Transcript Multielectrode Arrays - Florida State University

Multielectrode Arrays
Joe Kostansek, Greg Loney, Ariel
Simonton
•“Bundles” of electrodes
situated together to facilitate
recording of field potentials
and in some instances
stimulation of brain areas
•These bundles are capable of
recording from hundreds to
thousands of neurons at a time
and are typically considered
chronic
•Typically PCA is used to
segregated individual neurons
or individual neuronal
phenotypes
•1958 – Strumwasser
•Utilized single 80-mm stainless steel wires to
record from awake, behaving squirrels
•Recordings lasted for a week or longer
•Strumwasser concluded that the constant
waveform and amplitude implied that he was
recording from the same neurons repeatedly
•Techniques have largely remained unchanged
for the last several decades
•Due to the small diameter of microwires, position is not fixed and
fluctuates with movement, BP, etc…
•With the advent of silicon based electronics and reduced price in the
1970’s, fixed arrays now become possible
•Due to strength of silicon and thus increased surface tension, more
electrodes are able to be implemented and implanted
•Michigan-Array is one of the first silicon –based arrays. Due to increased
density of probes, researchers are now able to record from soma and
dendrites simultaneously
•Utah-array: Developed in the late 1980’s-early 1990’s
•Each Individual silicon ‘tong’ contains a platinum insulated microelectrode
thus allowing for higher resolution at each individual recording point
•Notice the scaled position of each tong. Due to increased resolution,
researchers are know able to record from various layers of cortex, which
may contain different cell types, with a greater degree of accuracy
•Rigidity of array causes some problems in vivo
•Polymide-based electrode
cuff
•Highly flexible
•Rings of electrodes allow for
continuous recordings,
especially of peripheral nerves
•Sieve electrode
•Nerve is cut and allowed to
regenerated through the holes
in the ‘sieve’
•Often coated with BDNF or
other NTF’s in order to
facilitate regeneration and
approach
Multielectrode Arrays
in vivo
Advantages
• Ability to record or stimulate hundreds to thousands of
neurons
• Can correlate activity across many neurons
• Can be used in awake, behaving animals to study higher
processes:
– Memory formation
– Sensory integration
• Chronic or Acute
• Potential human clinical uses:
– Deep brain stimulation
– Seizure control
– Restoring motor control
Disadvantages
• Cannot look at single ion channels
– Better suited for recording action potentials
• Typically implanted chronically
– Can cause immune response and inflammation
– May lead to neurodegenerative effects
• Not a disadvantage, but: Provides a LOT of
data, it’s necessary to have sophisticated
sorting software to properly analyze results
Analysis of Multielectrode Array Data
Sample recording from the amygdala
• Individual spikes are
recorded into a
program and can be
bundled (as seen
here) depending on
waveform properties.
• Some arrays have the
ability to individually
move each wire.
– Beneficial for
recording from
multiple layers of
cortex
Single Neuron Analysis
• First step: analysis of individual
recorded neurons
– Commonly used: raster plots and
peristimulus time histograms
(PSTH)
• These figures graph individual
firing patterns over time so that
they may be correlated to either
a stimulus or behavioral event.
