Neocortical cells integrate inputs from thousands of presynaptic neurons distributed along their dendritic arbors. Propagation of postsynaptic potentials to the soma is crucial in determining neuronal output. Using intracellular recordings in anesthetized and non-anesthetized, naturally awake and sleeping cats, we found evidence for generation of fast, all-or-none events recorded at the soma in about 20% of regular-spiking and intrinsically-bursting neurons. These events, termed fast prepotentials (FPPs), were suppressed by hyperpolarizing the neurons or by inhibiting synaptic transmission with perfusion of Ca2+-free artificial cerebrospinal fluid. FPPs could be evoked by activation of specific cortical inputs and allowed neurons to fire at more hyperpolarized levels of membrane potentials. Thus, FPPs represent a powerful mechanism to boost the output of neocortical neurons in response to given inputs. We further found evidence for modulation of FPPs generation across the waking–sleep cycle, indicating important changes in the integrative properties of neocortical neurons in different states of vigilance. We suggest that FPPs represent attenuated spikes generated in hot spots of the dendritic arbor and constitute a powerful mechanism to reinforce the functional connections between specific elements of the cortical networks.
Neocortical neurons operate as integrators of presynaptic neuronal activity. Each of them receives thousands of excitatory and inhibitory inputs, the majority of which arise in local circuits, which are mainly targeting their dendritic arbors (Szentágothai, 1965; Cragg, 1967; Gruner et al., 1974; DeFelipe and Farinas, 1992; Peters, 1994). Postsynaptic potentials (PSPs) from many dendritic branches propagate to, and are integrated at, the soma and axon hillock, where they are transduced into output spike-trains (Yuste and Tank, 1996; Magee, 2000; Reyes, 2001; Williams and Stuart, 2003). Pyramidal-shaped neurons possess branched dendrites that can span virtually all layers, so that some presynaptic terminals are located close to the soma whereas others are situated in remote locations. An important issue concerning synaptic integration is how distant PSPs propagate to, and impact on, the soma.
The propagation of PSPs along the dendrites of cortical cells has been extensively studied during the last decade (Johnston et al., 1996; Yuste and Tank, 1996; Magee, 2000; Reyes, 2001; Hausser and Mel, 2003; Williams and Stuart, 2003). According to the passive cable theory, synaptic inputs would attenuate in a distance-dependent manner along the dendritic arbor, so that proximal inputs have a stronger somatic impact than distal ones (Rall, 1967, 1977; Spruston et al., 1994). This attenuation is partially compensated by an increase in the local amplitude of EPSPs, resulting from both an increase in local input resistance and a decrease in local capacitance with distance from the soma (Jaffe and Carnevale, 1999; Magee, 2000; Magee and Cook, 2000; Williams and Stuart, 2002). Although a relative increase in AMPA receptor density has been found in the distal dendrites of neocortical pyramidal cells (Dodt et al., 1998), the somatic amplitude of unitary EPSPs decrease with the distance of their generation site to the soma (Magee and Cook, 2000; Williams and Stuart, 2002). Thus, the synaptic currents underlying EPSPs do not fully compensate for the filtering due to the passive properties of the dendritic membrane, as is the case in hippocampal pyramidal cells (Magee and Cook, 2000; Smith et al., 2003). On the other hand, the dendrites of pyramidal neurons are provided with voltage- and Ca2+-dependent conductances that actively participate in the propagation or attenuation of PSPs (Johnston et al., 1996; Reyes, 2001).
Regenerative potentials in the dendrites of neocortical neurons have been mainly studied in pyramidal cells in vitro. A wide variety of dendritic potentials has been found to be initiated in the apical arbor, including fast Na+ spikes, slow Ca2+ spikes or complex Na+–Ca2+ potentials (Kim and Connors, 1993; Schiller et al., 1997; Schwindt and Crill, 1997; Stuart et al., 1997; Schwindt and Crill, 1998; Larkum et al., 1999b, 2001). Simultaneous dendritic and somatic recordings have revealed that dendritic spikes generally attenuate when they propagate to the soma (Schiller et al., 1997; Stuart et al., 1997; Larkum et al., 2001). Recent studies have also revealed the generation of Na+ and Ca2+ spikes in apical dendrites of pyramidal neocortical cells in vivo (Helmchen et al., 1999; Svoboda et al., 1999; Zhu and Connors, 1999; Larkum and Zhu, 2002). Both experimental data (Larkum and Zhu, 2002) and modeling studies (Rhodes and Llinás, 2001; Rudolph and Destexhe, 2003) suggest that the in vivo condition could be more favorable for active propagation of dendritic regenerative potentials. Indeed, the active propagation of dendritic spikes in vivo could be facilitated by a more depolarized level of membrane potential (Vm) and depolarizing fluctuations due to synaptic background could occasionally boost the impact of dendritic spikes.
