Activity-dependent properties of synaptic transmission at two classes of connections made by rat neocortical pyramidal axons in vitro



  • 1To compare the dynamics of synaptic transmission at different types of connection, dual intracellular recordings were made from pairs of neurones in slices of adult rat neocortex. Excitatory postsynaptic potentials (EPSPs) were elicited by single spikes, spike pairs and brief spike trains in presynaptic pyramidal cells and responses recorded in postsynaptic pyramidal cells and in inter neurones.
  • 2Pyramid–pyramid EPSPs were strongly voltage dependent and this resulted in a range of paired pulse effects. At thirty-two of sixty-nine pyramid–pyramid connections, the 2nd EPSP was the same shape as the 1st, indicating minimal interaction between active synapses. In these thirty-two connections, paired pulse depression (PPD) was apparent (2nd EPSP integral 46 ± 21% of the 1st, at 5–20 ms), which recovered within 60–70 ms.
  • 3In eleven additional pyramid–pyramid pairs, the 2nd EPSP was also the same shape as the 1st, but paired pulse facilitation (PPF, 149 ± 32%) decaying within 50–60 ms was apparent. Even these connections displayed frequency-dependent depression, however, as 3rd EPSPs were smaller than 1st EPSPs at intervals < 100 ms.
  • 4At twenty-five pyramid–pyramid connections, 2nd EPSPs were broader than 1st EPSPs and in sixteen of these, voltage- and NMDA receptor-dependent enhancement was large enough to obscure the underlying PPD. PPD was revealed by postsynaptic hyperpolarization (4 pairs), N-methyl-D-aspartate (NMDA) receptor blockade (3 pairs), or if Mg2+ was removed (in the one case studied). If synapse location allowed significant depolarization of one active site by another, voltage-dependent enhancement could produce supralinear EPSP summation and overcome PPD. Third EPSPs were, however, consistently smaller than 1st EPSPs.
  • 5In striking contrast, profound frequency-dependent facilitation, independent of voltage or NMDA receptors was seen at fifteen connections involving two classes of postsynaptic inter neurones.
  • 6At these pyramid–interneurone connections, facilitation of the 2nd EPSP (655 ± 380% at 5–20 ms) decayed rapidly, within 50–60 ms. Third and fourth EPSPs showed additional facilitation which decayed more slowly, within 90 ms and 2 s, respectively. Facilitation due to five to six spike trains was still apparent at 3 s. Therefore, once initiated by a brief high frequency spike train, facilitation was maintained at lower frequencies.

Recent activity modifies synaptic responses to subsequent action potentials (e.g. Lundberg & Quilisch, 1953; Del Castillo & Katz, 1956; Katz & Miledi, 1968, 1970). Calcium entering during a single spike may be insufficient at all release sites to release transmitter, but may prime quiescent sites. Subsequent spikes at these primed sites can then activate release (Smith, Augustine & Charlton, 1985; Zucker & Fogelson, 1986; Stanley, 1989, 1993). The paired pulse facilitation (PPF) and post-tetanic potentiation that result are, of course, most apparent when the probability of release in response to single spikes is very low.

Presynaptic mechanisms that depress subsequent release are also proposed, e.g. release sites may be refractory for some tens of milliseconds (Betz, 1970; Stevens & Wang, 1995; Thomson & Deuchars, 1995). Those contributing to the 1st EPSP would be less available for the next and where release probability is high, this refractoriness would result in paired pulse depression (PPD). In several invertebrate preparations, synapses from a single presynaptic axon onto different postsynaptic targets display different release probabilities and therefore exhibit either tonic or phasic release patterns (Katz, Kirk & Govind, 1993; Walrond, Govind & Huestis, 1993; Davis & Murphey, 1993; Govind, Pearce, Wojtowicz & Atwood, 1994). Only with paired intracellular recordings where identification of pre- and postsynaptic neurones is unambiguous are such studies possible in the adult mammalian central nervous system (CNS; Thomson & Deuchars, 1995; Markram & Tsodyks, 1996).

