between the red and green ones did not produce
changes in fEPSP amplitudes (data not shown), suggesting that both pathways might
be partially stimulated, and cancelled each other’s effects. Lateral PP produced an
average of 113.58% (±4.1 SEM) pp facilitation in four similar experiments, while
medial PP yielded 86.14% (±1.6 SEM) pp depression in these same slices.
This experiment and the CA1 multi-point stimulation described in Section
3.3.2.4 demonstrate the ability to stimulated afferent pathways independently from
each other, which is important for interacting with tissues whether for biosensor or
neuroprosthetic applications.
Figure 3.6: Paired-pulse
facilitation and depression
in medial and lateral PP.
A. Photomicrograph of
DG on cMEA#3. The
electrodes highlighted in
green and red gray
correspond respectively to
lateral and medial PP
stimulation sites, while
the yellow box marks the
sub-array whose
responses are analyzed in
B and C. B. DG’s outer
blade response to pp
stimulation at lateral PP
(green dots). Colormap
of sub-array electrode
responses, interpolating
between ratios of second
pulse fEPSP amplitude
over first pulse. C. pp
stimulation at medial PP
(red dots). Colormap
interpolating between pp
ratio of fEPSP
amplitudes. Stimulations
were at 80 µA intensity
and 40 ms interstimulus
interval.
80
81
3.4. Discussion
In an effort to circumvent limitations resulting from fixed numbers of
channels currently available in pMEA recording systems, a few tissue-specific
cMEA configurations have been produced: a hexameric layout, commercially
available (Multi Channel Systems), has smaller and closer electrodes in its center for
retinal recordings; elliptic (Thiebaud et al., 1997) or circular (Duport et al., 1999)
electrode layouts have been designed for hippocampal slices with a single or double
layer of electrodes matching roughly the cytoarchitecture of hippocampal pyramidal
cells. These latter arrays, however, are hard to align with brain slices due to small
differences between slices. Additionally, none of these pMEAs are optimized for
CSD analysis. This chapter introduced three high-density pMEAs that are well
suited for CSD analysis, and two of which were conformally-mapped to
hippocampal slice cytoarchitecture. Each array was designed with specific
experimental goals which illustrate the advantages of high-density conformal
layouts.
The electrophysiological experiments presented above demonstrated the
functionality of high-density arrays for electrical stimulation and recording of
physiologically meaningful data. The arrays recorded fEPSPs and population spikes
with magnitudes greater than 10 times the background noise. Laminar profiles of
FPs and CSDs were generated in response to dendritic stimulation using cMEA#1
(Figure 3.2). The FPs and CSDs are very similar to those reported by others in vivo
(Richardson et al., 1987; Kloosterman et al., 2001). Typical monosynaptic
82
responses were observed in CA1 in response to dendritic stimulation. As previously
reported (Nicholson and Freeman, 1975; Buzsaki et al., 1986, Holsheimer, 1987
#1760; Kloosterman et al., 2001), CSD analysis narrowed the current sink and
uncovered an additional current source in the stratum radiatum, thus providing a
better spatial map of the synaptic current responses along the dendrites than the raw
FPs. The population spike sink was observed to initiate basal to the cell layer, which
is where the pyramidal axon hillocks are located and in agreement with published
findings (Richardson et al., 1987). The FP and CSD recording obtained from
cMEA#1 were consistent with the literature thereby demonstrating the functionality
of such high-density pMEAs.
The 50 µm inter-electrode distance made the pMEAs fit for CSD analysis, as
previous simulation results indicated that this is the minimum acceptable spatial
resolution for CSD analysis in hippocampus in vitro (Wheeler and Novak, 1986).
This theoretical limit on the inter-electrode distance was explored by comparing the
effects of lower spatial resolution on calculated CSD. CSD analyzed from data
obtained with 20 electrodes showed narrower sinks and sources when compared to
those calculated from data from two 10 electrodes subgroups (Figure 3.3). Slight
differences between Odd and Even subgroups are probably due to variations in
electrical activity along the extent of the pyramidal cells. This observation points to
the importance of electrode placement in FP recordings, and alludes to the variability
that can be expected from single electrode recordings. The location of synaptic
83
responses was thus more accurately delimited by mapping current sinks and sources
calculated from FP recorded by arrays with ≤50 µm inter-electrode distance.
CSD of multiple laminar profiles of pyramidal cells in CA1 were obtained
through sequential monopolar stimulations along one peripheral column of cMEA#1.
The profiles obtained mapped afferent input distribution onto CA1 pyramidal cells.
Basal side stimulation generated a sink at stratum oriens and a source around the
pyramidal layer, which was of the opposite polarity to when the stimulus was in
stratum radiatum. Exciting the cell body layer directly evoked a smaller response,
which is in accordance with its lower number of synaptic inputs. These results are
consistent with the neuroanatomy of the CA1 region (Andersen, 1975), where SchC
and commissural fibers project to both stratum radiatum and oriens forming two
excitable pathways (Buzsaki and Eidelberg, 1982; Richardson et al., 1987).
To investigate the independence of these two pathways, pp stimulations were
performed on the same dentritic side (cis configuration) or at locations on opposite
sides of the cell body layer (trans configuration). Apical pulses did not have a
facilitatory effect on basal stimulations, and vice versa, suggesting that the two
afferent pathways have independent effects at the pp interval used. (Figure 3.5)
While this observation may not be true for all sites, time intervals or in vivo, it
indicates that apical and basal dendrites can be easily stimulated independently using
this high-density array for different investigational purposes (Cash and Yuste, 1999;
Courellis et al., 2000).
84
Section 3.3.3 demonstrated such fine control over the stimulation paradigm
through an experiment with cMEA#3. There two adjacent afferent pathways were
selectively stimulated and differentiated using high-density arrays designed to match
the anatomical area. Each of these sub-pathways encompasses approximately 100
µm, therefore according to the Nyquist theorem, the 50 µm inter-electrode spacing
enabled selection of accurate stimulation sites in each sub-pathway. Medial and
lateral PP were selectively stimulated and discriminated based on their pp responses.
Lateral PP exhibited 113% pp facilitation, while medial PP showed pp depression of
87%. The high-density cMEA thus provided an easy and rapid method to selectively
stimulate adjacent pathways or their subdivisions in the same slice. This ability was
not previously possible with sharp electrode recordings or lower density MEAs, and
therefore enables more careful pathway selections in future experiments
(McNaughton and Miller, 1984; Dahl et al., 1990).
3.5. Conclusion
Chapter 2 illustrated advantages of pMEAs methodologies for extracting rich
data from brain slices. However, rapid transitions of cell types and nonsymmetrical
organization in the brain mandate high spatial resolution sampling to obtain an
accurate image of activity. The limited number of recording channels currently
available requires conformal MEA designs that place dense electrode patches at
areas of interest. Experiment specific cMEAs can be designed to conform to any
tissue cytoarchitecture. This chapter described the design and application of
conformal high-density cMEAs for in vitro electrophysiological experiments with
85
acute hippocampal slices. These cMEAs enabled precise and convenient selection of
stimulation and recording sites, while providing sufficient spatial resolution for CSD
analysis. Combining the methodologies of chapter 2 with the capabilities of
conformal high-density cMEAs provides a more versatile and powerful tissue
interface for biosensor and neuroprosthetic applications (Berger et al., 2001;
Gholmieh et al., 2001; Shimono et al., 2002).
86
4. SUBSTRATE COATINGS CONTROL NEURON MORPHOLOGY AND
PHYSIOLOGY ON MEAS
4.1. Introduction
Control over the interface between neurons and MEAs is crucial for the
development of viable electrode-based neural prostheses (Rutten et al., 1999; Berger
et al., 2001), cell-based biosensors (Gross et al., 1995), and the creation of designer
neuronal circuits for the study of network dynamics (Stenger et al., 1998; Maher et
al., 1999a). For these purposes, a diverse set of MEAs has been developed for
recording the activity of neuronal populations, including microwire arrays
(Deadwyler et al., 1996; Williams et al., 1999), silicon or ceramic based implantable
arrays (Hetke et al., 1994; Nordhausen et al., 1994; Bragin et al., 2000; Burmeister
et al., 2000), and planar MEAs (Gross et al., 1982; Egert et al., 1998; Oka et al.,
1999). These neuron-silicon interface devices have to overcome several mechanical
challenges from material disintegration in the saline oxidative environment, to
biocompatibility and tissue toxicity, to cellular attachment and encapsulation.
Various surface modifications are being investigated to reduce glial scarring and
increase neuronal interaction with implanted devices (Kam et al., 1999, 2002; Shain
et al., 2003). Most of these approaches involve selective patterning of biologically
active molecules on surfaces to spatially confine cellular attachment (Kleinfeld et al.,
1988; Stenger et al., 1998; James et al., 2000).
Cells must attach to, survive on, and form tight contacts with electrodes in
order to stimulate and record their activity. Dissociated neurons cultured on pMEAs
provide an ideal system for the investigation of this neuron-electrode interface. First,
pMEAs can stimulate and record electrophysiological activity of many neurons
simultaneously, while still enabling microscopic morphological observations and
biochemical studies (Pine, 1980; Jimbo et al., 1993; Wong, 1998; James et al., 2000;
Tateno et al., 2005). Second, dissociated neuronal cultures are simplified neural
tissue preparations that can be grown in a highly controlled environment and still
form synapses and develop into interconnected networks retaining physiological
activity and plasticity (Droge et al., 1986; Kamioka et al., 1996; Fitzsimonds et al.,
1997; Banker and Goslin, 1999). At the neuron-electrode interface, it is the
properties of the substrate covering the pMEAs that determine the degree of neural
attachment, motility, growth, and survival (Kleinfeld et al., 1988; Ignatius et al.,
1998).
Figure 4.1: Chemical structure of
polycationic substrates. a) PDL, b)
PO, and c) PEI. All three polymers
have amine groups (NH
2
) which are
positively charged at physiological
pH, however, both PDL and PO
additional have carboxyl groups
(C=O) which contribute a partial
negative charge.
This paper describes the relationship between cell attachment, network
morphology, and electrophysiological activity on pMEAs. To this end, dissociated
hippocampal neurons were cultured on top of pMEAs coated with three commonly
used biochemical substrates: Poly-D-Lysine (PDL), PolyOrnithine (PO), and
87
88
PolyEthylenImine (PEI) (Figure 4.1). PDL and PO are polycationic molecules that
were first used to allow cells, which have an overall negative surface charge, to
attach to similarly charged glass surfaces (Yavin and Yavin, 1974; Letourneau, 1975;
McKeehan and Ham, 1976). Their synthetic nature allowed them not to interfere
with cell physiology, while PDL’s D- enantiomeric configuration enabled it to
withstand proteolytic cleavage. PO is similar to PDL in structure and charge, but has
the amine group closer to the backbone. PEI, lacking the slightly negatively charged
carbonyl moieties of PDL and PO, is therefore a more positively charged
polycationic organic molecule that has been reported to allow neurons to mature
faster (Lelong et al., 1992). In addition to these three synthetic molecules, a
solubilized basement membrane (BM) extract was also used as a culture substrate.
BM is a reconstituted extracellular matrix that consists of laminin (56%), collagen IV
(31%), entactin (8%), heparan sulfate proteoglycan, matrix metalloproteinases and
various growth factors: EGF, bFGF, NGF, PDGF, IGF-1, and TFG-B. The structure
and activity of reconstituted BM is reported to be similar to in vivo extracellular
matrix (Kleinman et al., 1986), and its laminin is known to promote axonal growth
(Sanes, 1989; Matsuzawa et al., 1996b; Esch et al., 1999). These four substrates
were combined to yield a total of eight different substrate conditions (Figure 4.2),
onto which hippocampal neurons were cultured. The effects of these substrate
combinations on neuron physiology was assessed first morphologically, by
comparing cell attachment, and branching patterns, then immunohistochemically, by
staining axons and dendrites, and finally electrophysiologically, by analyzing
89
neurons’ spontaneous activity patterns. Finally, network morphological features and
activity parameters were correlated to establish relationships useful for
understanding effects of surface modification in developing coating materials for
implantable electrodes and for creating designer neuronal networks on MEAs.
4.2. Materials and Methods
4.2.1. Substrate Preparation
Glass coverslips and pMEAs were cleaned with a mild detergent, thoroughly
rinsed with distilled water, and sterilized with alcohol and UV light. The following
substrates were dissolved in 1 mM sodium borate buffer pH 8.5: 1-10 mg/ml PDL, 1-
10 mg/ml PEI, 1-10 mg/ml PO. The coverslips or MEAs were coated with 0.5 ml of
these substrates per cm
2
. After an overnight incubation, surfaces were rinsed with
borate buffer several times and either coated with 250 µg/ml BM (Matrigel, Beckton
Dickinson), or kept in buffer overnight. Surfaces were rinsed again after this second
coating and stored in buffer or culture medium until shortly before culture, at which
time they were dried until cell seeding. Eight surface conditions were thus created
(Figure 4.2).
4.2.2. Cell Culture
Neurons are cultured according to a slightly modified protocol from the one
described in (Banker and Goslin, 1999).
Pregnant Sprague-Dawley rats were sacrificed by CO
2
inhalation 18 days
after plugging. Following decapitation, the embryonic brains were removed from the
skull and placed in cold calcium and magnesium-free Hank’s salt solution (HSS)
90
buffered with HEPES pH 7.3. Hippocampi were then dissected under a stereo
microscope and kept in cold HSS. Meningeal membranes were removed during
three HSS rinses, and then the hippocampi were mechanically dissociated by
repeated pipetting through a series of fire-polished Pasteur pipettes with decreasing
tip diameters. Dissociated cells were then passed through a 40 µm cell strainer to
remove large clumps. The cells were mixed in Neurobasal medium (NBM,
InVitrogen, Carlsbad, USA) supplemented with 0.5 mM glutamine, B-27 supplement
(InVitrogen), Penicillin-Streptavidin, and 25 µM glutamate. Cells were then plated
at a density of ~100,000 cells/cm
2
on coated coverslips or MEAs. Cultures were
kept in a humidified incubator at 37 °C in 10% CO
2
and medium was replaced
biweekly with NBM without additional glutamate. Phase contrast photomicrographs
were taken before feedings, and recording started after two weeks in culture. Cells
survived for 3-4 weeks under these conditions. Cell survival was monitored by
conducting LDH assays (Boehringer Mannheim, Germany) to obtain a measure of
cell death.
4.2.3. Image Analysis
Cell cultures were taken out of the incubator periodically, and phase contrast
photomicrographs were digitally acquired (DMIRB microscope, Leica, Germany;
SPOT RT camera, Diagnostic Instruments, Michigan, USA) before feeding or
recording. Pictures were renamed to blind the conditions and visually scored
according to the following morphological features on scales from 1 to 5, with half
point increments:
91
Individual Cells: This feature of cell cultures ranked the degree to which cell bodies
remained individually distinguishable. A score of 5 indicated that cell bodies were
completely separate from each other, whereas 1 reflected that no single cell body
was discernable.