In the present study, we show that ∼20% of regular-spiking (RS) or intrinsically bursting (IB) neocortical neurons display fast and high-amplitude events recorded at the soma, which were different from EPSPs. These sharp events, previously described as fast prepotentials (FPPs) in the hippocampal (Spencer and Kandel, 1961), neocortical (Deschênes, 1981) and thalamocortical (Steriade et al., 1991; Timofeev and Steriade, 1997) neurons presumably represent fast Na+ dendritic spikes that could be evoked by activation of specific cortical inputs. FPPs were able to enhance significantly the somatic impact of such inputs, thus resulting in an important boosting of the neuronal output. We also show that, in chronically implanted animals, the generation of FPPs is actively modulated across natural states of vigilance.
Materials and Methods
Experiments were performed in adult cats under anesthesia as well as in chronically implanted, naturally waking and sleeping adult cats. The experimental protocols were approved by the Committee for Animal Care and Protection of Laval University (permission no. 2002–007) and also conformed to the policy of the American Physiological Society. Every effort was made to minimize animal suffering, and only the minimum number of animals necessary to obtain reliable data was utilized.
For acute experiments, 34 adult cats (2.5–4 kg) were anesthetized with pentobarbital (Somnotol, 35 mg/kg, i.p.) (n = 29) or ketamine–xylazine (10–15 mg/kg and 2–3 mg/kg i.m., respectively) (n = 5). The animals were paralyzed with gallamine triethiodide after the electroencephalogram (EEG) showed typical signs of deep general anesthesia and supplementary doses of anesthetics were administered at the slightest changes toward activated EEG patterns. The cats were ventilated artificially with the control of end-tidal CO2 at 3.5–3.7%. The body temperature was maintained at 37–38°C and the heart rate was ∼90–100 beats/min. Stability of intracellular recordings was ensured by hip suspension, drainage of cisterna magna, bilateral pneumothorax and filling the hole made in the scull with a solution of 4% agar.
Experiments on non-anesthetized animals were conducted on eight adult cats, chronically implanted as previously described (Steriade et al., 2001; Timofeev et al., 2001). Briefly, surgical procedures for chronic implantation of recording and stimulating electrodes were carried out under deep barbiturate anesthesia (Somnotol, 35 mg/kg, i.p.), followed by two to three administrations, every 12 h, of buprenorphine (0.03 mg/kg, i.m.) to prevent pain. Penicillin (500 000 units i.m.) was injected during three consecutive days. During surgery, the cats were implanted with electrodes for electro-oculogram (EOG), electromyogram (EMG) from neck muscles, and intracortical EEG recordings. In addition, one to three chambers allowing the intracellular penetrations of micropipettes were placed over various neocortical areas. Acrylic dental cement was used to fix on the skull the electrodes and recording chambers, and a previously described system (Steriade and Glenn, 1982) allowed head fixation without pain or pressure during recording sessions.
At the end of experiments, the cats were given a lethal dose of pentobarbital.
Intracellular Recordings and Stimulation
Intracellular recordings were performed using glass micropipettes filled with a solution of 3 M potassium acetate (KAc) and direct current (DC) resistances between 25 and 75 MΩ. A high-impedance amplifier with active bridge circuitry was used to record the Vm and inject current into the neurons.
In acute experiments field potentials were recorded at 3–6 mm from the impaled neurons, using bipolar coaxial electrodes, with the ring at the pial surface and the tip at the cortical depth, separated by 0.8–1 mm. Stimulating electrodes (one or two, similar to those used for field potential recordings) were inserted in the vicinity of micropipettes and into related thalamic nuclei. Cortical neurons were recorded from areas 5, 7 or 21 of intact cortex and in small isolated cortical slab (7 × 12 mm) from suprasylvian association areas 5 and 7. The preparation of cortical slabs is described elsewhere (Timofeev et al., 2000).
Chronically implanted cats were trained to sleep in the stereotaxic apparatus for 4–5 days after surgery. Intracellular recordings began when the cat displayed normal sleep–waking cycles. After small perforation of the dura, a glass micropipette was inserted in the cortex and the recording chamber was filled with warm sterile solution of 4% agar. Two to three recording sessions, lasting for 1–3 h, were performed daily; 7–10 days of recordings could be made in each chamber.