Postsynaptic receptor and dendritic cable properties further complicate the picture in central neurones. Glutamate, acting at AMPA, kainate and/or NMDA receptors, mediates most fast excitatory transmission in the CNS. AMPA/kainate receptors desensitize rapidly (e.g. Trussell, Thio, Zorumski & Fischbach, 1988) and at synapses involving multiple release sites this may be a significant factor in PPD (Trussell, Zhang & Pvaman, 1993). NMDA receptor/channels activate and desensitize less rapidly and are voltage dependent when extracellular Mg2+ is present (Nowak, Bregestovski, Ascher, Herbert & Prochiantz, 1984; Mayer, Westbrook & Guthrie, 1984). Depolarization reduces Mg2+ blockade and EPSPs utilizing these receptors increase in amplitude and duration. In addition, a wide range of voltage-gated currents that shape postsynaptic responses have been described in cortical pyramidal dendrites (e.g. Markram & Sakmann, 1994; Andreasen & Lambert, 1995; Schwindt & Grill, 1995; Magee, Christofi, Miyakawa, Christie, Lasser-Ross & Johnston, 1995; Magee & Johnston, 1995). Postsynaptic responses to transmitter release are therefore an amalgam of postsynaptic receptor responses and the conductances activated by that response. Somatic responses to these dendritic events would in turn depend also on neuronal cable properties and synaptic position(s).

Previous publications (Thomson & West, 1993; Thomson, Deuchars & West, 1993a) described PPD at pyramid–pyramid connections (see also Markram & Tsodyks, 1996, for similar data in the juvenile), contrasting with powerful PPF at pyramidal inputs to some classes of interneurones (Thomson, Deuchars & West, 1993b; Thomson, West & Deuchars, 1995). However, paired recordings in which brief train effects could be studied in any detail were relatively few in each of these earlier studies and the time course even of simple, presynaptically mediated PPD and PPF at adult synapses was not determined. To obtain a more complete picture of paired pulse and brief train effects and their time, voltage and NMDA receptor dependence, a larger sample and more detailed analysis were required. This paper therefore compares data from wider samples of pyramid–pyramid and pyramid–interneurone connections in superficial and deep layers of three regions of adult rat neocortex.


In the early series of experiments, male Sprague–Dawley rats, 120–250 g in body weight, were anaesthetized with Muothane (ICI) and decapitated. Coronal shoes of neocortex, 400 μm thick, were prepared in ice-cold standard artificial cerebrospinal fluid (ACSF; see below) equilibrated with 95% O2-5% CO2 and incubated in warmed (34–36 °C) ACSF. In a later series of experiments, the rats were anaesthetized with Sagatal (pentobarbitone sodium, Rh&#x006f;&#x030c;ne Mérieux, 60 mgkg−1i.p.) and perfused transcardially with an ice-cold ACSF in which the NaCl had been replaced with 248 mm sucrose and to which Sagatal (60 mg l−1) was added, and then decapitated. In the later experiments 500 μm thick shoes were used and during shoe preparation in ice-cold medium and for the first hour of incubation in the interface recording chamber (34–36 °C), the shoes were maintained in the sucrose-containing medium (without Sagatal) equilibrated with 95% O2–5% CO2. This was then replaced with standard ACSF containing (mm): 124 NaCl, 25.5 NaHCO3, 3.3 KCl, 1.2 KH2PO4, 1 MgSO4, 2.5 CaCl2, and 15 D-glucose, in which all recordings were performed. Recordings commenced after a further hour in this medium. The only relevant significant difference observed between earlier and later protocols was the higher incidence of connections in later experiments and an increase in spontaneous activity in the shoes. Data were therefore pooled.

Electrophysiological recordings

Intracellular recordings were made from pairs of neurones with conventional sharp electrodes filled with 2 M KMeSO4 and, in the majority of experiments, 2% w/v biocytin (100–150 MΩ) using an Axoprobe (Axon Instruments). Single spikes, pairs of spikes, or spike trains were elicited in the presynaptic neurone by injection of ramp and/or square wave current pulses delivered once every 3 s. The shape and amplitude of the pulse and its repetition rate could be modified according to the experimental design. Continuous analog recordings from both neurones for the entire duration of the paired recording were made on magnetic tape (Racal Store 4). Postsynaptic membrane potential was maintained within 2 mV of a preset value and after sufficient data had been collected at one potential, could be changed. During continuous current injection, electrode balance was monitored by observing voltage responses to small, brief current pulses injected before or after responses to presynaptic spikes.

