Synaptic Contributions to Focal and Widespread Spatiotemporal Dynamics in the Isolated Rat Subiculum In Vitro

Temporal dynamics of population activity in the isolated subiculum in vitro

Lowering extracellular Mg²⁺ induced spontaneous epileptiform activity in the isolated subiculum after 10-20 min (∼70% minislices) (Behr and Heinemann, 1996; Harris and Stewart, 2001). Recordings were performed once activity stabilized, between 30 min and 2 hr after washout of Mg²⁺. Epileptiform activity was recorded at both the proximal (near CA1) and distal (near pre-subiculum) borders (Fig. 1A,B), indicating the widespread activation of all subicular cells. Local field activity was also detected (see below). The intervals between epileptiform events were widely irregular (mean frequency, 0.26 ± 0.19 Hz; CV, 48 ± 15%; n = 13) (Fig. 1B, right traces). The GABA_A receptor antagonist PTX (100 μm) increased 1.3 ± 0.4 times the frequency of epileptiform activity (0.21 ± 0.07 Hz; CV, 46 ± 23%; n = 9), which was blocked by AP-5 (100 μm; n = 8) and strongly reduced by NBQX (n = 2) or CNQX (n = 6) at 10 μm (Fig. 1C).

To investigate whether there is a clear leading region of epileptiform activity, we performed cross-correlation analysis of dual field recordings. We looked for the appropriate slice width (Fig. 1A) for which epileptiform activity was spontaneously generated with minimal temporal variations. Minislices 900-1300 μm wide showed large variability of latencies between dual field recordings separated ∼300 μm (CV, 76 ± 33%) (Fig. 1E). Epileptiform activity was not recorded in minislices <500 μm wide (n = 9) (Fig. 1F). Minislices 500-750 μm wide (n = 15) were less excitable than minislices >750 μm wide (n = 15) (Fig. 1F). Interestingly, latency variations were significantly reduced in smaller minislices, and mean latency significantly decreased to -1.4 ± 1.9 msec (versus -14.6 ± 4.4 before cutting; same interelectrode distance before and after cutting; n = 4; p < 0.005) (Fig. 1G).

Therefore, similar to the hippocampus (Miles et al., 1984), the site of origin of population activity in the subiculum is spatially constrained in minislices 500-750 μm wide (latency, ∼0 msec). This also guaranteed that latencies are conserved among different episodes of field activity. We used this approach to better understand how different cell types interact to generate field activity in the isolated subiculum.

Focal and widespread field activity in the isolated subiculum

To examine the relationship between single-cell firing and population activity, cell-attached recordings were initially obtained. Whole-cell configuration in current and voltage clamp was then established to characterize membrane properties and synaptic activity. We took special care to avoid artificial delays between patch and field recordings by minimizing their relative distance (∼20-50 μm). The maximal error introduced was estimated from the delay between the cell and field spike evoked by local stimulation (0.1 ± 0.8 msec; n = 4 cells). A total of 54 cells were recorded: 17 were classified as IB+, 17 were IB-, 12 were RS, and 8 were of the FS type (Table 1; see Materials and Methods).

We classified field activity as desynchronized, widespread, and focal (Fig. 2). During periods of desynchronized activity, irregular cell firing was present mostly in FS and bursting cells (Fig. 2A; Table 1). The nature of this firing could be intrinsic or synaptic, because spontaneous EPSCs were evident in voltage-clamp configuration (Fig. 2A, vc trace) and EPSP-driven firing was present in current-clamp configuration (Fig. 2A, cc trace). Periods of desynchronized field activity were characterized by a flat cross-correlation of dual field recordings separated 450 ± 69 μm (n = 7) (Fig. 2D).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Simultaneous field (fp) and patch recordings. Cell responses were examined using cell-attached (ca) and whole-cell (wc) configurations in the current-clamp (cc) and voltage-clamp (vc) modes. Field recordings corresponding to ca, cc, and vctraces are depicted. A, Simultaneous field and patch recordings during desynchronized field activity. This cell was classified as weak bursting (IB-). B, Simultaneous field and patch recordings during widespread field activity. C, Simultaneous field and patch recordings during focal field activity. D, Cross-correlation analysis of simultaneous field recordings. Results are shown from three consecutive episodes of desynchronized field activity. E, Widespread activity. F, Focal activity. Insets show examples of dual field recordings (fp1 and fp2).