Cell clusters: Almost corollary to the “Individual Cells” measure, this feature ranked
the degree to which cells aggregated into clusters and the latter’s size. In a culture
with a score of 5, most cell bodies have clumped into very large clusters, whereas
one with a score of 1 had no clumps at all.
Thin branches: This measure assessed the abundance of fine branches in the culture.
A 5 was assigned to cultures where the entire surface was covered with thin
branches, and 1 was for cultures which had no thin branches.
Fasciculation: Almost corollary to the “Thin Branches” measure, this feature ranked
the degree to which branches aggregated into fascicles and the latter’s thickness. A
score of 5 was assigned to cultures that had many thick fascicles, whereas 1 was
given to samples that had no fascicles.
The phase micrographs were scored blindly according to these measures by
the first two authors. After the blind was removed, Kruskal-Wallis statistical
analysis was conducted on the data sets, along with Wilcoxon rank sum pair-wise
comparisons to assess significant differences between conditions.
A linkage dendrogram was plotted in MATLAB based on similarity between
pairs of median values of each morphological feature for each of the conditions. The
92
dendrogram was constructed from a linkage matrix generated by calculating the
distance between the centroid of the Euclidian distances of each parameter.
Surface areas covered by branches and cell bodies were measured with
SigmaScan Pro software (SPSS, Illinois, USA). Since branches are dark in phase
microscopy, and somas are circled by white halos, different cell parts were selected
by brightness values of individual pixels. The measure of branch surface coverage
was thus obtained by visually setting a dark intensity threshold and selecting and
counting all pixels that were darker. Soma surface area was calculated as the sum of
pixels that were brighter than another visually-defined threshold after applying
image analysis filters to remove small debris particles (2-3 pixels in size), expand
selection perimeter, and fill-in halos in order to delimitate the somas more
accurately. While threshold values were determined specifically for each image,
filters were applied uniformly to all. Statistical significance of measured values was
assessed by one-way ANOVA and pair-wise comparisons.
4.2.4. Immunohistochemistry
Dendrites and axons were immunohistochemically stained with antibodies
specific to microtubule associated protein-2 (MAP-2, clone AP-20 from Labvision,
CA, USA), and growth associated protein-43 (GAP-43, clone GAP-7B10 from
Sigma, MO, USA), respectively. Staining protocol was an adaptation of the one
described in (Banker and Goslin, 1999). Briefly, cells were fixed with 4%
paraformaldehyde for 10-15 min, and permeabilized with 0.25% Triton-X for 5 min.
Non-specific binding was blocked with 10% Bovine Serum Albumin (BSA, Roche
93
Applied Science, USA) in PBS for 30 min. Fixed cells were then incubated with
primary antibodies diluted in a 3% BSA in PBS solution for 2 hours at 37 °C. After
rinsing, secondary antibodies (FITC labeled anti-IgG1, and Texas Red labeled anti-
IgG2a from Rockland Immunochemicals, PA, USA) were applied to the cells for 1
hour in the dark at 37 °C. Following another rinsing, stained cells were finally
mounted with a polyvinyl alcohol and glycerol mountant containing 1% DABCO
(Sigma). Fluorescent photomicrographs were taken using a DMIRB inverted
microscope equipped with a Spot RT camera. This labeling protocol yielded green
fluorescent dendrites and red axons.
Surface coverage of fluorescent dendrites and axons was quantified using
SigmaScan Pro image analysis software. First, color channels were separated into
red and green images. Then, pixels brighter than a visually-set background threshold
were counted as dendrites or axons depending on their color channel. Lastly, one-
way ANOVA was run on measured surface area values to determine statistical
significance.
4.2.5. Electrophysiological Analysis
Two weeks after cell seeding, electrodes were taken out of the incubator and
photographed. They were set into the MEA-60 (Multi Channel Systems, Reutlingen,
Germany) pre-amplifier box, which allowed simultaneous recording from the 60
electrodes of a pMEA. Arrays were covered with a small cap traversed by a
perforated tube that provided a slow flow of humidified 10% CO
2
and 90% O
2
gas
94
mixture. The chamber was maintained at 37 °C. The first five minutes of recorded
spontaneous activity were analyzed for each pMEA cell culture.
Extracellular spike trains were detected by conservatively setting a threshold
value just above a visually determined background noise for each recording channel.
Whenever the field potential crossed that threshold, a timestamp was recorded along
with a 3ms spike waveform sampled at 20 or 25 kHz (1 ms pretrigger and 2 ms post-
trigger). Recorded waveforms were used for spike sorting using OfflineSorter
(PlexonInc, TX). Waveforms were also used to discard spikes that were less than
four times the signal-to-noise ratio. This re-thresholding was accomplished in a
custom Matlab script (Mathworks, MA). Noise was calculated as the smaller of the
root-mean-square value of all 1 ms pre- or all 1 ms post-trigger segments for a unit.
The contrasted signal was the peak-to-peak spike amplitude. This re-threshholding
script also allowed manual assignment of overlapping spikes to multiple units. The
last pre-processing step was elimination of all units that had less than 10 spikes over
the entire recorded period. This filtering reduced the effect of the large number of
silent channels on parameters of interest. Sorted spike timestamps were then
imported into NeuroExplorer (Nex Technologies, MA) for analysis. This program
calculated spontaneous firing and bursting parameters such as mean firing rates, the
median of ISI distributions, mean burst rates (where bursts are calculated according
to a surprise method described in (Legendy and Salcman, 1985)), and unit cross-
correlations. A normalized index of this last measure was used as a measure of
firing synchrony. Correlation indices were calculated as the mean of the cross-
95
correlation over a ±200 ms window minus the mean of the entire cross-correlation
multiplied by the number of bins and divided by the number of reference events.
This normalization produces a non-symmetrical correlation matrix that is however
more appropriate for large spike trains. Distributions of these five
electrophysiological parameters between cultures on PEI and PO+BM were
compared with a Kolmogorov -Smirnov test. In addition, the relation between
morphological features and electrophysiological measures were statistically
investigated with Spearman rank correlation coefficients.
4.3. Results
Whether cultured on glass coverslips or MEAs, hippocampal neurons
attached and developed morphologies that were reproducibly characteristic of the
coating substrate on which they were cultured. Figure 4.2 shows photomicrographs
of several cultures on different substrates after 17 days in culture. These cultures
were plated at a density of ~100,000 cells/cm
2
. This density increased the likelihood
of cells settling on top of electrodes, and the lifespan of the cultures, since
dissociated cells do not survive as well at lower densities. Nonetheless, this density
does not create a confluent monolayer of cells, thus enabling the distinction of cell
bodies and processes.
Figure 4.2: Phase micrographs of neurons cultured on MEAs coated with the eight different substrate
conditions: a) MEA with NO substrate yielded no cell attachment; these white balls were floating
clumps weekly anchored to the surface by frail branches. This image scored a 1.5, 4, 1.5, 1.5 on the
individual cell, cluster, thin branching, and fasciculation parameters, respectively. ( b) Cells on PDL
coated MEA had extensive thin branching, and a few small clumps and fasciculated branches. This
image scored a 4, 2, 4.5, 2.5. c) PEI scored 5, 1, 5, 1 as it had the most distinguishable individual
cells and thin branches while having no clumps or fascicles at all. d) PO got 1.5, 4, 3, 4, as it had
some clumps and fasciculation while maintaining considerable thin branching. e) BM alone formed
large cell clumps and fasciculated branches, with many thin branches and thus scored 1.5, 5, 4, 4 f)
PDL+BM had more clumps and fascicles than PDL alone and scored 2.5, 3.5, 4, 4 g) PEI+BM formed
a few clumps and fasciles and got 3, 3, 5, 3 while h) PO+BM had the most extensive fasciculation and
clumps while maintaining considerable amount of individual cells and extensive thin branching and
scored 3, 5, 4, 5. Scale bar 100 µm for all images.
Neurons did not attach at all, or very poorly at best, when the glass or MEA
surface was not coated with any substrate (Figure 4.2A). Any remaining cell bodies
clumped and floated, while the few branches that extended out of them were frail,
thin but fasciculated and elongated. Hippocampal neurons did however attach on
coated surfaces and develop networks varying in morphological features. Networks
gradually developed and accentuated their morphological characteristics over weeks
in culture. Branching patterns in the first week were not clearly distinguishable
between conditions, while cells bodies had already started to form clusters on most
substrates except PEI. In order to track cell growth and motility over time, pictures
96
97
of cells cultured on MEAs were taken at different DIV. Aligning and linking these
pictures into time-lapse movies revealed that cells on PEI did not move once settled
on the surface, while those on other conditions were more motile, and moved slowly
along fasciculating branches to form clusters. Development of neuronal cultures on
different substrates can be traced in supplementary movies. (PDL.mov, PEI.mov,
PO.mov, PDL+BM.mov, PEI+BM.mov, PO+BM.mov)
During the second week in culture, networks developed characteristic
morphological features: cells formed small clusters and some processes fasciculated
on PDL and PO (Figure 4.2B and D), whereas PEI kept cell bodies distinct and
prevented clustering and branch fasciculation (Figure 4.2C). Addition of a BM coat
to the surfaces (Figure 4.2E-H) induced more cell clustering and process
fasciculation than each of the underlying substrates alone. BM also supported
attachment and growth of small numbers of glial cells, which did not develop in non-
BM conditions. Though networks of cells varied in arrangements and local density
on different sections of the culture surfaces, overall morphological features examined
were very stereotypical for the conditions and reproducible between cultures.
Morphologies of networks growing on the various substrates were ranked on a scale
from 1 to 5 for each of four physical features: individual cells, clustering, thin
branches, and fasciculation.
Figure 4.3: Morphological feature
scores for the various substrates and
their grouping. Median and SEM of a)
Individual cells and Clumps, b) Thin
Branches and Fasciculation plotted for
each substrate. c) Linkage dendrogram
classifying the substrates into four
groups based upon significant
differences in their morphological
features. The height of the lines relates
to the degree of difference between
groups, and the horizontal dashed line
delimits statistical significance, where
all junctions occurring above the line
indicate significantly different
conditions on one or more
morphological features. The numbers
in c) are sample numbers for a) and b)
as well.
Though the sample images in Figure 4.2 and their scores are representative of
the morphological patterns observed on each of the substrate conditions, there was
some variability between images whether from the same dish or different cultures.
Therefore images taken between 13 and 18 DIV were grouped for statistical analysis,
as these had developed their characteristic morphologies. Scores were averages from
two or more scorers, since the difference between their scores was smaller than the
standard deviation of same substrate images. Median scores for each morphological
feature across conditions and their standard error of the mean are plotted for each
98
99
condition in Figure 4.3. Kruskal-Wallis statistical analysis of these ranked values
indicated that substrates did have a significant effect on each scored morphological
feature (p<0.01). Wilcoxon rank sum pair-wise analysis was then conducted to
compare the substrates to each other in each of the features. Uncoated samples and
PEI coated ones were the most different conditions from the rest. The uncoated
samples had significantly less cells and branches attached after two weeks than any
of the conditions. PEI samples, which also had almost no clumps or fasciculation,
did, however, maintain significantly more individual cells than all the other
conditions. The other conditions were significantly different from each other on
some features but not others. The most apparent trends were the effects of addition
of a BM coat. BM induced significantly more branch fasciculation when added to
any substrate, and yielded significantly more clumps in all but PDL coated samples,
where the trend was still evident though not significant. The corollary features were
also affected, where BM significantly decreased individual cells compared to the
same polycationic substrate alone, and it significantly decreased thin branching in all
but PEI coated samples. There were neither significant differences between PDL
and PO, nor between PDL+BM and PO+BM in any of the ranked features. These
pair-wise comparisons revealed that the eight substrate coatings yielded four
morphologically distinguishable groups.
Figure 4.4: Image analysis of
surface coverage of cells and
branches. Phase contrast micrograph
of neurons cultured on a) PDL and b)
PEI. Image analysis of the same
pictures is shown below with the
branches overlaid in green and the
cell bodies in red. The branch area
was 17% in a) and 29% in b), while
the cell area was 11% covered in a)
and 23% in b). Scale bar 100 µm c)
Bar graph of the median branch
(black) and cell (white) areas for four
substrates and SEM.
100
101
In order to visually classify the substrates according to their morphological
similarity, a linkage dendrogram was generated (Figure 4.3c). The dendrogram
arranged substrates into groups based on the closeness (Euclidian distance) between
median values of the morphological parameters. According to the dendrogram,
uncoated surfaces and PEI coated ones produce the most divergent morphologies.
The other six conditions are arranged into two families of three, where BM, PO+BM,
and PDL+BM form a group that is significantly different from the one consisting of
PDL, PEI+BM, and PO. The horizontal dashed line is drawn to indicate statistical
significance, where all junctions occurring above the line indicate significantly
different conditions on one or more morphological features.
Surface areas covered by cells and branches cultured on the various
substrates were quantitatively measured in phase contrast photomicrographs. This
analysis took advantage of the fact that cell bodies produce a white halo and
branches appear dark in these images. Two sample images are shown in Figure 4.4,
where the branch area selection is highlighted in green and the cell body area in red.
These image analysis examples illustrate this method’s ability to accurately and
reproducibly identify most of the branches and cell bodies in a phase-contrast image,
regardless of their morphologies and quantify their percent surface coverage.
Several pictures from different cultures on glass cover slips were analyzed for the
various non-BM conditions; BM conditions were not analyzed due to the appearance
of light gray deposits during coating. These deposits, assumed to be protein and
collagen clumps, interfered with branch area measurements. Median values of the
102
percent image area covered by branches and by cells are plotted in Figure 4.4E along
with their respective standard errors. ANOVA statistics yielded p<0.05 for a
significant difference between the four groups for the percentage coverage of
branches or cells, or their sum but not for their ratio. This group significance was
however due to the large difference between uncoated surfaces and those with
substrate. Pair-wise comparisons between PDL, PEI, and PO were not significant as
there was a large variability between surface values which was not related to
substrate, age, nor seeding density.
To investigate effects of the eight substrate conditions on branch growth and
differentiation, surface coverage of axons and dendrites was measured. Axons were
labeled with anti-gap-43 antibodies, dendrites with anti-map-2, and these primary
antibodies were counterstained with Texas Red- and FITC-ligated secondaries
respectively (Figure 4.5). Axons, in red, appeared to follow the morphological
trends described above, whereby they were fasciculated on PO+BM, very finely
branched on PEI, and intermediate on PDL. Dendrites appeared less affected by the
substrate in terms of thickness and fasciculation, but they did wrap around axonal
fascicles and intermingle in cell clumps. The median area coverage of dendrites and
axons on PDL, PEI, and PO is plotted in Figure 4.5E with the respective standard
errors. ANOVA statistics on the axonal or dendritic areas, and their sum or ratios
yielded p>0.05, indicating no significant difference between these groups, probably
due to a large variability between surface values which was not related to substrate,
age, nor seeding density. However, in all cases, axons occupied more surface than
dendrites.
Figure 4.5: Immunohistological
staining of axons and dendrites and
measurement of their surface areas.
