Content uploaded by Guenter W Gross
Author content
All content in this area was uploaded by Guenter W Gross
Content may be subject to copyright.
1512 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 56, NO. 5, MAY 2009
Magnetic Stimulation and Depression of Mammalian
Networks in Primary Neuronal Cell Cultures
Jochen F. Meyer*, Bernhard Wolf, and Guenter W. Gross
Abstract—For transcranial magnetic stimulation (TMS), the
coupling of induced electric fields with neurons in gray matter is
not well understood. There is little information on optimal stimula-
tion parameters and on basic cellular mechanisms. For this reason,
magnetic stimulation of spontaneously active neuronal networks,
grown on microelectrode arrays in culture, was employed as a test
environment. This allowed use of smaller coils and the continual
monitoring of network action potential (AP) activity before, during,
and for long periods after stimulation. Biphasic, rectangular, and
500 µs long pulses were used at mean pulse frequencies (MPFs)
ranging from 3 to 100 Hz on both spinal cord (SC) and frontal
cortex (FC) cultures. Contrary to stimulation of organized fiber
bundles, APs were not elicited directly. Responses were predomi-
nantly inhibitory, dose dependent, with onset times between 10 s
and several minutes. Spinal networks showed a greater sensitivity
to activity suppression. Under pharmacological disinhibition, some
excitation was seen at low pulse frequencies. FC cultures showed
greater excitatory responses than SC networks. The observed pri-
mary inhibitory responses imply interference with synaptic exocy-
tosis mechanisms. With 20 000 pulses at 10 Hz, 40% inhibition was
maintained for over 30 min with full recovery, suggesting possible
application to nonchemical, noninvasive pain management.
Index Terms—Biomedical measurements, magnetic field ef-
fects, microelectrode array recording, nervous system, neuronal
networks.
I. INTRODUCTION
MUCH speculation has been raised about the mechanisms
underlying the effects of transcranial and peripheral
magnetic nerve stimulation. However, in the transcranial do-
main, the complexity of the tissue structures has prevented a
good understanding of mechanisms and considerable uncer-
tainty exists about biochemical and functional changes in cells
while alternating electromagnetic fields are imposed. Further-
more, there is little data on which stimulation patterns optimize
responses. TMS stimulators and coils for human application
cannot repeat their stimulation pulses at frequencies exceeding
30–40 Hz, and there is a restriction on the maximum total ap-
plication time of these high-intensity stimulations, due to coil
heating [1]. Currently, behavioral TMS effects on the brain are
usually short lived and their underlying mechanisms are not un-
derstood [2]–[4]. Different types of TMS (single pulse, paired
Manuscript received June 25, 2008. First published February 6, 2009; current
version published May 22, 2009. This work was supported in part by the Heinz
Nixdorf Foundation and in part by the Charles Bowen Memorial Endowment to
the Center for Network Neuroscience (CNNS). Asterisk indicates correspond-
ing author.
*J. F. Meyer is with the Department of Medical Electronics, Technical Uni-
versity of Munich, Munich 80333, Germany (e-mail: meyer@tum.de).
B. Wolf is with the Department of Medical Electronics, Technical University
of Munich, Munich 80333, Germany (e-mail: wolf@tum.de).
G. W. Gross is with the Department of Biological Sciences, University of
North Texas, Denton, TX 76203 USA (e-mail: gwgross@cnns.org).
Digital Object Identifier 10.1109/TBME.2009.2013961
pulse, repetitive) affect relatively large brain areas and complex
neuronal networks, making a determination of mechanisms dif-
ficult. Kobayashi and Pascual-Leone point out that “studies to
date have not provided enough data to establish the clinical
indication for a systematic application of TMS as a diagnos-
tic or therapeutic tool in any neurological or psychiatric dis-
ease” [5], [6]. The present study reports the use of spontaneously
active primary neuronal cultures grown on microelectrode ar-
rays as the target for magnetic stimulation. The neuronal cell
layer with multidirectional axonal connections partially mimics
gray matter in the cortex or spinal cord [7], and modulation of
neuronal network activity can be used as a direct measurement
of the magnetic field impact. Such changes in the spontaneous
activity allow a quantification of the effects of alternating mag-
netic fields by plotting the temporal evolution of spike activity
before, during, and after the stimulation, and by allowing nu-
merous applications to the same network over a period of days.
Magnetic field strengths were in the range of those produced by
commercially available TMS stimulators and coils [up to 0.6 T
at the point of stimulation (see Fig. 4)]. However, a greater vari-
ety of stimulation patterns and longer application times could be
used. Maximum applied pulse frequencies were 300 Hz, which
represent a large change over conventional TMS stimulation.
II. MATERIALS AND METHODS
A. Neuronal Cell Cultures
Frontal cortex or spinal cord tissues were dissected from
mouse embryos at 15–16 days gestation. Cells were mechan-
ically and enzymatically dissociated with papain (Roche Ap-
plied Science, Indianapolis, IN) and then seeded on poly-d-
lysine-coated microelectrode array (MEA) plates made of glass,
indium-tin-oxide (ITO) conductors, and a methyltrimethoxysi-
lane insulation (Center for Network Neuroscience, University
of North Texas, Denton, TX). The 64-electrode recording ma-
trix forms a 1-mm square area in the middle of the MEA plate.
Electrode recording sites are 15 µm in diameter, spaced equidis-
tantly 125 µm apart, and are electrochemically gold plated
to reduce electrode impedances to 0.8 ±0.1 (SD) MΩ.For
spinal cord cells, minimum essential medium (Sigma Aldrich,
St. Louis, MO) with 10% horse serum (Sigma Aldrich) was
used. Frontal cortex cultures were raised in Dulbecco’s mod-
ified eagle medium with 5% horse serum. Cell cultures were
incubated at 37 ◦C under a 10% CO2atmosphere for at least
four weeks, until mature networks had developed. During that
time, half of the cell medium was exchanged with fresh medium
every three to four days [8]. Maturity is defined in terms of op-
timization of active electrodes, stability of activity patterns, and
minimal morphological changes in cell positions [9].
0018-9294/$25.00 © 2009 IEEE
Authorized licensed use limited to: University of North Texas. Downloaded on July 13, 2009 at 13:35 from IEEE Xplore. Restrictions apply.
MEYER et al.: MAGNETIC STIMULATION AND DEPRESSION OF MAMMALIAN NETWORKS IN PRIMARY NEURONAL CELL CULTURES 1513
Fig. 1. Life-support system for cell culture chambers (bottom middle). Line
1: 10% CO2and 90% air supply for pH stability in the medium reservoir.
Line 2: medium supply. Line 3: sterile water supply for osmolarity control
in the medium reservoir. CO2-impermeable Pharmed tubing was used. The
supply medium was maintained between 39 and 40 ◦C to eliminate gas bubble
formation in the culture chamber maintained at 35 ◦C.