During widespread population activity (Fig. 2E), cell firing was recorded simultaneously with epileptiform field events (Fig. 2B). Interestingly, some cells (mostly bursting cells) increased their firing 50-100 msec before the onset of field events. Recordings under current and voltage clamp showed that this was associated with an increase of synaptic activity (Fig. 2B, cc and vc traces). This buildup period has been described in other in vitro models of epilepsy (Chamberlin et al., 1990; Traub and Dingledine, 1990) and in immature population activity (Menendez de la Prida and Sanchez-Andres, 1999; Wenner and O'Donovan, 2001).

The combination of cell-attached and field recordings facilitated the identification of focal activity (Fig. 2C), which was characterized by low-amplitude changes in the field that could be associated with cell firing. The mean, but not the variance, of the distribution of field event amplitude was different in focal (74 ± 28 μV) versus widespread episodes (131 ± 54 μV; data from 21 events from n = 7 minislices; p = 0.05). Once in whole-cell configuration, barrages of synaptic activity were evident in current- and voltage-clamp simultaneous to focal field activity (Fig. 2C, cc and vc traces). Not all the cells recorded exhibited firing in correlation with focal activity. Those that did included 5 of 17 IB+ cells, 5 of 17 IB-cells, 5 of 12 RS cells, and 6 of 8 FS interneurons. This suggests that only a fraction of cells contribute to the averaged field changes (see below). Focal activity occurs erratically, with no clear periodicity (mean frequency, 0.36 ± 0.16 Hz; CV, 58 ± 18%; n = 22 minislices). The absence of peaks in the auto-correlation (data not shown) and cross-correlation functions (Fig. 2F) from dual field recordings confirmed its local non-oscillatory nature and suggested that focal activity was not related with waves of activity across the tissue.

Role of bursting and FS cells in the generation of focal activity

We first examined the contribution of each cell type to focal activity using field-triggered data segments from cell-attached recordings (see Materials and Methods). Both IB+ (n = 5) and IB-(n = 5) cells that participated in focal activity typically began to fire before the field activity [-13 ± 35 msec (n = 41 episodes from five IB+ cells) and -12 ± 41 msec (n = 40 episodes from five IB+ cells)] (Fig. 3A,B). In contrast, RS cells were activated after the field activity (12 ± 25 msec; n = 40 episodes from five cells) (Fig. 3A,B), and a same cell could remain silent during some episodes (Fig. 3A, RS cell). In addition, bursting cells typically fired bursts, whereas RS cells fired single spikes (Fig. 3A). We also examined the behavior of FS interneurons. Similar to RS cells, FS interneurons began to fire 10-12 msec after focal field activity (11 ± 15 msec; n = 64 episodes from six cells) (Fig. 3B), but in contrast to the glutamatergic cell type, this was followed by high-frequency firing (Fig. 3A,C).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Cellular and synaptic behavior underlying focal activity. A, Cell responses in cell-attached (ca) configuration during focal activity were examined using field-triggered data segments (fp trace). Two consecutive episodes are shown for each cell (RS, IB+, IB-, and FS cells). B, Histograms of cell firing from each cell type. C, Representative response from another FS interneuron (ca trace) triggered by the onset of focal field activity (fp trace). Note the different time scale compared with A. D, Field-triggered synaptic currents from an IB-cell. E, Field-triggered synaptic currents from the cell shown in D at three different holding potentials (-58, -80, and -95 mV). F, Field-triggered synaptic currents from a different IB-cell at two holding potentials. G, Field-triggered composite reversal potential from cells shown in E (black trace) and F (gray trace). E-G share the same time scale (shown in G).

These data suggest that both IB+ and IB-cells constituted the first glutamatergic type that initiates a focal recruitment process that subsequently activates RS and FS cells. FS interneurons would in turn suppress activity. To test this idea, we looked at the field-triggered synaptic changes in 20 cells (5 IB+, 5 IB-, 5 RS, and 5 FS) and examined the temporal correlation between excitation and inhibition using the composite reversal potential (see Materials and Methods).