Local change in extracellular Ca2+ concentration ([Ca2+]o) in the cortex was achieved using the reverse microdialysis method. The membrane of the microdialysis probe (2 mm length, 0.22 mm diameter, from EICOM, Kyoto, Japan) was inserted in the cortex and the recording micropipettes were placed at 0.2–0.3 mm from the membrane. The microdialysis probe was perfused with the following artificial cerebrospinal fluids (ACSF) (concentration in mM): control (NaCl 124, KCl, 2.5, NaHCO3 26, NaH2PO4 1.25, MgSO4 2, MgCl2 1, CaCl2 1); high Ca2+ (NaCl 124, KCl, 2.5, NaHCO3 26, NaH2PO4 1.25, MgSO4 2, MgCl2 0, CaCl2 5); and Ca2+ free (NaCl 125, KCl, 2.5, NaHCO3 26, NaH2PO4 0, MgSO4 2, MgCl2 1, CaCl2 0, MnCl2 1). The actual [Ca2+]o at the recording site was measured with a Ca2+-sensitive microelectrodes (Diamond General, Ann Arbor, MI).
Signals were recorded with a 16-channel Vision data acquisition system (Nicolet, Madison, WI) at a sampling rate of 20kHz. Analyses were performed off-line on a personal computer using IgorPro software (WaveMetrics, Lake Oswego, OR). For the analysis of spontaneous synaptic activity, intracellular recordings were filtered off-line between 0 and 1000 Hz to eliminate high-frequency electronic noise. The peak-to-peak amplitude of residual noise in the intracellular recordings was measured during periods of absence of spontaneous synaptic activity (i.e. during inter-spindle lulls under barbiturate anesthesia). The maximal noise amplitude ranged between 0.1and 0.2mV. All spontaneous depolarizing events (presumably EPSPs) with amplitude >0.2 mV were then extracted automatically using a custom-written routine in IgorPro. The events were detected as positive peaks in the first derivative of the intracellular signal. The detection of non-monotony in the rising phase of each event allowed the discrimination of compound events that appeared as single peaks in the intracellular signal. The depolarizing slope of each event was fitted with a sigmoid function. The amplitude and maximum slope were determined between the beginning and the end of the rising phase of detected events. The rise-time was measured between 10 and 90 % of peak amplitude. For determination of FPPs and action potentials (APs) threshold, we measured the membrane potential (Vm) for each individual FPP and AP at theinitiation point, determined in the averaged traces. For APs, the initiation point was determined in averaged APs that were not initiated by FPPs. The distinction between the initial FPP and the following AP was very clear in the first derivative of the signal as shown in Fig. 7A. The initiation Vm was then measured for all APs at the same time before the peak of the AP. Pooled data are expressed as mean ± SEM. When relevant, statistical analysis was carried out using two-tailed t-test.
Characterization of FPPs in Neocortical Neurons
We analyzed the features of spontaneous postsynaptic events in electrophysiologically identified RS and IB neocortical neurons (McCormick et al., 1985; Connors and Gutnick, 1990; Nuñez et al., 1993) recorded under barbiturate anesthesia. Depolarizing events of amplitude >0.2 mV were extracted automatically from intracellular recordings (see Materials and Methods). The number of events extracted varied from cell to cell between 7000 and 21 000 (mean 10654 ± 411) for 1 min of recording.
(i) Spontaneous EPSPs were characterized by low-amplitude and slow-rising phase. Overall, 78.7 ± 1.4% of events had amplitude <0.5 mV and 87.6 ± 2.0% had maximal slope <1 V/s. In 78% of cells (65/83) no event with amplitude >5 mV or maximum rising slope >6 V/s was detected.
(ii) However, in 22% of the cells (18/83), some high-amplitude and fast-rising events were extracted. These events appeared as a population distinct from spontaneous EPSPs, as revealed by plotting the amplitude versus the maximal slope for all events (Fig. 1, middle panel right). Their mean amplitude ranged from cell to cell between 3.6 and 10.2 mV, with an average value of 6.16 ± 0.14 mV, and their mean maximal slope ranged from 8 to 20 V/s, with an average value of 12.97 ± 0.30 V/s. The rise-time of these events (0.49 ± 0.009 ms) was significantly shorter than the rise time of the faster EPSPs (1.09 ± 0.041 ms, P < 0.001) and significantly longer than the rise time of full-blown APs (0.28 ± 0.008 m, P < 0.001). Their falling phase was characterized by a biphasic decay best fitted by a double exponential: the τ of the initial fast decay was 0.79 ± 0.03 ms, while the τ of slower decay was 6.93 ± 0.29 ms. Because of these features, fast-rising events were very similar to the FPPs initially described in hippocampal neurons by Spencer and Kandel (1961), and hereafter we will call them FPPs.
All neurons in which FPPs were detected had APs >0.6 ms at half amplitude (mean = 0.81 ± 0.01 ms). The cortical depth distribution of neurons was calculated from the depth of the recording indicated by the micro-driver and corrected according to the angle of the pipette within the cortex (Fig. 2). The neurons displaying FPPs were distributed mainly between 200 and 800 µm, and between 1200 and 1600 µm, indicating that most of them were recorded in layers II/III and V/VI. The input resistance of neurons displaying FPPs (17.43 ± 0.95 MΩ) was not significantly different from that of neurons without FPPs (18.44 ± 1.00 MΩ, P > 0.8).