During experiments, cell cultures were kept alive by means of
a closed chamber setup (Fig. 1) with constant flow of heated, pH-
and osmolarity-controlled media, at flow rates of ∼25 µL/min,
allowing for continuous removal of metabolites with relatively
low shear forces [10]. This procedure maintained stable refer-
ence activity over a period of three to four days, so that long
recovery periods between stimulations could be used [10]. The
cell chamber consisted of an aluminum base plate with attached
heating resistors and a stainless steel chamber block with a
70-µm thick, 2-cm-diameter glass window glued to the bottom
of this component with biocompatible silicone sealer. This pro-
vided a vertical distance of 200 µm between the cell culture and
the window for the slow medium circulation. The cell chamber
setup was placed on a heated inverted phase contrast microscope
stage for optical control. Fig. 1 shows a schematic diagram of the
life-support components and medium flow. To minimize bubble
formation from outgassing in the closed chamber, the recording
temperature was maintained at 35 ◦C and the supply chamber
medium was raised to 39 ◦C–40 ◦C.
Disinhibition of frontal cortex (FC) and spinal cord (SC)
cultures was achieved by addition of 40 µM bicuculline and
40 µM bicuculline in combination with 1 µM strychnine, respec-
tively. These concentrations reliably produced stable, rhythmic
bursting activity in both tissue types without having deleterious
effects on the cells [11], [12].
B. Stimulation Environment
The source of the stimulation signals was a 200-W mono am-
plifier (Velleman, Fort Worth, TX) fed by a standard PC sound-
card output. The stimulation files were designed by the author
in the .wav format and replayed with the sound-editing program
Goldwave. Fig. 2 shows a screenshot of the induced stimuli as
picked up by the Plexon amplifier system and software. Tem-
plate matching software allowed rejection of stimulus artifacts
on the channels with neural activity. The design of the stimu-
lation coil was based on the intention of achieving a maximum
magnetic flux density (and thus a high electric field strength).
Several different coil geometries were simulated by means of
finite element electromagnetic CAD software (EM Studio, CST
Fig. 2. (a) and (b) Examples of several action potential waveshapes recognized
by the program based on the SSQD template matching. (c) Stimulation artifact:
Rectangular pulse, 3 mV amplitude induced in the ITO conductors. Artifacts
caused saturation of the Plexon amplifiers only above a repetition frequency of
350 Hz, which was not used here. Thus, directly evoked network action poten-
tials would have been recorded. Stimulation pulses were always recorded on
one channel for accurate control of stimulation start and end in the offline data
analysis.
Fig. 3. (Left) Schematic of coil made of pure iron with 0.22-mm diameter
copper wiring (200 turns). Assembly was designed to allow positioning below
and above the cell culture. “I” and “B” denote current and magnetic field
directions, respectively. (Right) Detail (cross section): (a) Glass substrate of
MMEP (1 mm), (b) neuronal network layer (10–15 µm), (c) culture medium
(200 µm), (d) glass coverslip (70 µm), (e) iron core (2 mm diameter), and (f)
copper wire (0.22 mm diameter).
GmbH, Darmstadt, Germany). According to the obtained sim-
ulation results, a coil design was chosen that allowed the cell
culture to be placed between the two poles of the coil. Thus, a
uniform high field exposure of the cells in the recording elec-
trode matrix could be assured. Fig. 3 shows a schematic of the
stimulation coil, designed and fabricated by the first author. The
wiring was applied in such a way that the current direction was
the same for both the upper and the lower poles of the coil, ensur-
ing that the alternating fields of both poles did not annihilate each
other. Fig. 3(b) shows details of the structures at the site of the
cell culture. Cells growing on the 1-mm2recording matrix were
penetrated by relatively homogeneous field strengths. The gra-
dient of the magnetic field was nearly linear in a 2-mm diameter
area centered on the 1-mm diameter-recording matrix (Fig. 4).
Experiments were performed with single networks that were
on the recording stations for time periods ranging from 8 h to
six days. Each experiment usually consisted of many exposure
episodes (maximum 27) that generally used different variables.
However, for each single episode, the exposure variables re-
mained constant. The exposure patterns consisted either of pulse
trains at 0.5–5 Hz or burst trains that consisted of 20 pulses per
burst. Pulse frequencies ranged from 30 to 300 Hz within one
Authorized licensed use limited to: University of North Texas. Downloaded on July 13, 2009 at 13:35 from IEEE Xplore. Restrictions apply.
1514 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 56, NO. 5, MAY 2009
Fig. 4. (a) B-field plotted as a set of curves: Each curve represents the z-
component of the B-field along a 3.5-mm long, vertical line at a radial distance
(y-direction) from the center of the coil core outward according to the number
adjacent to each curve (in mm). The zero-line corresponds to the location of
the neuronal cell layer in the z-direction. Plot is stretched in the z-direction
with respect to panel B for better visibility. (b) Contour plot of the B-field at
a cut through the coil, medium (M), glass (G), and cell layers at 0 mm in the
x-direction.
Fig. 5. Schematic of two primary methods of exposure: pulse trains and burst
trains. Pulse trains are defined by a single frequency that was kept constant
during a specific exposure episode. Burst trains are defined by two frequencies:
the burst frequency and the pulse frequency in the bursts. These frequencies
were also kept constant for specific exposure episodes.
burst. Bursts were applied at frequencies ranging from 0.5 to
5 Hz. Each single pulse was rectangular and biphasic, with a
duration of 250 µs per phase. Rise and fall times were approxi-
mately 5–10 µs. Fig. 5 illustrates these parameters graphically.
In temperature test runs, temperature elevations of more than
2◦C in the culture medium were only observed for exposure
doses of more than 30 000 pulses at a mean frequency of 100 Hz.
The temperature was measured using a miniature thermocouple
that was placed between a dummy MEA plate and the chamber
with the cover glass. All other experimental conditions were
the same. Medium temperature was recorded during exposure
to the entire range of pulse frequencies. Measurements of the
magnetic field induced currents in the ITO showed that induced
voltages did not exceed 5 mV (Fig. 2). Previous data [13] and
concurrent stimulation experiments with the same MEA and
recording equipment have determined that measurable network
responses did not occur below pulse amplitudes of 400 mV
(biphasic, 300 µs per phase).
C. Signal Acquisition
Analog electrical signals obtained from the neurons through
the microelectrode arrays were amplified with a total gain of
10K, digitized at 40 kHz and transformed into time stamps for
storage. Active units were discriminated by template match-
ing. A template is an average of a group of waveforms selected
from the action potentials (APs) of one class, or unit. Incoming
waveforms are compared to the template or templates if more
than one unit is recorded by a specific electrode, using a sum of
squared differences calculation (SSQD) over the points of the
waveform and template. The SSQD is compared to a tolerance
or tightness of fit value. If it is less than that value, the waveform
is assigned to the class (unit) represented by that template. The
tolerance value is manually adjusted by observation during the
selection of units. Waveforms are accepted or rejected on this
basis. Artifacts have a very large SSQD compared with wave-
forms of the unit classes and are rejected as an active unit. For
preamplification, signal A/D conversion, filtering and further
amplification with software adjustable gain, a Plexon MAP data
acquisition system was used (Plexon Inc., Dallas, TX). Plexon
Rasputin software performed real-time wave shape sorting. Ac-
tivity windows showed both the discriminated waveforms of the
currently chosen unit action potential, all the active waveforms
recorded from the current multi-microelectrode plate (MMEP),
and a real-time raster window of adjustable time scale showing
all the recorded action potentials as timestamps. Fig. 2 depicts
two typical screenshots of sorted action potential waveforms
and one of an unsorted stimulation artifact. Examples of typical
raster plots are shown in Figs. 7 and 8.
