Action potentials are the end product of synaptic integration, a process influenced by resting and active neuronal membrane properties. Diversity in these properties contributes to specialized mechanisms of synaptic integration and action potential firing, which are likely to be of functional significance within neural circuits. In the hippocampus, the majority of subicular pyramidal neurons fire high-frequency bursts of action potentials, whereas CA1 pyramidal neurons exhibit regular spiking behavior when subjected to direct somatic current injection. Using patch-clamp recordings from morphologically identified neurons in hippocampal slices, we analyzed and compared the resting and active membrane properties of pyramidal neurons in the subiculum and CA1 regions of the hippocampus. In response to direct somatic current injection, three subicular firing types were identified (regular spiking, weak bursting, and strong bursting), while all CA1 neurons were regular spiking. Within subiculum strong bursting neurons were found preferentially further away from the CA1 subregion. Input resistance (R N), membrane time constant (τm), and depolarizing “sag” in response to hyperpolarizing current pulses were similar in all subicular neurons, while R N and τm were significantly larger in CA1 neurons. The first spike of all subicular neurons exhibited similar action potential properties; CA1 action potentials exhibited faster rising rates, greater amplitudes, and wider half-widths than subicular action potentials. Therefore both the resting and active properties of CA1 pyramidal neurons are distinct from those of subicular neurons, which form a related class of neurons, differing in their propensity to burst. We also found that both regular spiking subicular and CA1 neurons could be transformed into a burst firing mode by application of a low concentration of 4-aminopyridine, suggesting that in both hippocampal subfields, firing properties are regulated by a slowly inactivating, D-type potassium current. The ability of all subicular pyramidal neurons to burst strengthens the notion that they form a single neuronal class, sharing a burst generating mechanism that is stronger in some cells than others.
Subiculum, along with presubiculum and parasubiculum, forms an anatomical transition area between the Ammon's Horn region and entorhinal cortex (EC). Subiculum primarily receives axonal input from CA1 and EC layer II/III neurons, and its activity has been shown to be correlated with theta and gamma oscillations as well as with sharp waves of the CA1 region (Chrobak and Buzsaki 1996; Colling et al. 1998). Subiculum has been shown to be the major output center of the hippocampus, to which it belongs, sending parallel projections from various subicular regions to different cortical and subcortical areas (Naber and Witter 1998). Functionally, the principal neurons of the subiculum appear to be “place cells,” similar to those of CA1 (Sharp and Green 1994); and in a human magnetic resonance imaging (MRI) study, subiculum was shown to be an area involved in memory retrieval (Gabrieli et al. 1997). Therefore both CA1 and subiculum have been implicated in memory processes and spatial coding, and both are major gateways from the hippocampus to the rest of the brain; however, subiculum has received far less experimental attention.
A major difference between subicular and CA1 pyramidal neurons resides in their firing properties. In response to direct somatic current injection, CA1 pyramidal neurons generate a regular train of action potentials. In the adjacent subiculum, however, most pyramidal neurons have been shown to burst in response to current injection (Behr et al. 1996; Greene and Totterdell 1997;Mason 1993; Mattia et al. 1993;Stewart and Wong 1993; Taube 1993), a phenomenon that has also observed in vivo (Sharp and Green 1994). These different firing properties can have profound implications for the efficacy of information transfer between neurons as well as signaling within a given neuron. For example, given the low probability of synaptic neurotransmitter release, a high-frequency burst entering a presynaptic terminal increases the likelihood of synaptic release per burst to near one (reviewed in Lisman 1997). Similarly, a high-frequency burst backpropagating into the dendrites likely influences synaptic integration and pre-/postsynaptic coincidence detection in ways distinct from a single action potential.
Understanding all the mechanisms by which neurons generate their precise firing properties is a difficult task. The bewildering array of voltage-gated ion channels in the CNS that are activated during action potential generation certainly plays the predominant role in determining a neuron's firing properties. The context in which these ion channels interact, however, is also crucial. Modeling studies have shown that the morphology of a neuron can drastically influence a cell's firing properties. By varying the dendritic arborization while maintaining a steady channel density, a neuron may exhibit either regular spiking or burst firing (Mainen and Sejnowski 1996). Additionally, subthreshold conductances can contribute significantly to understanding firing properties. For example, in thalamus, the timing of bursts as well as the depolarization leading to the calcium-dependent burst is determined byI h, a current that becomes inactive during action potential generation (reviewed in Steriade et al. 1993).