FPPs appeared as all-or-none events that crowned spontaneously occurring EPSPs. They appeared either in isolation or led to full-blown APs (see overlay in Fig. 1). The essential condition for FPP generation was a relatively depolarized Vm and the presence of synaptic activity. Actually, hyperpolarizing the cell by intracellular injection of current suppressed FPPs (Fig. 3). However, depolarizing current pulses could not elicit FPPs during periods without spontaneous synaptic activity (i.e. interspindle lull in barbiturate anesthesia) but, instead, induced trains of APs. In Fig. 3, the cell was held at two Vm levels by injection of hyperpolarizing current pulses. The histogram of Vm (gray) shows a bimodal distribution resulting from current pulses, with FPPs initiated only at Vms more depolarized than –70 mV. In this neuron, the mean threshold for the generation of FPPs was –66 mV. In the 18 neurons with FPPs, we compared the Vm at the initiation of FPPs and the Vm at initiation of APs. The threshold for FPPs’ generation was different from cell to cell and varied from –72 to –62 mV (mean 67.2 ± 0.2 mV), but was always more hyperpolarized than the threshold for axonal APs initiation (55.9 ± 0.1 mV, P < 0.001) (see below). In all 18 neurons, holding the cell below the threshold for FPPs’ generation completely suppressed their appearance.
In order to determine the dependence of FPPs’ generation on synaptic transmission, we used the reverse microdialysis technique to change the extracellular concentration of calcium ([Ca2+]o) in the vicinity of the recorded neurons. The actual [Ca2+]o at the recording site was measured with a Ca2+-sensitive electrode. The baseline control [Ca2+]o was 1.2 mM, as previously described in vivo (Massimini and Amzica, 2001); 10–15min after perfusion of the microdialysis probe with high Ca2+ ACSF, the [Ca2+]o reached a stable level of 3 mM; after 20min of perfusion with the Ca2+-free ACSF, the [Ca2+]o stabilized at 0.6 mM. Two RS and one IB neurons displaying FPPs could be recorded under different conditions of [Ca2+]o. An example is depicted in Figure 4. Increasing [Ca2+]o markedly enhanced the background of synaptic activity and the incidence of FPPs, from 0.68 ± 0.04 Hz to 1.18 ± 0.07 Hz (P < 0.05). In low Ca2+ condition, the spontaneous synaptic activity was strongly reduced and FPPs were almost completely suppressed (0.03 ± 0.00, P < 0.01). The same effect was observed in the three neurons. The change in FPPs incidence was not related to change in neuronal firing, since the mean firing rate of neurons (calculated for 10 neurons recorded in three conditions) was slightly, but not significantly, decreased in high Ca2+ condition (0.75 ± 0.23, P > 0.3) and increased in low Ca2+ condition (1.41 ± 0.33, P > 0.6), as compared to control condition (1.08 ± 0.15). No significant change in Vm was observed between the three conditions (Fig. 4). The slight change in firing rate was due the change in the firing threshold for APs, as shown in Figure 4. Thus FPPs’ generation appeared to be strongly dependant on synaptic transmission.
We examined the possibility of generating FPPs within local cortical circuits and analyzed the spontaneous activity of RS or IB neurons recorded in isolated cortical slabs from the suprasylvian association cortex. FPPs were found in 15% (3/20) of such neurons, thus indicating that FPPs could be generated in response to activation of inputs arising from local cortical networks. The FPPs recorded in isolated cortical slabs shared common features with those recorded in intact cortex, i.e. they were all-or-none events, associated with intense synaptic activity during spontaneous or elicited active periods (Fig. 5A), and their generation was voltage dependent (Fig. 5B).
Selective Amplification of Cortical Inputs
In 13 out of 18 neurons, we tested the efficacy of cortical and/or thalamic inputs to evoked FPPs. FPPs could be elicited by cortical simulation in 8 of the 13 neurons. In general, responses to cortical stimulation consisted in a depolarization-hyperpolarization sequence, the first depolarization being composed of a summation of EPSPs. The early depolarization was crowned by an FPP when the response reached the threshold for their generation (Fig. 6). Interestingly, the generation of FPPs selectively depended on the stimulated pathway. At variance with an earlier study that pointed to thalamic inputs generating FPPs’ in cortical neurons (Deschênes, 1981), cortical stimulation evoked FPPs while thalamic stimuli, applied to intralaminar centrolateral or lateral posterior nuclei, did not elicit FPPs in the same neuron, even when the response reached the threshold for FPP generation (Fig. 6A). This was observed in the 4 neurons in which we tested both thalamic and cortical stimulation.