Online analysis was performed with the Plexon “Rasputin”
software package and Neuroexplorer software (Nex Technolo-
gies, Littleton, MA). Signals from a total of 64 analog chan-
nels had to be selected and assigned to 32 available digital
signal processors (DSP). Up to four units per channel could
be discriminated in real time if waveform shapes allowed a
clear separation. The mean number of selected active units was
31 ±9 (SD) per stimulation episode for the experiments under
disinhibited conditions and 29 ±12 (SD) for the experiments
under native conditions.
D. Data Analysis
Data analysis was performed offline using Neuroexplorer and
custom programs to process ensemble averaged data as well as
individual unit spike rate profiles. Although the primary activ-
ity parameter used in this study was network spike production,
burst rates were also monitored to supplement activity decay
(or increase) information. Bursts were identified by digital RC
integration with rise times of 70 ms. A low threshold was used
to determine the start and stop times for the bursts, and a higher
threshold to confirm that the integrated profile represented a
high-frequency action potential discharge. A gap of 100 ms
devoid of spikes was used to separate complex bursts into indi-
vidual bursts events. Both thresholds were set by inspection of
the Neuroexplorer raster display with the intent to capture major
bursts and ignore spike clusters of two to three spikes [14]. This
operational definition allowed a survey of burst patterns within
the same culture if thresholds were not changed.
Magnetic pulse artifacts were not sorted and, therefore,
not recorded. One channel was selected for stimulus artifact
Authorized licensed use limited to: University of North Texas. Downloaded on July 13, 2009 at 13:35 from IEEE Xplore. Restrictions apply.
MEYER et al.: MAGNETIC STIMULATION AND DEPRESSION OF MAMMALIAN NETWORKS IN PRIMARY NEURONAL CELL CULTURES 1515
Fig. 6. Network responses to three consecutive stimulations at the same stimulation pulse and burst frequencies (see text) but increasing numbers of applied
pulses. Data are plotted in 1 min bins for spike (S, left ordinate) and burst (B, right ordinate) rates averaged across all channels. “R” denotes reference spiking
baseline activity. Both inhibition and rebound excitation are seen (a). The percent spike reduction increased almost linearly with the number of pulses applied
between 2500 and 15 000 pulses (insert). (b) Expanded view of the three depression episodes during (shaded areas) and after the magnetic field application.
monitoring to provide exact time points on start and end of
exposure episodes. Using crosscorrelograms between activity
channels and the channel that the stimulation artifacts were
recorded on, a preamplifier dead period of 1.75 ms was found
that was not shown directly by the raster plots. At a 10-Hz mean
pulse frequency, this translates to a 17.5-ms blockage per sec-
ond (or 1.75%). At 100 Hz, a recording time block of 17.5% oc-
curs. This suggests that the global activity decreases per minute
is enhanced by the amplifier saturation. All figures showing
quantitative data have been adjusted by a frequency-dependent
factor that reduces the activity reduction per minute. In the case
of activity decrease, this compensation factor (cf) also depended
on the respective maximum level of decrease, since the stronger
the activity decay the more units remained silent for some time
after the end of the stimulation period. Empirically, we found
that the compensation factor had to be adjusted using the formula
cf(adjusted)[%] = 107.11 −cf[%] ×0.7115.
For activity increase responses, the unadjusted factor was
used.
III. RESULTS
A. Network Responses to Number of Pulses
For short duration exposures (<5 min), stimulation effects
were generally dose-dependent. Fig. 6 shows three consecutive
stimulation episodes from the same SC network using mean
network data averaged across all channels in 1 min bins. All
stimulations used the same field exposure: 0.6 T, 5 Hz burst
frequency, 200 Hz pulse frequency, but with increasing exposure
times and, therewith, number of applied pulses (2500, 10 000,
15 000). The maximum inhibitory effect rose nearly linearly
with increasing dose (inset), and the excitatory rebound effect,
after the end of the stimulation, increased in amplitude and
duration with dose. All three exposures were 100% reversible,
and activity returned to the reference value (R) of approximately
380 mean spikes per minute.
The respective raster plot diagrams of the 10K and 15K stim-
ulation episodes show that the networks respond and recover
slowly (Fig. 7). There is no immediate response to the onset
of stimulation, but a gradual, slow decay in activity until many
channels fall silent. The recovery at the end of the stimulation
(right side of panels) is also gradual. Heterogeneous reactions
of the separate units can also be seen. Some neurons quit fir-
ing at 10 000 pulses, some only after 15 000. The onset and
recovery times also differ from unit to unit and depend on the
applied number of pulses. To show this more clearly, activities
from the same channels shown in (b) and (c) are plotted as
counts per minute for each discriminated, selected unit in (d)
and (e), respectively, with an expanded time window of 30 min.
Even excitation development in some channels (DSPs 27a/b) is
slow and not immediate. In contrast, electrical stimulation of
Authorized licensed use limited to: University of North Texas. Downloaded on July 13, 2009 at 13:35 from IEEE Xplore. Restrictions apply.
1516 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 56, NO. 5, MAY 2009
Fig. 7. Raster plot with action potential time stamps of all 29 recorded units
from the same network shown in Fig. 5. The stimulation periods are identified
by solid black bars at the bottom of the raster panels. (a) 2500 pulses (60 s
window); (b) 10 000 pulses (207 s), (c) 15 000 pulses (260 s). There was no
immediate network response as seen with electrical stimulation, which elicits
APs in the millisecond range. Here, we can see that at least a few seconds elapse
before the first cells show a reaction. Another 10–20 s later, the mean activity
of all the recorded units decays by more than 10%, which is our definition of
the onset time. (d), (e): Spike rate quantification of individual units showing
heterogeneous responses of different cells in one network to the same magnetic
field exposure. Note that panel (d) corresponds to raster plot (b), and (e) to (c).
Time period: 30 min (d), 86 min (e). The “PopVector” window (lower right-hand
corner) shows the average network response. The experimental episodes (b) and
(c) were separated in time by 138 min. The impression that activity decrease
is due to artifacts blocking AP recording is negated by the observation of slow
recovery after stimulation.
neuronal cell cultures usually elicits an immediate action po-
tential response in many or all units of a network (Fig. 8).
These reactions were never seen after application of the mag-
netic fields we used. It is also of importance to note that the
individual units maintain their general response profiles over
two exposure episodes separated by 138 min [Fig. 7(d) and (e)].