In addition to their effects on firing properties, dendritic morphology and subthreshold conductances play a key role in the process of synaptic integration. A single pyramidal neuron may receive thousands of excitatory and inhibitory inputs, whose synaptic potentials are sculpted by the neuron's intrinsic morphology and ionic conductances. This transmuted signal is then deciphered near the axon hillock where it becomes the output of the neuron in the form of the action potential. Action potentials backpropagating into the somatodendritic region serve to inform the rest of the neuron about its action potential output, and these are also influenced by cell morphology and ionic conductances. Therefore studying the context in which action potentials occur also provides valuable information about how a neuron integrates its inputs.
To gain an understanding of the intrinsic firing and integrative properties of subicular neurons, we employed patch-clamp techniques and measured resting membrane and firing properties of both subicular and CA1 pyramidal neurons. Although CA1 pyramidal neurons do not burst in response to somatic current injection in vitro, their proximity to subiculum and the abundance of data gathered on these neurons help provide a framework within which burst firing may arise in subiculum.
Patch-clamp recording and analysis
Transverse hippocampal slices (300 μm) were made from 2- to 9-wk-old Wistar rats. The animals were anesthetized with halothane, and perfused intracardially with ice-cold, oxygenated (bubbled with 95% O2-5% CO2) artificial cerebrospinal fluid (ACSF), containing (in mM) 125 NaCl, 2.5 KCl, 2 CaCl2, 1 MgCl2, 25 NaHCO3, 1.25 NaH2PO4, and 25 glucose. Throughout the dissection the brains were maintained in ice-cold, oxygenated ACSF. After decapitation, the brains were rapidly removed, and a cut was made on the dorsal surface at a 60° angle to the horizontal plane. The brains were then mounted ventral-side up, and 300-μm slices (including Ammon's Horn, subiculum and cortex) were prepared using a vibratome (Leica VT 1000S). Slices were incubated for 30 min to 1 h in a holding chamber at 35°C and then stored at room temperature. Individual slices were then transferred to a submerged recording chamber (35 ± 2°C) and visualized with differential interference contrast optics and a Newvicon camera system (Stuart et al. 1993). Somata were selected for recording based on their pyramidal shape and smooth, low-contrast appearance.
Whole cell current-clamp recordings were made from cells using an electrode connected to a bridge amplifier (Dagan BVC-700). Patch-clamp electrodes were pulled from glass capillary tubes, fire polished, and filled with intracellular solution containing (in mM) 115 potassium gluconate, 20 KCl, 10 sodium phosphocreatine, 10 HEPES, 2 EGTA, 2 MgATP, 0.3 NaGTP, and 0.1% biocytin for subsequent determination of morphology. Electrode resistance in the bath ranged from 2 to 7 MΩ, and series resistance ranged from 9 to 35 MΩ. No correction was made for the theoretical junction potential between internal and external solutions. Capacitance compensation and bridge balance were performed. In some experiments, K-MeSO4 was substituted for K-gluconate, and no qualitative difference in burst firing was noted (n = 3). Thirty-nine of the 171 subicular neurons were recorded with 10 mM EGTA in the patch solution. The percentage of each of the firing subtypes was equivalent in this group, so all cells were grouped together for use exclusively in the firing property description as well as the anatomical mapping experiments.
In some experiments, perforated-patch recordings were made, for which the pipette solution contained 150 mM KCl, 20 mM tetraethylammonium chloride (TEA), and gramicidin (9 μg/ml). Perforated-patch recordings were characterized by a gradual reduction in the measured resistance from >1 GΩ (initial seal resistance) to <150 MΩ (input resistance). A precipitous drop in resistance was considered to be indicative of breaking into whole-cell configuration, in which case the recording was discarded. Monitoring spike half-width with inclusion of TEA in the pipette provided an additional means of assessing subtle perforated patch rupture into the whole-cell configuration. In synaptic experiments, stimulation was evoked by current delivered into the alveus through a broken patch pipette filled with ACSF.
Data were transferred during the experiment to a MacIntosh Power PC computer via an ITC 16 digital-analog converter (Instrutech), which was controlled by IGOR Pro Software (Wavemetrics). Subsequent analysis was done off-line using IGOR Pro Software and statistical tests were performed using Excel software (Microsoft). Statistical analysis of multigroup data was performed using a single-factor ANOVA. When there was a significant difference between the groups, Tukey's multicomparison tests were performed to determine the level of significance for each pairwise comparison. All measurements are provided as means ± SE.
