Factors determining the efficacy of distal excitatory synapses in rat hippocampal CA1 pyramidal neurones

Abstract

1. A new preparation of the in vitro rat hippocampal slice has been developed in which the synaptic input to the distal apical dendrites of CA1 pyramidal neurones is isolated. This has been used to investigate the properties of distally evoked synaptic potentials.

2. Distal paired-pulse stimulation (0.1 Hz) evoked a dendritic response consisting of a pair of EPSPs, which showed facilitation. The first EPSP had a rise time (10–90 %) of 2.2 ± 0.05 ms and a half-width of 9.1 ± 0.13 ms. The EPSPs were greatly reduced by CNQX (10 μm) and the remaining component could be enhanced in Mg²⁺-free Ringer solution and blocked by AP5 (50 μm). In 70 % of the dendrites, the EPSPs were followed by a prolonged after-hyperpolarizarion (AHP) which could be blocked by a selective and potent GABA_B antagonist, CGP 55845A (2 μm). These results indicate that the EPSPs are primarily mediated by non-NMDA receptors with a small contribution from NMDA receptors, whereas the AHP is a GABA_B receptor-mediated slow IPSP.

3. With intrasomatic recordings, the rise time of proximally generated EPSPs (3.4 ± 0.1 ms) was half that of distally generated EPSPs (6.7 ± 0.5 ms), whereas the half-widths were similar (19.6 ± 0.8 ms and 23.8 ± 1 ms, respectively). These results indicate that propagation through the proximal apical dendrites slows the time-to-peak of distally generated EPSPs.

4. Distal stimulation evoked spikes in 60 % of pyramidal neurones. At threshold, the distally evoked spike always appeared on the decaying phase of the dendritic EPSP, indicating that the spike is initiated at some distance proximal to the dendritic recording site. Furthermore, distally and proximally generated threshold spikes had a similar voltage dependency. These results therefore suggest that distally generated threshold spikes are primarily initiated at the initial segment.

5. At threshold, spikes generated by stimulation of distal synapses arose from the decaying phase of the dendritic EPSPs with a latency determined by the time course of the EPSP at the spike initiation zone. With maximal stimulation, however, the spikes arose directly from the peak of the EPSPs with a time-to-spike similar to the time-to-peak of subthreshold dendritic EPSPs. Functionally, this means that the effect of dendritic propagation can be overcome by recruiting more synapses, thereby ensuring a faster response time to distal synaptic inputs.

6. In 42 % of the neurones in which distal EPSPs evoked spikes, the relationship between EPSP amplitude and spike latency could be accounted for by a constant dendritic modulation of the EPSP. In the remaining 58 %, the change in latency was greater than can be accounted for by a constant dendritic influence. This additional change in latency is best explained by a sudden shift in the spike initiation zone to the proximal dendrites. This would explain the delay observed between the action of somatic application of TTX (10 μm) on antidromically evoked spikes and distally evoked suprathreshold spikes.

7. The present results indicate that full compensation for the electrotonic properties of the main proximal dendrites is not achieved despite the presence of Na⁺ and Ca²⁺ currents. Nevertheless, distal excitatory synapses are capable of initiating spiking in most pyramidal neurones, and changes in EPSP amplitude can modulate the spike latency. Furthermore, even though the primary spike initiation zone is in the initial segment, the results suggest that it can move into the proximal apical dendrites under physiological conditions, which has the effect of further shortening the response time to distal excitatory synaptic inputs.

The excitatory synaptic activation of hippocampal CA1 pyramidal neurones is primarily mediated by pyramidal neurones in the ipsilateral and contralateral CA3 region, axons of which form the Schaffer collateral and commissural pathways, respectively. Terminals of these fibres make synaptic contacts on most of the highly arborized dendritic tree of the CA1 pyramidal neurones (Andersen & Lømo, 1966). A second minor excitatory input comes from neurones in layer III of the entorhinal cortex via the entorhinal pathway, which is thought to form synaptic contacts exclusively on the distal apical dendrites (Colbert & Levy, 1992).

Theoretical analyses have indicated that the anatomical and electrotonic structure of the dendritic tree results in a marked attenuation of distally generated excitatory postsynaptic potentials (EPSPs) and alters their time course and efficacy, in terms of their ability to evoke spikes (Turner & Schwartzkroin, 1980; Mainen, Carnevale, Zador, Claiborne & Brown, 1996). However, experimental investigations of the influence of dendritic propagation on EPSP efficacy have so far not been conclusive. In one study, no difference was found between distally and proximally generated EPSPs, indicating that dendritic propagation has no effect on distal EPSPs (Andersen, Silfvenius, Sundberg & Sveen, 1980). In another study, distally generated EPSPs were found to be about twice as slow as proximally generated EPSPs (Turner, 1988). This difference was largest for small EPSPs (< 0.5 mV) and decreased as the amplitude of the EPSPs was increased, indicating that there may be an amplitude-dependent compensation for dendritic propagation. Recently, however, dual patch-clamp recordings from the soma and apical dendrites of the same pyramidal neurone have indicated that dendritic propagation also affects threshold EPSPs (see Fig. 1A in Spruston, Schiller, Stuart & Sakmann, 1995).

The reasons for these different findings and discrepancies between experimental and theoretical investigations are presently not clear. One explanation could be that the theoretical analyses are usually based on the assumption that the dendrites are passive structures, which is now known not to be the case (for reviews see Johnston, Magee, Colbert & Christie, 1996; Stuart, Spruston, Sakmann & Häusser, 1997). There is now evidence that dendritic currents are involved in the amplification of distally evoked EPSPs (Magee & Johnston, 1995a; Lipowsky, Gillessen & Alzheimer, 1996; Gillessen & Alzheimer, 1997). However, the activation of these currents seems to enhance the amplitude of the EPSPs without changing the overall time course of distally evoked EPSPs at the somatic level (Lipowsky et al. 1996; Gillessen & Alzheimer, 1997). It therefore seems unlikely that the activation of dendritic currents can fully explain the different experimental findings and the discrepancies between theory and experimental observations. Another possibility concerns the dendritic location of the synapses. It is generally assumed that the Schaffer collateral fibres follow a laminar course parallel to the stratum pyramidale (SP; Andersen et al. 1980). However, Turner (1988) suggested that the observed unification of distally and proximally generated EPSPs when the stimulation was increased could be the result of a progressive activation of non-laminar afferents which form synapses in overlapping regions of the apical dendrites. Recent morphological studies of single CA3 pyramidal neurones have supported this by showing that Schaffer collateral fibres deviate substantially from a laminar course and that they often have an extensive system of collaterals within the CA1 region (Li, Somogyi, Ylinen & Buzsáki, 1994). It is therefore possible that the similar time course and efficacy of distally and proximally generated EPSPs reported by Andersen et al. (1980) results from stimulation of non-laminar fibres which activate synapses at more proximal locations, resulting in an overestimation of the efficacy of distally generated EPSPs.

Another unresolved issue concerns the initiation of spikes by synaptic activation. Intradendritic recordings have indicated that apical dendrites of CA1 pyramidal neurones are capable of generating Na⁺- and Ca²⁺-dependent spikes (for review see Stuart et al. 1997). Extracellular analyses have suggested that spikes evoked by distally generated synaptic potentials are primarily initiated in the proximal apical dendrites (Andersen & Lømo, 1966; Herreras, 1990). However, intracellular and patch-clamp recordings have indicated that the site of initiation of synaptically evoked spikes is located primarily in the initial segment, although it can shift to the proximal dendrites under extreme conditions (Turner, Meyers, Richardson & Barker, 1991; Spruston et al. 1995).

Therefore, some basic questions concerning the impact of dendritic propagation and synaptic spike initiation in CA1 pyramidal neurones still remain unresolved. (1) How does dendritic propagation affect distally generated EPSPs? (2) Is there a relationship between EPSP amplitude and dendritic modulation? (3) Where is the spike initiated? (4) Is there a shift in the initiation zone and, if so, under what conditions does it occur? To address these questions, we have developed a new experimental preparation, which provides a more complete isolation of the synaptic input to the distal parts of the hippocampal CA1 apical dendrites. This preparation has enabled us to assess the effect of dendritic propagation on distally evoked EPSPs. Moreover, the preparation has allowed us to gain insight into the spike generating capabilities of distally located excitatory synapses in mature animals.

METHODS

Experiments were performed on hippocampal slices prepared from sixty-four male Wistar rats (250–300 g). Each rat was anaesthetized with chloroform, after which it was decapitated and the brain quickly removed and placed in a standard Ringer solution (see below) at 4°C. The hippocampus was dissected free and slices (400 μm thick) were cut on a McIlwain tissue chopper. The slices were immediately transferred to the recording chamber, where they were placed on a nylon-mesh grid at the interface between warm (31–33°C) standard Ringer solution (pH 7.3) and warm humidified Carbogen (95 % O₂-5 % CO₂). Perfusion flow rate was 1 ml min⁻¹.