B. Control of Activity Suppression
Network responses were a function of stimulation pulse and
burst frequency and showed both increases and decreases in
the spontaneous network activity. However, excitation was usu-
ally of a low level and was seen primarily at pulse frequen-
cies below 50 Hz. The dominant effect was clearly inhibition,
which reached levels of more than 80% at 200 Hz [Fig. 6(b)].
For this reason, we prefer to describe the interaction of the
magnetic fields with the tissue as “magnetic pulse exposure”
(MPE) instead of “magnetic stimulation.” Because of concerns
for coil temperature and concomitant medium temperature in-
creases, high-frequency pulses (>50 Hz) were applied in bursts.
Clearly, activity suppression can be plotted as a function of mag-
netic burst frequencies. However, to avoid confusion between
the applied magnetic pulse bursts and “burst” of spike activity,
which is a general feature of the spontaneous activity pattern,
and to simplify quantification, we are representing the magnetic
exposure in terms of “mean pulse frequency (MPF).” Since the
number of pulses per stimulation burst was kept constant at 20, a
burst frequency of 0.5 Hz always translated to a mean frequency
of 10, 1–20 Hz, 2–40 Hz, and so on.
Fig. 9 shows activity suppression as a function of MPE in
terms of MPF. The data show that MPF determines the de-
gree of activity decrease, which can be maintained for extended
periods of time [maximum: 35 min in (d)]. 20–40% activity
decreases are stable during MPE and generally reproducible for
the same stimulation patterns and the same type of tissue. Panel
(a) shows the strong effect of pulse frequency with episodes
1, 2, and 3 representing responses to 20, 3, and 5 Hz mean
frequencies. Fig. 9(b) shows average network activity from a
75-days in vitro frontal cortex culture, (c) and (d) from a 62-
days in vitro spinal cord culture. The mean frequencies used
were 20 Hz each in (b), 10 Hz in (c) and (d), and 80 Hz in (e).
These panels demonstrate that the total number of pulses does
not determine the percent activity decay, as saturation is reached
before the end of the exposure period. Continued MPE merely
maintains a specific level of activity suppression. Total activ-
ity suppression was achieved at high frequencies [Fig. 9(e)],
although longer exposures could not receive adequate atten-
tion because of the lack of coil cooling. Rebound excitation
[dashed ellipses in Fig. 9(c) and (d)] was seen in 100% of SC
and 66% of FC cultures at low MPF (10 Hz) and decreased
to less than 10% of all experimental episodes at 100 Hz MPF.
Measurements of duration and amplitude showed high variabil-
ity in both parameters. Closer examination of panels (b)–(d)
shows a sensitivity difference between SC and FC networks.
Whereas the SC networks show a 22% and 35% decrease at
10 Hz [(c) and (d)], the FC network (b) reaches that level (23%)
at 20 Hz. These differences are represented more quantitatively
in Fig. 10.
C. Network Responses Under Disinhibited Conditions
A seemingly paradoxical effect of magnetic stimulation was
found in FC networks that were under the influence of bicu-
culline, a GABAAreceptor blocker [15], and in SC networks
under bicuculline and strychnine. In Fig. 10, we show satura-
tion response data from 16 networks. Saturation is defined as
the maximum response obtained during an exposure episode.
This response is independent of dose but sensitive to the mean
pulse frequency applied. The percent changes represent maxi-
mum decreases (or increases) achieved at the mean pulse fre-
quency indicated. A clear correlation between the mean fre-
quency and the maximum effect is seen. Under native condi-
tions, spinal cord tissue seems to respond with greater activity
Authorized licensed use limited to: University of North Texas. Downloaded on July 13, 2009 at 13:35 from IEEE Xplore. Restrictions apply.
MEYER et al.: MAGNETIC STIMULATION AND DEPRESSION OF MAMMALIAN NETWORKS IN PRIMARY NEURONAL CELL CULTURES 1517
Fig. 8. Comparison of network response characteristics to electrical (a) and (b) and magnetic (c) stimulation. Single electrical pulses above 0.6 V elicit almost
immediate responses, which is not the case for the magnetic stimulation. (a) Three simultaneous oscilloscope traces at a sweep of 50 ms per division show responses
to a 0.8-V, 300-µs biphasic pulse within 20–40 ms. (b) Multiple single channel responses to repetitive electrical pulses at 0.9 V. Except for the second stimulation
pulse that arrived during a spontaneous activity burst, the network responds reliably and almost immediately. Noise lines in (a) and (b) are approximately 40 µV
p/p. The stimulus electrode is not shown in (a) or (b). (c) Multichannel raster plot of an 8-s panel showing no overt responses to the magnetic stimulation pulses
(bars at bottom of panel).
decreases than frontal cortex tissue [Fig. 10(a) and (b)]. Under
pharmacological disinhibition, excitation was seen primarily at
low burst frequencies [(c) and (d)]. FC cultures showed greater
excitatory responses than SC networks (c). Again, the inhibition
was greater in spinal cord than in frontal cortex networks. For
(a), (b), and (d), an exponential decay fit was chosen:
y=A1·e(−x/t1)+y0.
For c, a Gaussian fit was chosen
y=y0+A
w·π/2
·e−2((x−xc)/w)2.
The fit values were: A, y0=−30.6±4.3;A1=
115.4±121.6;t1=7.3±4.9;R2=0.26;B,y0=−87.3±5.5;
A1=96.8±9.4;t1=23±5.9;R2=0.92;C,y0=−35±18.3;
xc=45.6±2.2;w=46.2±8.4; A= 7959.5±2203; R2=
0.84;D,y0=−83.9±30.9;A1= 181.4±42.8;t1=29±19.5;
R2=0.77.
The time to saturation of the activity decay as a function of
mean stimulation frequency could also be quantified (Fig. 11).
These panels contain data from all recorded episodes and show
the almost linear relation between the number of pulses required
to reach saturation, the time to saturation, and the mean stimu-
lation frequency. Saturation data was derived from overall spike
production per minute. Burst per minute values are shown in
Authorized licensed use limited to: University of North Texas. Downloaded on July 13, 2009 at 13:35 from IEEE Xplore. Restrictions apply.
1518 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 56, NO. 5, MAY 2009
Fig. 9. Long duration (>5 min) exposure, showing pulse frequency-dependent
activity decreases, saturation, and maintenance of activity suppression.