The properties of recorded cells were assessed from the voltage responses to 100 or 200 ms current pulses between −2.0 and +1.0 nA in amplitude, delivered from membrane potentials ranging from −65 to −80 mV. In early experiments, these data were stored on analog tape and analysed using a digital storage oscilloscope. In later experiments, neuronal responses to current injection were controlled and collected to disk using pCLAMP software (Axon Instruments).

Data analysis

Dual recording data were collected from tape to disk and analysed off-line (using in-house software; Thomson et al. 1993a; Thomson, West, Hahn & Deuchars, 1996). Each single sweep (100, 200 or 400 ms in duration, depending on the duration of the postsynaptic responses from that pair) was then observed and either accepted, edited, or rejected according to the trigger point(s) that would trigger measurement or averaging of postsynaptic data during subsequent analysis. Acceptable, or adequately edited sweeps were those in which the rising phase of each presynaptic spike was recognized by the software as a trigger point and where spike-elicited EPSPs were not contaminated by artifacts, postsynaptic spikes or large spontaneous postsynaptic potentials. Editing consisted of removing false trigger points, or adjusting inaccurate trigger points.

EPSPs elicited by the 1st action potential in a train were analysed using the rising phase of the 1st spike to trigger data collection and analysis. Analysis of 2nd, 3rd, 4th, etc. spike responses required triggering of data analysis from the 2nd, 3rd, or 4th, etc. spike. Numbers of spikes and interspike intervals were varied from sweep to sweep in some experiments. Sweeps were therefore selected into data subsets during off-line analysis according to the number of presynaptic spikes, the interspike interval(s) and/or the duration of the presynaptic spike train. To analyse paired pulse effects for each paired recording, subsets of single sweeps in which two presynaptic spikes occurred were selected according to the presynaptic interspike interval. Where 50–200 sweeps with a similar interspike interval (e.g. between 5 and 8 ms or between 25 and 30 ms) had been recorded, these sweeps were averaged twice, once using the rising phase of the 1st spike and once using the rising phase of the 2nd spike to trigger analysis. The interval chosen to represent these averaged responses was the most commonly recorded interval within that subset. The amplitude and shape of the averaged 1st EPSP were compared with those of an averaged single-spike EPSP (interleaved records). Only if the two superimposed was the integral of the single-spike EPSP used to normalize the integral of the two-spike EPSP. For some pairs only one interspike interval was adequately studied; in others a range of intervals was available.

Illustrated averages of responses to brief trains of spikes are composites of (i) averaged responses to single presynaptic spikes, (ii) averaged responses to pairs of presynaptic spikes at one or more preset interspike interval(s), and (iii) averaged responses to 3rd spikes, again at a preset interval, and so on. Two-, three-, four-, five- and six- spike train EPSP integrals were measured as the area under the average EPSP obtained under stable conditions and normalized to the area under a single-spike EPSP recorded under identical conditions. These measurements were controlled by hand and artifacts excluded. To obtain an indication of the shape and time course of the 2nd EPSP, the averaged single-spike EPSP was subtracted digitally from the averaged two-spike EPSP, the two EPSPs superimposed and, if required, scaled for comparison of shape.

EPSP 10–90% rise time was measured as the time taken for the single-spike EPSP to rise from 10 to 90% of its peak amplitude. EPSP width at half-amplitude was the time interval between the EPSP rising to 50% and then falling again to 50% of its peak amplitude. In the figures, spike artifacts have simply been removed graphically after averaged responses had been computed and composites produced. In some of the smaller EPSPs in which the EPSP decay was noisy, three- to five-point averaging was employed to smooth later parts of the decay. Dotted fines in the composite averages illustrated indicate probable decay phase(s) of the single-, two-, three- or four-spike EPSPs that would have occurred had the spike train ended there, assessed from the decay of briefer trains of EPSPs. Numbers given in the text and tables are means ±s.d.