In four other neurons, we compared the potency of cortical stimuli applied at different sites to evoke FPPs. Stimuli were applied alternatively through two electrodes (see Materials and Methods) inserted in the ipsilateral cortex, at 1–2 mm from the recorded neuron and separated by 2–3 mm. The intensity of stimulation was adjusted for each electrode to induce EPSPs of similar amplitude, when subthreshold for FPP. Figure 6B shows the responses of another cortical neuron to stimuli applied through two different cortical electrodes. Stimuli delivered to the first electrode evoked an EPSP that was crowned by an FPP when it reached the threshold (left), whereas stimuli applied to the second electrode evoked a similar responses but no FPP (right). To be sure that the second stimulation did not fail to evoke FPPs because of more depolarized threshold for FPP generation, we held the cell at a more depolarized level of Vm while stimulating through the second electrode. At more depolarized Vm, the EPSP eventually evoked an AP (not shown) but no FPP was observed. The consequence of FPPs’ generation in response to cortical stimulation was a non-linearity in the voltage–response relationship, as revealed by plotting the maximum Vm reached by the evoked response versus the Vm before stimulation. As shown in Figure 6B (bottom), the progression of the Vm reached by the response increased linearly with the initial Vm (blue open triangles) until it reached the threshold for FPP generation. Then, the FPPs induced a jump in the Vm reached by the response (blue filled triangles). On the other hand, the Vm reached by the response to stimuli applied through the second electrode increased linearly even above the threshold for FPP generation (green open circles).
Besides its dependence upon synaptic inputs, the generation of FPPs was also dependent on the temporal sequence of the inputs. In 2 out of 5 neurons, a single cortical stimulus was ineffective in initiating an FPP but, when paired at short time interval (∼10 ms), an FPP crowned the second EPSP (Fig. 6C). This effect was not only due to a more depolarized Vm since a single stimulus could not evoke an FPP even when the response passed over the threshold for FPP generation.
An important function of FPPs appeared to be their contribution to the output of cortical neurons, i.e. the generation of full-blown APs. As shown in Figure 7A, APs could be initiated in the same neuron following the summation of several EPSPs, whereas only one FPP was necessary to reach the firing threshold from the same Vm level. However, as can be see in Figure 7 (expanded traces in A and averaged traces in B), the firing threshold for APs was the same whether the AP was initiating from summated EPSPs or from one FPP. Indeed, FPPs are events of high amplitude generated at Vms 5–10 mV more hyperpolarized than the firing threshold for full-blown APs (see histograms in Fig. 7B). Consequently, FPPs facilitate neuronal firing at more hyperpolarized levels. This is of special importance for the integrative properties of cortical neurons.
Since FPPs are generated in response to the activation of specific inputs (see again Fig. 6A,B), they can boost selectively the output of the neuron in response to those inputs. In the neuron depicted in Figure 8, cortical stimulation evoked a sequence of depolarizing–hyperpolarizing potentials and, when the neuron was more depolarized, early EPSPs were crowned by FPP. At slightly more depolarized potential, the FPP led to full-blown AP and, when the cell was more depolarized, the AP rode directly on the EPSPs. We calculated the probability for the evoked response to elicit an AP (output) as a function of the initial Vm when FPPs were generated. We then extrapolated the output–Vm function for the same response when no FPP was generated. In both cases, the output–Vm relation was best fitted by a sigmoid function. As seen in the bottom graph of Figure 8, the generation of FPP shifted the output–Vm curve to more hyperpolarized levels. Thus, FPPs allowed the neuron to fire in response to the same synaptic input at Vms more hyperpolarized by ∼5 mV.
Modulation of FPP Generation across Behavioral States
Finally, we investigated the possibility that the incidence of FPPs is modulated by natural states of vigilance and analyzed intracellular recordings of RS and IB neurons in awake and sleeping cats. FPPs were found in 10 out of 56 neurons (18%) in different states of vigilance. FPPs were present in all behavioral states but their frequency was different and depended on the Vm variations (Fig. 9). During slow-wave sleep (SWS), the mean Vm was more hyperpolarized and the incidence of FPPs was the lowest. During waking and paradoxical sleep (PS; or REM sleep), the Vm was equally more depolarized and the frequency of FPPs was higher in both these states than in SWS, reaching the highest values in PS. We measured the incidence of FPPs in eight neurons, during different states of vigilance. The incidence of FPPs was the lowest during SWS (4.4 ± 0.95Hz) and was significantly higher during waking (15.02 ± 1.18 Hz, P < 0.01) and PS (28.97 ± 4.40 Hz, P < 0.01). Despite a similar Vm level, the incidence of FPPs during PS was significantly higher than during waking (P < 0.05), suggesting that neuromodulators implicated in the regulation of behavioral states may modulate the generation of FPPs.