Double-stained images showing
axons counter-stained with Texas
Red (red), and dendrites with FITC
(green), for neurons cultured on a)
PDL, b) PEI, c) PO, and d) PO+BM.
Each image was separated into its red
and green color channels (below each
original) in order to measure the
surface area brighter than the black
background. Scale bar 100 µm. e)
Bar graph of the median percentages
of image surface area covered with
dendrites (white) and axons (black)
and SEM for four substrates.
Spontaneous electrophysiological activity of hippocampal neurons growing
on the various substrate-coated MEAs was recorded during their third week in
culture. The two conditions that produced the most different morphological features,
103
namely PEI and PO+BM, were compared. Inspection of a few overall activity
measures revealed consistency with some of the above morphological observations
(Table I). 21% of all electrodes recorded on either substrate had spike activity,
which is in agreement with their similar cell surface coverage. After noise filtering
and spike sorting, the PO+BM condition had more units per channel, consistent with
its large cell clusters on top of multiunit electrodes, whereas cell bodies were more
dispersed on PEI (Fig 6a and 6f).
Figure 4.6 illustrates these observed differences with two units from each
condition. Figure 4.6B and G contrast spike waveforms and amplitudes, where
PO+BM had more units per channel and their amplitude was often larger. Whereas
average firing rate was slightly greater on PEI, firing patterns on the two substrates
were qualitatively very different. Rastergrams (Figure 4.6C and H), revealed the
essence of the difference in activity: neurons cultured on PO+BM tend to fire in
synchronous bursts, whereas those grown on PEI tend to fire to the beat of their own
rhythm sporadically or continuously. ISI histograms (Figure 4.6D and I) supported
this bursting observation by showing a bimodal ISI histogram for PO+BM consisting
of a narrow distribution of short ISIs (mode 0.0095 and 0.0045 sec for 48c and 61 a
respectively), accounted for by high intra-burst firing rates, followed by a smaller
and broadly spread peak of longer ISIs corresponding to inter-burst rates (mode
104
0.0435 and 0.0455 sec). By contrast, PEI’s histograms are monophasic and represent
sporadic (45b, mode 0.21 sec) or continuous (83a, mode 0.041) firing with spreads
intermediate to PO+BM’s dual peaks. Lastly, synchrony between active units on a
particular MEA is revealed with color plot of correlation indices (Figure 4.6E and J).
A high index value meant that a channel on an array (y-axis) fires often within 200
ms of the other (x-axis), whereas a low index, in blue, reflects asynchronous firing.
Negative correlation indices, which reflected the likelihood of units to fire outside a
200 ms interval from the other (i.e. asynchronously), were always small in
magnitude (<0.25) indicating that there were no delayed correlations. The color
graph plots high correlation indices in hot red, and low valued ones in blue, at the
intersection between their respective units which are on the x- and y- axes. The
activity of networks exhibited more correlated firing on PO+BM than on PEI.
105
106
Figure 4.6: Analysis of electrophysiological activity parameters from two sample PO+BM and PEI
cultures. a) Neurons cultured on MEA coated with PO+BM and f) with PEI. Scale bars 100 µm. b)
Spike waveforms of sorted units recorded at channels 61 and 48 (3 units) in a), and g) from channels
45 (2 units) and 83 in f). c) and h) Raster plots of spike timestamps of all the units for 2 minutes in a)
and f) respectively. d) and i) Interspike Interval (ISI) histograms of the activity of units 61a and 48c
(green) shown in b) and 45b (red) and 83a shown in g) Insets are same ISI histograms extended to 10
sec. e) and j) Correlation index matrix for all the units in a) and b) where the index of each
correlation pair is plotted as a color according to the color bar at the coordinates specified by the two
units being compared. Autocorrelations are set to 0. Selected channels are highlighted in yellow.
107
Variability between unit activity patterns, even on a same MEA, was large
and its distribution too skewed to enable mean or median based statistics.
Distributions of electrophysiological parameters were therefore plotted for both
substrate conditions, and statistically compared with Kolmogorov-Smirnov tests
(Figure 4.7). The distributions were significantly different for all examined
parameters: mean spike amplitudes (Figure 4.7A), mean firing rates (Figure 4.7B),
median ISIs (Figure 4.7C), mean burst rates (Figure 4.7D), and correlation indices
(Figure 4.7E).
Figure 4.7: Distribution of electrophysiological parameter values for PO+BM and PEI. Distributions
of a) unit mean spike amplitude distribution, b) units’ mean firing rate, c) units’ median ISI, d) units’
mean burst rate, and e) correlations indices are all significantly different between PO+BM and PEI
(p<0.05).
Finally the question of whether morphological features are related to
electrophysiological parameters was addressed with a Spearman rank correlation
matrix. Figure 4.8 is a color plot of Spearman’s correlation index (rho), where
positive correlations are red and negative correlations are blue, and statistically
insignificant correlations are hashed. The more clumping and fasciculation a
neuronal network had, the more synchronous firing and faster bursting it showed,
although its mean firing rate dropped. For the case of individual cells and thin
branching, they were negatively correlated with synchronous activity and bursting,
indicating that individual cells were less likely to burst in synchrony, despite having
108
higher mean firing rates. In addition, thin branches were also negatively correlated
with smaller spike amplitudes.
Figure 4.8: Correlation between
morphological features and
electrophysiological parameters. Color
graph of Spearman Rank Correlations’
rho values between the five
electrophysiological parameters (mean
spike amplitude, mean firing rate, median
ISI, mean burst rate and correlation
indices) and the four morphological
features (individual cells, clumps, thin
branches and fasciculations). The colors
indicate the strength and direction of
correlation, where blue values are
negatively correlated and yellow to red
values are positively correlated. Green
colors represent weak correlations, but
only the hashed squares were not
statistically significant (p<0.05).
4.4. Discussion
Brain cytoarchitecture is closely related to its physiological function. Some
nuclei have densely packed neurons while others are sparsely populated, and some
areas are highly interconnected with extensive branching while others pass
fasciculated axonal bundles across long distances. (REFs) The electrophysiological
activity of these neurons is explored in terms of their intrinsic activity or in relation
to their position in the network, or as a function of a behavior. In this manuscript,
the relationship between network architecture and activity was explored in a
simplified dissociated neuron system with similar cell populations and no
confounding behavior.
Cultured dissociated hippocampal neurons have proven very useful in
numerable microscopic, physiological, and biochemical studies of cell survival,
109
110
growth, morphology, and electrophysiology. Cell cultures provide a simplified
system in a highly controlled environment amenable to specific manipulations.
Hippocampal neurons cultured from E18 embryonic rats are reported to form a
mostly homogeneous population consisting of 85-95% pyramidal neurons in the first
week in culture (Banker and Goslin, 1999), however, cursory staining for inhibitory
interneurons revealed a constitution of more than 20% interneurons at the third week
in culture (data not shown). Nonetheless, by using a serum-free growth medium
developed for neuronal cultures (Brewer, 1995), fixing cell density and maintaining
regular feeding schedules, culture variables were kept identical across all
experimental conditions. Substrate concentration was observed to affect morphology
slightly, but not in the range used (data not shown).
Culturing dissociated hippocampal neurons on the eight different substrates
preparations yielded networks with different morphological features. While it was
expected that neurons would not attach to uncoated glass surfaces, the difference in
adhesion to substrates was not anticipated. Most notably, cell somas hardly moved
after seeding on PEI whereas they exhibited varying degrees of motility on other
substrates (as seen in the supplemental time-lapse videos online), presumably due to
PEI’s strongest cationic charges. This variation in attachment and motility led to
network morphological features that were classified into four significantly different
groups (Figure 4.3E). Cells did not attach well nor developed branches in the first
group cultured without any substrate. The second group, with only PEI in it, formed
networks whose neurons remained separate and whose branches were thin but
111
extensive. The third group consisted of BM, PO+BM, and PDL+BM, and yielded
large clumps of cells and fasciculated branches. PDL, PO, and PEI+BM formed the
last group, which yielded intermediate networks with both small clumps and
individual cells, and thin branches along with fascicles. These groups demonstrated
a strong relationship between culture morphology and adhesion to substrate.
Basement membrane increased motility, fasciculation, and clumping on all
polycationic substrates, presumably by shielding their charge and providing a more
natural surface for neurons to attach to (Goldberger et al., 1997; Dityatev and
Schachner, 2003). Few glial cells were also observed in BM containing conditions,
probably supported by its growth factors. These quantified morphologies were very
consistent and became apparent in the second week in culture and were independent
of variations in cell density, within the range investigated. This dependence of
morphology on substrate enables control of network architectural types through the
selection of underlying substrates.
The development of neurons and differentiation of neurites into axons and
dendrites was examined immunohistochemically and by image analysis. This
analysis provided quantitative measures of surface coverage, which however had
several challenges and limitations. First, intensity thresholds had to be visually set
for each picture, whether for dark branches, white somatic halos, or fluorescent
branches, as lighting conditions varied depending on the depth of medium over cells
and on exposure settings. The value of this threshold affected the measurement in
direct relation to the steepness of the image intensity distribution histogram at that
112
point. Comparison of two experimenters’ selections, however, showed less
difference in the results between images than between conditions (data not shown).
In addition, light intensity across imaged fields of view needed to be even, otherwise
the selection thresholds would be too stringent on one side of the image and not
selective on the other. Several images were thus analyzed in sections in order to
correct for uneven lighting. Local contrast and simple edge detection algorithms
were extremely hard to implement for these images as they had both large and small
features that varied in contrast gradients. Furthermore in phase contrast pictures,
fasciculated branches often had white halos that were selected as cell surface,
whereas some cells bodies were dark and flat and were selected as branches, along
with the grey deposits BM produced. In fluorescent image analysis, double-staining
was successful in selectively distinguishing axons from dendrites in all substrate
conditions, however, somas were often stained by one or both antibodies. The large
variability in fluorescent area measurements could be due to the 40x magnification
used for photomicrography, which created brighter pictures but did not sample large
areas of the cultures. Finally, all collected measurements were from two-
dimensional pictures, while cultures had cell clumps and fasciculated branches that
were stacked in three dimensions.
These challenges and limitations contributed to a large variability in surface
measurements, which therefore did not allow rejection of the null hypothesis: There
was no difference in surface areas covered by somas or branches, whether axons or
dendrites, between the four non-BM containing conditions. This negative result
113
however suggested that substrates did not affect cellular development or
physiological properties of neurons, but merely their arrangements. It has been
shown that laminin, which constitutes 70% of BM, promotes axonal differentiation
and growth in young culture (Sanes, 1989; Matsuzawa et al., 1996b; Esch et al.,
1999), however, this effect has not been described in older preparations. Preliminary
results for BM-containing conditions did indicate such a trend of increased axon to
dendrite ratio, however data were not significant.
Electrophysiological activity of the two most morphologically different
networks was compared, those coated with PEI and PO+BM. Despite extremely
divergent morphologies, both conditions had an average of 21% of the MEA
electrodes recording activity. This was in agreement with surface measurement
results which indicated no difference in surface coverage between substrate
conditions. In contrast, channels of PO+BM were more likely to have multi-unit
activity, consistent with the observation that most of the activity came from
electrodes that were directly under large clumps. Spike sorting was therefore
required in order to extract meaningful measure from the recorded data. The quality
of spike sorting was evaluated by looking for the distinctive 1 ms refractory period in
unit autocorrelograms. The five electrophysiological parameters examined for all
units, namely mean spike amplitude, mean firing rate, median ISI, mean burst rate,
and correlation index, all had considerable variability and were not normally
distributed. Data distributions were therefore compared and indicated that all five
parameters were significantly different between the two substrate conditions.
114
Neurons grown on PEI formed networks that fired sporadically or continuously at a
faster mean rate, but always asynchronously, whereas networks formed on PO+BM
tended to fire in synchronous bursts.
These different firing patterns can be explained in the context of their
networks morphological features. Recorded extracellular spike amplitudes depend
on the proximity of axon hillocks to recording electrodes and on the coupling
between somata and electrode tips (Claverol-Tinture and Pine, 2002). Though units
on PEI and PO+BM had significantly different mean spike amplitude distributions,
only thin branching was significantly correlated to smaller spike amplitudes. This
might be due to currents carried by thinner branches being smaller, or to incomplete
electrode coupling. Though large clumps would provide a better seal with
electrodes, the gain might be countered by the separation of active neuron in the
clump from the electrode surface by other non-active cells. The increased mean
firing rate associated with PEI is mostly due to a few very rapid firing units.
However, even when these outliers were eliminated, the distributions of mean firing
rates were still significantly different. This finding therefore suggests that clumping
somehow inhibits the otherwise higher intrinsic firing rate of neurons, while
inducing synchronized bursting. The distribution of median ISIs is bimodal and
wider for PO+BM than for PEI (Figure 4.7C), as bursting creates stretched ISI
histrograms for each unit. Figure 4.6D and I show that the PO+BM histogram peak
is at shorter intervals due to high firing rate inside bursts, but the histogram extends
to longer intervals that correspond to interburst intervals. This latent ISI distribution
115
extends beyond the one second range (inset Figure 4.6D), and Figure 4.7C indicates
that many median ISIs are greater than 30 seconds long. The median ISI is therefore
related to the mean bursting rate which is positively and significantly correlated to
clumps and fasciculation, and negatively correlated to thin branching, and individual
cells (although the correlation between individual cells and median ISI is not
significant). Similar correlations are also observed for correlation indices, which
were measures of synchronous firing of units within a ±200 ms window. Therefore,
cell clumping and fasciculation led to networks that burst synchronously, whereas
individual cells with extensive thin branching fired asynchronously. This suggests
that signals were able to traverse longer distances when reinforced in bundles and
linked large clusters of neurons.
There were two biases in the data. PO+BM coated MEAs were sometimes
recorded few days later than those coated with PEI, and some PO+BM cultures were
seeded at a slightly higher density. Partial correlations were thus run with reduced
data sets that eliminated outlying cases and normalized the data for age and density.
Correlation strength was slightly reduced in these partials, however significances
were not affected. This indicated that though density and age could indeed affect
firing patterns, the correlations described here were independent of these two
variables.
In order to determine whether the observed correlations between activity
parameters and morphology features are causal or spurious, partial correlations were
run alternatively eliminating either substrate condition entirely. Both of these
116
partials yielded no significant relations, specifying that the observed correlations
were indeed due to differences between substrates. However, variability within each
substrate condition might not have been sufficient to reveal correlations between
morphological features and electrophysiological parameters. Therefore, in order to
resolve whether activity patterns are due to network morphology or to some other
property of the substrates, recordings from MEAs coated with another substrate are
needed.
These correlational results, however, provide information on the effect of
different substrates on the spatial distribution of neurons and their branches, as well
as a relation between wiring and firing in neuronal networks. Understanding these
relations is important for the design of patterned neuronal cultures. Patterning
enables the alignment of cells on planar electrode arrays, which simplifies long-term
electrophysiology, microscopic observation, and data analysis (Matsuzawa et al.,
2000). Furthermore, refined control of neurite pathways and connections allows the
creation of defined circuits for the study of the computational abilities of biological
neuronal networks (Stenger et al., 1998; Esch et al., 1999).