• Example data: Shows an
individual neuron’s response to
mechanical stimulation of a digit
on the hand of an owl monkey
Neural Ensemble Response
• Next step: Visualizing neural activity as a
whole
– Population Peristimulus Time Histograms (PPSTH)
– Spatiotemporal Maps
– Linear Time Series Analysis
– Artificial Neural Networks
Population Peristimulus Time
Histogram
• X-axis
– Individual neurons
arranged
rostral/caudal or
some other method
of organization
• Y-axis
– Peristimulus time
• Z-axis
– Instantaneous firing
rate (spikes/sec)
– Spike counts per
time bin
Spatiotemporal Population Map
• Magnitude of neuronal firing = # of standard
deviations away from spontaneous firing rate
Linear Time Series Analysis
• Best for continuous
stimuli or behavioral
variables
• Uses neuronal inputs
and behavioral outputs
in a modified linear
regression equation
• X(t) matrix
– Columns: Single
neurons
– Rows: Time segments
• If data is 3D, then x,y,z
is analyzed in a separate
matrix and incorporated
with X(t) matrix
Artificial Neural Networks (ANNs)
• Very important tool to look at neural
ensemble activity in relation to sensory input
or behavioral output
• Require no a priori assumptions
• Useful for both categorical and continuous
stimuli
Optimized Learning Vector
Quantization
• Three layers: input, hidden, output
– Input: Data
– Hidden: Two artificial neural units
– Output: Prediction
• A method of analysis in which the program
“learns” which neural patterns are associated
with a given output
• Can follow with principal component analysis
Summary
• Multielectrode arrays are very useful tools for
recording data from many neurons
• Creates a spatiotemporal summary of neural
activity
• Information recorded from multielectrode arrays
may be analyzed for individual cells as well as
populations
• Advances in understanding of sensory
perception, learning and memory, and other
higher processes can be contributed to the
introduction of multielectrode arrays
Nucleus accumbens neurons are
innately tuned for rewarding and
aversive taste stimuli…
Roitman, MF, Wheeler, RA & Carelli,
RM
Methods
• Rats were implanted with intraoral catheters, and
microelectrode arrays in the anterior digastric
muscle and nucles accumbens
• Rats received 30-cue trials (light & tone) of both
sucrose and QHCl, presented after a variable
delay
• EMG activity recorded from anterior digastric
muscle
• Individual neuronal activity recorded from nucles
accumbens
A & B: EMG activity data in response to intraoral
infusions of sucrose and QHCl, respectively
C & D: Sample EMG traces indicating “learning”
of taste-cue pairings for sucrose and QHCl,
respectively
E & F: Average latency to first EMG burst for cuepaired sucrose and QHCl, respectively. Bars below
the line indicate that first burst occurred during
cue, bars above the line indicate that first burst
occurred during infusion
G: Note large amplitude bursts following QHCl
infusion
H: EMG activity displayed a trend to decrease as
a function of infusion duration C
Four classes of cells: Raster plot and spike
frequency bins (histograms) of four
representative cells
A: Sucrose inhibitory
B: Sucrose excitatory
C: QHCl inhibitory
D: QHCl excitatory
Average firing rates of the 4 identified classes of
cells:
Sucrose inhibitory (39 0f 102)
Sucrose excitatory (13 of 102)
QHCl inhibitory (10 of 98)
QHCl excitatory (30 of 98)
Representative sucrose inhibitory cell: Note the opposite pattern of firing for
both types of stimuli
Firing rates for all four identified cells
plotted against EMG activity for each
100 ms bin of pre-infusion (6 s; black)
and post-infusion (6 s; gray).
Sucrose inhibitory: negatively correlated
Sucrose excitatory: positively correlated
QHCl inhibitory: negatively correlated
QHCl excitatory: positively correlated
Four new classes of cells: Raster plot and spike
frequency bins (histograms) of four
representative cells
A: Sucrose-cue inhibitory
B: Sucrose-cue excitatory
C: QHCl-cue inhibitory
D: QHCl-cue excitatory
Average firing rates of the 4 identified new
classes of cells:
Sucrose inhibitory (16 0f 102)
Sucrose excitatory (26 of 102)
QHCl inhibitory (12 of 98)
QHCl excitatory (27 of 98)
Cue-invoked firing is significantly correlated with “learning”
A: Firing rate increases as a function of trial repetitions
B: Firing rate is negatively correlated with latency to first burst
Conclusions
• Individual, naïve neurons in the nucleus accumbens demonstrate unique
patterns of firing to prototypically rewarding an aversive stimuli and are hihgly
correlated with reflexive behaviors associated with these stimuli
• These same neurons demonstrate a reversed pattern of activation in
response to stimuli of the opposite valence
• Neuronal activation demonstrates a pattern of adaptation reminiscent of
learning
• Nucleus accumbens neurons may be innately tuned to encode predictions
and aggregate motor output associated with rewarding and aversive taste
stimuli
Cortical Excitation and Inhibition
following Focal Traumatic
Brain Injury
Ming-Chieh Ding, Qi Wang, Eng H. Lo, and
Garrett B. Stanley
Background
• Brain injuries causing swelling of the cortex leads
to:
–
–
–
–
Increased extracellular K+
Altered firing rates
Neuronal injury/death
Stroke
• Changes from brain injury can lead to overall
changes in network inhibition and excitation
• Purpose: To assess the effects of compression
injury on excitatory and inhibitory networks in
vivo.