In the present paper we report the presence of fast and high-amplitude events, FPPs, recorded from the soma of ∼20 % of RS and IB neocortical neurons. The FPPs were all-or-none events, suppressed by hyperpolarization, and always associated with synaptic background activity. They were evoked by activation of specific local cortical inputs and exerted a powerful and selective boosting of cortical neuronal output. Finally, we found that the incidence of FPPs was modulated by natural states of vigilance.
Nature and Origin of FPPs
FPPs were first described in the hippocampus (Spencer and Kandel, 1961) and later also found in neocortical (Deschênes, 1981), thalamocortical (Steriade et al., 1991; Timofeev and Steriade, 1997) and other central neurons. They were interpreted to either represent attenuated dendritic spikes (Turner et al., 1993; Ariav et al., 2003) or reflect APs in electrotonically coupled neurons (Llinás et al., 1974; MacVicar and Dudek, 1981; Valiante et al., 1995; Gibson et al., 1999; Hughes et al., 2002; Landisman et al., 2002). Although in vitro studies have revealed abundant electrical coupling via gap-junctions between both pyramidal neurons and interneurons in immature neocortex (Peinado et al., 1993; Bittman et al., 2002), electrophysiological evidence for electrical coupling in adult animal has been only found between local interneurons but not between pyramidal neurons (Connors et al., 1983; Galarreta and Hestrin, 1999; Gibson et al., 1999). In the present study, FPPs were recorded in RS and IB neurons with relatively wide APs (mean duration at half amplitude: 0.82 ± 0.01 ms), indicating that these neurons were most likely not interneurons. In a few experiments, we stained the recorded neurons with intracellular injection of Neurobiotine. One neuron that displayed FPPs was stained and was identified as a pyramidal-shaped layer III neuron (data not shown). Recent experiments in the hippocampus demonstrated the generation of spikelets of 1–10 mV in amplitude mediated by axo-axonal gap junctions between pyramidal cells (Schmitz et al., 2001). However, to the best of our knowledge, there is no evidence for electrical coupling between axons of pyramidal neurons in the neocortex. On the other hand, the possibility that electrical coupling between neocortical pyramidal neurons may persist in adult animals has been raised by molecular studies. Thus, connexin36 might be expressed by a subset of pyramidal neurons (Deans et al., 2001) and the expression of mRNA for another type of connexin (43) has been found in pyramidal cells of mature neocortex (Simburger et al., 1997).
In the present study, the generation of FPPs was voltage-dependent and always associated with synaptic activity. Similarly, the generation of dendritic spikes recorded from pyramidal neocortical or hippocampal neurons in vitro has been shown to be voltage-dependent (Kim and Connors, 1993; Schwindt and Crill, 1997; Schiller et al., 2000; Ariav et al., 2003). Finally, we found that FPPs were suppressed when synaptic transmission is inhibited by perfusion of Ca2+-free ACSF and their incidence increased with perfusion of high-Ca2+ ACSF. These observations do not support the idea that FPPs represent APs in electronically coupled neurons, since electrical coupling is not affected or even enhanced in low-Ca2+ conditions (Perez-Velazquez et al., 1994; Valiante et al., 1995; Draguhn et al., 1998; Mann-Metzer and Yarom, 1999). Thus, although we cannot completely discard the possibility that FPPs represent APs in electrotonically coupled neurons, our results strongly suggest that they are more likely dendritic regenerative potentials. It is also unlikely that the neurons that displayed FPPs in our study were recorded at dendritic levels: (i) none of them showed characteristic features of established dendritic recordings in vitro or in vivo, such as plateau potentials or complex spikes (Kim and Connors, 1993; Steriade et al., 1996; Helmchen et al., 1999; Zhu and Connors, 1999; Larkum and Zhu, 2002); (ii) all cells generated trains of full-blown, overshooting APs of constant amplitudes in response to depolarizing current pulses, whereas backpropagating axonal APs recorded in the dendrites of cortical neurons in vivo show strong attenuation and broadening with distance from the soma (Svoboda et al., 1999; Waters et al., 2003). We therefore think that our recordings were performed at somatic levels of cortical neurons and thus suggest that FPPs represent attenuated dendritic spikes.