Biocytin reconstruction and subiculum map
Neurons filled with biocytin were processed for visualization using an avidin-horseradish peroxidase (HRP) 3,3′-diaminobenzidine reaction. Morphological reconstructions were made using a camera lucida system, whereby each neuron was confirmed to be pyramidal. Morphological analyses were performed on a subset of reconstructed neurons that were examined closely to exclude cells that had obvious dendritic cuttings at the slice surface. Scholl concentric ring analysis was used to provide an estimate of dendritic arborization (Scholl 1953). Twenty-micrometer-diameter concentric rings were superimposed on the camera lucida reconstruction, and numbers of dendritic crossings were counted along both the basal (negative numbers) and apical (positive numbers) dendrites. Measurements of apical dendritic diameter were also performed under ×50 oil magnification at 50-μm increments from the soma.
The anatomical distribution of subicular neurons within the hippocampal formation was determined for 101 biocytin-filled neurons. Since the neurons were obtained from different slices of different aged rats, a normalized scale corresponding to a subicular map was developed. One axis of the map extended from the CA1-subiculum border to a tangential line connecting the alveus (at the end of subiculum) and dentate gyrus. The other axis extended from the alveus to the hippocampal sulcus (see Fig. 2). Each neuron was then given a normalized xycoordinate that corresponded to its relative position within the subiculum. During anatomical measurement, the experimenter was blind to the firing properties of the neuron. Twenty cells processed for morphology were discarded due to unresolvable anatomical borders.
Classification and anatomical distribution of subicular pyramidal neurons
Whole-cell patch-clamp recordings were obtained from 171 subicular neurons. These neurons were categorized based on their firing response to a just-suprathreshold, 1-s current injection from the resting potential. Three subtypes were discerned: strong bursting (SB, 88 cells, 51%), weak bursting (WB, 29 cells; 17%) and regular spiking subicular (RS, 54 cells; 32%) neurons (Fig.1). A burst was defined as two or more action potentials occurring at a high frequency (>200 Hz), followed by an afterhyperpolarization lasting tens of milliseconds. Single bursts were also elicited by short (10 ms) current injections, which resulted in bursts that outlasted the current injections (Fig. 4 B). SB neurons fired more than one burst (2–7 times) during a 1-s current pulse, whereas WB neurons fired one burst at the beginning of the current pulse and afterward generated a train of single action potentials. RS neurons fired trains of action potentials in response to DC injection. For comparison, recordings were made from 20 CA1 pyramidal neurons, all of which fired regular trains of action potentials in response to DC injection.
Perforated-patch recordings were made to test whether dialysis of the neurons in the whole-cell configuration grossly affected the electrophysiological characteristics of the neurons (see Fig. 7). All three subicular neuronal subtypes were found in the perforated-patch configuration (SB: n = 7, WB: n = 5, RS: n = 1).
To determine whether any differential distribution of the subicular firing types exists within the subiculum, 101 stained subicular neurons (53 SB, 17 WB, 31 RS) were mapped onto a two-dimensional axis system (Fig. 2, see methods for description of anatomical measurements). There was a statistically significant effect showing that recordings further away from CA1 (nearer presubiculum) were more likely to be from SB neurons than RS neurons (ANOVA P < 0.003). SB and RS neurons were most segregated from each other (P < 0.005), while WB neurons overlapped each of the other firing subtypes (P> 0.2). Conversely, there was no pattern of distribution of different subicular subtypes with respect to the alveus-hippocampal sulcus axis (P > 0.7).
Of the 171 recordings, 121 were processed with a biocytin-avidin-HRP reaction and confirmed to have pyramidal morphology by microscopic examination. Detailed camera lucida drawings were made of six SB, six CA1, five WB, and five RS pyramidal neurons to analyze dendritic morphology. Measurements of apical dendritic diameters were made in 50-μm increments from the soma, but no differences were found between all pyramidal cell types (P > 0.7; grouped data: soma = 15.5 ± 0.4 μm, 50 μm = 3.8 ± 0.2 μm, 100 μm = 3.65 ± 0.18, 150 μm = 3.6 ± 0.13 μm). Scholl analysis, however, revealed a significantly increased amount of dendritic arborization in the proximal apical dendrite of CA1, compared with subicular neurons (Fig.3, P < 0.001).