Isolation of the distal synaptic input to CA1 pyramidal neurones

The afferent input to the distal third of the apical dendrites of a group of CA1 pyramidal neurones was isolated by a double cut in stratum radiatum (SR) 250–300 μm from the superior border of SP. The cuts left a small ‘bridge’ approximately 200 μm wide at right angles to the dendritic axis (Fig. 1A and B). The cut was made with a custom-made knife consisting of two razor blade chips mounted together in the same plane but separated by a gap of 200 μm. The knife was mounted on a micromanipulator and lowered into the slice parallel to a small region of SP in area CA1b. Because this procedure always left some uncut fibres, the incisions on each side of the ‘bridge’ were completed using a microdissection knife (Fine Science Tools Inc., Heidelberg, Germany). This procedure ensured that only the apical dendrites of pyramidal neurones whose somata were located directly above the ‘bridge’ will extend into the superficial part of SR and stratum lacunosum-moleculare (L-M) (Fig. 1B). Structures near the edges of the ‘bridge’ are likely to have been damaged by compression with the knife thereby decreasing the functional width of the ‘bridge’ to less than 200 μm. However, the lateral extension of the dendritic tree at the level of the ‘bridge’ is usually less than 200 μm (Bannister & Larkman, 1995), so the dendritic tree of a large number of pyramidal neurones would be expected to be intact.

Figure 1. Isolation of the distal synaptic inputs.

A, schematic illustration of the slice preparation and the electrode placement for testing the effectiveness of the isolation of distal synaptic inputs. In each slice, two cuts were made which left a small ‘bridge’ in SR. B shows a drawing of the region containing the ‘bridge’ with a schematic drawing of a CA1 pyramidal neurone superimposed (A/O, alveus/oriens; SP, stratum pyramidale; SR, stratum radiatum; L-M, stratum lacunosum-moleculare). C, extracellular field responses, recorded at locations I-III in A, to stimulation of A/O (Stim. 1) and afferent fibres near the SR/L-M border (Stim. 2). At all locations, stimulation of the A/O evoked an antidromic spike (arrow) followed by an orthodromic field EPSP and a population spike. However, stimulation of the distal afferent fibres (Stim. 2) only evoked a response at location II opposite the ‘bridge’. The response consisted of a field EPSP with a small inflection indicative of a population spike.

Following the microdissection, the slice was allowed to rest for at least 1 h before recordings were started.

Stimulation and recording procedure

Intracellular recordings from CA1 pyramidal neurones were obtained using borosilicate glass microelectrodes (1.2 mm o.d., Clark Electromedical) filled with 4 M K⁺ acetate (tip resistances: 60–80 MΩ). Penetrations of the distal apical dendrites were made at the centre of the ‘bridge’ as indicated in Fig. 2A. Intradendritic recordings were identified by their similarity to those reported previously from histochemically verified dendritic recordings (Andreasen & Lambert, 1995). Dendritic penetrations were accepted for analysis if the resting membrane potential (RMP) was stable and more negative than -50 mV and the membrane input resistance (R_in) was ≥ 10 MΩ. Intrasomatic recordings were obtained from SP directly above the ‘bridge’ as indicated in Fig. 9A. The criteria for accepting intrasomatic recordings were similar to those used earlier (Andreasen, Lambert & Jensen, 1989). Extracellular recordings were obtained using 1 M NaCl-filled glass microelectrodes (tip resistances: 15–30 MΩ).

Figure 2. Dendritic responses to distal synaptic activation.

A, schematic illustration of the experimental setup for intradendritic recordings of synaptic responses. B, example of a typical dendritic response to distal paired-pulse stimulation (300 μA, interpulse interval 50 ms) recorded about 275 μm distally to the superficial border of SP. The response, consists of a cEPSP and a tEPSP, the latter being facilitation and followed by a prolonged AHP. Unless otherwise noted, the responses in this and the following figures are the mean of 4 to 5 individual recordings. C shows the dendritic response to distal stimulation of increasing intensity. PPF is present at each of the intensities illustrated and there is a progressive increase in peak amplitude of the EPSPs until spikes are initiated. Note that in this dendrite, the EPSPs were not followed by an AHP. Four individual responses to 250 μA are shown superimposed while the other traces are means. D, the spike generating tEPSPs in response to 250 μA on an expanded time scale. Note the spikes are evoked on the decaying phase of the EPSP at variable latency. RMP: in B, - 64 mV; and in C and D, -71 mV.

Figure 9. Somatic recordings of distally evoked EPSPs.

A, schematic illustration of the experimental setup for somatic recordings of distally and proximally evoked EPSPs. To the right is shown the responses to distal stimulation of increasing intensity in the presence of BIC (10 μm), CGP 55845A (2 μm) and AP5 (50 μm). B, distally evoked EPSPs recorded in a dendrite (Distal dendritic, RMP -68 mV) and a soma (Distal somatic, RMP -69 mV) of two different pyramidal neurones. The responses are superimposed to the right, to highlight the difference in time course. C, somatic recordings in response to proximal (Proximal somatic) and distal (Distal somatic) threshold-straddling stimulation (95 μA and 500 μA, respectively). The subthreshold EPSPs (top) showed that the rising phase of the proximally evoked EPSP is faster than that of the distally evoked EPSP, whereas the decaying phases are similar. This was also the case for the threshold EPSP (bottom). The thresholds for the two spikes (marked by arrows) were, however, similar. All spikes have been truncated. The somatic recordings in A, B and C are from the same pyramidal neurone.

Bipolar Teflon-insulated platinum electrodes (50 μm in diameter) were used for orthodromic and antidromic stimulation of the CA1 pyramidal cells with constant-current pulses (50–500 μA; 50 μs duration; 0.1 Hz). In order to activate the distal afferent fibres, a stimulation electrode was placed on the slice close to the border between SR and L-M on the subicular side of the ‘bridge’ (Stim. 2 in Fig. 1A). This placement minimized the activation of non-laminar afferents. For stimulation of the proximal afferent fibres close to the superficial border of SP, a monopolar electrode consisting of a blunt glass pipette (tip diameter 10–20 μm) filled with 1 % agar dissolved in 0.9 % NaCl was used. For antidromic activation, a bipolar stimulation electrode was placed in SP (Fig. 6C). Unless otherwise noted, a paired-pulse stimulation protocol was used with an interpulse interval of 50 or 100 ms. The two responses to paired-pulse stimulation are termed the conditioning (c) and test (t) response, respectively.

Figure 6. Properties of distally evoked dendritic spikes.

Aa, dendritic recordings of individual threshold responses from two different dendrites. With threshold-straddling distal stimulation, spikes always appeared on the decaying phase of the EPSP. A typical example is shown to the left (spike latency from peak: 3.3 ms) and an extreme example to the right (spike latency: 18 ms). Ab, dendritic recording of a proximally evoked threshold spike which rides on the peak of the EPSP. B, superimposed dendritic responses to distal stimulation at the same intensity. Note the variation in peak amplitude of the EPSP and spike latency. C, recordings from the same dendrite as in B. Ca, superimposition of two dendritic responses evoked with the same stimulating intensity. In each case a single fast spike was initiated, though one appeared on the decaying phase whereas the other arose from what appeared to be a prolonged peak. Cb, the prolongation of the EPSP peak (arrow) becomes more evident when superimposed on a subthreshold EPSP of similar amplitude. RMP: in Aa, -71 and -68 mV; in Ab, -70 mV; and in B and C, -68 mV.

Conventional recording techniques were employed, using a high input impedance amplifier (Axoclamp 2A, Axon Instruments Inc.) with bridge-balance and current injection facilities. Results were digitized on-line using a Labmaster A/D converter and pCLAMP acquisition software (Axon Instruments Inc.) on a 486 PC computer and recorded for off-line analysis using a modified digital audio processor (Sony PCM-701es) and a video tape recorder.

EPSP amplitude, rise time (10–90 %) and half-width were measured with respect to the pre-stimulus baseline. Time-to-peak and time-to-spike were measured from the beginning of the EPSP (i.e. the first deviation of the voltage from the baseline following the stimulation). Spike amplitude was measured from the first inflection point to the peak. All analyses were performed using pCLAMP analyses software.

Values are given as means ± s.e.m. unless otherwise indicated. For statistical analyses, Student's t test was used as appropriate.

Drugs and solutions

The composition of the standard Ringer solution was (mm): NaCl, 124; KCl, 3.25; NaH₂PO₄, 1.25; NaHCO₃, 20; CaCl₂, 2; MgSO₄, 2; D-glucose, 10; bubbled with Carbogen (pH 7.3). Some of the experiments were performed in the presence of dl-2-amino-5-phosphonovaleric acid (AP5, 50 μm), bicuculline methobromide (BIC, 10 μm) and CGP 55845A (2 μm) in order to block N-methyl-D-aspartate (NMDA), γ-aminobutyric acid (GABA)_A and GABA_B receptors, respectively. In some experiments, nominally Mg²⁺-free Ringer solution was used.

For local application of tetrodotoxin (TTX), a blunt glass pipette (o.d. 10 μm), was filled with standard Ringer solution containing 10 μm TTX and connected to a pressure injection system (PV830 Pneumatic PicoPump, WPI). The tip of the pipette was placed on the surface of the slice close to, and upstream to, the site of interest, thereby ensuring that most of the TTX would be washed away by the continuous flow of Ringer solution. TTX was applied by a short pressure pulse of 100–300 ms (10–20 psi) after which the TTX-containing pipette was withdrawn.