Cell patterning relies on the creation of a contrast between repulsive and
attractive surfaces for cells to attach and grow on. Neurons have a weak negative
overall surface charge and thus prefer positively charged surfaces or extracellular
matrix proteins to adhere to. Using repulsive molecules such as PolyEthylene Glycol
(PEG) or proteins such as Bovine Serum Albumin (BSA) to coat surfaces between
attractant patterns increases neural compliance stringency (Branch et al., 2001).
117
Various silicon microfabrication techniques are being adapted for patterning
molecules on culture surfaces. These methods, which include photolithographic
techniques (Corey et al., 1996), silicone channels (Walter et al., 1987; Baier and
Bonhoeffer, 1992), or stamps (Branch et al., 1998; James et al., 2000; Chang et al.,
2003), are applied to precisely deposit spatial patterns of bio-organic substrates on
culture surfaces. Cultured cells would only attach to areas coated with attractant
molecules. Such cell patterning methods have already benefited the study of
chemical and mechanical effects of specific substrates on in vitro neuronal growth
and axonal guidance (Stenger et al., 1998; Esch et al., 1999). However, whereas
patterning neurons has already been shown to produce synaptically active cultures
(Kleinfeld et al., 1988; Ma et al., 1998; Liu et al., 2000; Matsuzawa et al., 2000) and
to increase the recording yield of neurons on MEAs (Chang et al., 2001), there are
reports that patterning can affect cell function (Singhvi et al., 1994), neurite
outgrowth (Tai and Buettner, 1998), and apoptosis (Chen et al., 1997, 1998). This
report described the effects of surface coating materials on neuronal network
morphology and electrophysiological activity, and revealed relations between them.
The relations between substrate induced network morphological features and
electrophysiological activity described here will therefore enable design of patterned
network with specific activities and prediction of the effects of coating implantable
devices with biologically adhesive or repulsive molecules. For example, patterning
electrode tips with PEI and the surrounding electrode shafts or surfaces with PDL or
PO+BM would allow neurons to attach to MEAs whether planar or implantable, and
118
somas would migrate on the surface until they reached PEI, where they would adhere
tightly and stop moving. This strong coupling at the electrode would also increase
recorded spike amplitude. In addition, branch fasciculation would promote
synchronization of bursting activity which might allow electrically stimulated
responses to propagate further into the network. Patterning of substrates on MEAs
could therefore be used to enhance both recorded and stimulated signals from
neuroprosthetic devices.
4.5. Conclusion
Neurons cultured on different substrates develop networks with distinct
morphologies and activity patterns. Network with neural clusters and axonal
fascicles fire in synchronized bursts whereas those with more individual neurons and
thin branches fire in asynchronous sporadic or continuous bursts. These relations in
morphological and electrophysiological properties can be harnessed for the design of
substrate patterns on MEAs that would enhance recorded and propagated signals. In
the future, the causality between these correlations will be investigated, as well as
their direct relation to electrical excitability and applicability to patterned neuronal
networks, biosensors, and implantable neuron-silicon interface devices.
119
5. CONCLUSION
The evolution of brain prosthetics, neuron-based biosensors, and bio-artificial
networks has focused on the development of an interface between nervous tissue and
computers. MEAs were presented above as ideal candidates for this neuron-silicon
interface. Fundamental to the progress of these MEA applications is an
understanding and control of the interaction between neuronal cells and MEAs’
inorganic materials. The quantitative physiological measures in chapter 4 should
provide some insight into optimal coatings for neuroprosthetics and biosensors, as
well as information on the relationship between morphology and electrophysiology
in two dimensional neural networks.
Neuroprosthetic implantable devices have to overcome several mechanical
challenges from material disintegration in the saline oxidative environment, to
biocompatibility and tissue toxicity, to cellular attachment and encapsulation. After
an electrode implantation, the adjacent neurons appear to have some degree of
mobility, as their recorded signals are reported to appear or disappear with time
(Kipke, 2001). If the electrodes are coated with a highly adhesive substrate, moving
neurons might attach themselves to them and become anchored (Ignatius et al.,
1998). Optimizing the interface between neurons and MEAs for these applications
would consist of enhancing cell viability and survival time. Prolonging attachment
of cells onto the surface is critical, yet even more significant is the development of
cell patterning methods that direct neurons to electrodes. This would increase the
120
yield of active electrodes, as well as provide more reproducible or consistent
neuronal networks on the sensors (Chang et al., 2001).
This principle can be tested in a cell culture system. The neurons and their
processes are able to move on the surface. They can however be guided to attach
and grow along specific patterns. Cell patterning relies on the creation of a contrast
between repulsive and attractive surfaces for the cells to attach and grow on.
Neurons have a weak negative overall surface charge and thus prefer positively
charged surfaces such as the cationic polymers discussed above or some basement
membrane proteins. Other extracellular matrix molecules such as tenascin are
however axonal repellents, and along with hydrophobic and anionic molecules such
as Polyethelene Glycol are considered repulsive substrates, since they discourage cell
attachment and axonal growth (Kleinfeld et al., 1988; Tessier-Lavigne and
Goodman, 1996). While some uncoated substrates are repellents by nature of their
surface charge, placing repulsive and attractive substrates adjacent to each other
increases the stringency of the pattern. This is probably because opposite coatings
prevent uncontrolled reattachment of desorbed molecules or proteins from the
medium. However, in order to confine cells and their processes to specific areas on
the surface of the pMEA, it is usually sufficient to deposit the attractant substrate on
those areas, and leave the rest untreated. Figure 5.1 shows an example of dissociated
hippocampal neurons patterned on top of a pMEA. The neurons attach mostly along
the substrate tracts that are aligned to the electrodes.
Figure 5.1: Photomicrographs of a culture patterned on a pMEA after 11 days in vitro (DIV). PDL
substrate was deposited in parallel lines using a channel matrix aligned on pMEA. The cell density of
this culture is approximately 125,000 cells/cm
2
, and the cells thus tend to cluster and override the
pattern.
Cell-based biosensor could also greatly benefit from such patterning. Beyond
the mentioned increase in recording yield, sensor reproducibility could be
significantly enhanced. In unpatterned cultures, the degree of reproducibility
between different cultures is limited to a measure of network homogeneity, which in
itself is controlled by variables such as cell density, surface coating, and medium
content. The production of highly reproducible biosensors mandates directed cell
growth at a sub-millimeter resolution, as the relative positions of the cells and
electrodes could affect the readouts of the sensor. Patterning enables the alignment
of cells on planar electrode arrays, which simplifies electrophysiology, microscopic
observation, and data analysis. Additionally, since the movement of neurons on the
121
122
surface would be restricted, the system will be more stable over time, and this will
enable longer-term comparisons.
Patterning neurons and their branches has been accomplished by many very
different methods: Several photolithographic approaches are used to deposit specific
patterns of self-assembling organosilanes, which can be derivatized into repellents or
attractants (Corey et al., 1996; Matsuzawa et al., 1996a; Stenger et al., 1998);
Patterns can be etched into deposited substrates by UV irradiation or laser ablation
(Stenger et al., 1992; Matsuzawa et al., 1993; Corey et al., 1997; Stenger et al.,
1998; Sabanayagam et al., 2000); The properties of the surface materials themselves
can be harnessed to create attractive or repelling zones (Torimitsu and Kawana,
1990; Ignatius et al., 1998), or to control surface reactions for covalent linkage of
biomolecules which also slows down desorption of substrates from the surface
(Detrait et al., 1998; Saneinejad and Shoichet, 1998; Bledi et al., 2000); Patterns can
also be made by rubber stamping adhesive proteins on surfaces (Branch et al., 1998;
Martinoia et al., 1999; James et al., 2000), or spatially confining their flow with
channel matrices (Walter et al., 1987; Vielmetter et al., 1990; Baier and
Klostermann, 1994; Folch and Toner, 1998) (Figure 5.1); Surface structure and
topology can be controlled during manufacture to create obstacles, channels, wells,
walls, or roughness that affect cell attachment, function and survival as well
(Morikawa et al., 1991; Wojciak-Stothard et al., 1996; Rajnicek et al., 1997); Even
more elaborate micromanufacture techniques have been employed to build wells or
microcages for the confinement of neurons on top of electrodes, and tunnels for the
123
guidance of their dendrites (Maher et al., 1999b; Maher et al., 1999a). There are
also methods that position cells without modifying their contact surface: for
example, small applied DC electric fields can induce directional neuritic growth
(McCaig and Rajnicek, 1991; Rajnicek et al., 1992), or laser tweezers can be used to
place the cells at chosen locations (Odde and Renn, 1999).
Beyond the control over the location of the neuronal soma, the attachment,
growth and differentiation of the neuronal process can also be controlled by similar
patterning methods. Cells bodies tend to move to coated areas that are wide enough
for them (>20 um), and thus network nodes can be created by making connecting
pathways 10 microns or less (Corey et al., 1996). In addition, uni-directional growth
was achieved by patterning axonal growth promoters or directionally facilitated
pathways, which took advantage of the neuronal development property that the
longest neurite becomes the only axon (Stenger et al., 1998; Esch et al., 1999).
Furthermore, patterned networks have been demonstrated to produce highly ordered
cell cultures that can remain synaptically active (Kleinfeld et al., 1988; Ma et al.,
1998; Liu et al., 2000; Matsuzawa et al., 2000)
Such refined control over the network of neurons can be useful for
neuroprosthetics, cell-based biosensors, and would open the door for the field of
biological artificial networks. Neuroprosthetics would benefit from the ability to
guide regenerating axons to their proper targets on the implants. The patterned
molecules in this case would be different axonal guidance molecules, but some of the
principles and methods would be similar. Biosensors would have even more
124
reproducible networks whose more consistent and predictable activity would
therefore enable sensor-to-sensor comparisons. Additionally, uni-directional
connections between individual neurons or neuron groups will enable the creation of
designer biological neuronal networks, which can be used to test hypotheses of
neuronal network computations that cannot be easily addressed in other biological
systems (Poirazi and Mel, 2001). Figure 5.2 shows an plausible application of
patterning a neuronal network with specific axonal directionality. The patterning
would take advantage of the differential adhesivity of substrates, and their axonal
growth promoting features. PEI can be applied at the nodes on top of electrodes, this
would cause any cells that settle on nodes to stick and remain there. PDL
connections allow neurite growth between nodes, and the 10µm width would force
cell bodies to migrate towards the wider nodes where they would stick. BM patterns
could be used to control the directionality of axonal growth in specific directions,
based on the principle that BM promotes axonal development. Similar patterning
could be designed and implemented for neuroprosthetic or biosensor applications.
Figure 5.2: Neuronal network circuit design
through adhesive biomolecules. Adhesive
substrates can be patterned on top of electrode
arrays with 30µm wide nodes 200µm apart and
10µm thick connections. Nodes can be coated
with PEI (black) to irreversibly attach neuronal
cell bodies to the electrode sites. Connections
can then be coated with either PDL (red) to
enable neurite attachment or BM (green) to
promote axonal growth along the specific
direction (arrows).
The objective of neuronal network analysis is the elucidation of neuronal
information processing codes. Correlations of activity between neurons have long
been used to predict their connections (Gerstein and Perkel, 1972; Perkel et al.,
1975). More recently, correlations of activity between neurons have been used to
predict their activity dependent plasticity (Jimbo et al., 1999). And finally,
connections between neurons can be used to predict their plasticity (Fitzsimonds et
al., 1997; Tao et al., 2000). It follows that the potential plasticity of a network can
be predicted from its connectivity. Therefore, it is hypothesized that network
connectivity can predict plasticity in response to stimulation. Small confined
connectivity networks are required to begin testing this hypothesis. Patterned
neuronal networks aligned on pMEAs provide such networks whose nodes would be
addressable through the electrodes.
125
126
In the slice and in vivo, sequential single electrode recordings were used
diligently to study network connectivity, (Andersen et al., 1971; Yeckel and Berger,
1990; Yeckel and Berger, 1995), examining spatio-temporal correlations was not
possible with this methodology. MEA technology offers several advantages over
traditional single electrode recording in the areas of high-throughput testing, spatial
mapping of electrical activity, temporal information processing, spatio-temporal
activity monitoring, long-term physiological investigations, and neuroprosthetic-
driven network dynamics elucidation. While technological advances have increased
the production and use of pMEAs, the limited number of electrodes has led to the
creation of generic arrays consisting of symmetrical matrices of electrodes that suffer
from misalignment with tissue cytoarchitecture due to the nonsymmetrical anatomy
of the brain.
In all in vitro electrophysiological experiments, considerable care and time is
spent locating the appropriate stimulation and recording sites. Herein, the first
advantage of pMEAs presents itself for high-throughput experiments, by expediting
the process of localizing the optimal stimulation and recording sites in a target area.
While the sequential approach is reminiscent of micro-advancing single electrodes, it
is different in that it does not induce incremental damage to the tissue, and in being
automatable in software without the need for manual repositioning. Beyond the gain
in speed and efficiency, the ability to stimulate different sites in close proximity to
each other without disturbing the tissue by repeated electrode insertions will enable
more accurate and concise stimulations. The advantage of such fine spatial control
127
over the stimulation site was illustrated with the selective excitation of PP sub-
pathways. The versatility of custom-designing such high-density pMEAs will enable
and speed up new electrophysiological experiments in various tissues where there is
a need for selective stimulation of adjacent pathways. Research on brain slices with
sharp cytoarchitectural boundaries such as the rat barrel cortex (Wirth and Luscher,
2004), or defined fiber tracts such as the cerebellum (Heck, 1995; Egert et al., 2002),
would benefit significantly from high-density pMEAs that conform to their
geometries.
pMEAs are, however, more than just a sum of many electrodes that allow
rapid accumulation of many channels of electrophysiological data from a single
slice. pMEAs allow the creation of spatio-temporal activity maps which enable
many new applications not previously possible with single electrode recordings.
Spatially, the activity maps allow tracing of the spread of a response across the
tissue. Figure 2.5 illustrated how this information can be useful in an LTP
experiment in the hippocampal slice. The area responding to the stimulation can be
traced by color-mapping the fEPSP amplitude on the slice image (Figure 2.5A,B),
and similarly the amount of potentiation can be mapped (Figure 2.5D) in order to
visually determine the extent and spread of these parameters over the entire slice.
Mapping various activity parameters spatially onto the slice is especially useful for
studying topographically organized brain structures such as cortical areas (visual,
auditory, barrel …), which have functionally distinct subdivisions.
128
Information about the activity in different cellular layers (strata) can only be
mapped with higher density arrays that sample closely enough to enable CSD
analysis. This mathematical transformation is necessary to compensate for FP
intermixing in the conductive tissue, and to resolve current sources and sinks. The
electrode spacing requirement depends on the distance from the source and the size
of different spatial domains in the slice, and as such has been estimated at 50 µm.