(a1) Exposure to stimulation bursts at 1 Hz (20 pulses per burst at 100 Hz,
10 000 pulses total; mean pulse frequency: 20 Hz). The maximum decrease to
200 spikes per minute is highlighted by the horizontal arrow. (a2) Exposure to
single pulses at 3 Hz (2500 pulses total). (a3): Exposure to single pulses at 5
Hz (5000 pulses total). (b) Two consecutive stimulation episodes from a frontal
cortex culture. First episode: 15 000 pulses (13 min), second episode: 30 000
pulses (26 min). Mean frequency was 20 Hz for both episodes resulting in the
same decay level. The times to saturation are 11 and 12 min, respectively. (c),
(d) Two episodes from a spinal cord culture, separated by four other stimulation
episodes. (c) Mean frequency 10 Hz, 10 000 pulses (17 min), (d) mean fre-
quency 10 Hz, 20 000 pulses (34 min). (e) Example of 100% spike suppression
at 80 Hz mean pulse frequency with full recovery.
some of the graphs (Figs. 6, 9, 12, and 13) but were not used
for statistical analysis. Inserts 1 and 2 in panel (a) show exam-
ples of different times to saturation caused by MPFs of 20 Hz
(a1) and 60 Hz (a2). The times to saturation were 8 and 2 min,
respectively. Higher MPFs generally caused the activity de-
crease to saturate earlier than lower MPFs. It is also important
to note that the slopes are very similar, indicating that the time
to saturation does not appear to be a function of disinhibition
for the two different tissues used.
D. Changes in Activity Patterns Under Native and
Disinhibited Conditions
In an attempt to shed light on the mechanisms involved, we
observed activity burst patterns under normal and disinhibited
conditions. Fig. 12 shows responses to three consecutive MPEs
in a frontal cortex network under native (normal) conditions.
The total number of applied pulses (dose) was kept constant at
10 000, as was the mean frequency (100 Hz). Panel (b) shows the
raster plot corresponding to the third MPE with the solid black
bar, representing the exposure time period. Panels [(c), (d), and
(e)] indicate the changes of the activity burst pattern in terms
of mean burst duration, average spike frequency inside a burst,
and average number of spikes in a burst. In brief, activity bursts
Fig. 10. Maximum response amplitude (per cent spike rate decrease or in-
crease from reference), at different mean stimulation frequencies. (a) Data from
five frontal cortex (FC) cultures under native medium conditions, (b) data from
three spinal cord (SC) cultures, native medium, and (c) four FC cultures under
bicuculline. (d) Four SC cultures under bicuculline and strychnine. All panels
show activity inhibition at high burst frequencies. However, SC cultures (b),
(d) show greater sensitivity. In addition, excitation is seen when cultures were
disinhibited with bicuculline or bicuculline and strychnine (c), (d). The number
of applied pulses was at least high enough for the maximum response to satu-
rate. The pulse frequencies used were between 100 and 300 Hz. Data in (a), (b),
and (d) were fit with exponential functions (R2values: 0.26, 0.92, and 0.77,
respectively. Data in (c) were treated with a Gaussian fit (R2: 0.84).
Fig. 11. Summary data on the time needed for achieving the maximum effect
in one stimulation episode as a function of MPF. Data point symbols vary
with the respective mean stimulation frequency used. (a) and (c) show frontal
cortex tissue reactions, (b) and (d) responses from spinal cord. (a) and (b)
are reactions under native media conditions, whereas (c) and (d) networks
were pharmacologically disinhibited. (a): Insert 1: Response to a 20-Hz (MPF)
exposure episode with a time to saturation of 8 min. Insert 2: Response to a
60-Hz (MPF) episode, resulting in a time to saturation of only 2 min.
Authorized licensed use limited to: University of North Texas. Downloaded on July 13, 2009 at 13:35 from IEEE Xplore. Restrictions apply.
MEYER et al.: MAGNETIC STIMULATION AND DEPRESSION OF MAMMALIAN NETWORKS IN PRIMARY NEURONAL CELL CULTURES 1519
Fig. 12. Changes in activity burst parameters as a function of MPE under native (normal medium) conditions. (a) Spike and burst plots of three consecutive
exposure episodes in a frontal cortex culture. Stimulation parameters (mean stimulation frequency/dose): 100/10 000 for all episodes. (b) Raster plot for all recorded
units of the third episode. (c) Burst duration, (d) burst period, (e) mean spike frequencies in bursts, and (f) number of spikes in bursts. During activity suppression,
burst periods lengthen and all other measures decrease, resulting in a large decrease in network spike production. Values in (c),(d), (f) were calculated as averages
per minute from all recorded units.
become shorter and less frequent with a reduction of mean spike
frequencies during MPE.
Fig. 13 shows a similar FC experiment under bicuculline,
with identical doses but with ascending mean frequencies rang-
ing from 60 to 100 Hz in a frontal cortex culture. Spike and
burst rates (a) increase and almost double in value. The general
excitation effect is seen in the raster plot [(b), episode 1 in (a)]
with a clear indication of the response delay. The MPE duration
between the vertical lines is 5 min. It is interesting to note that
burst durations (c) decrease, but that the decrease in the burst
period [i.e. and increase in burst rate (d)] and the strong increase
in the mean spike frequencies in the burst account for the ex-
citation. More subtle pattern changes are also indicated, such
as biphasic responses for episode 3 in burst duration and burst
period. However, analyses of such observations must be left to
future studies.
IV. DISCUSSION
The use of TMS as a diagnostic and therapeutic tool for clini-
cal applications is primarily based on empirical data and in vivo
research on humans, utilizing relatively coarse fMRI monitoring
of brain activity in certain areas. In peripheral magnetic stimu-
lation, highly organized nerve bundles are stimulated and action
Authorized licensed use limited to: University of North Texas. Downloaded on July 13, 2009 at 13:35 from IEEE Xplore. Restrictions apply.
1520 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 56, NO. 5, MAY 2009
Fig. 13. Changes in activity burst parameters as a function of MPE under disinhibited conditions. (a) Spike and burst plots of three consecutive exposure
episodes in a frontal cortex culture under 40 µM bicuculline. Stimulation parameters (mean stimulation frequency/dose): episode 1: 60/15 000 (2): 80/15 000;
(3) 100/15 000. (b) Raster plot for all recorded units of the first episode. Stimulation period: horizontal bar and shaded vertical areas). (c) Burst duration, (d) burst
period, (e) mean spike frequencies in bursts, and (f) mean number of spikes in bursts. Values in (c) to (f) were calculated as averages per minute from all recorded
units. The increase in network spike production is caused by a faster burst rate, but also by an increase in spike frequencies in activity bursts.
potentials are elicited directly. The current flowing through the
stimulation coil induces a pair of virtual anodes and cathodes at
the sites of the electrical field gradient maximum multiplied by
the electrical field strength, thereby depolarizing a number of
axons. In contrast, less quantitative data exists on the mode of
action of low-frequency electromagnetic fields on morpholog-
ically less organized neural networks as found in mammalian
gray matter. Almost no information is available for magnetic
field effects on morphologically random networks as are found
in cell culture. Here, direct cellular effects may be observed
without being masked by excitation linked to oriented cellular
components such as apical dendrites in pyramidal cells of the
intact cortex. In addition, networks grown on microelectrode
arrays allow continual, multisite monitoring of action potential
activity patterns over long periods of time. This environment
is unique, as it provides an ability to quantify subtle effects on
neurons reflected in changes in network activity on the action
potential level.
