6798 Journal of Physiology (1997), 505.3, pp.

617-632
617
Action potential initiation and propagation in rat neocortical
pyramidal neurons
Greg Stuart *, Jackie Schiller and Bert Sakmann
Abteilung Zellphysiologie, Max-Planck-Institut fur medizinische Forschung, Jahnstrasse 29,
D-69120 Heidelberg, Germany
1. Initiation and propagation of action potentials evoked by extracellular synaptic stimulation was studied
using simultaneous dual and triple patch pipette recordings from different locations on neocortical layer 5
pyramidal neurons in brain slices from 4-week-old rats (P26-30) at physiological temperatures.
2. Simultaneous cell-attached and whole-cell voltage recordings from the apical trunk (up to 700 ,um distal to
the soma) and the soma indicated that proximal synaptic stimulation (layer 4) initiated action potentials first
at the soma, whereas distal stimulation (upper layer 2/3) could initiate dendritic regenerative potentials prior
to somatic action potentials following stimulation at higher intensity.
3. Somatic action potentials, once initiated, propagated back into the apical dendrites in a decremented manner
which was frequency dependent. The half-width of back-propagating action potentials increased and their
maximum rate of rise decreased with distance from the soma, with the peak of these action potentials
propagating with a conduction velocity of approximately 0 5 m s '.
4. Back-propagation of action potentials into the dendritic tree was associated with dendritic calcium
electrogenesis, which was particularly prominent during bursts of somatic action potentials.
5. When dendritic regenerative potentials were evoked prior to somatic action potentials, the more distal the
dendritic recording was made from the soma the longer the time between the onset of the dendritic
regenerative potential relative to somatic action potential. This suggested that dendritic regenerative
potentials were initiated in the distal apical dendrites, possibly in the apical tuft.
6. At any one stimulus intensity, the initiation of dendritic regenerative potentials prior to somatic action
potentials could fluctuate, and was modulated by depolarizing somatic or hyperpolarizing dendritic current
injection.
7. Dendritic regenerative potentials could be initiated prior to somatic action potentials by dendritic current
injections used to simulate the membrane voltage change that occurs during an EPSP. Initiation of these
dendritic potentials was not affected by cadmium (200 /aM), but was blocked by TTX (1 /sM).
8. Dendritic regenerative potentials in some experiments were initiated in isolation from somatic action
potentials. The voltage change at the soma in response to these dendritic regenerative events was small and
subthreshold, showing that dendritic regenerative events are strongly attenuated as they spread to the soma.
9. Simultaneous whole-cell recordings from the axon initial segment and the soma indicated that synaptic
stimulation always initiated action potentials first in the axon. The further the axonal recording was made
from the soma the greater the time delay between axonal and somatic action potentials, indicating a site of
action potential initiation in the axon at least 30 /am distal to the soma.
10. Simultaneous whole-cell recordings from the apical dendrite, soma and axon initial segment showed that
action potentials were always initiated in the axon prior to the soma, and with the same latency difference,
independent of whether dendritic regenerative potentials were initiated or not.
11. It is concluded that both the apical dendrites and the axon of neocortical layer 5 pyramidal neurons in
P26-30 animals are capable of initiating regenerative potentials. Regenerative potentials initiated in
dendrites, however, are significantly attenuated as they spread to the soma and axon. As a consequence,
action potentials are always initiated in the axon before the soma, even when synaptic activation is intense
enough to initiate dendritic regenerative potentials. Once initiated, the axonal action potentials are
conducted orthogradely into the axonal arbor and retrogradely into the dendritic tree.

* To whom correspondence should be addressed at Division of Neuroscience, John Curtin School of Medical Research,
Australian National University, Canberra, ACT 0200, Australia.
618 G. Stuart, J Schiller and B. Sakmann J Physiol. 505.3

The response of a neuron to a given synaptic input will METHODS

depend on where within the neuron this synaptic input is Slice preparation and electrical recording
integrated to initiate an action potential. Initial work Experiments were performed on 300 ,um thick, sagittal brain slices
suggested that action potentials in spinal motoneurons are from somatosensory neocortex of P26-30 Wistar rats prepared as
initiated in the axon initial segment (Eccles, 1964). Later described in the previous paper (Schiller, Schiller, Stuart &
work in hippocampal and cerebellar Purkinje neurons, Sakmann, 1997). During recording slices were continuously
however, suggested that action potentials can also be perfused with oxygenated solution of the following composition
initiated in the dendrites (see Johnston, Magee, Colbert & (mM): 125 NaCl, 25 NaHCO3, 25 glucose, 2-5 KCl, 1-25 NaH2PO4, 2
Christie, 1996). Similarly, some studies have concluded that CaCl2 and 1 MgCl2 (pH 7*4 with 95% 02-5% C02), and all
experiments were performed at 35 + 1 'C. Pooled data are
action potentials can be initiated in the dendrites of neo- expressed as means + S.D. No correction was made for the junction
cortical pyramidal neurons (Deschenes, 1981; Pockberger, potential between bath and pipette solutions, which was
1991; Kim & Connors, 1993; Regehr, Kehoe, Ascher & experimentally determined to be -12 mV.
Armstrong, 1993; Hirsch, Alonso & Reid, 1995). Other All chemicals were purchased from Sigma, except APV, which was
studies, however, have concluded that there is an axosomatic purchased from Tocris Cookson (Bristol, UK).
location of action potential initiation in neocortical pyramidal
neurons based on either single site dendritic recordings Layer 5 pyramidal neurons were identified using infrared
(Amitai, Friedman, Connors & Gutnick, 1993) or dual illumination combined with differential interference contrast optics
and video microscopy, and patch-pipette recordings (cell-attached
simultaneous somatic and dendritic or somatic and axonal voltage clamp or whole-cell current clamp; seal resistances > 1 GQ)
recordings (Stuart & Sakmann, 1994). The conclusions of were made from the soma, apical dendrite (up to 700 ,um from the
Stuart & Sakmann (1994), however, have been questioned, soma) and/or axon (up to 30 ,um from the soma) of the same neuron
and the possibility raised that capacitive loading or 'wash- using identical microelectrode amplifiers (Axoclamp-2A, Axon
out' of intracellular constituents via the whole-cell recording Instruments). Current and voltage were filtered at 10 kHz and
pipettes may have influenced the site of action potential sampled at 50 kHz using a VME bus computer (Motorola Delta
initiation (Regehr & Armstrong, 1994). Furthermore, it has series 1147, Tempe, AZ, USA). For both cell-attached and whole-
been suggested that a developmental increase in the cell recordings, patch pipettes (4-7 MQ for somatic recordings,
8-10 MQ for dendritic recordings, 10-12 MQ for axonal
dendritic sodium channel density may lead to the initiation recordings) were filled with a potassium gluconate-based intra-
of action potentials in the dendrites of layer 5 pyramidal cellular solution (mM: 120 potassium gluconate, 20 KCl, 10 Hepes,
neurons in mature animals (Mainen, Joerges, Huguenard & 10 EGTA, 2 Na2-ATP and 2 MgCl2; pH 7*3 with KOH). During
Sejnowski, 1995). These questions, together with the cell-attached recordings the patch pipette was held at -60 mV
suggestion that in hippocampal CAI pyramidal neurons the (i.e. transmembrane potential close to 0 mV) to inactivate voltage-
site of action initiation can shift into the dendrites during dependent currents. Measurements of resting membrane potential
high intensity distal synaptic stimulation (Turner, Meyers, were made immediately following break-in (usually with seconds),
Richardson & Barker, 1991; Spruston, Schiller, Stuart & before significant dialysis with the intracellular pipette solution,
and whole-cell recordings were terminated if the access resistance
Sakmann, 1995), prompted a reinvestigation of the site of exceeded 100 MQ. In some experiments biocytin (5 mg ml-') was
action potential initiation in mature neocortical pyramidal added to the pipette solution and cells subsequently stained with
neurons. the avidin-biotinylated horseradish peroxidase complex reaction to
Here we address the issue of action potential initiation in reveal the cell morphology (see Schiller et al. 1997). Recordings
were only made from cells where it was possible to follow the apical
electrophysiologically mature (P26-30) layer 5 neocortical dendrite or axon from the cell soma to the site of dendritic or axonal
pyramidal neurons at physiological temperatures during recording. Axons were identified by their emergence from the basal
initiation of action potentials by extracellular synaptic part of the soma, myelination (usually starting 25 to 50 ,um from
stimulation at different locations along the somato-dendritic the soma), projection to the white matter and antidromic activation
axis and with different stimulus intensities. Simultaneous (n = 3). That dendritic or axonal recordings were made from the
dual and triple patch pipette recordings from the soma, apical dendrite or axon was confirmed by the use of fluorescent
dendrites and axon of the same neuron were made to dyes. The distance of dendritic recordings from the soma was
determine the site of action potential initiation by measured from the centre of the soma, whereas axonal recordings
were measured from the edge of the soma at the beginning of the
measuring at which recording site action potentials were axon (i.e. from the axon hillock).
recorded first. By sampling the membrane voltage of the
same neuron at different locations it was also possible to Excitatory synaptic potentials (EPSPs) were evoked by 200 #us
study how, once initiated, regenerative events spread within pulses (up to 100 V in amplitude) applied to an extracellular
stimulating pipette placed either distally (top of layer 2/3) or
pyramidal neurons. The experiments showed that action proximally (layer 4) and approximately 100 #tm lateral to the apical
potentials were always initiated in the axon before the soma dendrite. This stimulation pipette was made from a fire-polished
even when synaptic stimulation was strong enough to patch pipette with a tip diameter of approximately 10 ,um filled
initiate regenerative potentials in dendrites. Once initiated, with oxygenated extracellular solution. 'Threshold' stimulation was
action potentials propagated orthogradely into the axonal defined as extracellular synaptic stimulation or current injection
arbor and back into the dendritic tree. Some of these results at an intensity sufficient to initiate somatic action potentials on
have been previously published in abstract form (Stuart & most trials, whereas 'high intensity' stimulation was defined as
Sakmann, 1996). stimulation at an intensity up to 5-10 times greater than that
J Phy8iol. 505.3 Action oet
Ato potentiial initiation 619