Synaptic responses of strong bursting subicular neurons
To determine whether bursting can be driven synaptically and to gain an understanding of the synaptic inputs onto subicular neurons, pharmacological dissection of excitatory and inhibitory transmission in SB neurons was employed (Fig. 4). All SB neurons exhibited bursting when stimulated synaptically to threshold (n = 5). In some cases, the evoked potential did not reach action potential threshold in control, but application of bicuculline (10 μM) allowed the synaptic response to reach threshold (n = 4) and in some cases increased the number of action potentials within the burst (n = 3). TheN-methyl-d-aspartate (NMDA)-antagonist AP-5 (50 μM) did not affect bursting (n = 4) whereas 6-cyano-7-nitroquinoxalene-2,3-dione (CNQX, 2 μM) mostly abolished excitatory transmission (n = 2). These data indicate that SB neurons receive both significant GABA and glutamatergic inputs and that synchronous excitatory stimulation triggers bursting at the threshold for action potential generation.
Resting membrane properties of pyramidal neurons
Resting membrane properties and firing characteristics were analyzed in more detail in a subset of neurons (see Figs.5 and 6for illustration and Table 1 for a statistical summary). The resting membrane potential for SB neurons on seal rupture was approximately −67 mV (n = 15). Input resistance (R N) was estimated by the slope of a line fitted to the current versus steady-state voltage relationship of current-clamp recordings with low series resistance (<15 MΩ). The average R N for SB neurons was 37 MΩ (n = 15). As illustrated in Fig. 5, during hyperpolarizing current pulses a pronounced depolarizing “sag” occurs (sag ratio = steady-state voltage/peak voltage; the voltage deflection measured is restricted to between 5 and 10 mV). The sag ratio for SB was determined to be 0.79 (n = 15). On addition of 5 mM CsCl to the bathing medium this sag is nearly completely abolished, presumably due to a block of the hyperpolarization-activated, nonselective cation conductance (I h). In 5 mM CsCl, the membrane time constant (τm) can be more accurately estimated. In SB neurons, τm = 19 ms (n = 5). The average input resistance in CsCl (R N(Cs+ )) accordingly increased to 57 MΩ (n = 5).
The resting membrane parameters in WB neurons were nearly identical to those found for SB neurons. The resting membrane potential was about −66 mV (n = 12). Input resistance was measured to be 36 MΩ (n = 12). WB neurons also had a Cs+-sensitive sag that was the same as in SB neurons (sag ratio = 0.80, n = 12). The time constant (τm), as measured in 5 mM CsCl was estimated to be 23 ms (n = 6), withR N(Cs+ ) rising to 75 MΩ.
While RS neurons do not fire bursts of action potentials, their resting membrane properties were the same as those of SB and WB neurons. The resting membrane potential was −67 mV (n = 13),R N was 38 MΩ (n = 13), R N(Cs+ ) was 75 MΩ (n = 5), and τm was 20 ms (n = 5). The Cs+-sensitive sag was equivalent as well (sag ratio = 0.79, n = 13). When comparing the resting membrane properties (V rest,R N,R N(Cs+ ), and τm) of all three firing subtypes of subicular neurons using a one-factor ANOVA, there were no statistically significant differences among the three firing cell types (P > 0.35).
Although CA1 pyramidal neurons have similar morphology, general firing properties, and resting membrane potential (−66 mV, n= 13, P > 0.8) as RS subicular neurons, their resting membrane properties were strikingly different. The input resistance in control and 5 mM CsCl as well as the membrane time constant in CsCl were all significantly higher in CA1 neurons compared with all subicular firing types [R N = 55 MΩ (P < 0.001),R N(Cs+ ) = 120 MΩ (P < 0.05), and τm = 40 ms (P < 0.001)]. Interestingly, the CA1 sag ratio was identical to the subicular sag (0.79, n = 15,P > 0.9), implying that the difference in resting conductances between subfields is not due predominantly to a hyperpolarization-activated depolarizing conductance.