As stated in Section II, we have used an amplifier saturation
compensation of 1.75% per 10 Hz MPF, adjusted according to
the respective maximum response level achieved. It is impossi-
ble to know the exact activity levels that are lost. However, it is
highly unlikely that more than 1 AP per channel are ever blocked
in the 1.75-ms time interval. Although this affects the absolute
value of the activity decrease, it does not affect the temporal evo-
lution of the relatively slow activity decreases followed by even
slower recoveries. As shown in Fig. 6, recovery time periods
of 60–150 s after cessation of magnetic pulse exposure are not
Authorized licensed use limited to: University of North Texas. Downloaded on July 13, 2009 at 13:35 from IEEE Xplore. Restrictions apply.
MEYER et al.: MAGNETIC STIMULATION AND DEPRESSION OF MAMMALIAN NETWORKS IN PRIMARY NEURONAL CELL CULTURES 1521
TAB L E I
PERCENT OF RESPONSE ONSET TIME IN <1, 1–3, 3–5, AND >5MIN BINS
uncommon. Despite the loss of the amplifier saturation effect
after stimulation episodes, activity often continues to decrease
[cf. Fig. 6(b)] before a subsequent slow recovery. Because there
is no immediate recovery of activity after cessation of magnetic
stimulation, it is obvious that the global network response is
robust and that the saturation effect produces only minor quan-
titative changes in the recorded network spike production.
The effects reported here were subthreshold. Unlike electrical
stimulation of neuronal cell cultures [13], the magnetic pulses
did not cause the network to respond with an immediate, sud-
den activity increase or entrainment of activity patterns. The
responses were relatively slow with onset times ranging from
10 s to several minutes (Table I). They were predominantly in-
hibitory (Table II) and generally exhibited dose dependence, in
terms of number of pulses per second applied to the cell cul-
tures. A strict dose dependence was always seen when the same
stimulation parameters were used with varying numbers of total
applied pulses in the same culture (Fig. 6). Fig. 11 shows that
the duration of the network response is clearly a function of the
number of pulses applied up to the attainment of response sat-
uration. The stimulation variables were pulse frequency, burst
frequency, and total number of applied pulses. The number of
applied pulses per burst (20), the shape and length of each
stimulation pulse (biphasic rectangular, 500 ms long), and the
stimulation amplitude (maximum B-field strength 0.6 T) were
kept constant. Stimulation bursts were used because they had
proven effective in former studies [16] and allowed application
of high doses in short periods without excessive heating of the
stimulation coil. However, to allow a simple quantification, we
used a mean exposure frequency that is linearly related to the
burst frequency.
From stimulation data, it is expected that the coil created a cir-
cular electrical field and thus circular currents in the cell medium
and tissue. However, this is only an approximation, since one
can expect weak field distortions from the ITO conductors of
the MEA, the gold-plated recording sites and heterogeneities
of the culture density. These currents changed their direction
with the rapidly alternating field of the respective pulse fre-
quency and thus moved charged particles in an oscillatory fash-
ion. Since the morphological features of the neurons are ex-
pressed randomly on the glass chip surface, the electric field
cannot affect each cell in the same way. This correlates with
the observation of partly nonhomogeneous, unit-specific reac-
tions [Fig. 7(d) and (e)]. However, it could also reflect different
sensitivities of different neuronal subtypes.
A large number of publications exist on the possible bio-
logical effects of electromagnetic fields. Essentially, what trig-
gers any kind of cellular reaction must be a combination of the
Lorentz force, which the magnetic component exerts on charged
particles, and the electric field component that attracts and repels
charges, and basically results in small electrical currents inside
and between cells. The reported effects of alternating magnetic
fields on neuronal activity range from long-term potentiation in
rat hippocampus [16], induction of “virtual lesions” [17], and
alteration of animal behavior [18], to increases in corticospinal
excitability [19], depressant effects [20], as well as disruption
of rhythmic slow activity in rat hippocampal slices [21]. Sev-
eral possible mechanisms of electromagnetic field interaction
with cellular substructures have been proposed; among these are
conformation changes of proteins [22], changes of the opening
probabilities of voltage-gated ion channels [23], and changes in
calcium concentrations and oscillations through redistribution
of calcium in intracellular stores [24]. Other possible mecha-
nisms include phosphorylation and activation of stress-activated
protein kinase [25], alteration of gap junction function [26], in-
duced membrane potential changes [27], cell-surface receptor
redistribution and cytoskeletal reorganization [28], radical-pair
recombination or increase in NO synthesis through calcium en-
try into the cells or increase in NO-synthase activity [29], and
enhancement of c-fos, glial fibrillary acidic protein and nerve
growth factor [30]–[32].
None of the proposed hypotheses could be definitely sustained
or disproved. However, the relatively slow time course of the
responses leads us to suggest a slowly developing interference
with the mechanisms that sustain the normal spontaneous net-
work activity. In light of the complexity of synaptic mechanisms,
the synapse is a possible target of MPE. Synaptic mechanisms
have been suggested in a paper that recorded action potentials
from the occipital cortex of a cat during TMS. Enhancement
and inhibition of spontaneous activity with rebound excitation
was reported for single pulses, but for time periods of 4 s [3].
Intracortical facilitation and inhibition in response to transcra-
nial magnetic stimulation was also seen in conscious humans
but over time periods of only milliseconds [4]. In contrast, we
have used multiple pulses (up to 15 000) and maintained net-
work activity suppression for over 30 min. If the activity of these
cells can be suppressed reversibly for a prolonged time, a new
tool for noninvasive, nonchemical pain management might be
developed.
If synaptic mechanisms play a major role in the network re-
sponses to MPE, then pharmacological manipulations can pro-
vide some insight into mechanisms. Bicuculline is a GABAA
antagonist and blocks the inhibitory subcircuits in neuronal cor-
tical networks [15]. The situation is more complex in the spinal
cord as such tissue also uses inhibitory circuits that depend on
the neurotransmitter glycine, which was only partially blocked
by a concentration (1 µM) that does not disrupt glycinergic
transmission entirely [11]. Both types of networks showed inhi-
bition at high pulse frequencies and excitation under bicuculline
at low magnetic pulse frequencies. Under bicuculline (FC, bicu-
culline and strychnine for SC), and between 10 and 60 Hz mean
pulse frequency, the network responses were 65% excitation,
Authorized licensed use limited to: University of North Texas. Downloaded on July 13, 2009 at 13:35 from IEEE Xplore. Restrictions apply.