All pharmacological compounds were made up in aqueous stock solutions of 100–1000 times the required final concentration and diluted in standard Ringer solution as appropriate. TTX and BIC were purchased from Sigma, AP5 and 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) were purchased from Tocris and CGP 55845A was provided by Ciba Geigy.

RESULTS

Isolation of the distal synaptic input to CA1 pyramidal neurones

The synaptic input to the distal part of the apical dendrites was isolated as described in Methods and the efficacy of this isolation was tested experimentally in twenty-two slices as outlined in Fig. 1A. A bipolar stimulation electrode (Stim. 2) was placed close to the border between SR and L-M on the subicular side of CA1. Stimulation here will activate the Schaffer collaterals in the superficial portion of SR, and, to a certain extent, the entorhinal pathway in L-M. Responses were recorded extracellularly in SP at three different locations: I, located approximately 200 μm from the subicular edge of the ‘bridge’; II, corresponded to the middle of the ‘bridge’; III, approximately 200 μm from the CA3 edge of the ‘bridge’. Stimulation of the alveus/oriens (A/O) area (Stim. 1) was used to control for the viability of the CA1 region.

At all recording sites, stimulation of the A/O evoked a complex field potential (Fig. 1C) consisting of an initial antidromic population spike (PS) followed by a field EPSP on which an orthodromic PS of variable amplitude was riding. This shows that there were viable pyramidal neurones in all three areas. In the example shown in Fig. 1C, high intensity distal stimulation evoked a response at location II consisting of a field EPSP on top of which is a small PS. The lack of response at locations I and III indicates that the cuts had effectively separated the superficial SR and L-M from the rest of area CA1, and that few, if any, non-laminar fibres run through the ‘bridge’ into the deeper portion of SR. Although this was the case for the majority of slices tested, in a few instances very high intensity stimulation (800- 1000 μA) could evoke a small response at location III.

The passive membrane properties of CA1 pyramidal neurones

Intracellular recordings were obtained from seventy-six apical dendrites (see Methods). The mean RMP was -69.1 ± 0.3 mV (range -74 to -63.5 mV, n = 76). The input resistance (R_in) and the membrane time constant (τ_m) were determined from the response to a small hyperpolarizing current pulse as described earlier (Andreasen & Lambert, 1995). The mean R_in was 17.5 ± 0.5 MΩ (range 10 to 28.7 MΩ, n = 76) and τ_m was 5.8 ± 0.6 ms (range 3 to 10.8 ms, n = 48). In the remaining twenty-eight dendrites the presence of an inwardly rectifying current prevented determination of τ_m. All the basic membrane parameters of the apical dendrites were similar to those we have reported previously (Andreasen & Lambert, 1995).

Intracellular recordings were obtained from twenty-nine somata in SP, with a mean RMP of -69.4 ± 0.3 mV (range -73 to -64 mV, n = 29) and R_in of 33.5 ± 1.3 MΩ (range 21 to 50 MΩ, n = 29). The somatic voltage response to a hyperpolarizing current pulse was well described by a single exponential function giving a mean τ_m of 14.7 ± 1.1 ms (range 6.8 to 32.9 ms), which was nearly threefold slower than the dendritic τ_m.

Distally generated synaptic potentials recorded in the apical dendrites

In order to investigate the effect of dendritic propagation on distally generated EPSPs, it is necessary to characterize these at the dendritic level. Dendritically recorded synaptic potentials (Fig. 2A) result from the summation of unitary synaptic potentials generated on parts of the apical dendrites which are distal to the ‘bridge’ (Fig. 1B). In standard Ringer solution, distal paired-pulse stimulation evoked two fast EPSPs with marked paired-pulse facilitation (PPF) of the tEPSP (Fig. 2B). Although no detailed analysis of PPF was carried out, maximal PPF was seen to occur with interpulse intervals between 25 and 100 ms and decreased towards zero at intervals of 150 to 200 ms (n = 13, of which 8 were in the presence of 10 μm BIC, which did not affect the time course of PPF). Increasing the stimulating intensity resulted in a progressive increase in the amplitude of both the c- and tEPSP until a maximal amplitude was reached (Fig. 2C). When spikes were evoked, they initially appeared on the decaying phase of the EPSP and at a variable latency from the peak (Fig. 2C and D). In most cases, it was not possible to obtain acceptable exponential fits of the decaying phase of the EPSPs. The time course of the EPSPs (≥ 1 mV) was therefore represented by measuring the half-width and the rise time (10–90 %). Because there was no correlation between peak amplitude and rise time or half-width of the cEPSPs (or the tEPSP) (Fig. 3) all control data were analysed collectively, irrespective of their amplitude. The mean rise time of the cEPSPs and tEPSPs were 2.2 ± 0.05 and 2.1 ± 0.04 ms, respectively, (n = 23) while the half-widths were 9.1 ± 0.13 and 9.6 ± 0.13 ms, respectively, (n = 23).

Figure 3. Relationship between peak amplitude and time course.

There is no correlation between rise time (10–90 %) and peak amplitude (A), or between half-width and peak amplitude (B) of dendritic EPSPs (> = 1 mV) recorded in standard Ringer solution. 129 individual cEPSPs were included in the above plots.

The dendritic EPSP was greatly reduced by 10 μm CNQX (n = 5), leaving only a small residual EPSP of a few millivolts (Fig. 4A), which was enhanced in nominally Mg²⁺-free medium (Fig. 4B, n = 4). Note that the residual EPSP also showed PPF, particularly in Mg²⁺-free medium. In the presence of both CNQX and AP5 (50 μm), the EPSP was totally blocked (Fig. 4C).

Figure 4. The glutamatergic component of the dendritic synaptic potential.

A, the non-NMDA receptor antagonist CNQX (10 μm) greatly reduced the dendritic EPSPs leaving only a small component of 1–2 mV in amplitude. B, in another dendrite the CNQX-resistant component was greatly potentiated by perfusion with nominally Mg²⁺-free Ringer solution. C, addition of both CNQX (10 μm) and the NMDA-receptor antagonist AP5 (50 μm) completely blocked the dendritic EPSP. RMP: in A, -74 mV; in B, -66 mV; and in C, -71 mV.

None of the dendritic recordings showed any evidence of a fast inhibitory postsynaptic potential (IPSP) at RMP (Fig. 2B and C). Furthermore, changing the membrane potential (V_m) did not disclose a fast IPSP (n = 12, Fig. 5A). In the presence of BIC, however, the rise time (2.6 ± 0.1 ms, n = 22) and half-width (11.7 ± 0.3 ms) were significantly different from control values (P < 0.005 and P < 0.001, respectively), indicating that GABA_A receptors are activated to some extent.

Figure 5. The GABAergic component of the dendritic synaptic potential.

A, the voltage dependency of the distally evoked dendritic response. Before afferent stimulation V_m was stepped to different potentials by injecting 200 ms duration current pulses (0, ± 0.4, ± 0.8 nA). The amplitude of the EPSPs decreased with depolarization and increased with hyperpolarization, but there was no indication of a fast IPSP. B, response to paired-pulse stimulation at two intensities. C, in the same dendrite, application of the selective GABA_B-receptor antagonist CGP 55845A (2 μm) completely blocked the AHP and, in this case, also reduced the peak amplitude of the EPSPs. D, in another dendrite, the AHP was replaced by a prolonged depolarizing afterpotential in the presence of CGP 55845A. Note the very marked increase in peak amplitude of both EPSPs in the presence of CGP 55845A. RMP: in A, -71 mV; in B and C, -71 mV; and in D, -63 mV.

In 70 % (16/23) of dendritic recordings the EPSPs were followed by a prolonged afterhyperpolarization (AHP) of a few millivolts in amplitude and lasting several hundred milleseconds (Fig. 2B). The amplitude of the AHP increased with stimulating intensity (Fig. 5B). The AHP was completely blocked by 2 μm CGP 55845A (n = 6, Fig. 5C and D), a selective and potent GABA_B receptor antagonist, which has been shown to completely block both pre- and postsynaptic GABA_B receptors at concentrations below 2 μm (Davies, Pozza & Collingridge, 1993). The effect of CGP 55845A was, however, complex. In three cases, the EPSP decayed back to baseline following the block of the AHP. In the remaining three cases, the EPSP was followed by a prolonged depolarizing afterpotential (Fig. 5D). Furthermore, the amplitude of the EPSP was reduced in four cases, with the greatest effect being on the tEPSP, resulting in a concomitant reduction in PPF (Fig. 5C). In the remaining two cases, CGP 55845A induced a large increase in the amplitude of both c- and tEPSPs without any significant change in half-width (Fig. 5D). In the presence of both BIC and CGP 55845A, the mean rise time and half-width of the cEPSP was 2.6 ± 0.09 ms and 11.1 ± 0.3 ms, respectively (n = 16), which is similar to EPSPs in the presence of BIC alone.