As to the mechanisms underlying the generation of FPPs, it is well established that the dendrites of cortical pyramidal cells can sustain active regenerative potentials under certain conditions. Na+, Ca2+ and NMDA-generated spikes have been demonstrated in the apical and basal dendrites of cortical pyramidal neurons (Magee et al., 1998; Hausser et al., 2000; Schiller and Schiller, 2001; Hausser and Mel, 2003; Williams and Stuart, 2003). Because of the slow time course of Ca2+ and NMDA spikes (Kim and Connors, 1993; Schiller et al., 1997; Schwindt and Crill, 1997; Larkum et al., 1999a; Schiller et al., 2000; Larkum et al., 2001; Larkum and Zhu, 2002), we assume that at least the earliest sharp component of the FPPs represent a Na+ spike whose generation has been shown in the apical dendrites of neocortical and hippocampal pyramidal cells (Kim and Connors, 1993; Stuart et al., 1997; Golding and Spruston, 1998; Schwindt and Crill, 1998) and in the basal dendrites of hippocampal pyramidal neurons as well (Ariav et al., 2003). However, Na+ spikes generated in apical dendrites are most of the time associated with Ca2+ spikes and strongly attenuate while propagating to the soma (Schiller et al., 1997; Stuart et al., 1997; Larkum et al., 2001). As a result, when simultaneously recorded at the dendrite and soma in vitro, dendritic Na+ spikes do not appear as fast potentials at the soma. Moreover, we were able to block the generation of FPPs by injection of hyperpolarizing current at the soma. In vitro data demonstrate that alteration of somatic membrane potential by direct current injection can affect the membrane potential of the dendrites at least out to 300–400 µm from the soma (Schwindt and Crill, 1995). The high conductance state of cortical neurons in vivo probably further limits the dendritic impact of somatic current injection, suggesting that FPPs are not generated far from the soma, i.e. in proximal apical or basal dendrites of pyramidal cells. Another possibility is that FPPs were generated in distal apical dendrites and then propagated actively to the soma. This possibility was suggested in modeling studies showing that under in vivo-like conditions, a depolarized Vm increases the probability of generation and propagation of dendritic spikes (Rhodes and Llinás, 2001; Rudolph and Destexhe, 2003). In this condition, hyperpolarizing the soma could block the propagation FPPs (Larkum et al., 2001). However, their relatively small amplitude at the soma, compared to the amplitude of regenerative potentials generated in the dendrites, does not support the idea of an active propagation in this case.
In hippocampal pyramidal cells in vitro, coactivation of clustered neighboring inputs on basal dendrites can elicit a local spike that consists in fast Na+ and slow NMDA components (Ariav et al., 2003). These dendritic spikes, recorded at the soma, have a shape very similar to the FPPs described here and are similarly suppressed by injection of hyperpolarizing current at the soma. But in layer V neocortical pyramidal cells in vitro, only the slow NMDA spike was evoked by the activation of basal dendrites (Schiller et al., 2000). However, in our experiments, many neurons with FPPs were localized in layers II/III or VI. It is thus possible that Na+ spikes, similar to those generated in the basal dendrites of hippocampal neurons, can be generated in the basal dendrites of some subpopulation of neocortical pyramidal cells in other layers. Therefore, we suggest that FPPs represent dendritic Na+ spikes generated in the basal dendrites of neocortical pyramidal cells. The biphasic decay of the FPPs may reflect an underlying slow component due either to Ca2+ or NMDA conductances.
The generation of dendritic spikes is supposed to occur when synaptic inputs are activated simultaneously and/or converge at the same dendritic region. In neocortical pyramidal neurons, supralinear summation of EPSPs due to active dendritic processes has been found to occur in a time window <30 ms (Nettleton and Spain, 2000). Local spikes in the basal dendrites of both neocortical or hippocampal pyramidal cells can be initiated by co-activation of clustered neighboring basal inputs (Schiller et al., 2000; Ariav et al., 2003). Based on the local amplitude of EPSPs recorded in vitro, Williams and Stuart (2002) have found that the synchronized activation of 4–30 presynaptic neurons is required for the initiation of dendritic spikes. Thus, initiation of local spikes is expected to occur either if inputs carrying related information selectively innervate the same dendritic segments (Archie and Mel, 2000; Poirazi and Mel, 2001) or if the network activity is synchronized (Kamondi et al., 1998; Helmchen et al., 1999). In keeping with this idea, we found that FPPs could be evoked by gross cortical stimulations, which activate simultaneously a pool of neurons and fibers or by paired stimuli at high frequency. Thus, only converging coincident inputs or closely time-spaced inputs could generate FPPs, suggesting that FPPs are likely generated in response to an important depolarization of a particular dendritic segment. Interestingly, the activation of different cortical inputs had various effects on the generation of FPPs, even when the subthreshold evoked responses were very similar (see Fig. 6). This observation point out to the input-specific generation of FPPs that might be due either to the location of these inputs (apical or basal dendrites) or to the convergence of the simultaneously activated inputs (clustered or spread on the dendritic arbors). The input-dependent generation of FPPs may also be due to non-uniformly distributed glutamate receptors that form hot spots of glutamate responsivity on dendritic arbors. Synaptic inputs impinging upon these regions would have a higher probability to elicit dendritic spikes (Frick et al., 2001).