1522 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 56, NO. 5, MAY 2009
TAB L E I I
RESPONSE TYPES LISTED BY CATEGORIES
25% depression and 5% no effect (FC) and 41.7% excitation,
50% depression and 8.3% no effect (SC), as opposed to 3.2%
excitation, 71% depression and 25.8% no effect (FC) and 0%
excitation, 94.7% depression and 5.3% no effect (SC) in normal
medium (Table II). This includes those trials where a saturation
of the effects was not achieved. The excitation was greater in the
FC networks where essentially all inhibition was blocked. This
could be explained by a biphasic magnetic effect: general weak
excitation of all cells between 20 and 60 Hz MPF and general
strong inhibition thereafter. The net excitatory effect is revealed
when GABA receptors are blocked and is stronger in tissue that
uses only GABA inhibition. Simultaneous stimulation of exci-
tatory and inhibitory circuits in normal medium tends to mask
this effect. However, above 20–60 Hz, activity decreases are
generated that can even lead to total activity loss. In light of the
slow recovery from saturation, interference with synaptic exo-
cytosis via general synaptic disorganization, in terms of nonspe-
cific multiple interference with many mechanisms responsible
to support exocytosis, is a likely mechanism. Our observations
that inhibition is characterized by reduced burst duration, re-
duced spike frequencies in bursts, and longer burst periods are
also compatible with the hypothesis of reduced exocytosis due
to general and possibly even nonspecific interference with the
complex mechanisms of chemical synapses. Furthermore, the
time to saturation of activity suppression is not a function of
disinhibition for the two different tissues used (Fig. 11) and
depends solely on pulse frequency. All synapses seem to be
affected by MPE, implying a general interference with the effi-
ciency of common synaptic mechanisms. The greater sensitivity
of spinal cord networks to MPE (Fig. 10) may be a reflection of
remaining glycine inhibitory circuits rather than synaptic tissue
specificity.
The results of this study indicate that the use of primary neu-
ronal cultures on multielectrode arrays as target for MPE, serves
as an effective new tool to investigate and quantify changes in
spontaneous neural activity during and after exposure to alter-
nating electromagnetic fields. With the experimental configura-
tion presented, these fields can be of virtually any pulse shape,
bursting configuration and dose up to a certain limit at which
thermal effects obscure the electromagnetic field effects. This
is possible because of the considerably lower energy needed
to penetrate the thin cell culture as opposed to cortical tissue
that is surrounded by bone and thick connective tissue. We be-
lieve that this new experimental system will contribute to future
developments and optimizations of TMS.
ACKNOWLEDGMENT
All experiments were conducted at the University of North
Texas.
REFERENCES
[1] T. Weyh, K. Wendicke, C. Mentschel, H. Zantow, and H. R. Sieb-
ner, “Marked differences in the thermal characteristics of figure-of-eight
shaped coils used for repetitive transcranial magnetic stimulation,” J.
Clin. Neurophysiol., vol. 116, no. 6, pp. 1477–1486, 2005.
[2] S. Rossi and P. M. Rossini, “TMS in cognitive plasticity and the potential
for rehabilitation,” Trends Cogn. Sci., vol. 8, no. 6, pp. 273–279, 2004.
[3] V. Moliadze, Y. Zhao, U. Eysel, and K. Funke, “Effect of transcranial mag-
netic stimulation on single-unit activity in the cat primary visual cortex,”
J. Physiol., vol. 553, no. 2, pp. 665–679, 2003.
[4] H. Nakamura, H. Kitagawa, Y. Kawaguchi, and H. Tsuji, “Intracortical
facilitation and inhibition after transcranial magnetic stimulation in con-
scious humans,” J. Physiol., vol. 498, no. 3, pp. 817–823, 1997.
[5] M. Kobayashi and A. Pascual-Leone, “Transcranial magnetic stimulation
in neurology,” Lancet Neurol., vol. 2, no. 3, pp. 145–156, 2003.
[6] Y. Terao and Y. Ugawa, “Basic mechanisms of TMS,” J. Clin. Neurophys.,
vol. 19, no. 4, pp. 322–343, 2002.
[7] P. G. Nelson, R. D. Fields, C. Yu, and E. A. Neale, “Mechanisms involved
in activity-dependent synapse formation in mammalian central nervous
system cell cultures,” J. Neurobiol., vol. 21, no. 1, pp. 138–156, 1990.
[8] G. W. Gross, “Internal dynamics of randomized mammalian neuronal
networks in culture,” in Enabling Technologies for Cultured Neural Net-
works, D. A. Stenger and T. M. McKenna, Eds. New York: Academic,
1994, pp. 277–317.
[9] G.W.Gross,E.Dian,E.W.Keefer,A.Gramowski,andS.Stuewe,
“Long-term functional contact between nerve cell networks and micro-
electrode arrays,” in Toward Replacement Parts for the Brain,T.Berger
and D. Glanzman, Eds. Cambridge, MA: MIT Press, 2005, pp. 177–203.
[10] G. W. Gross and F. U. Schwalm, “A closed flow chamber for long-term
electrophysiological and microscopical monitoring of monolayer neuronal
networks,” J. Neurosci. Methods, vol. 52, no. 1, pp. 73–85, 1994.
[11] G. W. Gross, B. K. Rhoades, H. M. E. Azzazy, and M. C. Wu, “The use
of neuronal networks on multielectrode arrays as biosensors,” Biosens.
Bioelectron., vol. 10, pp. 553–567, 1995.
[12] G. W. Gross, A. Harsch, B. K. Rhoades, and W. G ¨
opel, “Odor, drug
and toxin analysis with neuronal networks in vitro: Extracellular array
recording of network responses,” Biosens. Bioelectron., vol. 12, no. 5,
pp. 373–393, 1997.
[13] G. W. Gross, B. K. Rhoades, D. L. Reust, and F. U. Schwalm, “Stimula-
tion of monolayer networks in culture through thin-film indium-tin oxide
recording electrodes,” J. Neurosci. Methods, vol. 50, no. 11, pp. 131–143,
1993.
[14] M. Parviz and G. W. Gross, “Quantification of zinc toxicity using neuronal
networks on microelectrode arrays,” Neurotoxicology, vol. 28, no. 3,
pp. 520–531, 2007.
[15] D. R. Curtis, A. W. Duggan, D. Felix, and G. A. Johnston, “Bicuculline,
an antagonist of GABA and synaptic inhibition in the spinal cord of the
cat,” J. Brain Res., vol. 32, no. 1, pp. 69–96, 1971.
[16] M.Ogiue-Ikeda, S. Kawato, and S. Ueno, “The effect of repetitive transcra-
nial magnetic stimulation on long-term potentiation in rat hippocampus
depends on stimulus intensity,” J. Brain Res., vol. 993, pp. 222–226,
2003.
[17] C. C. Hilgetag, H. Th´
eoret, and A. Pascual-Leone, “Enhanced visual
spatial attention ispilateral to rTMS-induced ‘virtual lesions’ of human
parietal cortex,” Nature Neurosci., vol. 4, no. 9, pp. 953–957, 2001.
[18] M. N. Zhadin, O. N. Deryugina, and T. M. Pisachenko, “Influence of com-
bined DC and AC magnetic fields on rat behaviour,” J. Bioelectromagn.,
vol. 20, pp. 378–386, 1999.
[19] A. Peinemann, B. Reimer, C. L¨
oer, A. Quartarone, A. M¨
unchau, B.
Conrad, and H. R. Siebner, “Long-lasting increase in corticospinal ex-
citability after 1800 pulses of subthreshold 5 Hz repetitive TMS to the pri-
mary motor cortex,” J. Clin. Neurophys., vol. 115, no. 7, pp. 1519–1526,
2004.