The spike generating properties of distally evoked EPSPs

The spike generating properties of the distal dendrites were unaltered in the presence of BIC, CGP 55845A and AP5, either individually or in different combinations. Therefore, all data concerning spike generation have been analysed collectively. Of the sixty-five dendrites in which the stimulus- response correlation was investigated, 60 % (39/65) showed spike firing. In twenty-two of these cases, spikes were only evoked by the tEPSP, even though the amplitude of the cEPSP in many instances exceeded the threshold level for spiking. In the remaining seventeen cases, spikes were also evoked by the cEPSP. In most cases the spikes evoked were single or double fast spikes, with a mean amplitude of 51.5 ± 0.6 mV (range 34.6 to 84 mV, n = 35) and were similar in all respects to the Na⁺-dependent spikes evoked by intradendritic current injection (Andreasen & Lambert, 1995). Compound spiking, which consists of an initial fast spike followed by one or more broad Ca²⁺-dependent spikes (Andreasen & Lambert, 1995) was only evoked in four cases (Fig. 10). The threshold, defined here as the absolute V_m at the peak of EPSPs which intermittently evoked spikes, was -54.1 ± 1 mV (range -61.7 to -44 mV, n = 21). There was no correlation between the dendritic RMP and the threshold level (not shown). In the 40 % (26/65) of dendrites in which stimulation of the distal synapses did not evoke spikes, V_m at the peak of the tEPSP nevertheless reached -55.9 ± 1.1 mV (range -66.8 to -44.8 mV, n = 26). In all cases, however, spikes could be evoked by intradendritic current injection.

Figure 10. Analyses of dendritic spike latency.

A, the initial phase of a dendritic threshold response evoked by distal stimulation. The inset shows the full course of the response with compound spiking. The time-to-peak (t_p) and time-to-spike (t_s) were both measured from the first inflection at the beginning of the EPSP. V_m at the peak of the EPSP (V_Thr) was also measured and represents the threshold level. B, the ratio between t_s and t_p plotted as a function of V_Thr. A ratio of 1.0 (dashed line) indicates no change in the rising phase of the EPSP at the spike initiation zone. Values > 1.0 indicate a slowing of the rising phase at the spike initiation zone. Note that, with one exception (ratio 6.4), the ratios are independent of the threshold level. C, superimposition of two maximal responses with compound spiking from the same dendrite as in A. For maximal responses, only t_s was measured. D, the relationship between t_s and t_p of subthreshold EPSPs (n = 11). The dotted line shows a correlation of 1.0 between t_s and t_p. Except for one case, the points are clustered around the dotted line. RMP: in A and C, -68 mV.

At threshold, spikes always appeared on the decaying phase of the EPSPs (Figs 2D and 6Aa). Spike latency, from the peak of the EPSP, was found to be independent of the threshold level, and could be the same whether this was, for example, -62 or -44 mV (Fig. 10B). However, at a given threshold, there was some variation in spike latency between different dendrites. This is illustrated in Fig. 6Aa, which shows a typical (left) and an extreme example (right). In some dendrites, small variations in amplitude of suprathreshold EPSPs were associated with a marked change in the spike latency (Fig. 6B) while in others, the spike latency varied noticably, even though the peak amplitude was unchanged (Fig. 2D).

In eight dendrites, the peaks of some of the EPSPs evoked by threshold-straddling stimulation were markedly prolonged (Fig. 6Ca). This prolongation of the EPSP was associated with a reduction in threshold. This is highlighted in Fig. 6Cb, in which a subthreshold EPSP is superimposed on a prolonged threshold EPSP. EPSPs with prolonged peaks were also seen in the presence of AP5 (not shown), indicating that the prolongation was not caused by activation of NMDA receptors.

The efficacy of distally evoked EPSPs in generating spikes was very sensitive to changes in the dendritic V_m. At threshold, a hyperpolarization of only -3.7 ± 0.37 mV (n = 10) was sufficient to completely block spike generation. For suprathreshold responses, hyperpolarization of the dendritic membrane by a few millivolts increased the spike latency and changed the EPSP to a threshold response. Further hyperpolarization completely blocked spike generation (Fig. 7A). The amplitude of the EPSPs was relatively insensitive to the small changes in dendritic V_m. This could be due to the small space constant of the distal part of the apical dendrites (Andreasen & Nedergaard, 1996) which would limit current spread in the distal direction from the recording site. Threshold spikes evoked by distal and proximal stimulation (see Methods) showed similar sensitivity to dendritic hyperpolarization (Fig. 7B).

Figure 7. Voltage-dependency of distally evoked dendritic spikes.

A, individual dendritic recordings showing the effect of changing the dendritic V_m on the spike generating properties of a distally evoked EPSP. RMP is the top trace. Modest hyperpolarization (-2.5 mV) increased the spike latency while larger hyperpolarization (-5.3 mV) blocked spike generation. Note that the amplitude of the EPSP is little effected by the hyperpolarization. B, the effect of hyperpolarization on distally (D) and proximally (P) evoked threshold responses in a different dendrite. A small hyperpolarization of only -1.5 mV was sufficient to prevent both EPSPs from generating spikes.

In contrast to distally evoked spikes, dendritic recordings of proximally evoked spikes showed that spikes always arose from the peak of the EPSPs (Figs 6Ab and 7B). The spike initiation zone for proximally evoked spikes is likely to be the initial segment, which is only a short distance from the active synapses. That distally evoked threshold spikes appeared after the peak of the EPSP would indicate that they are initiated at some distance proximal to the active synapses. Simultaneous patch-clamp recordings from dendrites and somata from young animals have recently led to the suggestion that synaptically evoked spikes are primarily initiated in the initial segment and then propagated back into the dendrites (Stuart & Sakmann, 1994; Spruston et al. 1995). To test whether the distally evoked dendritic spikes are somatic in origin, we conducted a series of experiments (n = 5) using local application of TTX (10 μm) to the somatic region (see Methods). Recordings were made from the distal dendrites and paired-pulse stimulation was used to activate the distal afferent fibres (Fig. 8A). The intensity was adjusted so that spikes were only evoked consistently by the tEPSP. The cEPSP therefore served as a control for diffusion of TTX into the area of active terminals. Antidromic stimulation of SP served as a control that TTX had reached the pyramidal neurone whose apical dendrite had been impaled. After several stable controls, a 300 ms pulse of TTX was applied onto the surface of the slice, following which the TTX-pipette was withdrawn. After 20 s, the antidromic spike had been blocked, but spikes could still be evoked by distal stimulation (Fig. 8B). After 180 s, however, the distally evoked spike was also blocked, revealing a tEPSP with an amplitude substantially greater than the spike threshold (Fig. 8B). At the same time, the cEPSP was unchanged, indicating that blockade of the distally evoked spike was not due to TTX diffusing into the recording area. Similar results were obtained in four other experiments. In all cases there was a delay (32 ± 11.1 s; n = 5) before the antidromic spike was blocked, which was about three times shorter than the delay between the block of the antidromic spike and the distally evoked spike (92 ± 24.6 s, range 20 to 160 s).

Figure 8. Distally evoked dendritic spikes are blocked by somatic application of TTX.

A, the experimental setup for local application of TTX (10 μm) during intradendritic recordings. One stimulation electrode was used to activate the distal afferent fibres (stim. D) and another was placed over SP to activate the pyramidal neurones antidromically (stim. A). The tip of the TTX-containing pipette was placed on the surface of the slice close to and downstream from the antidromic stimulation electrode. Flow direction is indicated to the right. Ba, control responses to paired-pulse stimulation of the distal afferent fibres (closed arrows) with an intensity that consistently generated spikes on the tEPSP. This was followed, 100 ms later, by a single antidromic stimulation (open arrow) at suprathreshold intensity. All spikes have been truncated. Eight control responses were collected to ensure their consistency. A single pulse (300 ms, 10–30 psi) of TTX was applied under visual control and the pipette immediately withdrawn from the slice. Bb, after 20 s, the antidromic response was completely blocked while the orthodromic response was still unaffected. Bc, after 180 s, the distally evoked spike was also blocked. Superimposition of the responses in Ba and Bc to the right shows that the cEPSP is unchanged while the tEPSP following TTX has an amplitude which clearly exceeds the pre-TTX threshold (dotted line). RMP: -67 mV.

Somatic recordings of distally generated EPSPs

We then used intrasomatic recordings to investigate how propagation through the proximal apical dendrites influences the time course and efficacy of distally generated EPSPs. To eliminate distortion due to variable activation of GABA and/or NMDA receptors, the experiments were carried out in the presence of CGP 55845A (2 μm), BIC (10 μm) and AP5 (50 μm). Under these conditions, distally generated EPSPs recorded at the somatic level had a very slow time course (Fig. 9A), which is in marked contrast to EPSPs recorded in dendrites under similar conditions (Fig. 9B). The mean rise time and half-width of the somatic EPSPs were 7.06 ± 0.4 and 25 ± 0.8 ms, respectively, (n = 18), which are significantly different from the rise time (2.4 ± 0.07 ms, n = 20, P < 0.001) and half-width (11.3 ± 0.3 ms, P < 0.001) of the dendritic EPSPs. As with the dendritic EPSPs, the amplitude of the somatically recorded EPSPs increased with the stimulating intensity (Fig. 9A). However, the change in amplitude was usually not very marked and reached a maximum value of 4.1 ± 0.5 mV (range 1.3 to 12.3 mV). Furthermore, at RMP, distal EPSPs evoked spikes in only 22 % (4/18) of the somatic recordings (Fig. 9A) compared with 60 % of the dendritic recordings. The small amplitude, slower time course and decreased safety factor for spike initiation of the distally generated EPSPs recorded at the somatic level indicates that propagation through the main proximal dendrites does indeed have a substantial impact on the EPSPs.