Methods
• Male Long-Evans rats
• Microarray is implanted into the barrel cortex (primary
somatosensory cortex)
– 90 minute recovery time
• Array
–
–
–
–
8 x 8 silicon electrodes
1mm length
400um spacing
100-400 kohm impedance
• Stimulus
– Mechanical stimulation of vibrissae
• Compression
– 1mm steel cylinder, 1mm of compression
A. Experimental setup
B. PSTH of all electrodes with stimulus of C2 vibrissae
C. Recordings from two single electrode channels
D. Cortical activation
A: Stimulus delivery
B: Channels
responsive to
vibrissae
deflection at
different interdeflection
intervals
D: Attenuation of
second stimulus
response
depending on IDI
Neurons have an attenuated
response to the second
stimulus at shorter IDIs
Neural response after compression
After compression, there is a slow recovery of
baseline neural activity.
Postcompression intensity often exceeded
precompression intensity
Neural response after compression
Increased activity post-compression, after a
specific amount of time.
Neuron Response Profiles
• Principal Channel
– Channel with the largest response magnitude to a
given stimulus
• Precompression Significant
– Significant channel before compression
• Postcompression Significant
– Significant channel post-compression, but not precompression
• Spike magnitudes significantly different between
principal, pre-compression and post-compression
significant channels
• Post-compression significant channels showed the
largest relative change in spike magnitude over time
• “Paired pulse” whisker stimulation after
compression, 50ms IDI
• Before compression, response at this IDI is
suppressed
• After compression, response to the same stimulus
IDI eventually leads to excitation
A: Principal channel
latency does not
change before and
after stimulus
C: Vector strength – a
measure of temporal
precision – does not
change in principal or
precompression
neurons
D: Vector strength
increases in
postcompression
significant channels
Conclusions
• After compression in the rat brain, neurons
displayed a change in response to vibrissae
stimulation
– Neurons displayed a period of inhibition after
compression, followed by excitation greater than seen
pre-compression.
– Some neurons were not responsive at all before
compression, but became active after compression
• Na+ and K+ levels unbalance after injury
– May cause lower threshold for depolarization
– Hyperexcitability
In vitro Multi(Micro)-Electrode Arrays (MEAs)
What is it?
• Instead of implanting into the organism and dealing with the difficulties of live animals, the in vitro approach
allows for cultured cells/tissues to be used.
• First done in myoneural junctions and gastropods (1980s, linear method)  technology has improved
technique dramatically (planar, 3D, perforated, thin, etc.). Extracellular recordings  field potentials, spikes
• Two types:
•Acute slices  neurons dissociated, spontaneously form networks
•Organotypic slices **  network integrity remains
The cells/tissues grow
directly onto the recording
electrodes
Advantages/Disadvantages
•Long term recordings (weeks to months if done carefully)
•Works like most electrophyisology  differences = array, analysis of data
•Multiple electrodes – some experimental, some controls, simultaneously stimulate/record from
different sites
•Non-invasive to the cell (no rupturing of cells)
•High spatial resolution (very low for single cells)
•Expensive, tough to maintain/clean
Want to start using this technique?
