Title: Micropatterned Neuronal Networks in Culture: Simplicity or Complexity
1Micropatterned Neuronal Networks in Culture
Simplicity or Complexity
- Bruce C. Wheeler
- Department of Bioengineering
- University of Illinois at Urbana-Champaign
2Recording Multichannel Neuronal Network Activity
Arrays from Multichannel Systems
Localization to Electrodes
Example 92 Neurons are Active
Automated Spike Detection and Sorting (Plexon
Inc.)
3Automated Analysis of Data from Neuronal Networks
Example 92 Neurons are Active
Automated Connectivity Analysis and Diagrams
(under development)
NN Cross-Correlations
Color Coded Activity Maps
4Confluence of Technologies
- Microelectrode Array Technologies
- Thomas (c. 1970), Gross (1977), Pine (1980)
- Used with Cell Culture, Isolated Neural Ganglia
- Novak (c. 1985) Brain Slice
- Perforated, flexible arrays (Boppart 1990)
- Micropatterning Technologies
- Letourneau (1980 -- UV), others -- scratching
glass etc. - Kleinfeld 1989 (photolithography)
- Corey 1991 (laser ablation, photolithography),
- Whitesides (microprinting) 1990s Corey/Branch
(direct printing) - Patterns and Arrays
- Technology Push
- Some of this technology push has become
commercialized and productive for biomedical
research
5Analysis of Array Data -- A Sampling
6Brain Slices Acute and Cultured
Novak, J. L., Wheeler, B. C. (1989).
Two-dimensional current source density analysis
of propagation delays for components of
epileptiform bursts in rat hippocampal slices.
Brain Research, 497, 223-230.
Acute Hippocampal Slice
Cultured Hippocampal Slice
Novak Wheeler
Acute Cerebellar Slice
Ulrich Egert http//www.brainworks.uni-freiburg
.de/ projects/mea/meatools/moviedemo/hcotc
7Tensor Biosciences collaborates with Panasonic, a
division of Matshushita, and the University of
California at Irvine
8Cultured Neurons as Sensors
Evidence of Dose/ Response Relationship (Bursts
per minute)
Guenter Gross U. North Texas http//www.cnns.org/
9Drug Analysis by State Space
Burst Duration vs. its Coefficient of Variation
Gross, Harsch, Rhoades, Goepel Odor, drug toxin
analysis ... Biosens. Bioelect. 12 (5), 373, 1973
10Statistical Properties of Interspike Intervals
Statistical distribution of interspike interval
times (top), interburst interval (left) and power
spectrum (right) for three sizes of networks
Segev, Beneviste, Hulata, Cohen, Palevski, Kapon,
Shapira, Ben-Jacob Long term behavior of
lithographically prepared in vitro neuronal
networks Phys Rev Lett 88(11), 118102, 2002.
11Gravitational Clustering
- Gerstein Aertsen
- Goal Graphical Representation of Functional
Relationships Among a Population of Neurons - Each neuron is represented by an atom in N space
- When a neuron fires, its atom gains charge, then
slowly discharges. - Like charges attract.
- Positions / trajectories are indicative of cell
assemblies.
Gerstein, G. L., Aertsen, A. M. H. J. (1985).
Representation of cooperative firing activity
among simultaneously recorded neurons. J.
Neurophysiol., 54, 1513-1528.
12 Inferred Connections from Spike Trains
Assumed Simulated Network
Projection into Principal Components Plane from
N-1 Space
Gerstein, G. L., Perkel, D. H., Dayhoff, J. E.
(1985). Cooperative Firing Activity in
Simultaneously Recorded Populations of Neurons
Detection and Measurement. J. Neurosci, 5(4),
881-889.
13Multi Dimensional Scaling to Create Maps
(millimeters on a map)
Adapted from Kruskal, J. B., and Wish, M.,
Multidimensional Scaling, p. 31-32, Sage
University paper series on Quantitative
Applications in the Social Sciences, 07-011.
Beverly Hills and London Sage Publications.
Adapted from Borg, I., and Lingoes, J. (1987)
Multi-dimensional Similarity Structure Analysis,
p. 2-4, Springer Verlag, New York
14Goal Map from Neural Correlation
Cross-Correlograms from Simulation
15ISOMAPbuilds map following geodesic distances
Here four simultaneous networks were simulated.