Penetration with a microelectrode will introduce a leak conductance which can have a substantial effect on the membrane properties (Spruston & Johnston, 1992) and therefore on the amplitude and time course of the EPSPs. Because the size of this leak conductance is likely to be different in dendritic and somatic recordings, a direct comparison between EPSPs recorded at the two sites is likely to be biased. Since the size of the leak conductance is not known, we cannot estimate the effect on the EPSPs or how it will bias the results. However, a somatic leak would be expected to affect synaptic potentials to a similar extent, irrespective of whether these are generated proximally or distally. Therefore, a direct comparison of EPSPs generated by proximal and distal synapses will provide a better estimate of the influence of dendritic propagation. One obvious, but important, assumption is that proximal and distal excitatory synapses are identical in all respects except for their location. Whether this is the case is currently unknown, but there is evidence that dendritic and somatic glutamate receptor-ionophore complexes are identical with respect to their kinetics and conductance (Spruston, Jonas & Sakmann, 1995). This means that the rising phases of the EPSPs in the vicinity of the synapses should be comparable. This is important since changes in the rising phase, together with amplitude changes, are sensitive indicators of electrotonic attenuation (Spruston, Jaffe & Johnston, 1994). EPSPs evoked by very localized proximal stimulation (Fig. 9A, see Methods) were compared with distally evoked EPSPs recorded in the same pyramidal neurone. As shown in Fig. 9C, the rise time of proximally generated EPSPs (3.4 ± 0.1 ms, n = 8) was nearly twice as fast as that of the distally generated EPSPs (6.7 ± 0.5 ms, n = 8), while the half-width of the proximal EPSPs (19.6 ± 0.8 ms) was somewhat shorter than that of the distal EPSPs (23.8 ± 1 ms). These results confirm that distally generated EPSPs are affected by the dendritic propagation.

Threshold spikes evoked by either distal or proximal EPSPs always appeared on the peak of the EPSPs recorded in the soma (Fig. 9C). This indicates that the spike initiation site for both proximally and distally generated EPSPs is close to the recording site. Furthermore, when tested in the same cell, the threshold level for somatic spikes was similar for proximally (-58.7 ± 0.9 mV, n = 6) and distally (-57. 4 ± 1.1 mV, n = 6, P > 0.025) generated EPSPs (Fig. 9C). Together with their similar sensitivity to hyperpolarization of the dendritic membrane, these results point to a similar site for threshold spike initiation.

Analysis of dendritic threshold and suprathreshold responses

The above analysis of somatic EPSPs was limited to subthreshold responses. However, voltage-dependent currents are present in the apical dendrites (Johnston et al. 1996), and it is likely that their influence will depend on the amplitude of the EPSPs. Because of the difficulties in evoking somatic spikes with distally generated EPSPs and the possible presence of a somatic leak, we used an alternative approach to analyse threshold and supra-threshold responses. This analysis is based on the reasonable assumption that synaptically generated spikes are primarily initiated at the initial segment (Stuart & Sakmann, 1994; Spruston et al. 1995) which, at least for threshold responses, is supported by the present study. Spikes evoked by distal EPSPs always arise from the peak of the somatically recorded EPSPs (Fig. 9C). This means that when a spike first appears on the dendritic EPSP, the peak of the EPSP at the spike initiation zone will have just reached the local threshold. The time-to-spike at threshold (t_s, Fig. 10A) in the dendrites will therefore reflect the time-to-peak (t_p) of the EPSP at the spike initiation zone. Because it takes time for the spike to propagate back to the dendritic recording site, t_s will slightly overestimate t_p at the spike initiation zone. However, since the velocity of back-propagating spikes in cortical pyramidal neurones has been estimated to be 0.15–0.24 m s⁻¹ at room temperature and two to three times faster at 35°C (Stuart & Sakmann, 1994; Spruston et al. 1995), a back-propagating spike will take less than 1 ms to reach the dendritic recording site. We have chosen to ignore this small and (presumably) constant factor and set t_s equal to t_p of the EPSP at the spike initiation zone. A good estimate of the change in time course of the dendritic threshold EPSP at the spike initiation zone can therefore be obtained by expressing t_s as a ratio of t_p of the dendritically recorded EPSP (Fig. 10A). A ratio of 1.0 indicates that the rising phase of the dendritic EPSP is unchanged at the spike initiation zone, while ratios > 1.0 indicate that the rising phase of the EPSP has become slower. All measured ratios were > 1.0, and none were < 1.0 (which would be consonant with dendritic amplification). The mean ratio (t_s/t_p) was 2 ± 0.2 (n = 21), which is very similar to the ratio between the rise time of distally and proximally generated EPSPs measured in the soma (6.7/3.4 = 1.97). This means that the rate-of-rise of the threshold dendritic EPSP is halved by the time it reaches the spike initiation zone. In Fig. 10B, the calculated ratio for each dendrite is plotted as a function of the threshold level (V_Thr, Fig. 10A). Except for one case, (which is the extreme example from Fig. 6Aa, where the ratio was 6.4), all the points are scattered between 1 and 3. Interestingly, the ratio is independent of the threshold level, even though the latter varied greatly from cell to cell (range: -61.7 to -44 mV)

Increasing the stimulating intensity usually reduced the spike latency until, at maximum intensity, the spike arose from the peak of the dendritic EPSP (Fig. 10C). Measuring t_s of these maximal responses and comparing them with the mean t_p of subthreshold EPSPs (Fig. 10D) showed that they were very similar in most cases.

We then examined the relationship between changes in amplitude and spike latency of distally evoked suprathreshold EPSPs. If propagation in the dendrites is a passive process, EPSPs of different amplitudes will experience the same degree of electrotonic attenuation and a linear relationship between dendritic EPSP amplitude and spike latency should exist. We therefore made the following measurements: the time-to-spike for threshold responses (t_s1); the amplitude of the EPSP just before the threshold spike (X); the time-to-spike for maximal responses (t_s2); the difference in amplitude between the maximum EPSP and the peak of the threshold EPSP (d;Fig. 11A). The absolute V_m just before the dendritic threshold spike was taken to represent threshold at the initiation zone (Thr_S). Although this is a reasonably accurate approximation (Stuart & Sakmann, 1994; Spruston et al. 1995), as will be evident below, the precise value of Thr_S is not important for the calculations. On the basis of these measurements, it is possible to make predictions about the rising phase of the maximal EPSP at the spike initiation zone if dendritic propagation is a passive process. In Fig. 11A, the predicted rising phase of the maximal EPSP is indicated by line Y, which is constructed by joining a point from the start of the threshold EPSP to a point (X + d) at the time of the threshold spike. The intersection between line Y and Thr_S marks the time at which a passively propagating EPSP with an amplitude of X+d would be expected to reach threshold at the spike initiation zone. This should result in a reduction in t_s1 such that the spike occurs at the time of the Y/Thr_S intersection. An estimate of the actual rising phase is obtained by joining a point from the start of the maximal EPSP to Thr_S at the time of the maximal spike (line Z in Fig. 11A). We then calculated the slopes of lines Y (the predicted dV/dt) and Z (the observed dV/dt), respectively, using the following equations:

Figure 11. Amplitude-induced changes in spike initiation.

A, superimposition of dendritic responses evoked by distal stimulation with threshold and maximal intensity. B, dendritic responses to distal stimulation of increasing intensity. Note that time-to-peak, indicated by the dotted line, is nearly constant. C, plot of the observed dV/dt as a function of the predicted dV/dt. The dashed line indicates a correlation of 1.0 between the observed and predicted dV/dt with a limit of +10 % (dotted line). •: correlation > 1.1; ○: correlation < = 1.1. D, superimposition of three dendritic responses to distal stimulation: threshold response (1), submaximal response (2), maximal response (3). The predicted and observed dV/dt for the submaximal response were 0.96 and 1.15, respectively, whereas they were 1.1 and 2.37 for the maximal response. RMP: in A, -68 mV; in B, -71 mV; and in D, -68 mV.