Microscope - One challenge among in vitro MEAs has been imaging them with microscopes that use high
power lenses, requiring low working distances on the order of micrometers.
In order to avoid this problem, “thin”-MEAs have been created using cover slip glass.
These arrays are approximately 180 μm allowing them to be used with high-power lenses.
Want to start using this technique?
Temperature Controller
Amplifier
Pelltier device
Heating element
60 channel
Upright/inverted
Blanking circuit
64 Channel Stimulator
Use to stimulate your
cultures
With a variety of factors –
electricity,
Solutions, etc.
MEA and Base
PCI Data Acquisition Card
Capable of recording from up to 128 channels simultaneously
Software, Air tables, computers, oscilloscopes, audio, amplifiers,
perfusion
Exceeding transfer
rates of 6 MHz.
systems, analog/digital converters etc.
Channels sampled at 50 kHz
Want to start using this technique?
The MEA
Standard Set-ups:
8 x 8 or 6 x 10 electrodes. Titanium oxide electrodes that have diameters between 10 and 30 μm. These arrays are
normally used for single-cell cultures or acute brain slices.
60 electrodes are split into 6 x 5 arrays separated by 500 μm. Electrodes within a group are separated by 30 um with
diameters of 10 μm. These can be used to examine local responses of neurons while also studying functional
connectivity of organotypic slices **
Want good spatial resolution? HD-MEA is your answer. It allows signals sent over a long distance to be taken with higher
precision. These arrays usually have a square grid pattern of 256 electrodes that cover an area of 2.8 by 2.8 mm.
Other types:
Perforated: The perforated MEA design applies negative pressure to openings in the substrate so that
tissue slices can be positioned on the electrodes to enhance contact and recorded signals.
Thin, multi-welled, hexagonal, 3D (penetrates farther into cultures)
Want to start using this technique?
Too expensive to buy, you can make your own!
Want to start using this technique?
So what now?
neural networks
Animat
epilepsy
synaptic plasticity (LTP, PPF, etc.)
development
regeneration
biological rhythms **
network oscillators
cardiac physiology
robotics
Other techniques:
Histology/ICC
Calcium imaging
Patch-clamp
optogenetics
A computer generated animal, in a virtual world. Cortical neurons from rats are dissociated and placed on a MEA
capable of both recording and stimulating neural activity.
Distributed patterns of neural activity are used to control the animat’s behavior in a simulated environment.
The computer acts as its sensory system providing electrical feedback to the network about the Animat’s movement
within its environment. Changes in behavior  neural plasticity.
http://upload.wikimedia.org/wikipedia/commons/thumb/5/55/Circadian_rhythm_labeled.jpg/350px-Circadian_rhythm_labeled.jpg
Some Useful Background Knowledge
Primary endogenous oscillator that controls circadian rhythms of numerous behavioural, endocrine and
physiological processes.
The
basis
cell-autonomous
circadian
oscillations
are positive
andreinforces
negative feedback
loops as responds
shown to
The
SCNfor
network
synchronizes
its component
cellular
oscillators,
their oscillations,
here.
rhythms
in protein
expression of
several clock
lightThese
inputloops
(RHT)drive
by altering
their
phase distribution,
increases
their components.
robustness to genetic perturbations,
and enhances their precision.
http://people.usd.edu/~cliff/Courses/Behavioral%20Neuroscience/Biorhythm/BRfigs/BRAfferent%20SCN%20figures.html
Welsh et al, Suprachiasmatic Nucleus: Cell Autonomy and Network Properties. Ann. Rev. Physiol. 2010
Some Useful Background Knowledge
Glycine
Glycine
Welsh et al, Suprachiasmatic Nucleus: Cell Autonomy and Network Properties. Ann. Rev. Physiol. 2010
Glycine is present in the SCN  can act as a classical inhibitory NT and an excitatory neuromodulator
Circadian release of glycine
In slices, high concentrations of glycine can reset the clock
What is glycine’s function in the SCN?