[20] M. B. Iyer, N. Schleper, and E. M. Wasserman, “Priming stimulation
enhances the depressant effect of low-frequency repetitive transcranial
magnetic stimulation,” J. Neurosci., vol. 23, no. 34, pp. 10867–10872,
2003.
[21] S.M. Bawin, W. M. Satmary, R. A. Jones, W. R. Adey, and G. Zimmerman,
“Extremely-low-frequency magnetic fields disrupt rhythmic slow activity
Authorized licensed use limited to: University of North Texas. Downloaded on July 13, 2009 at 13:35 from IEEE Xplore. Restrictions apply.
MEYER et al.: MAGNETIC STIMULATION AND DEPRESSION OF MAMMALIAN NETWORKS IN PRIMARY NEURONAL CELL CULTURES 1523
in rat hippocampal slices,” Bioelectromagnetics, vol. 17, pp. 388–395,
1996.
[22] L.J. Challis, “Mechanisms for interaction between RF fields and biological
tissue,” Bioelectromagn. Suppl., vol. 26, no. S7, pp. S98–S106, 2005.
[23] W. Chen, “Supra-physiological membrane potential induced conforma-
tional changes in K+channel conducting system of skeletal muscle
fibers,” Bioelectrochemistry, vol. 62, no. 1, pp. 47–56, 2004.
[24] G. P. Pessina, C. Aldinucci, M. Palmi, G. Sgaragli, A. Benocci, A. Meini,
and F. Pessina, “Pulsed electromagnetic fields affect the intracellular cal-
cium concentrations in human astrocytoma cells,” Bioelectromagnetics,
vol. 22, no. 7, pp. 503–510, 2001.
[25] W. J. Sun, H. Chiang, Y. T. Fu, Y. N. Yu, H. Y. Xie, and D. Q. Lu, “Exposure
to 50 Hz electromagnetic fields induces the phosphorylation and activity
of stress-activated protein kinase in cultured cells,” Electromagn. Biol.
Med. (formerly Electr.- Magnetobiol., vol. 20, no. 3, pp. 415–423, 2001.
[26] C. M. Li, H. Chiang, Y. D. Fu, B. J. Shao, J. R. Shi, and G. D. Yao, “Effects
of 50 Hz magnetic fields on gap junctional intercellular communication,”
Bioelectromagnetics, vol. 20, pp. 290–294, 1999.
[27] C. Grosse and H. P. Schwan, “Cellular membrane potentials induced by
alternating fields,” Biophys. J., vol. 63, no. 6, pp. 1632–1642, 1992.
[28] M. R. Cho, H. S. Thatte, M. T. Silva, and D. E. Golan, “Transmembrane
calcium influx induced by AC electric fields,” FA S E B J, vol. 13, pp. 677–
683, 1999.
[29] M. Fujiki and O. Steward, “High frequency transcranial magnetic stimula-
tion mimics the effects of ECS in upregulating astroglial gene expression
in the murine CNS,” Brain Res. Mol. Brain Res., vol. 44, no. 2, pp. 301–
308, 1997.
[30] A. Hausmann, C. Weis, J. Marksteiner, H. Hinterhuber, and C. Humpel,
“Chronic repetitive transcranial magnetic stimulation enhances c-fos in
the parietal cortex and hippocampus,” Mol. Brain Res., vol. 76, no. 2,
pp. 355–362, 2000.
[31] M. B. M¨
uller, N. Toschi, A. E. Kresse, A. Post, and M. E. Keck, “Long-
term repetitive transcranial magnetic stimulation increases the expression
of brain-derived neurotrophic factor and cholecystokinin mRNA, but not
neuropeptide tyrosine mRNA in specific areas of rat brain,” Neuropsy-
chopharmacology, vol. 23, no. 2, pp. 205–215, 2000.
[32] F. M. Longo, T. Yang, S. Hamilton, J. F. Hyde, J. Walker, L. Jennes,
R. Stach, and B. F. Sisken, “Electromagnetic fields influence NGF activity
and levels following sciatic nerve transection,” J. Neurosci. Res., vol. 55,
no. 2, pp. 230–237, 1999.
Jochen F. Meyer wasborninHamburg,Germany,in
1979. He received the B.S. and Dipl.-Ing. degrees in
2004 and 2005, respectively, from the Technical Uni-
versity of Munich (TEM), Munich, Germany, where
he is currently working toward the Dr.-Ing. degree,
all in electrical engineering.
In August 2006, he joined the Assistant Director of
the IMETUM Neurological Laboratory, Munich. His
current research interests include novel techniques in
measuring fast and slow membrane potential changes
of electrically active and nonactive cells in vitro; elec-
tric and magnetic manipulation of neuronal and tumor cell culture; neurodegen-
erative disease models using MEA-based neuronal cultures; combining novel
engineering concepts and biological requirements for improvements in electro-
physiological measurement systems; neuronal cell culturing techniques.
Bernhard Wolf received the Diploma, the Ph.D. de-
gree, the Staatsexamen in physics, and the Habilita-
tion degree in medical physics and biophysics from
Albert-Ludwigs-Universit¨
at, Freiburg, Germany, in
1973, 1977, 1978, and 1988, respectively.
During 1977–1979, he was a Research Fellow in
the Department of Biochemistry, University of Hei-
delberg, Heidelberg, Germany. During 1980–1998,
he was at the Institute of Immunology, Albert-
Ludwigs-Universit¨
at, where he was an Associate Pro-
fessor at the Faculty of Medicine in 1995. In 1988,
he was a Privatdozent. From 1998 to 2000, he was a Professor (C3/C4) in the
Department of Biophysics, University of Rostock, Rostock, Germany. He is
currently a Professor (C4) in the Department of Medical Electronics, Technical
University of Munich, Munich, Germany.
Guenther W. Gross was born in 1939. He re-
ceived the B.S. degree in engineering physics
from the Stevens Institute of Technology, Hobo-
ken, NJ, in 1962, and the Ph.D. degree in bio-
physics/neurophysiology from Florida State Univer-
sity, Tallahassee, in 1973.
From 1974 to 1977, he was a Postdoctoral Fellow
at the Max Planck Institute for Psychiatry, Munich,
Germany. In 1977, he was also a Postdoctoral Fel-
low at Sandoz Ltd., Basel, Switzerland. From 1978
to 1985, he was an Assistant Professor and an Asso-
ciate Professor at Texas Woman’s University, Denton. During 1985–1988, he
was an Associate Professor of neuroscience at the University of North Texas
(UNT), Denton, where is currently the UNT Regents Professor of Neuroscience
at the Department of Biological Sciences, and also the Director of the Center of
Network Neuroscience. His current research interests include experimental and
theoretical neuronal network dynamics with applications to the field of toxicol-
ogy, drug development, biosensors, and modeling of self-organizing dynamic
systems; development of microelectrode arrays, life-support systems, and high
throughput multinetwork platforms using neuronal cell cultures.
Authorized licensed use limited to: University of North Texas. Downloaded on July 13, 2009 at 13:35 from IEEE Xplore. Restrictions apply.