If the predicted dV/dt is equal to or greater than the observed dV/dt, the relationship between changes in peak amplitude and spike latency can be fully accounted for by passive propagation of the distally generated EPSPs. If, on the other hand, the predicted dV/dt is smaller than the observed (e.g. Figure 11A), the EPSP must have reached threshold faster at the spike initiation zone than expected with passive propagation. The validity of these predictions and interpretations depends on two assumptions: (1) the time-to-peak of distally evoked EPSPs is independent of their amplitude; (2) Thr_s is constant and independent of the rising phase of the EPSP. Regarding the first assumption, analysis of subthreshold EPSPs in individual dendrites showed this to be the case (Fig. 11B). Importantly, this means that on reaching the soma, the time-to-peak of passively propagating EPSPs is also independent of their amplitude and similar to that measured from threshold responses. Regarding the constancy of Thr_s, somatic recordings of just-suprathreshold EPSPs or responses to small depolarizing current pulses have shown that even very marked variations in the rising phase of the responses do not affect Thr_s to any great extent (Hu, Hvalby, Lacaille, Piercey, Østberg & Andersen, 1992). Furthermore, in the present study, the spike threshold for distally and proximally evoked EPSPs was found to be similar even though there was a marked difference in rise time between the two.

Plotting the correlation between observed and predicted dV/dt, and arbitrarily allowing a margin of +10 %, divided the pyramidal neurones into two groups (Fig. 11C). In one group, comprising 42 % of the cells, the observed dV/dt was ≤ 110 % of the predicted dV/dt. In these neurones, the relationship between amplitude and spike latency is in agreement with passive propagation of the EPSPs. In the other group (58 %), the observed dV/dt was > 110 % of the predicted dV/dt. In these neurones, the EPSPs reach threshold faster than expected for passively propagating EPSPs. The calculated values for dV/dt in Fig. 11C are all from maximal responses. Submaximal responses gave similar results, even though the differences between the predicted and the observed dV/dt were not usually as large. In the example shown in Fig. 11D, the amplitudes of the submaximal and maximal responses are very similar. Nevertheless, the predicted and observed dV/dt s for the submaximal response were 0.96 and 1.15, respectively; whereas for the maximal response they were 1.1 and 2.37, respectively. In some cases, the spike latency could vary considerably with little or no change in amplitude (e.g. Figs 2D and 6B). This indicates that the enhancement of submaximal EPSPs can occasionally be of an all-or-none nature.

DISCUSSION

A new experimental preparation

The reason for isolating the distal synaptic input as described, instead of the more commonly used procedures of either making a small tissue ‘bridge’ parallel to the dendritic axis (Andersen et al. 1980) or by placing a stimulation electrode close to the L-M region, is that it gives a more complete isolation of the distal synaptic input. As mentioned earlier, the Schaffer collateral fibres from CA3 pyramidal neurones are not all laminar, as generally assumed, but often deviate from a parallel course. Furthermore, the Schaffer collateral fibres often have a high degree of branching within the CA1 region (Li et al. 1994). Using the previous methods it will be very difficult to limit the synaptic input to a selective region of the dendritic tree of CA1 pyramidal neurones. Histochemical labelling of the Schaffer collaterals showed that a tissue ‘bridge’ (50 to 75 μm across) placed distally in SR and parallel to the dendritic axis, did not result in the isolation of a parallel bundle of fibres on the subicular side of the tissue ‘bridge’. Rather, labelled fibres were observed to fan out on the other side of the ‘bridge’ and cover most parts of SR (authors’ unpublished observation). By using the present isolation procedure, all non-laminar afferent fibres except those which run through the ‘bridge’ will be cut. However, the electrophysiological tests indicate that the number of non-laminar fibres running through the ‘bridge’ and into the deeper parts of SR is small. Furthermore, because of the relatively small dimensions of the ‘bridge’, only non-laminar fibres which run nearly parallel to the dendritic axis will be able to form synaptic contacts with the proximal apical dendrites of the pyramidal neurones from which the recordings are made. However, the number of these fibres will be even smaller, limiting the problem even further. The cutting procedure also ensured that any rebound activity from the CA3 region is restricted to the distal part of the apical dendrites.

The properties and pharmacology of distally evoked excitatory synaptic potentials

Paired-pulse stimulation of the distal afferent fibres to the CA1 region evoked a dendritic response which consisted of two fast EPSPs followed by a slow AHP. The dendritic EPSP was greatly reduced by the non-NMDA receptor-antagonist CNQX and the residual small EPSP was markedly enhanced in a nominally Mg²⁺-free medium. Furthermore, in the presence of both CNQX and AP5, the dendritic EPSP was completely blocked. This indicates that the dendritic EPSP is mainly composed of a non-NMDA receptor-mediated component and a minor NMDA receptor-mediated component. Application of glutamate to outside-out patches isolated from proximal apical dendrites of CA1 pyramidal neurones has shown that the AMPA and NMDA receptors are similar to those found at the somatic level (Spruston et al. 1995). The present study extends this observation by showing that both types of receptor are present in the distal parts of the apical dendrites and are activated by low frequency stimulation. Even though the NMDA receptor contribution to the distal EPSP appears to be small under physiological conditions, its presence will provide distally located synapses with an intrinsic voltage-dependent amplifier which could greatly enhance their efficacy.

Intradendritic recordings have shown that stimulation in the middle of SR results in PPF of dendritic EPSPs (Leung & Fu, 1994). In the present study, paired-pulse stimulation of the distal afferent fibres also resulted in PPF of the dendritic EPSP, showing that this type of short-term plasticity is a universal property of excitatory synapses on CA1 pyramidal neurones, irrespective of their location. Even though PPF, which is similar to that reported by Leung & Fu (1994), was not studied meticulously, the results provide some insight as to its functional importance for distally located synapses. In the group of pyramidal neurones in which distal stimulation evoked spikes, it was the tEPSP that was most effective. PPF therefore appears to be a very important mechanism of enhancing the efficacy of distal synapses.

The properties and pharmacology of distally evoked inhibitory synaptic potentials

The distally evoked EPSPs were usually followed by a slow AHP, which was blocked by CGP 55845A showing that it is a GABA_B receptor-mediated IPSP. It has been suggested that the slow IPSP in CA1 is mediated by a separate group of interneurones located in L-M (Lacaille & Schwartzkroin, 1988). The location and morphology of these L-M interneurones (Lacaille & Schwartzkroin, 1988) means that the isolation procedure will conserve a large part of their dendritic and axonal systems. It is therefore likely that the slow IPSP is caused by orthodromic activation of these interneurones. We found that a small dendritic hyperpolarization could either prevent spike generation or increase spike latency. This demonstrates that, despite its modest size, the slow IPSP could still have a functional significance in modulating the spike generating capabilities of distal excitatory synapses. The reason why the slow IPSP was absent in about 30 % of the dendritic recordings is not immediately apparent. However, it is possible that the slicing procedure occasionally severed the interneurones or limited their excitatory input to a degree which prevented synaptic activation.

The presence of a slow IPSP did not seem to affect the time course of the dendritic EPSPs. However, CGP 55845A did in most cases (4/6) reduce the amplitude of the EPSPs, whereas in the remaining cases (2/6) a marked increase was seen. We are not presently able to provide an explanation for this variable effect of CGP 55845A on the EPSP amplitude.

In the intact slice preparation, stimulation of the distal SR evokes a dendritic response consisting of an EPSP followed by a fast IPSP (Masukawa & Prince, 1984; Taube & Schwartzkroin, 1988; our own unpublished observations). Following isolation of the distal afferents, however, no fast IPSP was seen, either at RMP or when the dendritic membrane was depolarized. This is somewhat surprising, considering that stimulation of the entorhinal pathway is reported to evoke an often marked GABA_A receptor-mediated fast IPSP (Pearce, 1993; Empson & Heinemann, 1995). On the other hand, Masukawa & Prince (1984) found that the fast dendritic IPSP disappeared following removal of the somata and the proximal 50–100 μm of the apical dendrites. This, and the present study, do not therefore support a local generation of the entorhinal pathway-evoked fast IPSP, but rather indicate that it is generated at more proximal locations. In fact, Empson & Heinemann (1995) found that the major site of entorhinal pathway-activated GABA_A inhibition was in SR. Pearce (1993) also reported that multiple applications of BIC near the L-M/SR border only partially blocked the entorhinal pathway-evoked IPSP. Most of the GABAergic interneurones which have been implicated in the mediation of the fast IPSP are located in, or near, SP. Several of these have dendrites which extend into L-M (Halasy, Buhl, Lörinczi, Tamás & Somogyi, 1996), thereby providing the structural basis for entorhinal pathway-mediated activation of these interneurones. The dendrites from these interneurones will be cut by the isolation procedure, which would explain the lack of a fast IPSP in our preparation. Furthermore, the entorhinal pathway was not physically isolated from the SR in any of the aforementioned studies, so it is possible that laminar and non-laminar Schaffer collateral fibres were also stimulated, which could have activated interneurones located near SP.