Figure 1. Voltage-clamp recording of glycine-induced current in neurons of acute SCN slices
A – Application of glycine (5 s) at a holding
potential of 0mV generated an outward current in
83% of neurons – remaining neurons were
insensitive.
Concentration dependent effects (threshold 10µM).
Characteristics of responses.
B – Concentration response curve. Fitted to Hill
equation – EC50 at 780µM
C - Extracellular recordings – cell-attached mode –
showed a concentration dependent suppression of
spontaneous firing activity in SCN neurons sensitive
to glycine.
D – Agonists of glycine receptors (beta-alanine and
taurine) induce currents with similar characteristics
to glycine-induced currents (applied to the same
cell)
Evidence that SCN neurons in acute coronal brain
slices of mice exhibit a glycine-induced current.
Figure 2. Glycine activates strychnine-sensitive GlyRs in SCN
A – Typical response to 1mM glycine (outward
current) at a holding potential at 0 mV (upper
trace).
Typical response is reduced by the
coapplication of strychnine (5µM) (middle
trace)
Recovery after washout (lower trace)
B – Extracellular recordings in an acute SCN
slice.
Strychnine reduced the duration of the
inhibition of spontaneous electrical activity
C – Glycine antagonists strychnine (5µM),
PMBA (100µM) and ginkgolide B (1µM) reduce
glycine-induced currents by 51, 56 and 34%
respectively.
D - Suppression of the amplitude of the current
decreases with increased concentrations of
strychnine.
E - Gabazine knocks out the GABA component
of the glycine responses.
Figure 3. Ion selectivity and specificity of the glycine-induced current
A & B - Distinguish currents induced by glycine
and GABAA receptors – application of 100µM
GABA and 1mM glycine.
I-V relationship – reversal potentials of the cells
used in A: GABA-induced: -50.1mV, glycineinduced: -49.2mV. Average glycine trace: 47.1mV
Nernst potential for chloride at their conditions:
-51mV.
E - Strychnine + GABA had no effect on GABA
 effects
strychnine
are caused
Cresponses
– Co-application
of of
saturating
amounts
of
by a specific
blockresulted
of GlyR.in currents that were
GABA
and glycine
Gabazine
suppressed
GABA
currents.
smaller
than
the sum of
glycine
+ GABA. Slow
deactivation – glycine current.
F – Glycine does not act on nAChRs – no
difference in current amplitude when
D - 3/71 neurons tested for GABA and glycine
tubocurarine (blocker) is applied.
were insensitive to GABA but yielded a glycineinduced current.
Methods – in vitro Multi(Micro) Electrode Arrays (MEA’s)
Organotypic Slices
250-350µm thick coronal slices containing the SCN from 2-5 day old animals
Placed in a culture dish with culture medium (1mL DMEM/F12 – supplemented with
10% fetal calf serum, 2.5mM glutamax, 10mM Hepes, and 100µg/mL penicillin-streptomycin) which was
exchanged 3x/week
Incubated @37C in 5%CO2-95% air for more than 2 weeks
Before recording, the slice was placed onto a nitrocellulose-coated MEA (Multichannel
Systems, Reutlingen)
Recording medium: same as culture but Hepes was elevated to 20mM and the NaHCO3 was reduced to
0.56 g/L. Exchanged continously at 20µL/min using SP 260PZ syringe pump (WPI).
Maintained on MEAs for up to 3 weeks under flow-through culture conditions
Can be kept for weeks in culture
Can monitor the output signal of the circadian clock (electrical activity) for periods up to 3 weeks
http://www.staff.uni-mainz.de/golbs/Methods.html
Methods – in vitro Multi(Micro) Electrode Arrays (MEA’s)
Multi-electrode Array Recordings
Recorded long-term firing rate from organotypic slices of SCN/PVN using a MEA-1060 recording system
(Multichannel Systems, Reutlingen)
Two types of HD-MEA used with two different layouts:
Two fields of 30 electrodes with a diameter of 10µm and 30µm spacing – fields separated by 500µm
Only one field covered by the SCN (rendering other tissue (PVN) covered by other field)
One field of 60 electrodes with 10µm diameter and 40µm spacing
SCN covered entire field in this case
Extracellular signals amplified 1200x and sampled at 32kHz on 60 channels simultaneously
Noise detected and removed by threshold algorithms
APs exceeding said voltage threshold were digitized and stored as time-stamped spike cut-outs using the
MC Rack software (Multichannel Systems).