required for somatic action potential initiation. Dendritic current RESULTS

injections used to simulate EPSPs were generated by injection of an All experiments were made on visually identified layer 5
exponentially rising and falling voltage waveform (r.n, 0-3 ins; Toff'
3 ins) into the current-clamp input of the Axoclamp amplifier used pyramidal neurons in brain slices from 4-week-old rats at
to make the dendritic recording. Data from these experiments were physiological temperatures (35T0). Recordings were only
only used if the dendritic access resistance wa.s less than 30 ML2, used if the somatic resting membrane potential was more
was stable and could be adequately compensated. negative that -60 mV and the amplitude of somatic action
Analysis potentials, measured from threshold, was greater than
All data analysis was performed with IGOR (Wavemetrics, OR, 80 mV. A total of forty-five cells met these criteria from
USA) on a Macintosh computer. Action potential amplitude was thirty-one simultaneous somatic and dendritic, six simul-
measured from a baseline set at threshold and action potential half- taneous somatic and axonal and eight simultaneous somatic,
width gives the duration at half amplitude. The 'onset latency' and axonal and dendritic recordings. The average somatic
'peak latency' between events recorded at different locations is resting membrane potential and action potential amplitude
defined as the time difference between the onset or peak of a for these cells was -64-3 + 2-2 mV and 96-8 + 5-7 mV
particular event relative to that of the action potential recorded by (n = 45), respectively. Somatic input resistance and apparent
the somatic pipette. The time of onset was defined as the time at
which the maximum rate of rise of voltage was greater that membrane time constant were 32 + 3-9 M.Q and 11 + 08 ins,
approximately 20-40 V s-' for dendritic events, or 50 V s-' for respectively (n = 5). In addition, bursts of action potentials
somatic and axonal events. in response to somatic current injection or synaptic
stimulation were observed in 36 % of the neurons. In some

A B C
threshold high intensity
1

dendrite dendrite
-FL

2/3

soma

4
dendrite

100 AtM

Figure 1. Action potential initiation during synaptic stimulation at different intensities

A, camera lucida drawing showing the approximate location of stimulation and somatic and dendritic
recording pipettes during simultaneous somatic and dendritic recording from a neocortical layer 5
pyramidal neuron. Numbers on the left refer to cortical layers. Synaptic stimulation was in upper layer 2/3.
Only the initial part of the axonal arbor is shown. B, somatic and dendritic (thicker traces) cell-attached
(top) and whole-cell voltage (bottom) recordings during threshold intensity synaptic stimulation in layer
2/3. C, somatic and dendritic (thicker traces) cell-attached (top) and whole-cell voltage (bottom) recording,s
during high intensity synaptic stimulation in layer 2/3. All recordings from the same cell. Dendritic
recording 175 ,sm from the soma. The inserts in the lower parts of B and C show the whole-cell recordings
on a reduced time base. Width, 20 ins; height, 125 mV.
620 G. Stuart, J Schiller and B. Sakmann J Physiol.505.3

A B
120 - 60 -
100-
E 40 -
E 80- 0 C 20 - S
a) 0-
60-
V : ~~~~~0
e -20-
CE 40- 0 ce -40 -
E
< 20- E -60 -
0- a) -80 -
0 200 400 600 0 200 400 600
Distance from soma (pm) Distance from soma (pm)

C D
5-
800-
- 4-
0
f 600-
-c 3 0 *0
B 2- 400
0 E 00
*0 0
I 1- 200-
.0.~~~~
0- 0-
I I
.