In the presence of BIC, dendritic cEPSPs had a slightly slower rise time and a longer half-width compared with control medium, indicating that there must be some activation of distal GABA_A receptors. Masukawa & Prince (1984) also found that BIC still had a marked effect on the firing properties of the apical dendrites following the removal of the proximal dendrites and soma. This activation of distal GABA_A receptors could be due to spontaneously released GABA (Alger & Nicoll, 1980) which could tonically activate dendritic GABA_A receptors (Otis & Mody, 1992). A possible source of GABA could be the GABAergic terminals of horizontal interneurones in A/O, which have a highly arborized axonal plexus in L-M (McBain, DiChiara & Kauer, 1994). Tonic GABA_A receptor activation will decrease R_in and thereby reduce τ_m. This will lead to a faster decay of the EPSPs and a shorter half-width, which explains why BIC increased the half-width. The distal apical dendrites, which lie beyond the ‘bridge’, mainly consist of highly branched small diameter dendrites (Bannister & Larkman, 1995). However, at the level of the ‘bridge’ there would still be a large number of primary and second order large diameter dendrites, and it is most likely that the intradendritic recordings were obtained from these. The EPSP recorded at the level of the ‘bridge’ is therefore likely to be a summation of multiple unitary EPSPs generated at various locations on the distal dendrites. This means that there will be some degree of passive propagation before these unitary EPSPs reach the dendritic recordings site. A change in τ_m will therefore be expected not only to affect the decay of the EPSPs, but also to shorten their rising phase. This would explain why the rise time becomes slightly longer in the presence of BIC.

Impact of dendritic propagation on the time course of distally generated EPSPs

The distal apical dendrites, which lie beyond the ‘bridge’ collectively have an electrotonic length which is equal to that of the rest of the pyramidal neurone (Andreasen & Nedergaard, 1996). This means that the unitary EPSPs will already have been affected to a certain extent by the properties of these small diameter dendrites by the time they reach the dendritic recording site. A comparison between distally generated non-NMDA receptor mediated EPSPs recorded in the dendrites and in the soma showed that the EPSPs were further changed at the somatic level, with the rise time and half-width being 3 and 2.4 times longer, respectively. This is in agreement with theoretical analyses of the electrotonic structure of CA1 pyramidal dendrites, which indicate that the time course of distally generated synaptic potentials will be markedly changed once they reach the somatic region (Turner & Schwartzkroin, 1980; Spruston & Johnston, 1992; Mainen et al. 1996). Furthermore, theory predicts clear differences in both peak amplitude and time course between distal and proximal inputs (Mainen et al. 1996), which was borne out by the present results. Even though the amplitude of the EPSPs ranged from 1 mV to threshold, our results are qualitatively similar to those of Turner (1988). On the other hand, our results are in marked contrast to those reported by Andersen et al. (1980), who found no difference in time course or efficacy between distal and proximal EPSPs. They suggested that this lack of difference was due to an equal attenuation of the synaptic current in long spine necks and an insignificant attenuation within the main apical dendritic shaft. However, it has been suggested that the dimensions of most spine necks impose very little restraint on the current transfer from the spine to the dendrite (Harris & Stevens, 1989). Based on the morphology of the Schaffer collateral fibres, our own histological results and the study by Turner (1988), we feel that the most likely explanation for the results reported by Andersen et al. (1980) is that their isolation procedure does not prevent the activation of a substantial number of non-laminar fibres which form synapses over a considerable length of the dendrites. This will lead to an overestimation of the efficacy of distal synapses. In a later study, in which the afferent fibres were stimulated closer to the penetrated pyramidal neurone in order to limit the simultaneous activation of non-laminar fibres, the same group did find a small but significant difference in rise time and time-to-peak between distally and proximally evoked threshold EPSPs (Andersen, Storm & Wheal, 1987). In summary, our data, together with those of Turner (1988), indicate that propagation through the main proximal dendrites affects distally generated EPSPs.

As mentioned in Results, a direct comparison between dendritic and somatic EPSPs does not provide a complete assessment of the dendritic influence on distally generated EPSPs. We therefore used two alternative approaches in order to obtain a more realistic estimate of the influence of dendritic propagation. The first approach was to compare the time course of somatically recorded distally and proximally evoked EPSPs, while the second was to calculate the ratio between t_s and t_p measured from dendritically recorded threshold responses. Both methods assess changes in the rising phase of the EPSP, which is a very sensitive indicator for electrotonic attenuation (Spruston et al. 1994). Both methods showed that the rising phase of distally evoked EPSPs had become about twice as slow by the time the EPSP reached the spike initiation zone. This is similar to the difference (6.66/3.2 = 2.08) in rise time between small distal and proximal EPSPs reported by Turner (1988). Together, these results show that distally generated EPSPs of up to threshold size are modulated by the dendrites to a similar extent, irrespective of their initial peak amplitude.

Studies of neuronal models have indicated that distally generated synaptic potentials will be significantly attenuated and distorted upon reaching the soma (Turner & Schwartzkroin, 1980; Mainen et al. 1996). The present results would indicate that these theoretical analyses have overestimated the dendritic influence. However, while modelling describes the electrotonic structure of the dendrites, the present experimental study cannot separate electrotonic influence from other factors. Evidence has recently been presented that non-inactivating Na⁺ currents and low threshold T-type Ca²⁺ currents located in the proximal dendrites serve to amplify EPSPs, thereby compensating for electrotonic attenuation (Lipowsky et al. 1996; Gillessen & Alzheimer, 1997). Activation of these currents could explain why the experimental data indicate less dendritic modulation than is predicted from theoretical models. Interestingly, however, activation of these dendritic currents does not seem to change the rise time of the EPSPs (Lipowsky et al. 1996; Gillessen & Alzheimer, 1997), in which case our estimate of the change in rise time or time-to-peak primarily reflects the influence of the electrotonic structure of the main dendrites. EPSPs with a prolonged plateau-like peak were, however, observed in several dendrites. While it is possible that this prolongation is due to an all-or-none activation of dendritic currents, it is still uncertain to what extent Na⁺ and Ca²⁺ currents in the distal parts of the apical dendrites (Magee & Johnston, 1995b; Andreasen & Nedergaard, 1996) are involved in synaptic amplification.

Two groups of pyramidal neurones distinguished on the basis of their efficacy

Activation of distal synapses could drive the cell to a firing threshold in 60 % of the neurones. There was otherwise no electrophysiological distinction between spiking and non-spiking neurones. The V_m at the peak of the EPSPs in the twenty-six non-spiking neurones varied between -66.8 and -44.8 mV, and in twenty-two of these was within the range of thresholds measured in the spike generating neurones (-61 to -44 mV). The first question which needs to be addressed is: where are the threshold spikes initiated? Even though the apical dendrites of CA1 pyramidal neurones are capable of generating both fast Na⁺-dependent and slow Ca²⁺-dependent spikes (Andreasen & Lambert, 1995; Andreasen & Nedergaard, 1996), it has been suggested that synaptically generated spikes are primarily initiated at the initial segment (Stuart & Sakmann, 1994; Spruston et al. 1995; see, however, Colbert & Johnston, 1996). The data presented here are essentially in agreement with this for the following reasons. (1) The composition of the dendritic threshold response is in accordance with spikes being initiated proximally to the recording site and not within the distal apical dendrites. The response is also similar to that recorded 500 μm from the spike initiation zone during stimulation of distal afferent fibres in young cortical pyramidal neurones (Stuart & Sakmann, 1994). (2) A small hyperpolarization moved the spike from the peak of the maximal dendritic EPSP to its decay phase, thereby increasing the spike latency. If the spike had been initiated in the distal part of the apical dendrites, it would still be expected to appear on the peak of the EPSP during the hyperpolarization. (3) The similar sensitivity of proximally and distally generated threshold spikes to dendritic hyperpolarization suggests that both are generated at the same location.

The reason why distally generated EPSPs did not evoke spikes in 40 % of the pyramidal neurones is presently unknown. There are, however, several possibilities. One is that slicing and microdissection reduced the number of afferent fibres to such an extent that insufficient distal synapses were available to generate a summated EPSP large enough to reach threshold. This would explain cases where the maximal amplitude of the dendritic EPSP was only 4–5 mV. However, it would not explain the lack of spikes in the majority of cases in which the maximal amplitude was from 16 to 25 mV, especially when the amplitude of threshold EPSPs was between 10 and 21 mV. In these cases it is possible that slicing had severed the initial part of the axon thereby preventing spike initiation. An alternative explanation is that the lack of spiking reflects differences in dendritic attenuation. CA1 pyramidal neurones can be divided into two groups on a morphological basis (Bannister & Larkman, 1995). In 46 % of neurones, the main apical dendrite is a single trunk which spans the entire SR and divides around the SR/L-M border. In the other group, the main apical dendrite bifurcates within the first 10 to 65 % of SR. At a given level in SR, the diameter of the secondary dendrites is significantly smaller in the bifurcated neurones (Bannister & Larkman, 1995) which could result in a larger degree of dendritic attenuation (Spruston et al. 1994). While it is possible that the spiking and non-spiking pyramidal cells reflect differences in morphology, this would need to be tested by correlating synaptic efficacy with morphology.