Functional Implication of FPPs for the Integrative Properties of Cortical Neurons
Active regenerative dendritic potentials have mainly been studied in the apical dendrites of neocortical and hippocampal pyramidal cells. Physiological and modeling studies have shown that forward-propagating dendritic spikes can boost the influence of synapses in distal dendrites on their way to the soma, thereby circumventing the attenuation produced by the passive cable properties of dendrites (Cauller and Connors, 1994; Schwindt and Crill, 1998; Larkum et al., 2001; Williams and Stuart, 2002; Rudolph and Destexhe, 2003). It is assumed that dendritic spikes strongly attenuate as they propagate forward and appear at the soma as slow events undistinguishable from other postsynaptic potentials (Schiller et al., 1997; Stuart et al., 1997; Golding and Spruston, 1998; Larkum et al., 2001). However, in the present study we found that FPPs were significantly higher in amplitude and had faster rising phase than other spontaneous depolarizing events recorded at the soma. FPPs induced a non-linearity in the integrative properties of cortical neurons (see Fig. 6B), thus providing additional computational properties to the neurons by markedly enhancing the somatic impact of cortical inputs. We further demonstrate that the boosting of EPSPs by FPPs results in a functional hierarchy of the inputs by a selective enhancement of the output of cortical neurons for given inputs. Thus, FPPs may reinforce the functional links between specific elements of the cortical network. This hypothesis is of interest since we demonstrated, using dual intracellular recordings of synaptically connected neurons in vivo, that the somatic impact of individual cortical inputs during active states of the network is very weak due to a high conductance state and high failure rate (Timofeev and Crochet, 2002). FPPs might also enhance neuronal output in response to simultaneous arriving inputs. Indeed, the generation of dendritic spikes might require coincident activation of different inputs and may thus play an important role for coincidence detection of cortical inputs.
As demonstrated recently in hippocampal neurons, Na+ dendritic spikes could serve to improve the precision and the stability of the timing of axonal APs, mainly by reducing the temporal jitter of evoked APs (Ariav et al., 2003). This was apparently not the case for FPPs in the present study because (i) the generation of FPPs themselves showed relatively high temporal variability (not shown); and (ii) once an FPP is generated, the timing of the AP was also variable (see Figs 1 and 5A). Physiological studies have also suggested that forward-propagating Ca2+ spikes could be involved in the generation of burst firing in cortical neurons (Larkum et al., 1999a,b; Schwindt and Crill, 1999; Williams and Stuart, 1999). In contrast to Ca2+ spikes initiated in apical dendrites, we found that FPPs did not evoke a burst of axonal APs, but rather single AP. This is expected from the fast decay of the FPPs, which are more likely generated by a Na+ conductance and which contrast with the broad Ca2+ spikes evoked in apical dendrites.
Dendritic excitability is regulated by neuromodulators such as serotonin, acetylcholine or noradrenaline (Hoffman and Johnston, 1999; Carr et al., 2002). These neuromodulators are released by ascending activating systems that originate in the brainstem and basal forebrain, and play a major role in the control of behavioral states (Steriade and McCarley, 1990). In line with these observations, we found evidence for a modulation of FPPs generation across behavioral states of vigilance. In all recorded neurons, we found changes in the incidence of spontaneous FPPs associated with shifts from a state of vigilance to another. The incidence of FPPs appeared to be the lowest during SWS and the highest during PS. The decrease in FPPs incidence during SWS was expected because of Vm hyperpolarization during this state (Steriade et al., 2001) due to decreased firing rates in thalamocortical (Glenn and Steriade, 1982) and other corticipetal systems during SWS, compared to both waking and PS. The increased incidence of FPPs in PS compared to waking (Fig. 9) fits in well with the higher firing rates, during PS, of mesopontine cholinergic neurons with identified projections to intralaminar and lateroposterior thalamic nuclei (Steriade et al., 1990), which are the main afferents of cortical association areas at which level the present recordings were made. Thus, neuromodulators are able to control the integrative properties of cortical neurons across behavioral states.
This study revealed that ∼20% of RS and IB neocortical neurons are able to generate FPPs. The electrophysiological features of these events strongly suggest that they represent forward propagated Na+ dendritic spikes attenuated at the soma. We propose that FPPs represent dendritic spikes generated by coincident inputs converging on the same dendritic segment. These dendritic spikes strongly affect the integrative properties of cortical neurons and may serve to reinforce the functional links between specific elements of the cortical network or for coincidence detection as well. Finally, the generation FPPs appeared to be regulated across behavioral states, indicating important changes in the integrative properties of cortical neurons during states of vigilance.
We thank P. Giguère and D. Drolet for technical assistance. This work was supported by grants from the Canadian Institutes for Health Research (MT-3689, MOP-36545, MOP-37862) and US National Institute of Health (RO1, NS-40522). S. Crochet is supported by the Pickwick Postdoctoral Fellowship program from the National Sleep Foundation. I. Timofeev is a Scholar of the Canadian Institutes of Health Research.