Action potential properties of pyramidal neurons
The properties of action potentials in bursting and regular spiking neurons were further analyzed (see Fig. 6 and Table2 for a statistical summary). Action potential initiation in both subicular and CA1 pyramidal neurons arose from very similar absolute voltages (averages: RS = −47 mV, WB = −45 mV, SB = −47 mV, CA1 = −46 mV,P > 0.7), suggesting that there is little difference in the precise mechanism of initial action potential initiation in the different neurons. The current injection required to reach action potential threshold (rheobase) was also similar between subicular and CA1 pyramidal neurons (P > 0.2), although there was much variability within and between firing subtypes.
Within a burst, whether coming from a SB or WB neuron, each action potential had stereotyped properties. The number of action potentials within a burst ranged from two to six, and as they progressed the trend was for each action potential to become broader, with slower rates of rise and decay and smaller amplitudes (see Table 2). The first action potential within a burst (either SB or WB) had a half-width of 0.8 ms, which broadened to 1.2 and 1.7 ms in the subsequent spikes. Additionally, as the half-width broadened, so did the interspike interval (Table 2).
Comparing the first action potential of a burst to an action potential in RS neurons again revealed that all subicular neurons had very similar properties yet were distinct from CA1 neurons. The first derivative of V m(dV m/dt) gives a more sensitive measurement of action potential rising and falling, and the dV m/dt during the action potential rising phase may correspond to activation of fast voltage-gated Na+ channels (Colbert et al. 1997; Coombs et al. 1957). Using this measure, subicular neurons were found to have similar peak dV m/dt (rise) (subicular average = 290 mV/ms, P > 0.4), while the peak dV m/dt (rise) of CA1 neurons was significantly faster (CA1 = 380 mV/ms,P < 0.05). CA1 neurons also tended to have larger amplitude action potentials than subicular neurons (subiculum average = 93 mV, P > 0.7; CA1 = 110 mV,P < 0.001). There was no statistically significant difference in the peak rate of action potential descent [peak dV m/dt (fall)] of all pyramidal neurons measured (P > 0.2). The combination of these influences resulted in CA1 neurons displaying a wider half-width than subicular neurons (subiculum average = 0.82 ms,P > 0.2; CA1 = 0.95 ms, P < 0.05; see Fig. 6, inset).
Voltage-independence of burst firing
Other groups have reported that burst firing could be abolished by depolarization of the membrane to near action potential threshold (Mason 1993; Mattia et al. 1997a;Stewart and Wong 1993). We examined this in our study and found that neither depolarization (just below action potential threshold) nor hyperpolarization changed the firing properties of subicular neurons (n = 30). This was confirmed in perforated-patch recordings, which obviates concerns about dialysis (n = 10, Fig. 7).
Regulation of firing mode by potassium currents
Given that all subicular neurons are alike with regards to all resting and active properties measured, we next tested whether the tendency to burst could be regulated by potassium channels. Bath-application of a low concentration (50 μM) of 4-AP (a blocker of slowly inactivating, D-type potassium channels) was able to convert four of six RS subicular neurons into strong bursting neurons (Fig.7); in the other two neurons tested, weak bursting was elicited by 50 μM 4-AP. WB neurons were converted into SB neurons by the same concentration of 4-AP (n = 3). Interestingly, although CA1 neurons are quite distinct from subicular neurons, they too can be converted into bursting neurons by 50 μM 4-AP (5 neurons bursted strongly, 3 bursted weakly, and 1 did not change its firing properties).
This study represents the first patch-clamp analysis of bursting subicular neurons and the first to describe subicular neurons that burst repetitively in response to somatic current injection. Subicular intrinsic bursting is not due to dialysis by whole cell pipette solution since these subtypes were also observed in the less invasive perforated-patch configuration. Although we have classified the neurons into three groups, it may be appropriate to describe the firing properties of subicular neurons as lying along a continuum of propensity to burst. We observed a wide assortment of firing properties, ranging from neurons that never burst to those that would burst repetitively (7 times in 1 s, with >3 spikes per burst), with nearly all varieties in between (an example being a neuron that burst twice and then spiked regularly, which was classified as SB). Despite the variety in firing modalities, however, subicular neurons belong to a homogenous group with respect to all other measured parameters (both resting and active, see Tables 1 and2).