Average amplitudes and time courses of some of the EPSPs obtained from the earliest paired recordings reported here were included in previous publications (15 of 69 postsynaptic pyramidal cells, 7 of 15 interneurones). However, except for Fig. 11, none of the illustrated examples has appeared previously and more extensive analysis than was originally possible has been undertaken in all cases.

Figure 11.

Time course of PPF and PPD at a pyramid-pyramid connection, in low and high extracellular Ca2+ respectively

A, upper traces illustrate typical single presynaptic sweeps containing spike pairs at 2 interspike intervals (thick and thin traces, respectively). The lower traces are composite averages of EPSPs triggered by single presynaptic spikes and EPSPs elicited by spike pairs (at the 2 interspike intervals, thick and thin traces, respectively), the averages triggered from the rising phase of the 2nd spike. Both PPF in low Ca2+ and PPD in high Ca2+ can be seen to depend on interspike interval. B, in low Ca2+, despite PPF, brief train facilitation is less pronounced than in pyramid–interneurone connections. In high Ca2+, there is significant enhancement of 1st EPSPs and PPD. 80 ms after the 2nd EPSP, 3rd EPSPs are not depressed. (Data from this pair are also illustrated in Thomson et al. 1993a.)

Drug solutions

NMDA receptor antagonists D-2-amino-5-phosphonovaleric acid (D-AP5; 20–50 μm; Tocris Cookson, Bristol, UK), D-2-amino-7-phosphonoheptanoic acid (D-AP7; 20–50 μm; Sandoz Pharma, Basel, Switzerland) were dissolved in the perfusion medium. Nominally Mg2+-free medium was identical to the control medium except that no Mg2+ was added.

Regression analysis

To determine whether the regression analysis in Figs 2 and 8 was appropriate, data were also transformed on a logarithmic scale. Regression analysis was performed on raw data and on transformed data and in Fig. 2 the correlations obtained were found to be equivalent whether linear or semi-log plots were used. In Fig. 8, the ‘best fit’ was obtained using a double-log plot.

Figure 2.

Time course of PPD at pyramid–pyramid connections

• indicate all layer V pyramid–pyramid pairs exhibiting PPD (paired pulse depression) in which the 2nd EPSP was similar in shape to the 1st (22 pairs). Open symbols represent some of the pairs in which apparent PPF (paired pulse facilitation) was due to ‘boosting’ of the 2nd EPSP and includes some layer III pyramid–pyramid pairs. Blocking NMDA receptors (▵), hyperpolarizing the cell (□) or removing extracellular Mg2+ (○) removes this ‘boosting effect’. The underlying PPD then becomes apparent. The integral of the averaged 2-spike EPSP (50–200 single sweeps) was normalized against the integral of the averaged single-spike EPSP, recorded under the same conditions, and plotted against the interspike interval. The linear regression line for the filled symbols (dashed line, r= 0.77, n= 53) indicates that this form of depression lasts for 60–70 ms.

Figure 8.

Time course of paired pulse and brief train facilitation at pyramid–interneurone connections

The integral of the averaged 2-(3-, 4-, 5- and 6-) spike EPSP was normalized against the integral of the averaged single-spike EPSP, recorded under the same conditions, and plotted against the time interval between the 1st spike of the train and the last. Double-log regression lines for 2-, 3- and 4-spike EPSP plots are indicated by the dashed lines. Blockade of NMDA receptors (see points labelled D-AP7) reduced slightly but did not eliminate PPF at these connections.

Biocytin filling

In most experiments, synaptically connected neurones were filled with biocytin (+0.5 nA in a 50% duty cycle at 1 Hz, 5–15 min; see Deuchars & Thomson, 1995; Thomson et al. 1996, for processing details and descriptions of interneurone classes). In many cases it was possible to confirm the position and identity of the post-synaptic neurone. Reconstructions of two of the earlier pairs involving interneurones have been published. The detailed analysis of more recently recorded pairs remains to be done and does not form part of the present report.