(Novak and Wheeler, 1989) The high-density cMEAs used above have electrodes
spaced 50 µm apart (center to center) specifically for this purpose. Figure 3.2
illustrated the ability of CSD analysis to reduce the spatial spread of FPs and
accurately localize current sources and sinks and even unmask some that may be
overpowered by larger potentials. CSD analysis produced a more accurate laminar
profile of the CA1 pyramidal cell fEPSP and population spike. The CSD profiles
could thus be used to map activity along axo-dendritic trees that densely orient
parallel to each other. Kim et al. used cMEA#1 to localize estradiol’s effect on CA3
pyramidal cells and compare the activity maps to the estrogen receptors’ distribution.
(Kim et al., in submission) High-density pMEAs can thus be used for generating
accurate high-resolution spatial activity maps of a tissue, thereby complementing the
larger scale maps generated by more widely spaced pMEAs.
There are many applications to purely temporal or long-term recordings from
pMEAs, which investigate circadian rhythmicity and slice cultures, or the temporal
dynamic investigations of Beggs et al. (Beggs and Plenz, 2003) and Jimbo et al.
(Jimbo et al., 1999). However, high-density conformal pMEAs do not confer any
129
significant advantages for solely temporal experiments, and in the case of long-term
cultures, the tissue changes its position relative to the electrodes during the span of
the culture in ways that are not controllable, and thus makes conformal pMEA
design obsolete. On the other hand, patterning adhesion molecules onto arrays, in
order to align network development to electrode locations, facilitates long-term
monitoring and investigations of activity.
Spatio-temporal information is however the most distinguishing advantage of
MEAs, as they permit the simultaneous recording of activity at many points across
the tissue, thereby revealing correlations and interactions between different regions.
The activity in a slice spreads from one area to another along excitatory pathways.
This propagation can be traced in space over time with pMEAs. The interaction
between hippocampal areas was more specifically explored in the TMPP
experiments. TMPP, a GABA-A inhibitor, reduces inhibition in hippocampal slices.
In CA3, the large number of inhibitory interneurons usually prevents repetitive
bursting due to activation of recurrent fibers. In the presence of TMPP, this
inhibition was blocked, and CA3 pyramidal cells would burst synchronously when
stimulated either directly or upstream, through PP stimulation. The burst of
population spikes was easily localized to area CA3 by the pMEA. A facilitated
propagation of the response to CA1 was also observed, and the delay in CA1 fEPSP
indicated the number of synapses traversed by the spreading activity. Monosynaptic
responses could be easily compared to di- and trisynaptic ones by overlaying
recorded potentials and comparing lag times. The effect of TMPP on the
130
hippocampal slice was thus more than just disinhibiting CA3 pyramidal cells to the
point of evoking epileptiform activity: it also facilitated propagation of this activity
to CA1. This network effect of TMPP on the hippocampal slice could not have been
observed with single electrode intracellular recordings, though the latter revealed
epileptiform discharges in CA1 that were not detected extracellularly (Lin et al.,
2001). MEAs are therefore extremely useful for investigating the propagation of
epileptiform activity in vivo and in vitro, spontaneous or drug-induced (Nagao et al.,
1996; Harris and Stewart, 2001). pMEAs have been used to localize the origin of
epileptic seizures (optical multisite recordings (Colom and Saggau, 1994)), to
determine the propagation direction (Harris and Stewart, 2001) and calculate changes
in propagation velocity across fibers compared to normal states. (Holsheimer and
Lopes da Silva, 1989) Beyond drug screening, such monitoring of multi-synaptic
activity in a tissue is thus useful for investigations of network connections and
dynamics in normal or diseased states. In these investigations, observing the entire
slice with pMEAs not only allows localization of the site of activity of the drug, but
can also unveil related effects on other areas.
pMEAs have been used to time spreading activity or trace waves in several
tissue preparations: retina (Syed et al., 2004), cerebellum (Egert et al., 2002), and
barrel cortex (Wirth and Luscher, 2004), ... The pMEAs allow easy quantification of
propagation velocity across the slice. Experiments monitoring monosynaptic
responses at CA1 recorded spread along SchC at 0.25 m/s (Soussou et al., 2005),
consistent with the 0.25 m/s reported average velocity for this weakly myelinated
131
fiber. (Andersen et al., 2000) There are however many other measurements of
axonal propagation speed that show different values, including the synaptic delay
times calculated from the TMPP experiment, which appears longer than those
reported in the literature. The speed of propagation of activity in a tissue preparation
is generally thought of as a function of the axon thickness, their myelination, length,
and also the number of traversed synapses. There are however numerous other
factors that can influence the propagation velocity across the multisynaptic
pathways, beyond simple experimental manipulations such as temperature or ionic
concentration differences (unpublished data). In particular for hippocampus, the
behavioral state (REM, Slow-wave sleep, or alert (Winson and Abzug, 1978)
(irregular or theta EEG (Herreras et al., 1987)), anesthesia (Buzsaki et al., 1986; Pare
and Llinas, 1994), pharmacological disinhibition or activation facilitation (Sirvio et
al., 1996), stimulation frequency (Herreras et al., 1987; Yeckel and Berger, 1990),
stimulation intensity and relative location to recording sites (Pare and Llinas, 1994),
and animal species (Andersen et al., 1978) have all been shown to affect activity
propagation in the tri-synpatic pathway. pMEAs are extremely well suited for
investigating these variables to elucidate the interrelations between connected
regions in multi-synaptic circuits.
There appears to be significant variability in the propagation velocity and
delays reported in the literature. Using pMEAs, Andersen et al. mapped the angle
and distribution of the SchC in a sheet-like rat hippocampal preparation, and found a
distribution of conduction velocities for different fibers in the hippocampus. They
132
calculated the average conduction velocity of the weakly myelinated Schaffer
Collateral fibers at 0.25 m/s, associational fibers averaged 0.39 m/s while axons in
the fimbria fell into two classes with average velocities of 0.99 and 0.37 m/s
depending on whether they were myelinated or not. (Andersen et al., 2000) While
the propagation speed in the TMPP experiment matches these values, fibers running
parallel to the pyramidal layer were previously reported to have a propagation
velocity of 0.3 m/s in the guinea pig hippocampus. (Andersen et al., 1978) In
dissociated hippocampal neuron cultures, conduction velocity was calculated to 0.12
m/s with multichannel optical recordings with voltage sensitive dyes, and the
synaptic delay was calculated at 1ms (Kawaguchi et al., 1996). These differences in
velocities are also reflected in the variability of response delays. Yeckel and Berger
previously reported on trisynaptically evoked population spikes in CA1 from PP
stimulations in vivo, where they recorded mono synaptic responses in CA1 at 6 ms,
di-synaptic ones at 11 ms and trisynaptic ones at 17 ms, considerably different from
the above reported delays (Yeckel and Berger, 1995). The monosynaptic and di-
synaptic responses were postulated to constitute a feed-forward circuit, whereby PP
feeds directly into CA3 and CA1 (Yeckel and Berger, 1990). Hippocampal
trisynaptic response onset delays reported with optical recordings of voltage
sensitive dyes from PP stimulations yielded 5.5ms delay for granule cells, 13.5ms for
CA3, and 19.2 for CA1. In addition, potentiation by tetanic stimulation decreased
the onset latency by 65% at CA1 (Nakagami et al., 1997). In the isolated guinea pig
brain in vitro preparation, Pare and Llinas measured the population spike latency in
133
CA1 in response to entorhinal cortex (EC) stimulation (Pare and Llinas, 1994). They
reported spike spatial propagation beyond the transverse axis of the hippocampus,
and temporal delays that varied by up to 8 ms depending on the relative recording
and stimulation sites. The recorded delays suggested that the topographical
relationship between EC and hippocampus, though not spatially confined in
lamellae, was preserved in the time domain. They further demonstrated that the
spike latency depended on the stimulation intensity. Furthermore, a 3 ms delay in
propagation was observed in CA2 using optical recordings of voltage sensitive dyes.
(Sekino et al., 1997) The transmission from CA3 to CA1 could proceed either
directly through the SchC or stop for 3 ms in CA2 before proceeding. The
coexistence of these two modes of propagation of activity leads to delays in
propagation from CA3 to CA1 that vary from 7 to 13.5 ms. The differences in
reported multisynaptic delays therefore depend on many factors which can be
investigated systematically with pMEAs. Of course, higher-density conformal
pMEAs could resolve propagation along dendritic trees at higher spatio-temporal
resolution, and allow examination of changes of electrical conductance velocity or
summation along single cells. In summary, however, the reported propagation
velocities and delays point to the power and usefulness of pMEAs for investigations
of multi-synaptic network activity.
Such an understanding of network dynamics is critical for building
biomimetic devices or neuroprosthetics. The pMEAs’ ability to stimulate specific
regions of a tissue and record its spatio-temporal dynamics makes them useful test
134
beds for neuroprosthetic devices. The retinal prosthesis project relies on pMEA
recordings from in vitro retinal preparations to develop stimulation algorithms for its
implanted MEAs (Meister et al., 1994; Warland et al., 1997; Chichilnisky and
Kalmar, 2003; Humayun et al., 2003; Frechette et al., 2004). In the project to
replace CA3 function in vivo, an in vitro model is needed upon which to test non-
linear mathematical models of the input/output properties CA3. High-density
cMEAs are being used to develop a proof-of-concept for a cortical prosthesis. The
goal of this study is to investigate the fundamental science and implement
technology that will enable a biomimetic electronic device to be implanted into the
hippocampus to replace damaged neuronal circuits, and thereby re-establish its
original functions. The proof-of-concept consists of replacing the biological CA3
sub-region with an FPGA/VLSI-based model of the non-linear dynamics of CA3,
such that the propagation of spatio-temporal patterns of activity in the DG Æ
FPGA/VLSI Æ CA1 network reproduces that observed experimentally in the
biological DG Æ CA3 Æ CA1 circuit. (Berger et al., 2001; Berger et al., 2005) The
feasibility of this ambitious endeavor is being first demonstrated with an in vitro
hippocampal slice preparation on pMEAs. The first step entailed development of a
slice-pMEA preparation that allows experimental characterization of the combined
non-linear dynamics of the intrinsic hippocampal trisynaptic circuit. Inherent to the
success of this project is the development of a stable biological preparation that
requires optimization of the anatomical and the electrophysiological conditions.
Current pMEAs are not optimized for such an application because the symmetrical
135
distribution of electrodes does not match the cytoarchitecture of hippocampal slices.
Therefore, a new generation of cMEAs has been designed and built to allow
recording of spatio-temporal activity along the trisynaptic pathway of hippocampus.
The conformality aspect of the newly designed pMEA was crucial to obtain
trisynaptic recordings from hippocampal slices because it is extremely difficult if not
impossible to obtain simultaneous dentritic and somatic responses in the three sub-
regions of the hippocampus using the commercially available MEAs with their
sparse and symmetrical matrices of electrodes. The novel pMEA has recording
electrodes in the DG, CA3, and CA1 regions as well as stimulating electrodes in the
DG and the CA1 regions. The DG stimulating electrodes were used to activate the
perforant path, while the CA1 stimulating electrodes were used to channel the
FPGA/VLSI output to the CA1 pyramidal cells.
In summary, MEAs are destined to be an integral part of the neuron-silicon
interface. They are capable of stimulating neural tissue at multiple sites and of
recording spatio-temporal activity. The rapid transition of cell types in the brain
mandates high spatial resolution sampling while the nonsymmetrical organization of
the brain in conjunction with the limited number of recording channels requires
conformal electrode designs. Depending on electrode arrangements, MEAs can
therefore provide either low resolution information which allows a wide, albeit
sparse, coverage of activity at various slice regions, or at higher density enable CSD
analysis for accurate mapping of currents and sources in smaller cellular sub-regions.
The high-density conformal designs confer new abilities to the planar MEA research
136
tool whether it is applied to rapidly and accurately localize target field responses, to
increase sampling sites for high-throughput drug investigations, or to trace spatio-
temporal network activity propagation and dynamics for biomimetic neuroprosthetic
applications. The work presented here lays the ground for the custom design of
conformal high-density pMEAs that meet the individual needs of their application.
For implantable neuroprostheses, cell-based biosensors, and neuronal network
applications, MEA surface coating is as important as electrode designs in enhancing
functionality. Biological coating substrates can promote neuronal attachment and
induce different morphologies and electrophysiological activity patterns. Patterning
biomolecules will therefore affect not only neuronal attachment and growth, but
network physiology as well, and should be considered at neuron-silicon interface
design stages. In conclusion, the presented methodologies demonstrated the
signficant role of MEAs as neuron-silicon interfaces for neuroprosthetic, biosensor
and neuronal network applications.
137
ALPHABETIZED BIBLIOGRAPHY
Amaral DG, Witter MP (1989) The three-dimensional organization of the
hippocampal formation: a review of anatomical data. Neuroscience 31:571-
591.
Andersen P (1975) Organization of hippocampal neurons and their interconnections.
New York, NY: Plenum Press.
Andersen P, Bliss TVP, Skrede KK (1971) Lamellar organization of hippocampal
excitatory pathways. Experimental Brain Research 13:222-238.
Andersen P, Soleng AF, Raastad M (2000) The hippocampal lamella hypothesis
revisited. Brain Res 886:165-171.
Andersen P, Silfvenius H, Sundberg SH, Sveen O, Wigstrom H (1978) Functional
characteristics of unmyelinated fibres in the hippocampal cortex. Brain Res
144:11-18.
Baier H, Bonhoeffer F (1992) Axon guidance by gradients of a target-derived
component. Science 255:472-475.
Baier H, Klostermann S (1994) Axon guidance and growth cone collapse in vitro.
Neuroprotocols 4:96-105.
Banker G, Goslin K, eds (1999) Culturing Nerve Cells, 2nd Edition Edition.
Cambridge: MIT Press.
Baudry M, Krause M, Taketani M (2005) Advances in Network Electrophysiology
Using Multi Electrode Arrays. NY, NY: Kluwer Academic.
Beggs JM, Plenz D (2003) Neuronal avalanches in neocortical circuits. J Neurosci
23:11167-11177.
Berger TW, Baudry M, Brinton RD, Liaw JS, Marmarelis VZ, Park AY, Sheu BJ,
Tanguay ARJ (2001) Brain-Implantable Biomimetic Electronics as the Next
Era in Neural Prosthetics. Proceedings of the IEEE 89:993-1012.
Berger TW, Ahuja A, Courellis SH, Erinjippurath G, Gholmieh G, Granacki JJ,
Marmarelis VZ, Srinivasan V, Dong S, Tanguay AR, Wills J (2005) Brain-
implantable biomimetic electronics as neural prostheses to restore lost
cognitive function. In: Synaptic Plasticity: Basic Mechanisms to Clinical
Applications. (Baudry M, Bi X, Schreiber SS, eds). Boca Raton: Marcel
Dekker.
138
Bi G, Poo M (1999) Distributed synaptic modification in neural networks induced by
patterned stimulation. Nature 401:792-796.
Bi GQ, Poo MM (1998) Synaptic modifications in cultured hippocampal neurons:
dependence on spike timing, synaptic strength, and postsynaptic cell type. J
Neurosci 18:10464-10472.