The variability in dendritic threshold

The threshold for initiating spikes with distally generated EPSPs varied considerably (Fig. 10B). Though the reasons for this are not presently clear, there could be a number of explanations. The first is that there is a difference in the degree of dendritic attenuation. Morphological studies have shown that the somata of some CA1 pyramidal neurones are located up to 100 μm from the superficial border of SP (Bannister & Larkman, 1995). Because intradendritic impalements of these neurones would be correspondingly further from the soma, the degree of electrotonic attenuation would be expected to be larger at the spike initiation zone. If the higher threshold seen in some neurones is due to greater electrotonic attenuation of the EPSPs, a positive correlation between the threshold level and the t_s/t_p ratio would be expected. With the exception of one dendrite, this was not the case (Fig. 10B). A second explanation is that V_m varies at the spike initiation zone. This would be reflected as changes in the size of the distally generated EPSPs needed to initiate spikes, and therefore in the threshold level. However, from figures presented of dual patch-clamp recordings from soma and dendrites of pyramidal neurones it would seem that RMP is similar at these two sites (Stuart & Sakmann, 1994; Spruston et al. 1995). It would therefore be expected that there was a correlation between the dendritic RMP and the threshold level, which was not the case. At present, the most plausible explanations seem to be that the different threshold levels are due to variable degrees of activation of dendritic non-inactivating Na⁺ channels and/or T-type Ca²⁺ channels which increase the amplitude of the EPSPs and therefore would reduce the apparent threshold (Lipowsky et al. 1996; Gillessen & Alzheimer, 1997). However, further investigations would be needed to confirm this and exclude other possibilities.

The relationship between EPSP amplitude and spike initiation

At threshold, distally generated spikes always appeared on the decaying phase of the EPSPs with a time-to-spike (t_s) which was on average twice the time-to-peak (t_p) of the EPSP. With maximal stimulation, however, t_s was similar to t_p. The functional consequence of this is that the effect of dendritic propagation on the time course of the EPSPs is overridden as the size of the EPSP increases. There is, therefore, a time window during which the response time to distal excitatory inputs can be changed. As discussed earlier, the dendritic modulation of EPSPs up to threshold level is likely to consist of both an electrotonic and an active component. It is the final result of this modulation that determines the upper limit for t_s.

In the 42 % of the neurones in which distally evoked EPSPs could initiate spikes, the decrease in t_s with increase in peak amplitude can be explained by constant dendritic modulation of the EPSP up to suprathreshold level. However, in the remaining 58 % of neurones our data indicate that additional factors are involved in determining the relationship between t_s and peak amplitude. Analysis of the maximum dendritic responses in these neurones revealed that the increase in peak amplitude was accompanied by a reduction of t_s that was larger than would be predicted for constant dendritic modulation. It therefore appears that these neurones possess a mechanism that can be activated by a slight increase in the amplitude of the EPSPs and which ensures a faster response than would otherwise be obtained. One explanation for this could be that the threshold is not constant, as we initially assumed, but depends on the rising phase of the EPSP. A fast rising EPSP would be expected to initiate spikes at a lower threshold simply because fewer Na⁺ channels directly enter the inactivated state than with a slowly rising EPSP. However, there are several observations which seem to argue against this. Firstly, a sudden change in t_s was occasionally seen without any change in the EPSP. Secondly, a small hyperpolarization of the dendritic membrane increased t_s even though there was no change in the amplitude or kinetics of the EPSP. A hyperpolarization would be expected to decrease the steady-state inactivation of the Na⁺ channels and thereby decrease the threshold and reduce t_s. Thirdly, a change in threshold would be expected to affect the relationship between spike latency and peak amplitude in all neurones, and not just about half these. Fourthly, if the change in t_s is due to a sudden decrease in threshold, application of TTX to the somatic region would be expected to block antidromically and distally evoked spikes simultaneously. However, TTX blocked the distally evoked spikes on average 92 s after the antidromic spike. Another explanation for the larger-than-expected reduction in t_s could be that the peak amplitude is amplified by a sudden increase in the activation of dendritic currents. Even though activation of dendritic non-inactivating Na⁺ and/or T-type Ca²⁺ currents does not seem to change the time course of the somatic EPSP (Lipowsky et al. 1996; Gillessen & Alzheimer, 1997) an increase in amplitude with a constant rise time will result in the EPSP reaching threshold earlier. The non-inactivating Na⁺ and T-type Ca²⁺ current both have activation thresholds at about -70 mV (Magee & Johnston, 1995b; Lipowsky et al. 1996). A sudden increase in these currents would therefore require a relatively large increase in the amplitude of the distal EPSP. However, the present analysis indicates that even a small increase in EPSP amplitude could activate this additional mechanism, and in some cases, no change in amplitude at all was needed. Furthermore, this explanation would also require that TTX blocks the distally and antidromically evoked spikes simultaneously. Both of these observations therefore seem to argue against an increase in the activation of these two voltage-dependent dendritic currents.

On the other hand, the present results are more in agreement with a shift in spike initiation zone from the initial segment to the proximal dendrites, which would result in the spike being initiated earlier than predicted by constant dendritic modulation. The density of Na⁺ channels in the initial segment and soma has recently been suggested to be the same (Colbert & Johnston, 1996) and similar to that of the apical dendrites up to 300 μm from the soma (Magee & Johnston, 1995b). Furthermore, blocking the initial segment of subicular pyramidal neurones by selective application of TTX does not prevent axonal spikes from invading the soma antidromically and here evoke spikes (Colbert & Johnston, 1996). This indicates that the density of somatic Na⁺ channels, and by inference dendritic Na⁺ channels, is high enough to support spike generation. Because the spike threshold depends not only on channel density, but also on structural parameters such as the diameter (Mainen, Joerges, Huguenard & Sejnowski, 1995), the spike threshold will be lower within the dendrites than the soma for a given density of Na⁺ channels. Furthermore, there is some evidence for the presence of localized areas of high Na⁺ channel density within the proximal apical dendrites which could represent ‘hotspots’ (Magee & Johnston, 1995b). We, therefore, propose that the sudden reduction in spike latency seen in more than half of the neurones at suprathreshold levels is due to a shift in the spike initiation from the initial segment to the proximal apical dendrites. This would explain the delay between the action of TTX on antidromically and distally evoked spikes. Because of its placement, the antidromic electrode will most likely initiate spikes at the initial segmental/axonal region and therefore be affected first by somatic TTX application. If, as we propose, spikes evoked by supramaximal distal stimulation are initiated in the proximal dendrites, TTX would have to diffuse into this region of SR, which accounts for the delay. As mentioned earlier, others have suggested that spikes are only initiated in the proximal dendrites under extreme conditions (Turner et al. 1991; Stuart & Sakmann, 1994; Spruston et al. 1995). However, the simultaneous activation of GABAergic interneurones forming synaptic contacts with the proximal dendrites (Halasy et al. 1996) would be expected to have increased the spike threshold in this region (Masukawa & Prince, 1984). The preparation used here, however, appears to prevent any substantial activation of fast GABAergic inhibition when distal afferent fibres are stimulated.

In contrast to earlier suggestions, the results presented here indicate that the proximal apical dendrites can, in fact, function as the spike initiation zone under physiological conditions. Once threshold is reached, slight adjustments of the amplitude of the distally generated EPSPs can induce this shift. Because a decrease in spike latency was occasionally seen without a concomitant change in amplitude, there also seem to be conditions under which little or no change in amplitude is required. One possibility could be a modulation of the A-type K⁺ current, which has been shown to be present in the apical dendrites and to be of significance in determining their excitability (Andreasen & Lambert, 1995; Hoffman, Magee, Colbert & Johnston, 1997). A decreased activation of the A-type K⁺ current would increase the excitability of the dendrites and increase the likelihood that an EPSP of a given amplitude will induce a shift in the spike initiation zone (Hoffman et al. 1997). Conversely, an increase in the activation of the A-type K⁺ current could prevent this shift from occurring. This could be the case in the 42 % of neurones in which no shift was observed. In an earlier study, we found that compound spiking was absent in 56 % of the apical dendrites, but could be revealed by blocking the A-type K⁺ current (Andreasen & Lambert, 1995). This indicates that there is a marked difference in the activity level of the dendritic A-type K⁺ current among pyramidal neurones.

Concluding remarks

The results presented here indicate that propagation through the main apical dendrites delays the time-to-peak of the distally generated EPSPs. This provides more opportunity for summation and could be an important factor in regulating the efficacy of more proximally located excitatory synapses. Activation of distal excitatory synapses could induce spike firing in 60 % of the pyramidal neurones and there are indications that these spikes originated in the initial segment: (1) the timing of the spike in relation to the peak of the EPSP; (2) the similar sensitivity of distally and proximally generated threshold spikes to changes in dendritic membrane potential; (3) the pronounced effect of a small dendritic hyperpolarization on the spike latency.

Although dendritic propagation delays spiking with submaximal stimulation, as the intensity was increased towards maximum, however, the spike latency was reduced and approached the time-to-peak of the dendritic EPSPs. The functional implication of this is, that the influence of dendritic modulation can be overcome thereby ensuring a faster response time. This adds an extra dimension to the way in which pyramidal neurones respond to distal synaptic inputs in that there is also a time factor which can be modulated. Analysis of the relationship between amplitude and spike latency indicated two mechanisms by which the spike latency could be reduced: by increasing the amplitude of the distal EPSP so that, despite the effect of dendritic propagation, it still reaches threshold faster; by shifting the spike initiation zone into the proximal apical dendrites. This shift in spike initiation zone could be the result of a small increase in EPSP amplitude or a local decrease in A-type K⁺ channel activation.