Ehab Tousson and Hilmar Meissl, 2004, J Neuroscience
Figure 4. Glycine induced changes in the firing rate of SCN neurons in organotypic cultures
A – An increase in firing rate of SCN neurons due
to application of 1mM glycine
C – A decrease in firing rate of SCN neurons due to
application of 1mM glycine
 Suggests a counteraction of the effect of
glycine on glycine receptors. Both responses are
found throughout the circadian cycle
B
– In
cells thatinwere
excited
glycine, application
Excitation
the SCN
andby
inhibition
in the PVN.
ofPossible
5µM strychnine
reducedintheir
spikingtoactivity
that differences
response
glycine
Slice
culture
has
rhythmic
neuronal
firing
and other hypothalamus cells could depend on
circadian time.23.86±0.37h
D – The proportion of cells that were inhibited by
glycine was more prominent at CT 4 than at CT 16
(24 vs 6%)
At CT 4 a small subset of SCN neurons (5%) had a
biphasic response to glycine
E & F – Simultaneous recording from the SCN and
one of its targets (PVN) using high-density MEA’s
with two recording fields revealed opposite
responses to glycine.
Figure 5. Glycine phase-shifts the circadian rhythm of the firing rate of SCN neurons
A – Vehicle application (aCSF) with no resulting
phase shift (0.2±0.1 h) in circadian oscillation of
the firing frequency.
B – A phase advance (1.7±0.2 h) resulting from
the application of 1mM glycine 3 h before (CT4)
activity peak (CT7).
Circadian activity
the firing
rate
of the cells
were from
C - Aofphase
delay
(-1.4±0.2
h) resulting
measured for the
3 days.
application of 1mM glycine at CT 16 –
Glycine
the ability
to phase-shift
rhythmic
shortlyhas
before
the nadir
of SCN neuronal
Then 1mM glycine
was
applied
to
the
bath
and
the
activity
neuronal
of
activity activity in the SCN by activation
was measured
for
4
days.
strychnine-sensitive glycine receptors.
D – Phase
response
histogram
for vehicleonand
Phase shifts were
calculated
for the
activity recorded
1mM glycine
atas
CTfor
4 and
16. Phase
advance
individual electrodes
as well
the CT
average
activity
of
at
CT
4
and
phase
delay
at
CT
16.
all electrodes (in gray)  both methods showed similar
results.
E – Phase response histogram for glycine
coapplied with glycine receptor antagonists
strychnine and PMBA at CT 4 and CT 16
(applied for 5 s before application of glycine)
Conclusions/Thoughts
Glycine is able to phase-shift rhythmic neuronal activity in the master clock by the activation of
strychnine-sensitive glycine receptors
Glycine can function as both an inhibitory and excitatory NT in the SCN depending on circadian time
(possible mechanism – circadian fluctuation of the chloride equilibrium potential)
A weak glycinergic innervation of the SCN, as well as intrinsic release of glycine from the SCN could
lead to a precise fine-tuning of GABA- and NMDA-mediated synchronization and influence phase
resetting of the clock.
MEA allowed the researchers to observe long-term oscillations in the SCN with/without the treatment of
the SCN with glycine/drugs
The organotypic slices of the SCN/PVN allowed the researchers to have a in vitro system that was very
close to the in vivo system
They were able to record throughout the entire SCN system (core/shell) and even extensions to the PVN to
see how one system affects the other