An interesting observation of this study is that neurons further away from CA1 are more likely to burst, although we did not reproduce a previous study that found more RS subicular neurons nearer the hippocampal sulcus (Greene and Totterdell 1997). Different subfields of subiculum (proximal to CA1 vs. distal to CA1) have been shown to preferentially receive input from different CA1 regions (reviewed in Amaral 1993). Specifically, proximal CA1 (with respect to CA3) innervates distal subiculum, whereas distal CA1 innervates proximal subiculum. Likewise efferent projections from restricted portions of subiculum (which can be further segregated into dorsal vs. ventral subiculum as well as proximal vs. distal) send parallel inputs to a wide array of cortical and subcortical structures (Naber and Witter 1998). It may be that the firing properties of subicular neurons serve to segregate subicular output one level further. In fact, there is some evidence that regular spiking and bursting subicular neurons may project differentially to entorhinal cortex and presubiculum, respectively (Stewart 1997).
Comparison of subicular and CA1 pyramidal neurons
Analysis of the dendritic morphologies of pyramidal neurons in subiculum and CA1 revealed that CA1 neurons have significantly more oblique branches arising from the proximal apical dendrite than subicular neurons. It could be that this difference contributes to the firing properties of the two hippocampal regions. In published models, however, increased dendritic membrane surface area augments the propensity to burst (Mainen and Sejnowski 1996). It may also be that the different levels of proximal apical oblique branches are related to differences in synaptic input into the subicular and CA1 regions.
In this study 68% of the 171 subicular neurons exhibited intrinsic bursting behavior. By contrast, none of the 20 CA1 neurons we recorded from exhibited bursting behavior. With respect to resting properties, all subicular neurons have nearly identicalR N,R N(Cs+ ), and τm, all of which are significantly different from those measured in CA1 pyramidal neurons.
Overall, subicular pyramidal neurons have greater resting conductances than CA1 pyramidal neurons as measured byR N,R N(Cs+ ), and τm. These conductances can be divided into Cs+-sensitive and -insensitive portions. In the presence of 5 mM CsCl, subicular neurons still have a much smaller input resistance than in CA1, demonstrating that there is a greater amount of resting Cs+-insensitive conductances in subiculum. Levels of the Cs+-sensitive conductance, on the other hand, appear to be equivalent between all cell types. A majority of the Cs+-sensitive conductance is carried by the hyperpolarization-activated, nonselective cation current,I h, which is active at resting membrane potentials and does not inactivate (Magee 1998).
In CA1, subthreshold synaptic integration has been studied extensively, and I h has been shown to be important in regulating temporal summation. There is an increasing density ofI h distally along the apical dendrite, and it has been shown that during excitatory postsynaptic potentials, the deactivation of I h delivers a net hyperpolarization helping to prevent temporal summation at frequencies <100 Hz (Magee 1998, 1999). It will be interesting to determine whether I h performs a similar function in subiculum, where its increased resting, non-I h-derived conductances also shape synaptic integration. Given that subicular neurons have a faster τm, the hyperpolarizing effects ofI h deactivation may not be as important since the membrane will already repolarize quicker than in CA1 neurons. Recording the currents along the somatodendritic axis, as well as performing dual recordings to examine synaptic integration, will certainly shed light on these issues.
During action potential generation, the fastest rising phase of the spike likely corresponds to a time of maximal Na+ channel activation, prior to significant Na+ channel inactivation and subsequent activation of repolarizing K+channels. By using the first derivative of the action potential rise [dV m/dt (rise)], one can more directly measure the fastest portion of the action potential and compare it between cell types. We observed that the first action potential of a subicular neuron (whether in a train or a burst) has a similar peak dV m/dt (rise) across firing types, suggesting that the density of fast voltage-gated Na+ channels does not account for differences in the tendency of subicular neurons to burst. This does not, however, rule out an influence of a persistent Na+ current on bursting, which would not likely affect measurement of dV m/dt (rise). As a caveat, it should be noted the action potentials recorded at the soma are likely backpropagated from the axon (Colbert and Johnston 1996; Spruston et al. 1995). Therefore the action potentials recorded in the soma have already been shaped by geometry and intrinsic conductances of the soma and do not necessarily reflect what action potentials look like in the axon (Williams and Stuart 1999). However, given that both the absolute voltage threshold for action potentials as well as the resting parameters were similar across all subicular neurons, it is likely that dV m/dt (rise) can be used for comparisons.