Glutamatergic- and GABAergic-mediated synaptic activity during focal activity

In most cells (12 of 19: 4 IB+, 4 IB-, and 4 RS), excitatory synaptic currents were recorded at the onset of focal field activity (Fig. 3D,E). The composite Vrev(t) typically showed a depolarized shift during the first 50-70 msec (peak value, -31 ± 11 mV; n = 12) (Fig. 3G, black trace), which suggests a dominance of glutamatergic excitatory currents. No differences were found between IB+ (-28 ± 17 mV; n = 4), IB-(-34 ± 17 mV; n = 4), and RS cells (-27 ± 5 mV; n = 4). All FS cells examined also exhibited a peak value of Vrev(t) that was even more depolarized than in glutamatergic cells (-7 ± 5 mV; n = 5; p < 0.005). Because both IB+ and IB-cells were the major glutamatergic type activated before the onset of focal activity, this analysis supports their role in the recruitment process. In addition, it also supports that FS cells experience strong glutamatergic activation at the onset of field activity.

Some bursting (IB+, one of five; IB-, one of five) and RS cells (one of five) exhibited inhibitory synaptic currents at the onset of focal activity (Fig. 3F, current trace at -58 mV holding potential). In these cases, the composite Vrev(t) showed a drop toward negative potentials ∼-75 mV, near the reversal potential of control GABA_A-mediated chloride currents (Fig. 3F,G, gray trace). Because most FS inter-neurons began to fire 12 msec after the field (Fig. 3B), this probably reflects the effect of a fast inhibitory surround. A second barrage of excitatory synaptic activity was evident in these cells (Fig. 3F), as showed by the shift of Vrev(t) toward positive values (Fig. 3G, gray trace). This glutamatergic component could be related to the secondary recruitment of RS cells or with sustained activity of bursting cells (Fig. 3B).

In some neurons (5 of 20), long-lasting synaptic currents were recorded (Fig. 3D, gray shading). Their GABAergic nature was suggested by comparison with the temporal course of interneuron firing during field-triggered data (Fig. 3A,C,D) and by its reversal potential (Fig. 3E). The composite Vrev(t) showed a negative shift 150-200 msec after the onset of focal activity (-88 ± 6 mV; n = 5) (Fig. 3G, black trace). This is consistent with a late prevalence of GABA_B-mediated currents.

Role of different cell types in the initiation of widespread epileptiform activity

We also examined the contribution of each cell type to the onset of widespread epileptiform activity using field-triggered averages. We found that both IB+ and IB-cells began to fire 50-100 msec before the onset of widespread epileptiform activity (Fig. 4A,B). In contrast, RS cells tended to fire after the onset of epileptiform activity (Fig. 4C). Most FS interneurons fired at the onset, although some firing increases were detected before (Fig. 4D).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Cell behavior underlying widespread activity. Field-triggered (fp) cell responses (5 consecutive events are depicted) in a cell-attached configuration (ca) from representative IB+ (A), IB-(B), RS (C), and FS (D) cells. Firing histograms from all cells in each group are shown at the bottom.

A second observation was evident when comparing histograms of single-cell firing during widespread and focal activity. During focal activity, bursting cells had a peak of firing before the field activity and subsequently decreased firing as the RS and FS firing increased (Fig. 3B). In contrast, generation of widespread activity was associated with a preceding monotonic increase of bursting cell firing that reached a peak at the onset of the widespread activity (Fig. 4A,B). Both RS and FS cells had similar spike time histograms in focal versus widespread field activity. This suggests that bursting cells are critical in initiating widespread forms of population activity and that FS cell firing, which is time-locked to leading cell firing, plays a role in controlling runaway excitation.