0 200 400 600 0 200 400 600

Distance from soma (pm) Distance from soma (pm)

E F
1.0o 2.0 -

0
E
E 0.8 E 1.5- 0
0 0.6 0
c 0 0
a) 1.0-
'z 0.4- 0

4-
0
c 0.2
cX
0.5-
0 0o
0.0 0.0

0 200 400 600 0 200 400 600

Distance from soma (pm) Distance from soma (pm)

Figure 2. Properties of back-propagating action potentials at different distances from the soma
All data were obtained during simultaneous somatic and dendritic, or somatic, dendritic and axonal
recordings at the resting membrane potential during threshold synaptic stimulation in layer 2/3 under
conditions where somatic action potentials were initiated prior to any dendritic regenerative response.
A, amplitude of somatic (0; mean + S.D.) and back-propagating action potentials at different distances from
the soma (0) measured from a baseline set at threshold. The data were fitted to a single exponential with a
distance constant of 155 ,um (asymptotic amplitude: 50 mV). B, membrane potential reached at the peak of
somatic (0; mean + S.D.) and back-propagating action potentials at different distances from the soma (0).
The data were fitted to a single exponential with a distance constant of 158 ,um (asymptotic amplitude:
9 mV). The resting membrane potential at the soma (O; mean + S.D.) and at the different dendritic
recording sites (*) is also indicated. These data were fitted with a linear regression, slope 9 ,uV ,um-1.
C, width at half-amplitude (half-width) of somatic (0; mean + S.D.) and back-propagating action potentials
at different distances from the soma (0). The data were fitted with a linear regression, slope 2 /ss sm-'.
D, maximum rate of rise (Vmax) of somatic (0; mean + S.D.) and back-propagating action potentials at
different distances from the soma (0). The data were fitted with a single exponential with a distance
constant of 246 ,um. E, time difference between the onset of somatic and back-propagating dendritic action
potentials (onset latency) at different distances from the soma. The data were fitted with a linear regression
(forced to go through zero onset latency at O sum) whose slope gave a conduction velocity of 1P2 m s 1.
F, time difference between the peak of somatic and back-propagating dendritic action potentials (peak
latency) at different distances from the soma. The data were fitted with a linear regression (forced to go
through zero peak latency at 0 sum) whose slope gave a conduction velocity of 0-5 m s-'.
J Physiol. 505.3 Action potential initiation

experiments, neurons were filled with biocytin and neurons, the observed site of action potential initiation
subsequent staining showed that these cells had apical obtained with the different recording configurations were
dendrites which extended to the pia forming an apical tuft identical (n = 7; see Fig. 1).
(n = 3; see Fig. lA). These basic electrophysiological and These data show that distal, but not proximal, synaptic
morphological properties are similar to those previously stimulation can initiate dendritic regenerative responses
reported for mature tufted layer 5 pyramidal neurons (see prior to somatic action potentials, and that the occurrence of
McCormick & Prince, 1987; Kasper, Larkman, Liibke & this is increased as the intensity of synaptic stimulation
Blakemore, 1994). increased. One interpretation of this result is that the site of
Action potential initiation during threshold and high action potential initiation is dependent on the intensity and
intensity synaptic stimulation location of the synaptic input, such that action potential
Simultaneous cell-attached and whole-cell recordings from initiation can shift from close to the. soma into the apical
the soma and apical dendrite of the same layer 5 pyramidal dendrites during high intensity distal synaptic stimulation,
neuron were used to assess the site of action potential as has been suggested to be the case in hippocampal CAI
initiation during either proximal (layer 4) or distal (top of pyramidal neurons (Turner, Meyers, Richardson & Barker,
layer 2/3) extracellular synaptic stimulation (Fig. IA). 1991; Spruston et al. 1995).
Extracellular (cell-attached) recording was used to avoid Properties of back-propagating action potentials
possible effects on action potential initiation of 'wash-out' of
intracellular constituents or capacitive loading by the
Following initiation near the soma, action potentials actively
recording pipettes.
propagated back into the dendrites (see Stuart & Sakmann,
1994). The properties of back-propagating action potentials
Proximal synaptic stimulation always evoked action evoked by synaptic stimulation are described in detail below
potentials which were observed to occur first at the somatic only for those cells where synaptic stimulation evoked
recording site, independent of whether action potentials somatic action potentials prior to any dendritic response.
were evoked by threshold or high intensity synaptic These cells were chosen for this analysis so as to isolate the
stimulation (n = 5). Distal synaptic stimulation, however, properties of back-propagating action potentials from those
could shift the apparent site of action potential initiation of dendritically initiated events.
from close to the soma into the apical dendrites, particularly
during high intensity synaptic stimulation (Fig. 1; 10 out of The dependence of back-propagating action potential
15 cells). In those cases where it was possible to obtain both
amplitude on the distance the dendritic recording was made
from the soma was similar to that described by Stuart &
cell-attached and whole-cell recordings from the same

dendrite
I

I 20 mV
200 ms

soma

Figure 3. Frequency-dependent attenuation of back-propagating action potentials

Train of action potentials evoked by a long somatic current pulse (1 8 s, 600 pA) during simultaneous
dendritic (top) and somatic (bottom) recording 580 4um from the soma.
622 0. Stuart, J Schililer and B. Sakmann J Physiol. 505.3

Sakmann (1994) in 2-week-old animals at room temperature depolarized 500 ,um from the soma. This relationship was
(Fig. 2A). Action potentials propagated back into the apical statistically significant (P < 0 05). The half-width of back-
dendrites in a decremental manner, and were still of propagating action potentials increased slightly and their
substantial amplitude 500-600 /tm from the soma. Perhaps maximum rate of rise (Vmax) decreased substantially with
more important than the amplitude of the voltage change distance from the soma (Fig. 2C and D). The latency
measured from threshold is the absolute dendritic membrane difference between the onset and peak of somatic and
potential reached as this will be the most important dendritic back-propagating action potentials increased
determinant of the extent to which back-propagating action approximately linearly with distance from the soma over
potentials activate dendritic voltage-dependent channels or the first 700 ,sm of the apical dendrite (Fig. 2E and F). The
shunt synaptic conductances. The membrane potential slope of the linear regression fitted to this data gave an
reached by back-propagating action potentials at different approximate conduction velocity of propagation of the onset
distances from the soma is shown together with the dendritic and peak of action potentials back into the dendritic tree of
resting membrane potential at each dendritic recording site 1 2 and 0 5 m s-', respectively (see Fig. 2E and F).
in Fig. 2B. At distances up to 700 ,um from the soma, the
Recent experiments in hippocampal CAI pyramidal cells
voltage change associated with the back-propagating action
have shown that the amplitude of back-propagating action
potential combined with the dendritic synaptic potential
potentials decreases during a high frequency train, and that
usually depolarized the dendritic membrane potential to or failure of back-propagation of action potentials can occur at
past 0 mV. Note also that the resting membrane potential in
dendritic branch points (Spruston et al. 1995). While no
the dendrites was slightly more depolarized than at the
soma (Fig. 2B; on average approximately 4-5 mV more
failure of back-propagation of action potentials was

A B
control control
soma
soma
I 20 mV I 20 mV
5 ms 5 ms
dendrite
dendrite

cobalt cadmium
soma
I 20 mV I 20 mV
1 ms 2 ms
dendrite

Figure 4. Calcium electrogenesis associated with back-propagating action potentials