ISOMAP separated the networks and identified
connections within each correctly.
16Analysis of Bursts Across Electrode Arrays
Grey scale, block clustered depiction of
correlation coefficients between all pairs of
electrodes under low (A), medium (B), high (C)
calcium.
Connectivity Diagram corresponding to A, B. Using
Affinities -- a normalized correlation measure
Segev, Baruchi, Hulata, Ben-Jacob Hidden neuronal
correlations in cultured networks Phys Rev Lett
92(11), 118102, 2004.
17Other Correlation Measures
- max correlation over a window
- correlation coefficients
- synchronization index with a reference signal or
reference neuron - information theoretic measures
- The difficult part is finding the right measures
l l l
18Different Learning of Frequent vs. Rare Stimuli
Paradigm Stimulate A often (1/5 Hz), B rarely
(1/50 Hz) Early response (dashed) is different
from later (conditioned) response Rare
response is larger GABA mediated i.e. in the
inhibitory network Single cell responses (left)
show same trend as total population response
(right)
Eytan, Brenner, Marom, Selective Adaptation in
networks of cortical neurons J. Nsci. 23(28)
9349, 2003.
19Learning Reinforcing Long Latency Responses
Paradigm Simulate a pair of electrodes at 1/3
Hz Stop when there is a response in 40-60 msec
window twice in ten trials Repeat after
rest Network learns to respond to that stimulus
Shahaf, Marom Learning in Networks of Cortical
Neurons J. Nsci. 21 (22), 8782. 2001
20Electrical Potentiation of Neurons
Stimulating electrode
potentiated
depressed
Recorded neuron
Jimbo, Tateno, Robinson Simultaneous induction of
pathway-specific potentiation ... Biophys J. 76,
670, 1999.
21Culture / Array as Image Processor / Perceptron
Dense Neuronal Culture on Electrode Array.
Arrows Stimulated electrodes. Right electrical
(action potential) responses
Neurons Learn the Letter L Before Training
little difference After Training recognizes L
as opposed to inverted L
Ruaro, Bonifazi, Torre, Toward the Neurocomputer
... TBME, 52, 3, 371, Mar 2005
Responses to Stimulus Along Top Moving Gaussian
Low Pass Filter
22Animats(Potter et al., GaTech, Caltech, U.
FL.http//www.neuro.gatech.edu/groups/potter
Detect spike rates on electrodes Map rate vector
to display position Map position to stimulus
parameter
Goal Directed Network Control Function
Non-goal directed
Trajectory in Display space
Spike Rate vs. Inter-Stimulus Interval
Not shown Map from clusters to 2-D space,
feedback function
Clusters of the rate vectors (in arbitrary
coordinate system)
23Designing Networks for Geometrical Simplicity and
Control
- E18 Hippocampal Neurons (Sprague-Dawley Rats)
- At 4 days in vitro nearly Glial Free,
- Brain BitsTM
- Serum Free Defined Medium (B27/Neurobasal)
- Low Density (usually 200 cells / mm2)
24Micropatterning Technologies
Printed polylysine
insulator (3um)
metal
neuron
Substrate
Metal electrode
Metals platinum, indium tin oxide, titanium
nitride, gold Insulators silicon nitride,
silicon dioxide, glass, polyimide,
PDMS Permissive polylysine, laminin,
Nonpermissive PEG, chondroitin sulfate,
Microcontact Printing Photoresist
Patterning Microfluidic Deposition Laser
Ablation Microchannels Covalently linked Or
physisorbed
25Cells migrate to the pattern
1 DIV
Number of cells
7 DIV
Days in vitro
14 DIV
Area 0.96 mm2
Arrows indicate cells growing in off-pattern
area
AR0201-G7637
26Electrical Activity Density Dependent, Emerges
with Glia Maturation
Date of 1st Onset of Activity
27Localization enhances recordability
92 discriminable neurons
10 Day Culture, initially localized to 50 um
circles around electrodes
28Recorded Spikes
Occasional high amplitude (100-500 uV) and field
potentials
Typical Mono-, Bi-, and Tri-phasic low amplitude
(20-100 uV)
29 1 2 3
4
1
2
3
4
2
1
3
4
30Fraction of Active Sites Suggests 10 to 20 um is
Close Enough -- for a Cell Body
31Patterned Neuronal Networks Show Normal Behavior
- Examples of
- Propagation of signals
- Variety of Input / Output functions
- Rudimentary learning or plasticity
- Engagement of different circuits with increasing
stimulation
32Basic Geometrical Pattern 40 um wide
lines(Numbers are count of neurons within 30?m
of each electrode)
11 10 5 9 6 5
1 4 10 3 2 8 4 4
1 9 1 9 6 7 2 10
2 10 1 0 3 5 5 5
33Lines Show Independent Network Activity
L I N E 3
Spike Rate (Hz)
L I N E 4
Time (sec)
34Cross Correlograms Suggest Action Potential
Propagation
Reference CH 85, bin 0.1ms, range ?2 ms
600 ?s
300 ?s
Ref.