However, the temporal correlation between cell firing and the field activity varied among different episodes of focal and widespread activity. For a given cell, firing could occur before or after the onset of field activity (Fig. 5A; see also Fig. 3A, IB+ cell), suggesting that they variably participated in correlated firing. We defined the driving index (σ) as a statistical measure of the probability of a given cell to fire before the field activity by examining its behavior in 10-13 consecutive episodes of population activity (see Materials and Methods).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Electrophysiological properties and the driving index. A, Example of firing histograms from one cell (IB+ cell) in cell-attached recordings during field-triggered activity (data from 13 consecutive episodes). The driving index measures the probability of a given cell to fire before the field (see Materials and Methods). B, Statistically significant correlation (p < 0.05; r =-0.58) between the firing threshold and the driving index (n = 46 glutamatergic cells). C, Field-triggered averages (black traces) of synaptic currents during focal activity at three holding potentials. SDs are shown in gray. D, Conductance changes (up) and composite reversal potential (bottom) computed from data in C. E, Vrev-ΔG phase plots (see Materials and Methods and Borg-Graham et al. 1998). ΔG(t) is plotted as a function of the Vrev(t) during the first milliseconds before reaching the conductance peak. Arrows mark this temporal evolution. Gray circles denote the Vrev,ΔG values at conductance peak. Open circles denote the Vrev,ΔG values at the Vrev peak. Note the shunting effect of peak conductance increases (gray dots).

Both IB+ and IB-bursting cells exhibited a higher driving index than RS cells (Table 1) (p < 0.05). Nevertheless, the mean σ from bursting cells was <0.5, a statistical limit separating the leader and follower firing pattern (see Materials and Methods). This suggests that in statistical terms, bursting cells did not define a glutamatergic leader population (although 15 of 34 cells had σ > 0.5). The RS group, in contrast, behaved as a follower (mean σ = 0.08 ± 0.13; n = 12). All but one FS interneuron (seven of eight) had a driving index < 0.5, indicating that they were typically activated afterward.

The driving capacity of a cell is mainly determined by electrophysiological and morphological factors

Intrinsic electrophysiological properties such as resting membrane potential, input resistance, membrane time constant, and firing threshold could affect the firing capacity of a cell on synaptic activation. We examined these electrophysiological parameters among different cell types and found the firing threshold to be the most critical factor (Fig. 5B) (correlation coefficient r = -0.58; p < 0.05). Interestingly, bursting cells had a lower firing threshold (relative) than RS cells (Table 1). A higher input resistance in RS cells would favor their synaptic recruitment, but these cells exhibited the highest firing threshold. FS cells had high input resistance and a lower firing threshold than RS cells.

To address the possibility that those cells displaying σ > 0.5 received stronger synaptic activation, we examined the temporal evolution of synaptic conductance changes during field-triggered focal activity in a subset of cells (n = 17; see Materials and Methods) (Fig. 5C,D). The composite Vrev(t) initially evolved toward depolarized values and then hyperpolarized as the conductance reached a peak (Fig. 5D). This evolution is better depicted in Fig. 5E, in which ΔG(t) is plotted as a function of the Vrev(t) during the first milliseconds before reaching the conductance peak (arrows). From these plots, clearly the conductance increases associated with the Vrev peak (open circles; -27 ± 15 mV) were lower than the conductance peaks (gray circles; p < 0.055), which are associated with a Vrev of -56 ± 8 mV. This suggests that during conductance peaks concurrent contribution of inhibition and excitation existed. Values of conductance peaks and conductance increases at Vrev peak were similar in the three glutamatergic cell types (Table 2). FS interneurons were subject to larger increases than RS and bursting cells (Table 2) (p < 0.005). We did not find a good correlation between conductance increases in glutamatergic cells and their driving index (r = 0.13; p > 0.05). Additional experiments were performed using a cesium-based patch solution (n = 6). The driving index of these cells (ranging from 0 to 1) was established in the cell-attached configuration. No correlation between conductance changes and the driving index was documented (r = 0.12; p > 0.05).

The subtle depolarized Vrev at peak conductance suggests a shunting GABAergic effect over excitatory inputs (Fig. 5D). A use-dependent shift of the GABA_A reversal potential could also contribute (see Materials and Methods). Therefore, it could be possible that proximal GABAergic inputs mask distal excitatory ones (Koch et al., 1990). Space-clamp errors of distal inputs could also contaminate these results (Spruston et al., 1993). We therefore explored whether differences in apical/basal dendritic arborization were correlated with the driving index in a group of morphologically reconstructed cells (n = 15).