A, top: somatic and dendritic (thicker trace) action potentials evoked by threshold synaptic stimulation in
layer 2/3. Note the shoulder on the falling phase of the dendritic action potential (*). Middle: same traces as
shown in the top on an expanded time scale (see scale bar at bottom). Bottom: effect of CoCl2 (2 mM) on
somatic and dendritic (thicker trace) action potentials evoked by a threshold somatic current pulse (200 ms,
500 pA). All recordings from the same cell. Dendritic recording 325 #sm fromn the soma. B, top: somatic and
dendritic (thicker trace) action potentials evoked by threshold synaptic stimulation in layer 2/3 which
evoked a somatic action potential burst. Note the increased amplitude of dendritic action potentials during
the burst, despite the decrease in somatic action potential amplitude. Middle: same traces as shown in the
top on an expanded time scale (see scale bar at bottom). Bottom: effect of cadmium (200 /,M) on somatic
and dendritic (thicker trace) action potentials during action potential burst firing evoked by a threshold
somatic current pulse (200 ms, 500 pA). All recordings from the same cell. Dendritic recording 390 #um
from the soma. Different cell from A.
J Physiol. 505.3 Action potential initiation 623

observed at the dendritic recording sites in the present shown in Fig. 4. Note the reduced amplitude and width of
study, there was a clear decrease in the amplitude of back- not the first, but subsequent back-propagating action
propagating action potentials during a high frequency potentials during a high frequency (300 Hz) action potential
(25 Hz) action potential train at distal dendritic recording burst in the presence of cadmium (Fig. 4B). These results
locations (Fig. 3). show that activation of dendritic voltage-activated calcium
Previous studies have shown that action potentials in the channels by back-propagating action potentials causes a
apical dendrites of layer 5 pyramidal neurons are associated substantial broadening of the dendritic spike. Furthermore,
with significant dendritic calcium electrogenesis (see Amitai during burst firing this calcium electrogenesis also increases
et al. 1993; Kim & Connors, 1993). Similarly, back- the amplitude of dendritic back-propagating action
propagating action potentials recorded in the present study potentials.
were followed by a shoulder on their falling phase (* in Properties of dendritic regenerative potentials
Fig. 4A), which was larger the more distal dendritic In those cases where dendritic regenerative potentials were
recordings were made from the soma, and was particularly initiated prior to somatic action potentials following distal
prominent during bursts of somatic action potentials synaptic stimulation the further the dendritic recording was
(Fig. 4B). Both the shoulder during the falling phase of made from the soma the longer the time between the onset
dendritic back-propagating action potentials and the larger of the dendritic regenerative potential and somatic action
dendritic response evoked by dendritic back-propagating potentials (Fig. 5). This relationship, together with the
action potentials during burst firing could be blocked by the finding that dendritic regenerative events only occurred
application of cadmium (200 j/M; n = 5), cobalt (2 mM; during distal, but not proximal, synaptic stimulation
n = 4) or nickel (200 ,UM; n = 4). An example of the effect of indicates that these events are initiated at a distal location,
cobalt and cadmium on dendritic calcium electrogenesis presumably in the apical tuft (see Schiller et al. 1997).
during single and bursts of somatic action potentials is

A proximal recording (175 usm)

soma

I 20 mV
dendrite
0.5 ms

Figure 5. Distal location of dendritic electrogenesis

A, initiation of a dendritic regenerative response (thicker trace)
prior to somatic action potentials during proximal dendritic B distal recording (440 /sm)
recording (175 ,um from the soma; synaptic stimulation in
layer 2/3). B, initiation of a dendritic regenerative response soma
(thicker trace) prior to somatic action potentials during distal
dendritic recording (440 ,um from the soma; synaptic stimulation
in layer 2/3). Different cell from A. C, relationship between the
time of onset of dendritic regenerative events relative to the onset dendrite
of somatic action potentials (onset latency) and the distance the
dendritic recording was made from the soma for recordings that
initiated dendritic regenerative events prior to somatic action
potentials. Onset latencies from 3 cells where dendritic
regenerative events occurred prior to, but in isolation of, somatic
action potentials (see Fig. 8C) are not included. C 2.0 -

la 1.5 -
E-151
011

ao0
05
0.0
0 100 200 300 400 500 600 700
Distance from soma (pm)
624 G. Stuart, J Schiller and B. Sakmann J Physiol. 505.3

In some cases the apparent site of action potential initiation other recordings a clear separation of dendritically initiated events
could fluctuate between the somatic and dendritic recording and back-propagating action potentials was not observed, with the
sites from trial to trial at the same stimulus intensity two events merging to form a smooth waveform (see Fig. 6A,
(Fig. 6A). Furthermore, whether dendritic regenerative bottom). Some evidence that dendritic regenerative events may
have inactivated dendritic sodium channels comes, however, from
potentials were initiated prior to somatic action potentials more proximal dendritic recordings where following initiation of a
or not could be modulated by depolarization of the somatic dendritic regenerative potential the dendritic component attributable
(Fig. 6B; n = 5) or hyperpolarization of the dendritic (not to the back-propagating action potential appeared to be absent (see
shown; n = 2) membrane potential. Thus, an additional Fig. 1 C, bottom).
factor in determining whether dendritic regenerative Conductances mediating dendritic regenerative
potentials will be initiated prior to somatic action potentials potentials
will be the level of background excitation and inhibition.
To address the issue of which dendritic conductances
Note back-propagation of somatic action potentials was still underlie initiation of dendritic regenerative potentials
observed in those cells where the dendrites initiated regenerative current injections into the apical dendrite (175-390 sum
events prior to somatic action potentials, despite the fact that the from the soma) via the dendritic recording pipette were
depolarization associated with dendritic regenerative events might
have been expected to inactivate dendritic sodium channels, used in an attempt to mimic the dendritic voltage change
reducing the ability of action potentials to propagate back into the that occurs during an EPSP (see Stuart & Sakmann, 1995).
dendrites. That this was the case can be seen in dendritic recordings At threshold, dendritic current injections always evoked
were the dendritic response was clearly biphasic, presumably action potentials first in the soma (Fig. 7; n = 8). Further
representing first the dendritically initiated regenerative event, and increasing the amplitude of the dendritic current injection
second the back-propagated action potential (see Fig. 6B; top). In in most cases resulted in the initiation of dendritic

A soma B resting membrane potential

soma

dendrite
dendrite {

soma
depolarized
soma

dendrite
dendrite
-J|25 mV
4 ms

Figure 6. Modulation of the apparent site of action potential initiation

Simultaneous somatic and dendritic (thicker traces) recordings during synaptic stimulation in layer 2/3
(dendritic recording 455 #tm from the soma). All recordings from the same cell. A, top: somatic action
potential initiation prior to dendritic regenerative response. Bottom: initiation of a dendritic regenerative
response prior to the first, but not the second, somatic action potential. B, top: simultaneous somatic and
dendritic (thicker traces) recordings during high intensity synaptic stimulation at the resting membrane
potential (different location of synaptic stimulation in layer 2/3). Note the biphasic nature of the
dendritically recorded response. Bottom: depolarization of the soma by constant current injection causes
the somatic action potential to be initiated prior to a dendritic regenerative response.
J Physiol. 505.3 Action potential initiation 625