Black RC, Clark GM, Tong YC, Patrick JF (1983) Current distributions in cochlear
stimulation. Annals of the New York Academy of Sciences 405:137-145.
Bledi Y, Domb AJ, Linial M (2000) Culturing neuronal cells on surfaces coated by a
novel polyethyleneimine-based polymer. Brain Res Brain Res Protoc 5:282-
289.
Boppart SA, Wheeler BC, Wallace CS (1992) A flexible perforated microelectrode
array for extended neural recordings. IEEE Trans Biomed Eng 39:37-42.
Borkholder DA, Bao J, Maluf NI, Perl ER, Kovacs GT (1997) Microelectrode arrays
for stimulation of neural slice preparations. J Neurosci Methods 77:61-66.
Bossi A, Piletsky SA, Righetti PG, Turner AP (2000) Capillary electrophoresis
coupled to biosensor detection. J Chromatogr A 892:143-153.
Boutelle MG, Allen H, Hill O, Berners M, John R, Dobson PD, Leigh P (1996) New
technologies for amperometric biosensors. Journal of Molecular Recognition
9:664-671.
Bragin A, Hetke J, Wilson CL, Anderson DJ, Engel J, Jr., Buzsaki G (2000) Multiple
site silicon-based probes for chronic recordings in freely moving rats:
implantation, recording and histological verification. J Neurosci Methods
98:77-82.
Branch DW, Wheeler BC, Brewer GJ, Leckband DE (2001) Long-term stability of
grafted polyethylene glycol surfaces for use with microstamped substrates in
neuronal cell culture. Biomaterials 22:1035-1047.
Branch DW, Corey JM, Weyhenmeyer JA, Brewer GJ, Wheeler BC (1998)
Microstamp patterns of biomolecules for high-resolution neuronal networks.
Med Biol Eng Comput 36:135-141.
Branner A, Normann RA (2000) A multielectrode array for intrafascicular recording
and stimulation in sciatic nerve of cats. Brain Research Bulletin 51:293-306.
Brewer GJ (1995) Serum-free B27/neurobasal medium supports differentiated
growth of neurons from the striatum, substantia nigra, septum, cerebral
cortex, cerebellum, and dentate gyrus. J Neurosci Res 42:674-683.
139
Brown TH, Zador AM (1990) Hippocampus. In: The Synaptic Organization of the
Brain, Third Edition (Shepherd GM, ed), pp 346-388. New York: Oxford
University Press.
Burmeister JJ, Moxon K, Gerhardt GA (2000) Ceramic-based multisite
microelectrodes for electrochemical recordings. Analytical Chemistry
72:187-192.
Buzsaki G, Eidelberg E (1982) Convergence of associational and commissural
pathways on CA1 pyramidal cells of the rat hippocampus. Brain Res
237:283-295.
Buzsaki G, Czopf J, Kondakor I, Kellenyi L (1986) Laminar distribution of
hippocampal rhythmic slow activity (RSA) in the behaving rat: current-
source density analysis, effects of urethane and atropine. Brain Res 365:125-
137.
Byfield MP, Abuknesha RA (1994) Biochemical aspects of biosensors. Biosensors &
Bioelectronics 9:373-400.
Canepari M, Bove M, Maeda E, Cappello M, Kawana A (1997) Experimental
analysis of neuronal dynamics in cultured cortical networks and transitions
between different patterns of activity. Biological Cybernetics 77:153-162.
Cash S, Yuste R (1999) Linear summation of excitatory inputs by CA1 pyramidal
neurons. Neuron 22:383-394.
Chang JC, Brewer GJ, Wheeler BC (2001) Modulation of neural network activity by
patterning. Biosens Bioelectron 16:527-533.
Chang JC, Brewer GJ, Wheeler BC (2003) A modified microstamping technique
enhances polylysine transfer and neuronal cell patterning. Biomaterials
24:2863-2870.
Chapin JK (1999) Population-Level Analysis of Multi-Single Neuron Recording
Data: Multivariate Statistical Methods. In: Methods for Neural Ensemble
Recordings (Nicolelis MAL, ed). New York: CRC Press.
Chapin JK, Moxon KA, Markowitz RS, Nicolelis MA (1999) Real-time control of a
robot arm using simultaneously recorded neurons in the motor cortex. Nature
Neuroscience 2:664-670.
Chen CS, Mrksich M, Huang S, Whitesides GM, Ingber DE (1997) Geometric
control of cell life and death. Science 276:1425-1428.
140
Chen CS, Mrksich M, Huang S, Whitesides GM, Ingber DE (1998) Micropatterned
surfaces for control of cell shape, position, and function. Biotechnology
Progress 14:356-363.
Chichilnisky EJ, Kalmar RS (2003) Temporal resolution of ensemble visual motion
signals in primate retina. J Neurosci 23:6681-6689.
Clark GM (1998) Research advances for cochlear implants. Auris, Nasus, Larynx
25:73-87.
Clark GM, Hallworth RJ (1976) A multiple-electrode array for a cochlear implant.
Journal of Laryngology & Otology 90:623-627.
Claverol-Tinture E, Pine J (2002) Extracellular potentials in low-density dissociated
neuronal cultures. J Neurosci Methods 117:13-21.
Colom LV, Saggau P (1994) Spontaneous interictal-like activity originates in
multiple areas of the CA2-CA3 region of hippocampal slices. J Neurophysiol
71:1574-1585.
Corey JM, Wheeler BC, Brewer GJ (1996) Micrometer resolution silane-based
patterning of hippocampal neurons: critical variables in photoresist and laser
ablation processes for substrate fabrication. IEEE Trans Biomed Eng 43:944-
955.
Corey JM, Brunette AL, Chen MS, Weyhenmeyer JA, Brewer GJ, Wheeler BC
(1997) Differentiated B104 neuroblastoma cells are a high-resolution assay
for micropatterned substrates. J Neurosci Methods 75:91-97.
Courellis SH, Marmarelis VZ, Berger TW (2000) Modeling event-driven nonlinear
dynamics in neuronal systems with multiple inputs. In: Annual conference of
the biomedical engineering society. Seattle, WA.
da Silva EM, Soares AM, Moreno AJ (1998) The use of the mitochondrial
transmembrane electric potential as an effective biosensor in ecotoxicological
research. Chemosphere 36:2375-2390.
Dahl D, Burgard EC, Sarvey JM (1990) NMDA receptor antagonists reduce medial,
but not lateral, perforant path-evoked EPSPs in dentate gyrus of rat
hippocampal slice. Exp Brain Res 83:172-177.
Deadwyler SA, Hampson RE (1995) Ensemble activity and behavior: what's the
code? Science 270:1316-1318.
Deadwyler SA, Hampson RE (1997) The significance of neural ensemble codes
during behavior and cognition. Annu Rev Neurosci 20:217-244.
141
Deadwyler SA, Bunn T, Hampson RE (1996) Hippocampal ensemble activity during
spatial delayed-nonmatch-to-sample performance in rats. Journal of
Neuroscience 16:354-372.
Detrait E, Lhoest JB, Knoops B, Bertrand P, van den Bosch de Aguilar P (1998)
Orientation of cell adhesion and growth on patterned heterogeneous
polystyrene surface. Journal of Neuroscience Methods 84:193-204.
Dityatev A, Schachner M (2003) Extracellular matrix molecules and synaptic
plasticity. Nat Rev Neurosci 4:456-468.
Droge MH, Gross GW, Hightower MH, Czisny LE (1986) Multielectrode analysis of
coordinated, multisite, rhythmic bursting in cultured CNS monolayer
networks. J Neurosci 6:1583-1592.
Duport S, Millerin C, Muller D, Correges P (1999) A metallic multisite recording
system designed for continuous long-term monitoring of electrophysiological
activity in slice cultures. Biosensors & Bioelectronics 14:369-376.
Duport S, Robert F, Muller D, Grau G, Parisi L, Stoppini L (1998) An in vitro blood-
brain barrier model: cocultures between endothelial cells and organotypic
brain slice cultures. Proceedings of the National Academy of Sciences of the
United States of America 95:1840-1845.
Egert U, Heck D, Aertsen A (2002) Two-dimensional monitoring of spiking
networks in acute brain slices. Exp Brain Res 142:268-274.
Egert U, Schlosshauer B, Fennrich S, Nisch W, Fejtl M, Knott T, Müller T,
Hämmerle H (1998) A novel organotypic long-term culture of the rat
hippocampus on substrate-integrated multielectrode arrays. Brain Res Brain
Res Protoc 2:229-242.
Eichenbaum HB, Davis JL (1998) Introduction: Biological Neural Networks. In:
Neuronal Ensembles: Strategies for recording and decoding (Eichenbaum
HB, Davis JL, eds). New York: Wiley-Liss.
Eriksson PS, Perfilieva E, Bjork-Eriksson T, Alborn AM, Nordborg C, Peterson DA,
Gage FH (1998) Neurogenesis in the adult human hippocampus. Nat Med
4:1313-1317.
Esch T, Lemmon V, Banker G (1999) Local presentation of substrate molecules
directs axon specification by cultured hippocampal neurons. J Neurosci
19:6417-6426.
142
Falkenrath RA, Newman RD, Thayer BA (1998) America's Achiles' Heel: Nuclear,
Biological, and Chemical Terrorism and Covert Attack. Cambridge: MIT
Press.
Fitzsimonds RM, Song HJ, Poo MM (1997) Propagation of activity-dependent
synaptic depression in simple neural networks. Nature 388:439-448.
Folch A, Toner M (1998) Cellular micropatterns on biocompatible materials.
Biotechnol Prog 14:388-392.
Frechette ES, Sher A, Grivich MI, Petrusca D, Litke AM, Chichilnisky EJ (2004)
Fidelity of the ensemble code for visual motion in primate retina. J
Neurophysiol.
Freeman JA, Nicholson C (1975) Experimental optimization of current source-
density technique for anuran cerebellum. J Neurophysiol 38:369-382.
Gentleman D (1994) Growth and repair after injury of the central nervous system:
yesterday, today and tomorrow. Injury 25:571-576.
Gerstein GL, Perkel DH (1972) Mutual temporal relationships among neuronal spike
trains. Statistical techniques for display and analysis. Biophysical Journal
12:453-473.
Gholmieh G, Soussou W, Han M, Tanguay JAR, Berger TW (In Preparation) High-
Density Conformal Planar Multielectrode Arrays For Hippocampal Slice
Stimulation And Recording.
Gholmieh G, Soussou W, Courellis S, Marmarelis VZ, Berger TW, Baudry M (2001)
A biosensor for detecting changes in cognitive processing based on nonlinear
systems analysis. Biosens Bioelectron 16:491-501.
Gholmieh G, Courellis S, Fakheri S, Cheung E, Marmarelis V, Baudry M, Berger T
(2003) Detection and classification of neurotoxins using a novel short-term
plasticity quantification method. Biosens Bioelectron 18:1467-1478.
Goldberger A, Flaherty P, Villalba R, Riblet LA, Cornell-Bell A (1997) Extracellular
Matrix Proteins Influence Cell Morphology and Function in Rat Neurl
Cultures. In. Bedford, MA: Becton Dickinson Labware.
Gramowski A, Jugelt K, Weiss DG, Gross GW (2004) Substance identification by
quantitative characterization of oscillatory activity in murine spinal cord
networks on microelectrode arrays. Eur J Neurosci 19:2815-2825.
143
Gross GW (1994) Internal dynamics of randomized mammalian neuronal networks
in culture. In: Enabling Technologies for Cultured Neural Networks:
Academic Press.
Gross GW, Schwalm FU (1994) A closed flow chamber for long-term multichannel
recording and optical monitoring. Journal Of Neuroscience Methods 52:73-
85.
Gross GW, Williams AN, Lucas JH (1982) Recording of spontaneous activity with
photoetched microelectrode surfaces from mouse spinal neurons in culture.
Journal Of Neuroscience Methods 5:13-22.
Gross GW, Rhoades BK, Reust DL, Schwalm FU (1993) Stimulation of monolayer
networks in culture through thin-film indium- tin oxide recording electrodes.
J Neurosci Methods 50:131-143.
Gross GW, Rhoades BK, Azzazy HM, Wu MC (1995) The use of neuronal networks
on multielectrode arrays as biosensors. Biosensors And Bioelectronics
10:553-567.
Gross GW, Keefer EW, Pancrazio JJ, Stenger DA (1999) Rapid determination of
toxic effects of trimethylol propane phosphate usina neuronal networks on
microelectrode arrays. Society for Neuroscience Abstracts 1.
Grumet AE, Wyatt JL, Jr., Rizzo JF, 3rd (2000) Multi-electrode stimulation and
recording in the isolated retina. J Neurosci Methods 101:31-42.
Guilbault GG, Schmid RD (1991) Biosensors for the determination of drug
substances. Biotechnology & Applied Biochemistry 14:133-145.
Gulflink (1997) Military plans to field new chemical detector/alarm starting this
October. In: Gulflink.
Hampson RE, Deadwyler SA (1999) Methods, results, and issues related to
recording neural ensembles. In: Methods for Neural Ensemble Recordings
(Nicolelis MAL, ed). New York: CRC Press.
Han M, Gholmieh G, Soussou W, Berger TW, Tanguay AR, Jr. (2002) Conformally-
Mapped Multielectrode Arrays for In-vitro Stimulation and Recording of
Hippocampal Acute Slices. In: The 2nd Joint Meeting of the IEEE
Engineering in Medicine and Biology Society and the Biomedical
Engineering Society, pp 2127-2128. Houston, TX.
Harris E, Stewart M (2001) Propagation of synchronous epileptiform events from
subiculum backward into area CA1 of rat brain slices. Brain Res 895:41-49.
144
Hatsopoulos NG, Ojakangas CL, Donoghue JP (1999) Detection and identification
of ensemble codes in motor cortex. In: Methods for Neural Ensemble
Recordings (Nicolelis MAL, ed). New York: CRC Press.
Heck D (1995) Investigating dynamic aspects of brain function in slice preparations:
spatiotemporal stimulus patterns generated with an easy-to-build multi-
electrode array. Journal of Neuroscience Methods 58:81-87.
Herreras O, Solis JM, Martin del Rio R, Lerma J (1987) Characteristics of CA1
activation through the hippocampal trisynaptic pathway in the
unanaesthetized rat. Brain Res 413:75-86.
Hetke JF, Lund JL, Najafi K, Wise KD, Anderson DJ (1994) Silicon ribbon cables
for chronically implantable microelectrode arrays. IEEE Transactions on
Biomedical Engineering 41:314-321.
Heuschkel MO, Fejtl M, Raggenbass M, Bertrand D, Renaud P (2002) A three-
dimensional multi-electrode array for multi-site stimulation and recording in
acute brain slices. J Neurosci Methods 114:135-148.
Higson SP, Vadgama PM (1994) Biosensors: a viable monitoring technology?
Medical & Biological Engineering & Computing 32:601-609.
Holsheimer J (1987) Electrical conductivity of the hippocampal CA1 layers and
application to current-source-density analysis. Exp Brain Res 67:402-410.