The action potentials recorded in the somata of subicular and CA1 pyramidal neurons were observed to be quite different, with subicular action potentials having slower rising rates, smaller amplitudes, and smaller half-widths than observed in CA1 neurons. Given that subicular axonal action potentials have greater attenuation measured at the soma than CA1 neurons, perhaps action potentials correspondingly do not backpropagate into distal dendrites as effectively in subiculum compared with CA1 (or perhaps the action potentials are different from the start in the axon, although they do arise from the same voltage threshold).
Bursting in subiculum
Using patch-clamp recording, we have observed three electrophysiological subtypes of subicular neurons based on their action potential firing in response to a just suprathreshold, 1-s somatic current pulse. These subicular subtypes are regular spiking (RS), weak bursting (WB; an initial burst followed by regular spiking action potentials), and strong bursting (SB; repetitively bursting). Others have recorded from subicular neurons with sharp microelectrodes but only observed WB and RS firing subtypes (Greene and Totterdell 1997; Mason 1993; Mattia et al. 1993; Stewart and Wong 1993; Taube 1993). All but one of those studies found similar resting membrane properties between bursting and nonbursting neurons. Our measurements of R N,R N(Cs+ ), and τm are larger than those found in earlier studies, which is likely due to the somatic shunt introduced by sharp microelectrode impalement (Spruston and Johnston 1992;Staley et al. 1992). Stewart and Wong (1993) described a type of regular spiking neuron that had markedly different properties (little sag and longer τm) than those described here and elsewhere. We did not find any neurons that fit the characteristics they described.
Our recordings show that subicular bursting is not affected by depolarization or hyperpolarization of the membrane potential, a finding observed both in whole-cell and perforated-patch configurations. This is in conflict with previous studies using sharp microelectrodes, which found that depolarization prevented burst formation in subicular neurons (Mason 1993;Mattia et al. 1997; Stewart and Wong 1993). It is hard to reconcile this difference, though we note that these studies did not observe SB neurons, so perhaps bursting was not as robust in their preparation and therefore easier to prevent.
Our recordings suggest that all subicular pyramidal neurons are able to burst. Two-thirds of subicular neurons burst intrinsically to a varying degree, while the remainder burst when exposed to the D-type potassium current blocker 50 μM 4-AP. Based on the remarkable similarity between groups, it seems unlikely that resting membrane properties contribute significantly to the differences in firing subtypes in subiculum. Additionally, we did not find any obvious morphological differences in the dendritic arborization that could readily account for different firing properties (Mainen and Sejnowski 1996).
The ionic mechanisms of bursting in subiculum have been debated in the literature. In a series of papers, Mattia and collaborators have proposed that a Na+ current is responsible for the slow depolarizing envelope that underlies a burst (Mattia et al. 1993, 1997a,b). On the other hand, Stewart and Wong suggested a Ca2+ current is necessary for the burst (Stewart and Wong 1993). To resolve this controversy, these implicated currents will need to be measured directly with voltage-clamp procedures (Jung et al. 1999).
Bursting in CA1
CA1 pyramidal neurons have been studied extensively in the whole-cell patch-clamp configuration in vitro, and they do not burst in response to just-suprathreshold somatic currents injection in normal ACSF (Golding et al. 1999). However, other groups have reported intrinsic bursting at threshold in the CA1 region using sharp microelectrodes (Jensen et al. 1994; Kandel and Spencer 1961; Masukawa et al. 1982;Schwartzkroin 1975). Using pharmacological dissection of Na+ and Ca2+ currents, Yaari and collaborators have suggested that this intrinsic “threshold bursting” is dependent on a persistent Na+current (Azouz et al. 1996). Furthermore this threshold bursting can be enhanced when external potassium is raised from 3.5 to 7.5 mM (Jensen et al. 1994). It is unclear what accounts for the discrepancy in intrinsic firing properties when comparing sharp microelectrode versus patch-clamp recording techniques, although perforated-patch recordings from CA1 neurons also do not show threshold bursting (Spruston and Johnston 1992).
It has been demonstrated that, in vivo, CA1 pyramidal neurons exhibit complex spike bursting, a type of bursting which has been correlated with sharp waves in field recordings. These complex spike bursts are quite different from threshold bursting. Complex spike bursts require either strong synchronous synaptic inputs to the dendrites or strong direct dendritic current injection (Golding et al. 1999). Threshold bursting in subicular neurons, by contrast, requires only a suprathreshold somatic current injection. Complex spike bursts have been shown both in vivo and in vitro to be derived from dendritic electrogenesis, in the form of large regenerative Ca2+ and Na+ spikes (Golding and Spruston 1998; Golding et al. 1999; Kamondi et al. 1998). Recordings at the soma during these complex spikes often show a burst of action potentials.