regenerative potentials prior to somatic potentials (Fig. 7; 5 Initiation of dendritic regenerative potentials in
out of 8 cells; see also Stuart & Sakmann, 1994). When this isolation of somatic action potentials
occurred the dendritic response was clearly biphasic and As shown in Schiller et al. (1997), in some recordings
similar to that sometimes observed during synaptic dendritic regenerative potentials were observed at stimulation
stimulation (compare Fig. 7 with Fig. 6B, top). Presumably intensities subthreshold for somatic action potential initiation
the early dendritic response represents a locally generated (Fig. 8B; n = 3), or could appear in apparent isolation of
regenerative potential initiated close to the dendritic somatic action potentials, preceding them by up to 10 ms
recording pipette in the proximal apical dendrite, and the (Fig. 8C; n = 3). In the same neurons at similar stimulation
later response the dendritic voltage change associated with intensities dendritic regenerative events could occur prior to
the back-propagating somatic action potential. The somatic action potentials, apparently synchronized to
application of cadmium (200 ,UM; n = 4) had no effect on the somatic action potential initiation (Fig. 8D; n = 3). When
initiation of these dendritic regenerative events (Fig. 7A), initiated in isolation of somatic action potentials, dendritic
whereas TTX (1 /,M; n = 3) completely blocked both regenerative potentials attenuated significantly as they
dendritic responses and somatic action potentials (Fig. 7B). spread to the soma, such that the voltage change at the soma
That the application of cadmium in these experiments in response to these events was difficult to distinguish from
blocked voltage-dependent calcium channels was confirmed that which occurred when distal synaptic stimulation failed
by the complete block of synaptic transmission in these to evoke a dendritic regenerative potential (compare Fig. 8A
experiments. and B). These results indicate that dendritic regenerative

A threshold (4 nA) B threshold (3 nA)

soma soma
dendrite dendrite

high intensity (6 nA) high intensity (5 nA)

cadmium (6 nA) TTX (5 nA)

-J 25 mV -J 25 mV
2 ms 2 ms
dendrite
-v soma

Figure 7. Action potential initiation by simulated EPSPs

Somatic and dendritic (thicker traces) recording of action potentials initiated by a dendritic current
injection with the shape of an excitatory postsynaptic current (see Methods). A, top: somatic action
potential initiation during threshold intensity dendritic current injection (4 nA). Middle: increasing the size
of the dendritic current injection (6 nA) initiates a dendritic regenerative response prior to somatic action
potentials. Bottom: effect of cadmium (200 /tM) on action potential initiation (dendritic current injection
6 nA). Dendritic recording 390 /tm from the soma. B, top: somatic action potential initiation during
threshold intensity dendritic current injection (3 nA). Middle: increasing the size of the dendritic current
injection (5 nA) initiates a dendritic regenerative response prior to somatic action potentials. Bottom: effect
of TTX (1 uM) on action potential initiation (dendritic current injection 5 nA). Dendritic recording 290 /um
from the soma. Different cell from A.
626 6. Stuart, J Schiller and B. Sakmann J Phy8iol. 505.3

potentials undergo significant attenuation as they spread to The rate of rise of axonal action potentials (Vmax) increased
the soma. with recording distance from the soma (Fig. 1OD; this
relationship was statistically significant, P < 0 05). On
Axonal action potential initiation average, the Vmax of axonal action potentials was
To determine the site of action potential initiation, 826 + 145 V s-' (n = 14) compared with 697 + 158 V s-' at
simultaneous recordings were made from the soma and axon the soma. Axonal action potentials always occurred before
(Fig. 9A) and action potentials were evoked by distal somatic action potentials, with the latency difference
synaptic stimulation. As previously reported for younger between the onset of axonal and somatic action potentials
animals (P12-14; Stuart & Sakmann, 1994), synaptic increasing as recordings were made more distal from the
stimulation evoked action potentials which were always soma (Fig. 10E). This suggests that the actual site of action
observed to occur first at the axonal recording site (Fig. 9B; potential initiation was in the axon at a distance greater
n= 14). than 30 jum from the axon hillock. The slope of the linear
Some of the properties of axonal action potentials recorded regression fit to this data gave an approximate conduction
during simultaneous axonal and somatic recordings are velocity of action potential onset of 0 4 m s-' (Fig. IOE).
shown in Fig. 10. In some cases action potentials were The latency difference between the peak of somatic and
larger in the axon than the soma (see Fig. 9B); however, on axonal action potentials also increased as axonal recordings
average the amplitude of axonal action potentials (recorded were made more distally from the soma (Fig. 1OF),
up to 30 ,um from the axon hillock) was similar to that of indicating an approximate conduction velocity of action
somatic action potentials (Fig. 1OA and B). The half-width potential peak of 0 3 m s-. In some cases (n = 2), however,
of axonal action potentials was also similar to that of the peak of somatic and axonal action potentials occurred
somatic action potentials (Fig. 1OC: 0-46 + 0-06 ms in the simultaneously, despite a clear difference in action potential
axon (n = 14) compared with 0-46 + 0 07 ms at the soma). onset (Fig. 10F). This may have been due to slight damage

A
dendrite soma

B
dendrite

soma Figure 8. Generation of dendritic electrogenesis

in complete and relative isolation of somatic
action potentials
Somatic and dendritic (thicker traces) recording
during synaptic stimulation in layer 2/3. All
C soma recordings from the same cell. Dendritic recording
440 ,um from the soma. A, subthreshold somatic and
dendritic EPSPs. B, initiation of a dendritic
regenerative potential in the absence of somatic
dendrite action potentials. C, initiation of a dendritic
regenerative potential in relative isolation from
somatic action potentials. D, initiation of a dendritic
regenerative potential prior to a somatic action
potential.

D soma

-J 20 mV
dendrite
5 ms
J Phy8iol. 505.3 Action potential initiation 627

of the axon during recording, which could also explain the potentials was unchanged (Fig. llC; 6 out of 6 neurons).
reduced size of axonal compared with somatic action These experiments therefore show that action potentials are
potentials in some cells. always initiated in the axon before the soma, independent of
whether dendritic regenerative potentials are initiated prior
To investigate the site of action potential initiation under to somatic action potentials or not.
conditions where distal synaptic stimulation initiated
dendritic regenerative potentials prior to somatic action
potentials, simultaneous recordings were made from the DISCUSSION
soma, dendrite and axon of the same neocortical layer 5 The experiments described here were designed to locate the
pyramidal neuron (Fig. llA). These experiments showed site of action potential initiation and propagation during
that at threshold action potentials were always observed synaptic stimulation of mature layer 5 pyramidal neurons at
first at the axonal recording site, and recorded subsequently physiological temperatures. The results show that the site of
by the somatic and then dendritic recording pipettes (Fig. action potential initiation in these neurons is always in the
1lB; n = 8). Distal synaptic stimulation at high intensity axon, despite the fact that distal synaptic stimulation can
was then used to evoke dendritic regenerative potentials initiate dendritic regenerative potentials prior to somatic
prior to somatic action potentials. Under these conditions, action potentials. Once initiated in the axon, action
the temporal relationship between axonal and somatic action

Figure 9. Site of action potential initiation during

simultaneous somatic and axonal recording
A, IR-DIC image during a simultaneous somatic and axonal
recording from the same layer 5 pyramidal neuron.
B, top: somatic and axonal (23 ,um from the edge of the soma)
recording during action potential initiation by threshold
synaptic stimulation in layer 2/3. Bottom: same recording on an
expanded time scale (thicker traces represent axonal recording).