Holsheimer J, Lopes da Silva FH (1989) Propagation velocity of epileptiform
activity in the hippocampus. Exp Brain Res 77:69-78.
Humayun MS, Weiland JD, Fujii GY, Greenberg R, Williamson R, Little J, Mech B,
Cimmarusti V, Van Boemel G, Dagnelie G, de Juan E (2003) Visual
perception in a blind subject with a chronic microelectronic retinal prosthesis.
Vision Res 43:2573-2581.
Huotari MJ (2000) Biosensing by insect olfactory receptor neurons. Sensors and
Actuators B 71.
Ignatius MJ, Sawhney N, Gupta A, Thibadeau BM, Monteiro OR, Brown IG (1998)
Bioactive surface coatings for nanoscale instruments: effects on CNS
neurons. J Biomed Mater Res 40:264-274.
Ishizuka N, Cowan WM, Amaral DG (1995) A quantitative analysis of the dendritic
organization of pyramidal cells in the rat hippocampus. J Comp Neurol
362:17-45.
145
Jahnsen H, Kristensen BW, Thiebaud P, Noraberg J, Jakobsen B, Bove M, Martinoia
S, Koudelka-Hep M, Grattarola M, Zimmer J (1999) Coupling of organotypic
brain slice cultures to silicon-based arrays of electrodes. Methods 18:160-
172.
James CD, Davis R, Meyer M, Turner A, Turner S, Withers G, Kam L, Banker G,
Craighead H, Isaacson M, Turner J, Shain W (2000) Aligned microcontact
printing of micrometer-scale poly-L-lysine structures for controlled growth of
cultured neurons on planar microelectrode arrays. IEEE Trans Biomed Eng
47:17-21.
Jimbo Y, Robinson HP (2000) Propagation of spontaneous synchronized activity in
cortical slice cultures recorded by planar electrode arrays.
Bioelectrochemistry 51:107-115.
Jimbo Y, Robinson HP, Kawana A (1993) Simultaneous measurement of
intracellular calcium and electrical activity from patterned neural networks in
culture. IEEE Transactions on Biomedical Engineering 40:804-810.
Jimbo Y, Robinson HP, Kawana A (1998) Strengthening of synchronized activity by
tetanic stimulation in cortical cultures: application of planar electrode arrays.
IEEE Transactions on Biomedical Engineering 45:1297-1304.
Jimbo Y, Tateno T, Robinson HP (1999) Simultaneous induction of pathway-
specific potentiation and depression in networks of cortical neurons. Biophys
J 76:670-678.
Kam L, Shain W, Turner JN, Bizios R (1999) Correlation of astroglial cell function
on micro-patterned surfaces with specific geometric parameters. Biomaterials
20:2343-2350.
Kam L, Shain W, Turner JN, Bizios R (2002) Selective adhesion of astrocytes to
surfaces modified with immobilized peptides. Biomaterials 23:511-515.
Kamioka H, Maeda E, Jimbo Y, Robinson HP, Kawana A (1996) Spontaneous
periodic synchronized bursting during formation of mature patterns of
connections in cortical cultures. Neuroscience Letters 206:109-112.
Kawaguchi H, Tokioka R, Murai N, Fukunishi K (1996) Multichannel optical
recording of neuronal network activity and synaptic potentiation in
dissociated cultures from rat hippocampus. Neurosci Lett 205:177-180.
Keefer EW, Gramowski A, Stenger DA, Pancrazio JJ, Gross GW (2001)
Characterization of acute neurotoxic effects of trimethylolpropane phosphate
via neuronal network biosensors. Biosens Bioelectron 16:513-525.
146
Kennedy PR, Bakay RA (1998) Restoration of neural output from a paralyzed patient
by a direct brain connection. Neuroreport 9:1707-1711.
Kharitonov AB, Zayats M, Lichtenstein A, Katz E, Willner I (2000) Enzyme
monolayer-functionalized field-effects transistors for biosensor applications.
Sensors and Actuators B 70:222-231.
Kim MT, Gholmieh G, Soussou W, Ahuja A, Tanguay JAR, Berger TW, Brinton RD
(in submission) 17-beta estradiol potentiates fEPSP within each subfield of
the hippocampus with greatest potentiation of the associational/commissural
afferents of CA3.
Kipke DR (2001) Advanced neural implants and neuroprosthetic systems. In: Special
Sloan Seminar. California Institute of Technology.
Kleinfeld D, Kahler KH, Hockberger PE (1988) Controlled outgrowth of dissociated
neurons on patterned substrates. Journal of Neuroscience 8:4098-4120.
Kleinman HK, McGarvey ML, Hassell JR, Star VL, Cannon FB, Laurie GW, Martin
GR (1986) Basement membrane complexes with biological activity.
Biochemistry 25:312-318.
Kloosterman F, Peloquin P, Leung LS (2001) Apical and basal orthodromic
population spikes in hippocampal CA1 in vivo show different origins and
patterns of propagation. J Neurophysiol 86:2435-2444.
Lambert PR, Ruth RA, Hodges AV (1991) Multichannel cochlear implant and
electrically evoked auditory brainstem responses in a child with labyrinthitis
ossificans. Laryngoscope 101:14-19.
Legendy CR, Salcman M (1985) Bursts and recurrences of bursts in the spike trains
of spontaneously active striate cortex neurons. J Neurophysiol 53:926-939.
Lelong IH, Petegnief V, Rebel G (1992) Neuronal cells mature faster on
polyethyleneimine coated plates than on polylysine coated plates. J Neurosci
Res 32:562-568.
Letourneau PC (1975) Possible roles for cell-to-substratum adhesion in neuronal
morphogenesis. Dev Biol 44:77-91.
Lin J, Cassell J, Ritchie GD, Rossi J, 3rd, Nordholm AF (1998) Repeated exposure
to trimethylolpropane phosphate induces central nervous system sensitization
and facilitates electrical kindling. Physiology & Behavior 65:51-58.
147
Lin J, Ritchie GD, Stenger DA, Nordholm AF, Pancrazio JJ, Rossi J, 3rd (2001)
Trimethylolpropane phosphate induces epileptiform discharges in the CA1
region of the rat hippocampus. Toxicol Appl Pharmacol 171:126-134.
Liu Q, Coulombe M, Dumm J, Shaffer KM, Schaffner AE, Barker JL, Pancrazio JJ,
Stenger DA, Ma W (2000) Synaptic connectivity in hippocampal neuronal
networks cultured on micropatterned surfaces. Brain Res Dev Brain Res
120:223-231.
Ma W, Liu QY, Jung D, Manos P, Pancrazio JJ, Schaffner AE, Barker JL, Stenger
DA (1998) Central neuronal synapse formation on micropatterned surfaces.
Brain Res Dev Brain Res 111:231-243.
Maeda E, Robinson HP, Kawana A (1995) The mechanisms of generation and
propagation of synchronized bursting in developing networks of cortical
neurons. Journal of Neuroscience 15:6834-6845.
Maher MP, Pine J, Wright J, Tai YC (1999a) The neurochip: a new multielectrode
device for stimulating and recording from cultured neurons. Journal Of
Neuroscience Methods 87:45-56.
Maher MP, Dvorak-Carbone H, Pine J, Wright JA, Tai YC (1999b) Microstructures
for studies of cultured neural networks. Medical & Biological Engineering &
Computing 37:110-118.
Martinoia S, Bove M, Tedesco M, Margesin B, Grattarola M (1999) A simple
microfluidic system for patterning populations of neurons on silicon
micromachined substrates. Journal of Neuroscience Methods 87:35-44.
Mashour GA, Walker EE, Martuza RL (2005) Psychosurgery: past, present, and
future. Brain Res Brain Res Rev 48:409-419.
Matsuzawa M, Liesi P, Knoll W (1996a) Chemically modifying glass surfaces to
study substratum-guided neurite outgrowth in culture. J Neurosci Methods
69:189-196.
Matsuzawa M, Potember RS, Stenger DA, Krauthamer V (1993) Containment and
growth of neuroblastoma cells on chemically patterned substrates. J Neurosci
Methods 50:253-260.
Matsuzawa M, Weight FF, Potember RS, Liesi P (1996b) Directional neurite
outgrowth and axonal differentiation of embryonic hippocampal neurons are
promoted by a neurite outgrowth domain of the B2-chain of laminin. Int J
Dev Neurosci 14:283-295.
148
Matsuzawa M, Tabata T, Knoll W, Kano M (2000) Formation of hippocampal
synapses on patterned substrates of a laminin- derived synthetic peptide. Eur
J Neurosci 12:903-910.
McCaig CD, Rajnicek AM (1991) Electrical fields, nerve growth and nerve
regeneration. Exp Physiol 76:473-494.
McDermott H (1989) An advanced multiple channel cochlear implant. IEEE
Transactions on Biomedical Engineering 36:789-797.
McKeehan WL, Ham RG (1976) Stimulation of clonal growth of normal fibroblasts
with substrata coated with basic polymers. J Cell Biol 71:727-734.
McNaughton N, Miller JJ (1984) Medial septal projections to the dentate gyrus of
the rat: electrophysiological analysis of distribution and plasticity. Exp Brain
Res 56:243-256.
Meister M, Pine J, Baylor DA (1994) Multi-neuronal signals from the retina:
acquisition and analysis. Journal of Neuroscience Methods 51:95-106.
Miyakawa H, Kato H (1986) Active properties of dendritic membrane examined by
current source density analysis in hippocampal CA1 pyramidal neurons.
Brain Res 399:303-309.
Morgan CL, Newman DJ, Price CP (1996) Immunosensors: technology and
opportunities in laboratory medicine. Clinical Chemistry 42:193-209.
Morikawa T, Odagawa Y, Keino K, Fukuda J (1991) New plastic plates which
enhance neurite extension in culture: roles of bisphenol-A and
tricyclodecanyl units for growth and orientation of neurites on plastic plates
with microstructures. Neurosci Lett 127:16-20.
Motta PS, Judy JW (2005) Multielectrode microprobes for deep-brain stimulation
fabricated with a customizable 3-D electroplating process. IEEE Trans
Biomed Eng 52:923-933.
MultiChannelSystems (2005) MEA Publication List. In: MultiChannelSystems.
Myszka DG (1999) Survey of the 1998 optical biosensor literature. Journal of
Molecular Recognition 12:390-408.
Nagao T, Alonso A, Avoli M (1996) Epileptiform activity induced by pilocarpine in
the rat hippocampal-entorhinal slice preparation. Neuroscience 72:399-408.
Nakagami Y, Saito H, Matsuki N (1997) Optical recording of trisynaptic pathway in
rat hippocampal slices with a voltage-sensitive dye. Neuroscience 81:1-8.
149
Narayanaswamy R (1991) Current developments in optical biochemical sensors.
Biosensors & Bioelectronics 6:467-475.
Nicholson C, Llinas R (1975) Real time current source-density analysis using multi-
electrode array in cat cerebellum. Brain Res 100:418-424.
Nicholson C, Freeman JA (1975) Theory of current source-density analysis and
determination of conductivity tensor for anuran cerebellum. J Neurophysiol
38:356-368.
Nordhausen CT, Rousche PJ, Normann RA (1994) Optimizing recording capabilities
of the Utah Intracortical Electrode Array. Brain Research 637:27-36.
Novak JL, Wheeler BC (1988) Multisite hippocampal slice recording and stimulation
using a 32 element microelectrode array. Journal of Neuroscience Methods
23:149-159.
Novak JL, Wheeler BC (1989) Two-dimensional current source density analysis of
propagation delays for components of epileptiform bursts in rat hippocampal
slices. Brain Res 497:223-230.
Odde DJ, Renn MJ (1999) Laser-guided direct writing for applications in
biotechnology. Trends in Biotechnology 17:385-389.
Oka H, Shimono K, Ogawa R, Sugihara H, Taketani M (1999) A new planar
multielectrode array for extracellular recording: application to hippocampal
acute slice. Journal of Neuroscience Methods 93:61-67.
O'Sullivan C, Alcock SJ (1999) BIOSET: Biosensors for Environmental Technology
EU Workshop 'Biosensors for Environmental Monitoring: Technology
Evaluation'. Biosensors and Bioelectronics 14:541-544.
Paddle BM (1996) Biosensors for chemical and biological agents of defence interest.
Biosensors & Bioelectronics 11:1079-1113.
Pancrazio JJ, Whelan JP, Borkholder DA, Ma W, Stenger DA (1999) Development
and application of cell-based biosensors. Annals of Biomedical Engineering
27:697-711.
Pancrazio JJ, Gray SA, Shubin YS, Kulagina N, Cuttino DS, Shaffer KM, Eisemann
K, Curran A, Zim B, Gross GW, O'Shaughnessy TJ (2003) A portable
microelectrode array recording system incorporating cultured neuronal
networks for neurotoxin detection. Biosens Bioelectron 18:1339-1347.
150
Pare D, Llinas R (1994) Non-lamellar propagation of entorhinal influences in the
hippocampal formation: multiple electrode recordings in the isolated guinea
pig brain in vitro. Hippocampus 4:403-409.
Perkel DH, Gerstein GL, Smith MS, Tatton WG (1975) Nerve-impulse patterns: A
quantitative display technique for three neurons. Brain Research 100:271-
296.
Pine J (1980) Recording action potentials from cultured neurons with extracellular
microcircuit electrodes. Journal of Neuroscience Methods 2:19-31.
Poirazi P, Mel BW (2001) Impact of active dendrites and structural plasticity on the
memory capacity of neural tissue. Neuron 29:779-796.
Potter SM, DeMarse TB (2001) A new approach to neural cell culture for long-term
studies. J Neurosci Methods 110:17-24.
Rajnicek A, Britland S, McCaig C (1997) Contact guidance of CNS neurites on
grooved quartz: influence of groove dimensions, neuronal age and cell type. J
Cell Sci 110:2905-2913.
Rajnicek AM, Gow NA, McCaig CD (1992) Electric field-induced orientation of rat
hippocampal neurones in vitro. Exp Physiol 77:229-232.
Richardson TL, Turner RW, Miller JJ (1987) Action-potential discharge in
hippocampal CA1 pyramidal neurons: current source-density analysis. J
Neurophysiol 58:981-996.
Roe JN (1992) Biosensor development. Pharmaceutical Research 9:835-844.
Rostker B (1997) Reported Detection of Chemical Agent, Camp Montery, Kuwait.
Gulflink http://www.gulflink.osd.mil/campmont/index.html.
Rousche PJ, Normann RA (1998) Chronic recording capability of the Utah
Intracortical Electrode Array in cat sensory cortex. Journal of Neuroscience
Methods 82:1-15.
Rousche PJ, Normann RA (1999) Chronic intracortical microstimulation (ICMS) of
cat sensory cortex using the Utah Intracortical Electrode Array. IEEE
Transactions on Rehabilitation Engineering 7:56-68.
Rush AM, Rowan MJ, Anwyl R (2001) Application of N-methyl-D-aspartate
induces long-term potentiation in the medial perforant path and long-term
depression in the lateral perforant path of the rat dentate gyrus in vitro.