As reported previously, no difference in dendritic morphology was found between bursting and RS subicular cells (Staff et al., 2000; Harris et al., 2001). However, when cells were grouped by the driving index (σ = 0, n = 5 cells; σ > 0.5, n = 4 cells), the difference in the number of distal dendritic crossings was evident at 250-450 μm from the soma (Fig. 6A, arrow). The mean number of distal crossings from all cells showed a significant correlation with the driving index (Fig. 6C) (r = 0.76; p < 0.05). No difference in the spatial distribution of cells was present (Fig. 6D). Therefore, the difference in distal crossings suggests that driving cells may receive more inputs at secondary distal dendrites than follower cells. The good linear regression of I-V data (regression coefficient, 0.98-0.99) suggests good voltage-clamp condition and that voltage-gated channels were not activated. However, although reliable estimates of the Vrev of synaptic currents located 250-400 μm from the soma can be obtained (2-3 mV error), conductance values are significantly reduced by >20% because of space-clamp errors (Spruston et al., 1993). Therefore, the lack of correlation with somatic conductance changes could be related to significant electrotonic filtering.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Morphological properties and the driving index. A, Scholl diagram of apical and basal dendritic arbors of cells exhibiting low (σ = 0.0 ± 0.0; n = 5) and high (σ = 0.7 ± 0.2; n = 4) driving index. Statistical difference was evident at distal apical dendritic crossings (arrow). B, Representative examples of cells in each category. C, Correlation between the mean apical dendritic crossings at 250-450 μm from the soma and the driving index in all cells examined (n = 15). D, Spatial distribution of cells in each category. Positions are normalized.

MUA recordings during desynchronized, focal, and widespread population activity

Together, these results suggest that although some bursting cells play a leader role, network activation is a probabilistic process that needs the local recruitment of several cell types. Periods of focal activity seem to reflect the functional architecture of the underlying network with bursting, RS and FS cells differentially sculpting the spatiotemporal patterning. To gain in spatiotemporal resolution of the network activity, we combined field and multi-unit recordings. We used pharmacological tools to dissect the underlying circuit mechanism.

Data collected using MUA recordings were consistent with the picture that emerged from cell-attached recordings. Desynchronized, widespread, and focal field activities were correlated with three different MUA patterns (Fig. 7). During desynchronized field states, multi-unit firing was recorded (Fig. 7A). Both single spikes and bursts (Fig. 7A, inset a) of different amplitudes were erratically detected, suggesting the spontaneous activation of different cells. During widespread epileptiform activity, high-frequency firing was recorded in correlation with large field epileptiform events (Fig. 7B). Interestingly, firing increases were detected 50-100 msec before the onset of field events, which matched the buildup period. Cell ensemble activity was evident in MUA recordings during focal field activity (Fig. 7C). The amplitude, number, and sequence of spikes recorded within these ensembles varied substantially, confirming that many cells were variable recruited during consecutive episodes. This was also manifested in the variability of ensemble duration (124 ± 34 msec; CV, 57 ± 16%; n = 11 minislices). Ensemble activity during focal activity was similar to MUA activity detected before the onset of widespread field activity (Fig. 7, compare B, C).

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

Simultaneous field and MUA recordings during desynchronized (desynch; A), widespread (B), and focal (C) field activity. Insets a-c show additional details of firing activity.

Pharmacological dissection of synaptic factors contributing to focal field activity

We took advantage of the higher spatiotemporal resolution of MUA recordings to pharmacologically dissect the synaptic contributions to focal field activity. For this purpose, we initially addressed the question of what are the sources of cell firing during desynchronized activity by examining changes of firing rate under different receptor antagonists (Fig. 8A). We first derived confidence intervals of a firing rate from MUA recordings of 1 hr duration. Changes of 14% were not statistically significant (n = 5 minislices).

• Download figure
• Open in new tab
• Download powerpoint

Figure 8.