B
soma

-j 20 mV
5 ms

-J 20 mV
0.5 ms
628 C. Stuart, J Schiller and B. Sakmann J Physiol. 505.3

A B
120- 5 60- 0
0 0
100 -0 0 0 1-i 40- so o -

S
.-I 0 0 * *= 20-
80
a) CDo 0-
60 - 0
:
'a 40 -
. -40 -
20 -
E -60
E -80 I I
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Distance from soma (pm) Distance from soma (pm)

C D
1.0 - 1200-
, 0.8 - 1000-
0 0
T 800- 0
= 0.6- 0
*0 0 2. 600
0-1-0's 60 me
._ 0
3 0.4- 0
0
x

jE 400
I 0.2- 200
0.0 - 0
I
1 1 1 I I

0 5 10 15 20 25 30 0 10 20 30
Distance from soma (pm) Distance from soma (pm)

E F
100 - 200 -

'. 80- 0 0 (n
-= 150_
O 60- 0
c
* 0
a) a) 100- 0 0
-% 40- 0 * Cu

cn 0
C 20- 50-
0 CO
00
0- 0-
I I I I
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Distance from soma (pm) Distance from soma (pm)

Figure 10. Properties of axonal action potentials at different distances from the soma
All data were obtained during simultaneous somatic and axonal, or somatic, axonal and dendritic
recordings at the resting membrane potential during synaptic stimulation in layer 2/3. A, amplitude of
somatic (0; mean + S.D.) and axonal action potentials at different distances from the soma (0) measured
from a baseline set at threshold. The data were fitted with a linear regression, slope -0-2 mV /,m-1.
B, membrane potential reached at the peak of somatic (0; mean + S.D.) and axonal action potentials at
different distances from the soma (0). The data were fitted with a linear regression, slope -0-2 mV um-'.
The resting membrane potential at the soma (O; mean + S.D.) and at the different axonal recording sites (*)
is also indicated. This data were fitted with a linear regression, slope -5 #uV um-'. C, width at half-
amplitude (half-width) of somatic (0; mean + S.D.) and axonal action potentials at different distances from
the soma (0). The data were fitted with a linear regression, slope -0-3 us ,sm-'. D, maximum rate of rise
(Vmax) of somatic (0; mean + S.D.) and axonal action potentials at different distances from the soma (0).
The data were fitted with a linear regression, slope 6-9 V s-' 4um-'. E, time difference between the onset of
somatic and axonal action potentials (onset latency) at different distances from the soma. The data were
fitted with a linear regression (forced to go through zero onset latency at 0 um) whose slope gave a
conduction velocity of 0-4 m s-'. F, time difference between the peak of somatic and axonal action
potentials (peak latency) at different distances from the soma. The data were fitted with a linear regression
(forced to pass through zero peak latency at 0 /sm) whose slope gave a conduction velocity of 0-3 m s-T.
J Physiol. 505.3 Action potential initiation 629

potentials propagated both orthogradely into the axonal potential initiation was the same during cell-attached as
arbor and retrogradely into the dendritic tree. with whole-cell recording (see Fig. 1) suggests this was not
To address the possibility that the site of action potential the case.
initiation may change during development (see Mainen et al. The main difference between the results of the present
1995) all experiments in the present study were conducted study and those of the earlier study on 2-week-old animals
on 4-week-old animals. Developmental studies show that (Stuart & Sakmann, 1994) was the enhanced dendritic
the electrophysiological properties of neocortical layer 5 excitability in 4-week-old animals at more physiological
pyramidal neurons are mature by this age (McCormick & temperatures, which could lead to the initiation of dendritic
Prince, 1987; Kasper et al. 1994). Furthermore, at 4 weeks regenerative potentials during distal synaptic stimulation
of age the density of both voltage-dependent sodium and (see also Schiller et al. 1997). This increased dendritic
calcium channels in neocortical pyramidal neurons has excitability may reflect a developmental increase in the
reached that expressed by adult neocortical pyramidal density of dendritic voltage-dependent conductances,
neurons (Cummins, Xia & Haddad, 1994; Lorenzon & although developmental changes in morphology and the
Foehring, 1995). The possibility that 'wash-out' of intra- passive membrane properties of layer 5 pyramidal neurons
cellular constituents or capacitive loading by the whole-cell (McCormick & Prince, 1987; Kasper et al. 1994), or the
recording pipettes may have affected the site of action density and strength of synaptic innervation of the distal
potential initiation (see Regehr & Armstrong, 1994) was apical dendrites may also contribute. In addition, the higher
addressed by comparing action potential initiation during temperatures used in the present experiments (35 °C
cell-attached recording with that observed during whole- compared with 22 °C) may also have contributed to the
cell recording from the same neuron. That the site of action observed increase in dendritic excitability.

A B threshold

soma 20 mV
0.5ms

2/3 f=D

C high intensity
4

soma

5
dendrite,

6 100 gm

Figure 11. Site of action potential initiation during high intensity synaptic stimulation
A, camera lucida drawing showing the approximate location of stimulation and recording pipettes during
simultaneous whole-cell voltage recording from the axon, soma and dendrite of the same layer 5 pyramidal
neuron and synaptic stimulation in upper layer 2/3. Numbers on the left refer to approximate borders of
cortical layers. B, simultaneous recording from the axon, soma (thicker trace) and dendrite after threshold
intensity synaptic stimulation in layer 2/3. Same neuron as shown in A. C, simultaneous recording after
high intensity synaptic stimulation which initiated a dendritic regenerative potential prior to the somatic
action potential. All traces are from the same experiment. Axonal and dendritic recordings were made 20
and 300 jum from the soma, respectively.
630 6. Stuart, J Schiller and B. Sakmann J. Physiol. 505.3