Reference
?200 ?s
Lag Time (-2 to 2 msec)
35Stimulation of Propagated Activity
4.2 ms
Delays suggest AP propagation
4.4 ms
4.8 ms
36Stimulated Activity Appears to Propagate with Two
Different Conduction Velocities
16
14
y 0.0117x 4
Electrically coupled?
12
2
R
0.9859
10
Velocity 0.09 m/s
Delay (0-16 ms)
8
6
y 0.0016x 3.8579
4
Action potential?
2
0.9538
R
2
Velocity 0.63 m/s
0
0
200
400
600
800
1000
1200
Distance (0 - 1200 um)
37Stimulation Evokes a Variety of Responses
Spike probability
Time (0 to 0.5 sec)
388 Days
Stimulation Shows Incremental Network Involvement
20 Days
39STM 0.6 V
40STM 0.8 V
S
41STM 1.0 V
S
42STM 1.2 V
S
43STM 1.4 V
S
44Input output relation
Fan Out
Fan In
4527 day old culture, stains for neurofilament,
GFAP, nuclei
46 Stimulating 41 Recording from 44
47 Stimulating 44 Recording from 41
48 Stimulating 43 Recording from 41
41
41
43
43
49 Stimulating 43 Recording from 44
3 ms
6 ms
43
44
50 Spontaneous Activity Cross correlogram (41 44)
Bin 0.5 ms
51Proposed Connectivity
Units milliseconds
52- Biological Complexity Complicates Design and
Interpretation
53Spontaneous Rates Vary Across Array and Age
X spike rate Y time in days position
corresponds to electrode
54Any Pattern is Possible
But Survival/Activity Appear to Depend On
Connectedness
2-25-05ilines1div21-3MPSPEG
55Glia are Essential to Neural Activity
56Glia Colocalize with Neurons
Versus days in culture Glia within 2 um of a
neuron of Glia on Neurons of Neuron area
covered by Glia
Ratio
26 days 87 of Glia stay within 2um from neurons
13 days 95.6 of Glia with neurons Glia/Neuron
0.2
- Red Glia, Green Neuron,
- Pink Glia Neuron
57Even Simple Networks Have High Synapse Count
22 days
15 days
8 days
Synapse/ Neuron
670
450
300
58Neuron/Glia Locations Vary with Respect to
Substrate/Electrodes
Red, yellow Glial layer Green, blue Neurites
59Apparent Neural Progenitor Cells Propagate from
the Neural Bundle
60Two Populations by Drug Response -- two cell
types, or -- two states of maturation?
Characterize by Spontaneous Activity
Response to Stimulation Depends More on Character
of Cells Stimulated
61Complexity Issues for Cultured Neurons
- How many bbiological features can be controlled?
- How many does one need to control?
- An Array makes 60 simultaneous experiments
possible - Combinatoric explosion with pairwise correlations
- Working to automate stimulation / recording
capabilities - Need to develop paradigms for looking for network
connectivity
62Thanks to ...
Greg Brewers Lab (Southern Illinois Univ. School
of Medicine) At the UIUC Yoonkey Nam, John
Chang, David Khatami, Rudi Scharnweber, Betty
Ujhelyi, Kate Musick, Faramarz Edalat, Kathy
Motsegood. Former Students Joe Corey, Darren
Branch NSF EIA 0130828 (BITS Program) NIH R21
NS 38617 R55 RR 13320 Subcontract to R01
EB000786 at Georgia Tech (BRP)
Go Illini!