Pharmacological control of MUA firing during desynchronized activity. A, Representative MUA traces under control, zero Mg²⁺, and zero Mg²⁺ plus CNQX/AP5/PTX. B, Summary of results from pharmacological control of MUA firing in juvenile (left) and adult (right) rats. The number of slices is indicated. Statistical differences are shown: ^**p < 0.001; +p < 0.05.

Lowering extracellular Mg²⁺ induced a subtle increase of firing during desynchronized field activity (n = 7) (Fig. 8B). Most of this firing had a glutamatergic origin because it was reduced to 29 ± 18% of control by suppressing glutamatergic transmission with CNQX (or NBQX) and AP-5 (n = 6) (Fig. 8B). This was confirmed by adding a mixture of antagonists containing MCPG and MPPG (1 mm), atropine (1 μm), mecamylamine (20 μm), and saclofen (200 μm; n = 4) (Fig. 8B).

We next wondered about the synaptic components of focal field activity. Data from patch recordings suggested an initial glutamatergic activation, followed by a composite GABAergic component. Using pharmacological tools, we confirmed that NMDA, but not non-NMDA, transmission was critical in the generation of focal activity that was blocked by AP-5 (Fig. 9A,C).

• Download figure
• Open in new tab
• Download powerpoint

Figure 9.

Pharmacological dissection of synaptic factors mediating focal activity. A, Field-triggered MUA ensembles during focal activity from juvenile tissue in zero Mg²⁺ media and after bath application of PTX, CNQX, and AP-5. B, Field-triggered MUA ensembles during focal activity from adult tissue in zero Mg²⁺ media. The inset shows additional details of firing activity. C, Summary of pharmacological results. The number of slices is indicated. D, Nissl-stained subicular minislice. A mark denoted the position of tungsten electrodes. E, Simultaneous MUA recordings during focal activity. F, Cross-correlation analysis of simultaneously recorded MUA firing rate. Results from five consecutive episodes are shown in gray; the mean is shown in black. G, Same as in F after 4-7 min of PTX application. Note the significant increase of cross-correlation. H, Summary of the cross-correlation analysis under different conditions: zero Mg²⁺, PTX (4-7 min), saclofen, and CNQX. The number of slices is indicated. I, Pharmacological dependence of the frequency of focal and widespread field events.

Consistent with a role of inhibition in firing cessation, both saclofen and PTX induced an increase of ensemble duration (n = 5; p < 0.05) (Fig. 9C). Interestingly, this effect was also present under CNQX and NBQX (Fig. 9C)(p < 0.05). The prolongation of ensemble activity under PTX was less pronounced than under saclofen and CNQX (Fig. 9C). To better understand the network mechanisms, we analyzed data from dual MUA recordings (Fig. 9D).

Correlated ensembles were detected in <35% cases from paired MUA recordings separated 285 ± 30 μm (n = 9) (Fig. 9E). PTX significantly favored the spread of ensemble activity for 4-7 min after application (Fig. 9F-H). After ∼7 min, local activity became unstable, and widespread epileptiform events were more frequent (Fig. 9I). The spread of local activity was also favored by saclofen (Fig. 9H) (p < 0.05), but in contrast to PTX, this did not change neither focal nor widespread event frequency (Fig. 9I). CNQX (or NBQX) did not affect the local spread of activity, although it substantially increased ensemble duration (Fig. 9C,H). Because non-NMDA receptors mediate activation of FS interneurons (Menendez de la Prida, 2003), this probably reflected a disinhibitory effect as opposed to firing cessation. CNQX did reduce the frequency of both local ensembles and epileptiform events (Fig. 9I). Together, these results suggest that the spatial specificity of focal activity is strongly determined by local inhibition because impairment of GABA_A,B synaptic transmission favored merging of formerly non-overlapping ensembles.

Model and developmental dependence of focal activity

Finally, we tested whether focal activity was specific of the zero Mg²⁺ model or whether it could be induced by other manipulations such as increasing cell excitability using high K⁺ media. Increasing [K⁺]_o to5mm induced focal ensemble activity in four of five minislices examined (data not shown). Focal events in 5 mm K⁺ were less frequent (0.08 ± 0.06 Hz; n = 4) than in the zero Mg²⁺ model (p < 0.05).