Similar to hippocampal CAl pyramidal neurons (Spruston neighbouring, electrically coupled neurons (MacVicar &
et al. 1995), back-propagation of action potentials into neo- Dudek, 1981; Valiante, Velazquez, Jahromi & Carlen, 1995).
cortical dendrites was dependent on somatic action potential Brief dendritic current injections were made into the
frequency, such that there was a decrease in the amplitude proximal apical dendrite in an attempt to simulate
of dendritic action potentials during a train of somatic synaptically evoked dendritic electrogenesis (Fig. 7). At the
action potentials (Fig. 3). This effect was particularly clear
dendritic recording sites where these current injections were
during a high frequency burst of somatic action potentials in
made (175-390 ,um distal to the soma), dendritic regenerative
the presence of calcium channel blockers (Fig. 4B). While potentials were only observed with current injections which
the mechanism(s) underlying this observation are unknown,
were suprathreshold for initiation of somatic action potentials
cumulative inactivation of dendritic sodium channels may (see Fig. 7). Furthermore, the regenerative potentials
be involved (Colbert & Johnston, 1996a). Failure of action initiated by these current pulses were completely blocked by
potential back-propagation (see Spruston et al. 1995) was TTX, showing that they were mediated by voltage-
not observed at the dendritic recording sites investigated in
dependent Na+ channels. These findings differ from those
the present study; however, Ca2P imaging experiments described in the preceding paper by Schiller et at. (1997),
suggest that this may occur in the distal dendrites of the who show that dendritic regenerative potentials evoked by
apical tuft (Schiller, Helmchen & Sakmann, 1995). more distally applied (550-940 ,um distal to the soma) and
Axonal action potential initiation longer dendritic current pulses can be evoked under
The experiments described here together with results from conditions where the soma always remains subthreshold,
simultaneous somatic and axonal recordings in cerebellar and which are mediated mostly by dendritic voltage-
Purkinje and hippocampal subicular pyramidal neurons activated Ca2+ channels. This difference is presumably due
(Stuart & Hiausser, 1994; Colbert & Johnston, 1996b), to the difference in how the regenerative events were
directly confirm conclusions based on somatic microelectrode initiated by dendritic current injection in the two studies. It
recordings that action potential initiation occurs in the axon seems likely that the short-duration, more proximal
(see Eccles, 1964). While the exact site of action potential dendritic current injections used in the present study would
initiation in the axon of neocortical pyramidal neurons is have initated Na+-dependent dendritic regenerative events
unknown, the increase in the maximum rate of rise and more proximally than the long-duration, more distal current
time of onset of axonal action potentials relative to somatic injections used in Schiller et al. (1997). Given that this was
action potentials as recordings were made more distal from the case, the results from both investigations suggest that
the soma (see Fig. 1OD and E) suggests that action potential dendritic regenerative potentials are mixed Ca+- and Na+-
initiation occurs at a site at least 30 #sm distal to the axon dependent potentials which are predominantly mediated by
hillock. Recent work in hippocampal subicular pyramidal Ca2+ channel activation in the distal portions and by Nae
neurons also suggests that action potential initiation occurs channel activation in the proximal portion of the apical
in the axon at a site distal from the soma, possibly at the dendrite.
first node(s) of Ranvier (Colbert & Johnston, 1996b; see also Relationship between dendritic regenerative
earlier work in motoneurons by Gogan, Gueritaud & Tyc- potentials and axonal action potentials
Dumont (1983) and Coombs, Curtis & Eccles (1957)). Further While there was, in most cases, a clear temporal
evidence is needed to establish if this is the case in relationship between the initiation of dendritic regenerative
neocortical pyramidal neurons.
potentials and the occurrence of somatic action potentials,
Dendritic regenerative potentials this was not always the case (Fig. 8; see also Schiller et at.
The experiments suggest that stimulation of synapses on the 1997). This, together with the finding that distally initiated
distal apical dendrite, particularly at high intensity, can regenerative potentials attenuate significantly as they
evoke dendritic regenerative potentials initiated in the distal spread to the soma, suggests that action potential initiation
apical dendrites (see also Schiller et al. 1997). Once initiated only occurs in the axon after summation with other synaptic
these events propagate to the soma, undergoing significant inputs from different parts of the dendritic arbor. That this
attenuation. In this respect these dendritic regenerative is the case is clearly demonstrated by the experiments
responses are similar to how Spencer & Kandel (1961) where triple whole-cell voltage recordings were made
originally described the so called fast prepotentials (FPPs) simultaneously from the dendrites, soma and axon of the
they observed in hippocampal pyramidal neurons. The same neuron (Fig. 11). These experiments showed that
dendritically initiated regenerative potentials described action potentials were initiated in the axon before the soma,
here and those in Schiller et al. (1997), however, differ from and with the same temporal relationship, independent of
FPPs in that FPPs occur spontaneously and have a fast rise whether dendritic regenerative potentials occurred prior to
time and decay. Such potentials were not observed. somatic action potentials or not. The marked attenuation
Furthermore, while it was originally thought that FFPs of dendritic regenerative potentials as they spread to the
represent initiation of dendritic action potentials, there is soma and the axon contrasts with the relative effective
now evidence that they are due to action potentials in propagation of somatic action potentials back into the
J. Physiol. 505.3 Action potential initiation 631

dendritic arbor. Such a unidirectional propagation of active et al. (1997) could enhance synchronization by increasing
potentials within the dendritic arbor is predicted by the precision with which cortical neurons respond to
simulation studies (Rall & Segev, 1987; Mainen et al. 1995), common excitatory synaptic input (Softky, 1994). Whether
and presumably occurs due to impedance mismatches at this is the case or not will depend on whether the distal
dendritic branch points and increase in dendritic diameter dendritic electrogenesis described above in vitro also occurs
encountered as a regenerative potential propagates from the during the normal in vivo operation of the cortex.
dendrites toward the soma (Goldstein & Rall, 1974; Jack,
Noble & Tsien, 1983; Rall & Segev, 1987). Non-uniform
distributions of dendritic voltage-dependent Na, Ca2+ and
K+ channels may could also contribute to this attenuation. AMITAI, Y., FRIEDMAN, B., CONNORS, B. W. & GUTNICK, M. J. (1993).
These findings, and those of Schiller et al. (1997), therefore Regenerative activity in apical dendrites of pyramidal cells in
question the interpretation of previous reports of a shift in neocortex. Cerebral Cortex 3, 26-38.
the site of action potential initiation during high-intensity BUHL, E. H., HALASY, K. & SOMOGYI, P (1994). Diverse sources of
synaptic stimulation (Turner et al. 1991; Spruston et al. hippocampal unitary inhibitory postsynaptic potentials and the
1995). They suggest instead that intense distal synaptic number of synaptic release sites. Nature 68, 823-882.
stimulation can evoke dendritic regenerative potentials COLBERT, C. M. & JOHNSTON, D. (1996a). A decrease in Na current
which attenuate substantially as they spread to the soma. contributes to loss of action potential amplitude in dendritic spike
trains. Society for Neuroscience Abstracts 22, 791.
Conclusions and physiological significance COLBERT, C. M. & JOHNSTON, D. (1996b). The site of action potential
The results from this and the preceding paper (Schiller et al. initiation and Nae channel densities in the initial segment and soma
1997) show that both the distal apical dendrites and the of subicular pyramidal neurons. Journal of Neuroscience 16,
6676-6687.
axon of layer 5 pyramidal neurons are capable of initiating
regenerative potentials. Dendritic regenerative potentials, COOMBS, J. S., CuRTIs, D. R. & EcCLES, J. C. (1957). The
interpretation of spike potentials of motoneurones. Journal of
however, spread only weakly to the soma and axon. As a Physiology 139, 198-231.
consequence, action potential initiation is in the axon even
CUMMINS, T. R., XIA, Y. & HADDAD, G. G. (1994). Functional
when synaptic input is intense enough to initiate dendritic properties of rat and human neocortical voltage-sensitive sodium
electrogenesis. Thus the axon is the final site where synaptic currents. Journal of Neurophysiology 71, 1052-1064.
integration takes place, providing neurons with a single site DESCHENES, M. (1981). Dendritic spikes induced in fast pyramidal
where synaptic inhibition will be most effective. Consistent tract neurons by thalamic stimulation. Experimental Brain Research
with this idea, anatomical studies show that some classes of 43, 304-308.
inhibitory interneurons form synapses specifically on the EcCLES, J. C. (1964). The Physiology of Synapses. Springer-Verlag,
axon initial segment (Buhl, Halasy & Somogyi, 1994; Miles, Berlin.
Toth, Gulyas, Hajos & Freund, 1996). ENGEL, A. K., KONIG, P., KREITER, A. K., SCHILLEN, T. B. & SINGER,
W. (1992). Temporal coding in the visual cortex: new vistas on
Once initiated, axonal action potentials propagate actively integration in the nervous system. Trends in Neurosciences 15,
both into the axonal arbor and back into the dendritic tree. 218-226.
Back-propagating action potentials constitute a rapid GOGAN, P., GUERITAUD, J. P. & TYC-DUMONT, S. (1983). Comparison
retrograde signal to the dendritic tree indicating that the of antidromic and orthodromic action potentials of identified motor
neuron has generated an action potential. This retrograde axons in the cat's brain stem. Journal of Physiology 335, 205-220.
signal is enhanced by dendritic calcium electrogenesis, in GOLDSTEIN, S. S. & RALL, W. (1974). Changes of action potential
particular during bursts of action potentials (Fig. 4). In this shape and velocity for changing core conductor geometry.
way back-propagating action potentials may control, via Biophysical Journal 14, 731-757.
changes in membrane voltage and [Ca2+]i transients, both HIRSCH, J. A., ALONSO, J.-M. & REID, R. C. (1995). Visually evoked
calcium action potentials in cat striate cortex. Nature 378, 612-616.
short- and long-term changes in the efficacy of synaptic
connections and in addition represent a 'binding signal' for JACK, J. J. B., NOBLE, D. & TSIEN, R. W. (1983). Electrical Current
Flow in Excitable Cells. Oxford University Press, Oxford.
the distributed synaptic contacts of a connection (for review
see Stuart, Spruston, Sakmann & Hiausser, 1997). The JOHNSTON, D., MAGEE, J. C., COLBERT, C. M. & CHRISTIE, R. (1996).
Active properties of neuronal dendrites. Annual Review of
regenerative dendritic potentials observed to occur prior to Neuroscience 19, 165-186.
somatic action potentials during synaptic stimulation of the
KASPER, E. M., LARKMAN, A. U., LUBKE, J. & BLAKEMORE, C. (1994).
distal dendrites may have a different function. It has been Pyramidal neurons in layer 5 of the rat visual cortex. II.
suggested that ensembles of distant cortical neurons can Development - of electrophysiological properties. Journal of
generate action potentials which are phase locked (Engel, Comparative Neurology 339, 475-494.
Konig, Kreiter, Schillen & Singer, 1992). While the cellular KIM, H. G. & CONNORS, B. W. (1993). Apical dendrites of the
mechanisms underlying this phase locking are not well neocortex: Correlation between sodium- and calcium-dependent
established, it seems possible that regenerative potentials in spiking and pyramidal cell morphology. Journal of Neuroscience
distal dendrites of the nature described here and by Schiller 13, 5301-5311.
632 G. Stuart, J Schilller and B. Salmann J Physiol. 505.3