Neurosci Lett 298:175-178.
151
Rutten WL, Smit JP, Frieswijk TA, Bielen JA, Brouwer AL, Buitenweg JR, Heida C
(1999) Neuro-electronic interfacing with multielectrode arrays. Ieee
Engineering In Medicine And Biology Magazine 18:47-55.
Sabanayagam CR, Smith CL, Cantor CR (2000) Oligonucleotide immobilization on
micropatterned streptavidin surfaces. Nucleic Acids Research 28:E33.
Saneinejad S, Shoichet MS (1998) Patterned glass surfaces direct cell adhesion and
process outgrowth of primary neurons of the central nervous system. J
Biomed Mater Res 42:13-19.
Sanes JR (1989) Extracellular matrix molecules that influence neural development.
Annu Rev Neurosci 12:491-516.
Sekino Y, Obata K, Tanifuji M, Mizuno M, Murayama J (1997) Delayed signal
propagation via CA2 in rat hippocampal slices revealed by optical recording.
J Neurophysiol 78:1662-1668.
Selinger JV, Pancrazio JJ, Gross GW (2004) Measuring synchronization in neuronal
networks for biosensor applications. Biosens Bioelectron 19:675-683.
Shain W, Spataro L, Dilgen J, Haverstick K, Retterer S, Isaacson M, Saltzman M,
Turner JN (2003) Controlling cellular reactive responses around neural
prosthetic devices using peripheral and local intervention strategies. IEEE
Trans Neural Syst Rehabil Eng 11:186-188.
Shimono K, Baudry M, Panchenko V, Taketani M (2002) Chronic multichannel
recordings from organotypic hippocampal slice cultures: protection from
excitotoxic effects of NMDA by non-competitive NMDA antagonists. J
Neurosci Methods 120:193-202.
Shimono K, Brucher F, Granger R, Lynch G, Taketani M (2000) Origins and
distribution of cholinergically induced beta rhythms in hippocampal slices. J
Neurosci 20:8462-8473.
Singer W (2000) Why use more than one electrode at a time? New Technologies for
life sciences: Trends Guide 2000:12-17.
Singhvi R, Kumar A, Lopez GP, Stephanopoulos GN, Wang DI, Whitesides GM,
Ingber DE (1994) Engineering cell shape and function. Science 264:696-698.
Sirkar K, Revzin A, Pishko MV (2000) Glucose and lactate biosensors based on
redox polymer/oxidoreductase nanocomposite thin films. Analytical
Chemistry 72:2930-2936.
152
Sirvio J, Larson J, Quach CN, Rogers GA, Lynch G (1996) Effects of
pharmacologically facilitating glutamatergic transmission in the trisynaptic
intrahippocampal circuit. Neuroscience 74:1025-1035.
Skaggs WE, McNaughton BL (1999) Neuronal Ensemble Dynamics in
Hippocampus and Neocortesx During Sleep and Waking. In: Methods for
Neural Ensemble Recordings (Nicolelis MAL, ed). New York: CRC Press.
Song D, Xie X, Wang Z, Berger TW (2001) Differential effect of TEA on long-term
synaptic modification in hippocampal CA1 and dentate gyrus in vitro.
Neurobiol Learn Mem 76:375-387.
Soussou W, G. G, Han M, Ahuja A, Song D, Hsiao M-C, Wang Z, Tanguay ARJ,
Berger TW (2005) Mapping Spatio-Temporal Electrophysiological Activity
in Hippocampal Slices with Conformal Planar Multielectrode Array. In:
Advances in Network Electrophysiology Using MultiElectrode Arrays
(Baudry M, Taketani M, eds). NY, NY: Kluwer Academic.
Stenger DA, Georger JH, Dulcey CS, Hickman JJ, Rudolph AS, Nielsen TB, McCort
SM, Calvert JM (1992) Coplanar molecular assemblies of amino- and
perfluorinated alkylsilanes: Characterization and geometirc definition of
mammalian cell adhesion and growth. Journal of the American Chemical
Society 114:8435-8442.
Stenger DA, Hickman JJ, Bateman KE, Ravenscroft MS, Ma W, Pancrazio JJ,
Shaffer K, Schaffner AE, Cribbs DH, Cotman CW (1998) Microlithographic
determination of axonal/dendritic polarity in cultured hippocampal neurons. J
Neurosci Methods 82:167-173.
Steward O (1976) Topographic organization of the projections from the entorhinal
area to the hippocampal formation of the rat. J Comp Neurol 167:285-314.
Stoppini L, Duport S, Correges P (1997) A new extracellular multirecording system
for electrophysiological studies: application to hippocampal organotypic
cultures. Journal of Neuroscience Methods 72:23-33.
Syed MM, Lee S, He S, Zhou ZJ (2004) Spontaneous waves in the ventricular zone
of developing mammalian retina. J Neurophysiol 91:1999-2009.
Tai HC, Buettner HM (1998) Neurite outgrowth and growth cone morphology on
micropatterned surfaces. Biotechnol Prog 14:364-370.
Tao H, Zhang LI, Bi G, Poo M (2000) Selective presynaptic propagation of long-
term potentiation in defined neural networks. J Neurosci 20:3233-3243.
153
Tateno T, Jimbo Y, Robinson HP (2005) Spatio-temporal cholinergic modulation in
cultured networks of rat cortical neurons: spontaneous activity. Neuroscience.
Tessier-Lavigne M, Goodman CS (1996) The molecular biology of axon guidance.
Science 274:1123-1133.
Thiebaud P, de Rooij NF, Koudelka-Hep M, Stoppini L (1997) Microelectrode
arrays for electrophysiological monitoring of hippocampal organotypic slice
cultures. IEEE Trans Biomed Eng 44:1159-1163.
Thiebaud P, Beuret C, Koudelka-Hep M, Bove M, Martinoia S, Grattarola M,
Jahnsen H, Rebaudo R, Balestrino M, Zimmer J, Dupont Y (1999) An array
of Pt-tip microelectrodes for extracellular monitoring of activity of brain
slices. Biosens Bioelectron 14:61-65.
Tien HT, Salamon Z, Ottova A (1991) Lipid bilayer-based sensors and biomolecular
electronics. Critical Reviews in Biomedical Engineering 18:323-340.
Tong G, Malenka RC, Nicoll RA (1996) Long-term potentiation in cultures of single
hippocampal granule cells: a presynaptic form of plasticity. Neuron 16:1147-
1157.
Torimitsu K, Kawana A (1990) Selective growth of sensory nerve fibers on metal
oxide pattern in culture. Brain Res Dev Brain Res 51:128-131.
Tucker JB, ed (2000) Toxic Terror : Assessing Terrorist Use of Chemical and
Biological Weapons (Bcsia Studies in International Security). Cambridge:
MIT Press.
Tucker JB (2001) The Proliferation of Chemical and Biological Weapons Materials
and Technologies to State and Sub-State Actors. In: Subcommittee on
Internation Security, Proliferation, and Federal Services. Washington, D.C.:
Chemical and Biological Weapons Resource Page
http://cns.miis.edu/research/cbw/ttuck2.htm.
Tuszynski MH, Gage FH (1995) Maintaining the neuronal phenotype after injury in
the adult CNS. Neurotrophic factors, axonal growth substrates, and gene
therapy. Mol Neurobiol 10:151-167.
Vielmetter J, Stolze B, Bonhoeffer F, Stuermer CA (1990) In vitro assay to test
differential substrate affinities of growing axons and migratory cells. Exp
Brain Res 81:283-287.
Vogt K, Streit J, Dityatev A, Luscher HR (1997) Synaptic plasticity in dissociated
hippocampal cultures: pre- and postsynaptic contributions. Eur J Neurosci
9:1078-1082.
154
Walter J, Kern-Veits B, Huf J, Stolze B, Bonhoeffer F (1987) Recognition of
position-specific properties of tectal cell membranes by retinal axons in vitro.
Development 101:685-696.
Waltz JM (1997) Spinal cord stimulation: a quarter century of development and
investigation. A review of its development and effectiveness in 1,336 cases.
Stereotactic And Functional Neurosurgery 69:288-299.
Warland DK, Reinagel P, Meister M (1997) Decoding visual information from a
population of retinal ganglion cells. J Neurophysiol 78:2336-2350.
Wessberg J, Nicolelis MA (2004) Optimizing a linear algorithm for real-time robotic
control using chronic cortical ensemble recordings in monkeys. J Cogn
Neurosci 16:1022-1035.
Wheeler BC, Novak JL (1986) Current source density estimation using
microelectrode array data from the hippocampal slice preparation. IEEE
Trans Biomed Eng 33:1204-1212.
Wijesuriya DC, Rechnitz GA (1993) Biosensors based on plant and animal tissues.
Biosensors & Bioelectronics 8:155-160.
Williams JC, Rennaker RL, Kipke DR (1999) Long-term neural recording
characteristics of wire microelectrode arrays implanted in cerebral cortex.
Brain Research Brain Research Protocols 4:303-313.
Winson J, Abzug C (1978) Dependence upon behavior of neuronal transmission
from perforant pathway through entorhinal cortex. Brain Res 147:422-427.
Wirth C, Luscher HR (2004) Spatiotemporal evolution of excitation and inhibition in
the rat barrel cortex investigated with multielectrode arrays. J Neurophysiol
91:1635-1647.
Wojciak-Stothard B, Curtis A, Monaghan W, MacDonald K, Wilkinson C (1996)
Guidance and activation of murine macrophages by nanometric scale
topography. Exp Cell Res 223:426-435.
Wolpaw JR, Birbaumer N, Heetderks WJ, McFarland DJ, Peckham PH, Schalk G,
Donchin E, Quatrano LA, Robinson CJ, Vaughan TM (2000) Brain-computer
interface technology: a review of the first international meeting. IEEE Trans
Rehabil Eng 8:164-173.
Wong RO (1998) Calcium imaging and multielectrode recordings of global patterns
of activity in the developing nervous system. Histochem J 30:217-229.
155
Yavin E, Yavin Z (1974) Attachment and culture of dissociated cells from rat
embryo cerebral hemispheres on polylysine-coated surface. J Cell Biol
62:540-546.
Yeckel MF, Berger TW (1990) Feedforward excitation of the hippocampus by
afferents from the entorhinal cortex: Redefinition of the role of the trisynaptic
pathway. Proceedings of the National Academy of Sciences 87:5832-5836.
Yeckel MF, Berger TW (1995) Monosynaptic excitation of hippocampal CA1
pyramidal cells by afferents from the entorhinal cortex. Hippocampus 5:108-
114.
156
APPENDICE A. PUBLICATIONS RESULTING FROM THIS WORK
Brinton RD, Soussou W, Baudry M, Thompson M, Berger TW (2005) The
Biotic/Abiotic Interface: Achievements and Foreseeable Challenges. In:
Towards Replacement Parts for the Brain: Implantable Biomimetic
Electronics as Neural Prostheses (Berger T.W., D.L. G, eds). Cambridge,
MA: MIT Press.
Soussou W, G. G, Han M, Ahuja A, Song D, Hsiao M-C, Wang Z, Tanguay ARJ,
Berger TW (2005) Mapping Spatio-Temporal Electrophysiological Activity
in Hippocampal Slices with Conformal Planar Multielectrode Array. In:
Advances in Network Electrophysiology Using MultiElectrode Arrays
(Baudry M., M. T, eds). NY, NY: Kluwer Academic.
Gholmieh G, Soussou W, Courellis S, Marmarelis VZ, Berger TW, Baudry M (2001)
A biosensor for detecting changes in cognitive processing based on nonlinear
systems analysis. Biosens Bioelectron 16:491-501.
Han M, Gholmieh G, Soussou W, Berger TW, Tanguay AR, Jr. (2002) Conformally-
Mapped Multielectrode Arrays for In-vitro Stimulation and Recording of
Hippocampal Acute Slices. In: The 2nd Joint Meeting of the IEEE
Engineering in Medicine and Biology Society and the Biomedical
Engineering Society, pp 2127-2128. Houston, TX.
Kim MT, Gholmieh G, Soussou W, Ahuja A, Tanguay JAR, Berger TW, Brinton RD
(In Submission) 17-beta estradiol potentiates fEPSP within each subfield of
the hippocampus with greatest potentiation of the associational/commissural
afferents of CA3.
Soussou W, Yoon G, Brinton RD, Berger TW (In Preparation) Correlations Between
Network Morphology and Electrophysiology of Dissociated Neuronal Cells
Cultured on Coated Multielectrode Arrays.
Gholmieh G, Soussou W, Han M, Tanguay JAR, Berger TW (In Submission) High-
Density Conformal Planar Multielectrode Arrays For Hippocampal Slice
Stimulation And Recording.
Abstract (if available)
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Cytoarchitecturally conformal multielectrode arrays for neuroscience and neural prosthetic applications
PDF
Parametric and nonparametric models of short-term synaptic plasticity
PDF
Cue combination and object boundary detection
PDF
Noninvasive automatic detection of epileptogenic regions and networks using MEG measurements
PDF
Three-dimensional reconstruction from serial non-contiguous sections using variational implicit techniques
PDF
Memory capacity of neurons with active dendrites
PDF
The role of reactive oxygen species in neuronal cell death
PDF
Role of protein kinase C in estrogen neuroprotection
PDF
Neural homologies: Principles, databases and modeling
PDF
The role of Schwann cells in synapse formation and synaptic transmission at developing neuromuscular junctions
PDF
The implementation of a DNA computer to solve n-variable 3 CNF SAT problems
PDF
Modeling the nonlinear dynamic interactions among information pathways in the hippocampal neural system
PDF
Glial activation and neurotoxicity: Age-related cytokine expression, ApoJ-induced microglial activation, and role of peroxynitrite in mediating neuron death
PDF
Role of reactive oxygen species in iron-induced and age-related neuronal damage
PDF
17beta-estradiol potentiates and progesterone depresses fEPSP within each subfield of the hippocampus with greatest potentiation of the association commissural afferents of CA3
PDF
Activation and regulation of the taste transduction channel TRPM5
PDF
Time-varying analysis of autonomic cardiovascular control during arousal from sleep
PDF
Functional regulation and trafficking mechanism of rat plasma membrane GABA transporter 1
PDF
Modeling the mirror: Grasp learning and action recognition
PDF
Building a biological hybrid biosensor using nonlinear system analysis of the CA1 hippocampal neural network
Asset Metadata
Creator
Soussou, Walid Victor
(author)
Core Title
Multielectrode arrays as neuron -silicon interfaces
Degree
Doctor of Philosophy
Degree Program
Neuroscience
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
biology, neuroscience,engineering, biomedical,OAI-PMH Harvest
Language
English
Contributor
Digitized by ProQuest
(provenance)
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c16-465914
Unique identifier
UC11340995
Identifier
3219796.pdf (filename),usctheses-c16-465914 (legacy record id)
Legacy Identifier
3219796.pdf
Dmrecord
465914
Document Type
Dissertation
Rights
Soussou, Walid Victor
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA
Tags
biology, neuroscience
engineering, biomedical