We also examined tissue from adult rats to discriminate possible developmental changes. Focal (Fig. 9B) and widespread (data not shown) population activity were also recorded in the adult isolated subiculum in zero Mg²⁺ (n = 11). Although the frequency of these events, and the duration of focal activity, was lower than in juvenile tissue, their pharmacology was similar (Fig. 9C). This probably reflects the lower susceptibility of adult tissue to epilepsy. Interestingly, the synaptic component of single-cell firing during desynchronized states was lower in adult rats than in juvenile rats (Fig. 8B). This further suggests developmental differences in local connectivity between juvenile and adult tissue.

Functional organization of local subicular circuits

To better understand the local operation of neuronal circuits, it is essential to dissect their cellular and synaptic components and to reveal their patterning of activity. One approach is to artificially enhance glutamatergic currents by using low Mg²⁺ media. There are several computational and technical reasons for this. First, by increasing time constant of excitation through NMDA receptors, the window for summation is enlarged, which facilitates the recruitment process (Herron et al., 1985; Jones and Heinemann, 1988). Second, it has been theoretically demonstrated that slow excitatory synapses favor sustained network activity (Wang, 1999). This would allow detection of formerly hidden functional circuits. Additionally, in the zero Mg²⁺ model (1) recurrent inhibition is not directly impaired (Ling and Benardo, 1995), and (2) NMDA receptors contribute a small tonic depolarizing current that inject firing into the network (LoTurco et al., 1990; Myme et al., 2003). We also tested the effect of increasing cell excitability under 5 mm K⁺ and the possible developmental changes.

We found that focal activity is sculpted by the dynamic interactions between some bursting cells leading other bursting cells, RS cells, and FS interneurons. Because the activation of FS inter-neurons is time-locked to the firing of leader cells, inhibition is always adjusted by excitation. This is consistent with previous studies in the subiculum using extracellular stimulation (Finch et al., 1988; Taube, 1993; Behr et al., 1998; Menendez de la Prida, 2003). Follower bursting and RS cells generate a secondary front of excitation that could further increase cell firing and compensate with additional inhibition. As a result, local ensembles of active cells develop. Ensemble activity is very variable, suggesting that it does not always involve the same cellular groups. Using cell-attached recordings, we showed that a given cell variably participates in different focal events and that the definition of leader/follower has to be based on statistical average.

A classical theoretical constraint for the generation of focal activity is short-range excitation and long-range inhibition (Hartline et al., 1956; Gutkin et al., 2001). No morphological data are yet available to draw the picture in the subiculum, although broad excitatory arbors have been described (Berger et al., 1980; Finch et al., 1983; Harris et al., 2001). Computer simulations have shown that, even with excitation wider than inhibition, focal activity could still develop (Xing and Gerstein, 1996; Kang et al., 2003). Under this configuration, however, this activity could easily become unstable and spread throughout (Golomb and Amitai, 1997).

This picture is in agreement with our results: focal activity could generate widespread population activity. This transition is mainly controlled by fast GABAergic transmission. During the first minutes under PTX, cross-correlation significantly increased and widespread events emerged more frequently. Transition from focal to widespread activity is also regulated by glutamatergic components because CNQX decreased the frequency of focal and widespread activity in parallel. Possible factors underlying this transition include runaway excitation driven by follower glutamatergic cells. Because the circuit will compensate for that increase with more inhibition, additional mechanisms are needed to explain this transition. For example, it is possible that plasticity of synaptic activity at moderate and high-frequency firing could differentially affect glutamatergic and GABAergic synapses (Thomson and Deuchars, 1994; Galarreta and Hestrin, 1998), or significantly shift reversal potential for GABA_A-mediated chloride currents (Thompson and Gähwiler, 1989; Staley and Proctor, 1999).