LORENZON, N. M. & FOEHRING, R. C. (1995). Characterization of TURNER, R. W., MEYERS, D. E., RICHARDSON, T. L. & BARKER, J. L.
pharmacologically identified voltage-gated calcium currents in (1991). The site for initiation of action potential discharge over the
acutely isolated rat neocortical neurons. II. Postnatal development. somatodendritic axis of rat hippocampal CAI pyramidal neurons.
Journal of Neurophysiology 73, 1443-1451. Journal of Neuroscience 11, 2270-2280.
MCCORMICK, D. A. & PRINCE, D. A. (1987). Post-natal development of VALIANTE, T. A., VELAZQUEZ, J. L. P., JAHROMI, S. S. & CARLEN, P. L.
electrophysiological properties of rat cerebral cortical pyramidal (1995). Coupling potentials in CAl neurons during calcium-free-
neurones. Journal of Physiology 393, 743-762. induced field burst activity. Journal of Neuroscience 15,
MACVICAR, B. A. & DUDEK, F. E. (1981). Electrotonic coupling 6946-6956.
between pyramidal cells: a direct demonstration in rat hippocampal
slices. Science 213, 782-785. Acknowledgements
MAINEN, Z. F., JOERGES, J., HUGUENARD, J. R. & SEJNOWSKI, T. J. G.S. gratefully acknowledges support from the Alexander von
(1995). A model of spike initiation in neocortical pyramidal neurons. Humboldt foundation and the NH&MRC of Australia.
Neuron 15, 1427-1439.
AIILES, R., TOTH, K., GULYAS, A. I., HAJOS, N. & FREUND, T. F. Authors' present addresses
(1996). Differences between somatic and dendritic inhibition in the G. Stuart: Division of Neuroscience, John Curtin School of Medical
hippocampus. Neuron 16, 815-823. Research, Australian National University, Canberra A.C.T. 0200,
POCKBERGER, H. (1991). Electrophysiological and morphological Australia.
properties of rat motor cortex neurons in vivo. Brain Research 539, J. Schiller: Department of Pharmacology, Guggenheim 7, Mayo
181-190. Foundation, Rochester, MN 55905, USA.
RALL, WV. & SEGEV, I. (1987). Functional possibilities for synapses on Author's email address
dendrites and dendritic spines. In Synaptic Function, ed. EDELMAN,
G. AM., GALL, W. E. & COWAN, W. M., pp. 605-636. John Wiley & G. Stuart: Greg.Stuart@anu.edu.au
Sons, New York.
REGEHR, W. G. & ARMSTRONG, C. M. (1994). Dendritic function.
Where does it all begin? Current Biology 4, 436-439. Received 10 April 1997; accepted 14 August 1997.
REGEHR, W., KEHOE, J., ASCHER, P. & ARMSTRONG, C. (1993).
Synaptically triggered action potentials in dendrites. Neuron 11,
145-151.
SCHILLER, J., HELMCHEN, F. & SAKMANN, B. (1995). Spatial profile of
dendritic calcium transients evoked by action potentials in rat
neocortical pyramidal neurons. Journal of Physiology 487,
583-600.
SCHILLER, J., SCHILLER, Y., STUART, G. & SAKMANN, B. (1997).
Calcium action potentials restricted to distal apical dendrites of rat
neocortical pyramidal neurons. Journal of Physiology 505,
605-616.
SOFTKY, W. (1994). Sub-millisecond coincidence detection in active
dendritic trees. Neuroscience 58, 13-41.
SPENCER, W. A. & KANDEL, E. R. (1961). Electrophysiology of
hippocampal neurons. IV. Fast prepotentials. Journal of
Neurophysiology 24, 272-285.
SPRUSTON, N., SCHILLER, Y., STUART, G. & SAKMANN, B. (1995).
Activity-dependent action potential invasion and calcium influx
into hippocampal CAl dendrites. Science 268, 297-300.
STUART, G. & HAUSSER, Ml. (1994). Initiation and spread of sodium
action potentials in cerebellar purkinje cells. Neuron 13, 703-712.
STUART, G. & SAKMANN, B. (1995). Amplification of EPSPs by
axosomatic sodium channels in neocortical pyramidal neurons.
Neuron 15, 1065-1077.
STUART, G. & SAKMANN, B. (1996). Action potential initiation in
neocortical pyramidal neurons - revisited. Society for Neuroscience
Abstracts 22, 794.
STUART, G., SPRUSTON, N., SAKMANN, B. & HAUSSER, M. (1997).
Action potential initiation and backpropagation in neurons of the
mammalian central nervous system. Trends in Neurosciences 20,
125-131.
STUART, G. J. & SAKMANN, B. (1994). Active propagation of somatic
action potentials into neocortical pyramidal cell dendrites. Nature
367, 69-72.