In a recent study, Golding et al. demonstrated that bath-applied 100 μM 4-AP permitted a just-suprathreshold somatic current injection to produce a burst of action potentials (in contrast with the synchronized dendritic depolarization normally necessary for CA1 complex spike bursting). In normal ACSF, bursts are not produced by somatic current injection; rather, somatic current injection stimulates backpropagating action potentials, which activate dendritic A-type (early in a train) and D-type potassium currents (later in a train) (Golding et al. 1999; Hoffman et al. 1997; Magee and Carruth 1999). These currents provide a shunt that prevents dendritic Ca2+ spikes, such as those described for complex spikes. When a low concentration of 4-AP is present, however, backpropagating action potentials provide sufficient dendritic depolarization to elicit a dendritic Ca2+ spike, which, in turn, produces a somatic burst of action potentials. In the present study, low concentrations of a D-type potassium current blocker (50 μM 4-AP) usually converted a regular spiking CA1 pyramidal neuron into a threshold bursting neuron. It may be that the CA1 threshold bursting reported in our study involves a similar dendritic Ca2+ electrogenesis, although we cannot rule out that there may be some contribution of a persistent-Na+ current (Azouz et al. 1996). Furthermore it remains to be seen whether dendritic calcium electrogenesis underlies the mechanism for bursting in subiculum, as it does in bursting layer 5 cortical pyramidal neurons (Larkum et al. 1999; Schwindt and Crill 1999; Williams and Stuart 1999, but seeGuatteo et al. 1996).
Functional implications of subicular intrinsic bursting
There has been much speculation about the functional relevance of action potential bursting. Primarily, it has been suggested that bursting increases the probability of synaptic vesicle release at the presynaptic terminal (reviewed in Lisman 1997). This has been demonstrated directly in paired recordings from neocortical neurons, where it was shown that each action potential within a burst is transmitted to the presynaptic terminal and may result in synaptic release (Williams and Stuart 1999). Using minimal stimulation techniques in the Schaffer collateral to CA1 pyramidal synapse, it has been shown that high-frequency stimulation mimicking a burst ensures neurotransmitter release (Dobrunz et al. 1997).
It has also been hypothesized that bursting provides specific information content that is not found in general action potential activity. For instance, in “place cells” of the CA1 region, action potential bursting defines a more restricted place field within that described merely by the cells firing rate (Otto et al. 1991). CA1 bursting has also been correlated to consummatory behavior and slow-wave sleep, where the function of bursting versus tonic firing is less clear (Buzsaki 1986). Another example of the information content of bursting is the thalamus, where bursting is a network-driven phenomenon that is seen predominantly during sleep, but it also may be involved in the trance-like condition of absence seizures (Steriade et al. 1993).
The significance of bursting in subiculum, a phenomenon observed both in vitro and in vivo, is still largely unexplored. Subiculum is involved in both spatial and declarative memory processing (Gabrieli et al. 1997; Sharp 1999). It has been described as a relay center for the processed information of the hippocampus and also as a possible location of epileptogenesis. Given its role as a major output of the hippocampus, in combination with its dual entorhinal and CA1 input, one idea is that subiculum may serve to gate information flow out of the hippocampus, akin to thalamic gating of cortical flow. Subicular bursting and the resultant unerring synaptic release would ensure that the processed information reliably passes onto the next level of circuitry. Understanding the basic mechanisms of how these neurons generate their bursts and how bursting can be manipulated will certainly be a key to understanding subiculum's place within the hippocampal formation and will provide valuable information about hippocampal processing as a whole.
We thank D. Cooper, N. Golding, and T. Mickus for critical reading of the manuscript.
This work was supported by grants from the National Institute of Neurological Disorders and Stroke (NS-35180), the National Science Foundation (IBN-9876032), the Human Frontiers in Science Program, and the Sloan and Klingenstein Foundations (N. Spruston).
Address for reprint requests: N. Spruston, Dept. of Neurobiology and Physiology, Northwestern University, 2153 N. Campus Dr., Evanston, IL 60208-3520 (E-mail:).
- Copyright © 2000 The American Physiological Society