Comparison with previous studies

Focal field activity in the subiculum differs from other types of correlated activity described in the neocortex and the hippocampus. First, subicular focal activity is local and non-oscillatory as revealed by cross-correlation analysis. This is in contrast to many types of correlated activity that typically show oscillatory and propagating patterns (Steriade et al., 1993; Sanchez-Vives and McCormick, 2000; Wu et al., 2001) or to sharp-wave activity (Chrobak and Buzsaki, 1996). Instead, subicular focal activity is similar to spatially organized cortical ensembles (Mao et al., 2001; Cossart et al., 2003), although on a shorter time scale. Second, during focal activity, bursting cells lead population activity, whereas FS interneurons constitute a follower group. Such a temporal correlation is preserved during the buildup of widespread epileptiform activity. This is an important computational difference in comparison with models of slow propagating activity in which inhibitory firing precedes excitatory firing (Golomb and Ermentrout, 2002; Compte et al., 2003). Finally, repetitions of a precise firing pattern are not a feature of subicular focal activity (Cossart et al., 2003). Instead, the configuration of focal ensembles greatly varies with cells being recruited dynamically.

The firing correlation between driver and follower cells defines a succession of excitation and inhibition. This sequence is common in many preparations and represents the fact that local inhibition is secondarily recruited by excitatory circuits (Buzsaki, 1984). In contrast to other situations, in which excitation and inhibition were balanced continuously (Gutkin et al., 2001; Shu et al., 2003), subicular focal activity exhibited an initial dominance of excitation and a late prevalence of inhibition. Our analysis of synaptic conductances associated with focal activity revealed that conductance peaks were related with a slightly depolarized potential that subsequently hyperpolarized. This strongly suggests that the largest conductance increases (100-400%) are dominated by GABA-mediated currents, which in concomitance with a net hyperpolarizing driving force would shunt excitatory inputs and suppress cell firing. This mechanism is similar to that reported in the visual (Borg-Graham et al., 1998; Hirsch et al., 1998; Anderson et al., 2000) and auditory cortices (Zhang et al., 2003; Wehr and Zador, 2003).

The role of bursting and RS cells

The leading role of bursting cells has been reported in different acute preparations (Miles and Wong, 1983; Changac-Amitai and Connors, 1989; Jensen and Yaari, 1997) and in chronic epileptic tissue (Sanabria et al., 2001; Cohen et al., 2002). Indeed, it has been shown that a permanent upregulation of intrinsic bursting contributes to hyperexcitability in chronic epilepsy (Wellmer et al., 2002). In contrast to single spikes, bursts may enhance the security of spike transmission, depending on the release probability and the facilitating nature of synapses (Miles and Wong, 1986; Thomson and Deuchars, 1994). Our results support a role of some bursting cells leading field activity, in agreement with previous reports in the subiculum (Harris and Stewart, 2001). However, we approached this problem using cell-attached recordings that do not alter the firing capability of cells. In addition, we examined the firing patterns in statistical terms and evaluated their impact in local firing control.

We found that bursting cells are not a leader population per se, although many bursting cells behaved like leaders (σ > 0.5). In contrast, RS cells mostly behaved as followers. However, focal and widespread activities constitute a probabilistic process, and a given cell could lead or follow the field during different episodes. Therefore, it is the cooperative interactions between different cell types that generate a population event. The positive correlation between the driving index and the firing threshold and the number of distal apical dendrites suggests that leader cells would be those that are more easily recruited. Also, cells having a statistical leader role were of the IB+ and IB-bursting type, which confirms the critical contribution of these firing patterns in recurrent activity (Miles and Wong, 1983). Therefore, bursting subicular cells having more apical dendrites better contribute to collect and distribute activity through the network. The distal dendrites of these neurons can be also the site of the initiation of regenerative calcium spikes, which favor cell recruitment (Yuste et al., 1994; Kamondi et al., 1998; Larkum et al., 1999).

The role of followers in this scheme should not be considered secondary. This is particularly the case of RS cells, which displayed a low firing rate and high firing threshold. This determines that RS cells were sharply recruited and, therefore, the contrast between response and no-response is enhanced at the onset of field activity. However, RS cells remained silent during some episodes of focal activity, which possibly leaves them available for being recruited by other ensembles. In contrast, because follower cells inject firing into the circuit at the onset of field activity, they can participate in runaway excitation.