Pyramid-pyramid connections

All illustrations and analysis of EPSP integrals involve averaged EPSPs (50–200 sweeps) to exclude complications arising from sweep to sweep fluctuations in EPSP amplitude. As previously described (Thomson & West, 1993; Thomson et al. 1993a), pyramid-pyramid EPSP amplitudes are approximately normally distributed, the coefficient of variation (c.v.) of single-spike EPSPs is typically < 0.35 and typically the standard deviation does not change significantly with the 2nd EPSP. When, therefore, n > 50, a 10% difference in pyramid-pyramid EPSP amplitude would become significant at the P < 0.05 level and a 20% change at the P < 0.005 level. From a total sample of > 150 pyramid–pyramid pairs recorded, sixty-nine resulted in sufficient data of adequate quality (see Methods). These were selected for further detailed analysis of responses to pairs and brief trains of presynaptic spikes. Data from the other pairs were consistent, but in these cases either too few sweeps containing presynaptic spikes of a given interspike interval were obtained for significance, or too few single-spike EPSPs were available for comparison.

Of the sixty-nine selected pairs, thirty-two were recorded in motor, twenty-two in somatosensory and fifteen in visual cortical regions (see Table 1 for details). Twenty-five of the postsynaptic neurones were located in layer II/III and forty-three in layer V/VI. In Table 1, postsynaptic layer V pyramidal cells that were classified either as regular spiking (RS) or burst-firing (BE) during the recording are also indicated. With a much larger population, differences between regions, type of postsynaptic pyramidal cell or layer might become apparent, but within this limited sample paired pulse effects were not found to correlate significantly with pyramidal cell type or position, or with EPSP amplitude or time course.

Table 1. Properties of pyramid—pyramid EPSPs
Paired pulse effectLayer and firing characteristics (n cell pairs)EPSP amplitude (mv)EPSP rise time (ms)EPSP width at half-amplitude (ms)2nd EPSP integral as %1st EPSP integral*
  1. Values are means ±s.d. with the range shown in parentheses. Pyramid—pyramid cell pairs (57) are subdivided according to the paired pulse effects elicited: simple PPD and simple PPF in which the 2nd EPSP was the same shape as the 1st, and boosting PPD and PPF in which the 2nd EPSP was broader than the 1st at membrane potentials between −70 and −60 mV. The pairs are then further subdivided by the cortical region and layer and for layer V cells, by whether the postsynaptic neurone was a regular spiking (RS) or a burst firing (BF) pyramidal cell (cells that were not adequately classified are excluded). The amplitude of the 1st EPSP (after a 3 s interval), 10—90% rise time, and width at half-amplitude were recorded at membrane potentials between −65 and −75 mV (EPSPs recorded only at other membrane potentials are excluded). The 2nd EPSP integral, obtained by subtracting the single-spike EPSP integral from the two-spike EPSP integral, is expressed as a percentage of the 1st EPSP integral.* Interspike interval, 5—20 ms.

Simple PPD
  SomatosensoryLayer III (3)0.8 ± 0.9 (0.7—0.9)1.9 ± 0.8 (0.8—2.5)15.3 ± 8.2 (5—25)36 ± 16 (20—58)
 Layer V RS (3)0.6 ± 0.3 (0.4—1.0)1.3 ± 0.1 (1.2—1.4)7.3 ± 2.6 (5—11)58 ± 13 (40—70)
 Layer V BF (3)2.9 ± 2.0 (0.5—5.5)2.5 ± 0.5 (1.8—3.0)14.5 ± 0.4 (14—15)59 ± 17 (42—82)
  MotorLayer III (7)0.9 ± 0.4 (0.2—1.5)1.7 ± 0.5 (0.8—2.4)12.2 ± 4.5 (8—22)46 ± 21 (24—80)
 Layer V RS (5)2.1 ± 1.0 (0.4—3.1)1.9 ± 0.4 (1.4—2.6)14.5 ± 5.4 (10—25)33 ± 5 (25—39)
 Layer V BF (3)1.9 ± 0.4 (1.5—2.3)2.2 ± 0.6 (1.4—2.8)16.3 ± 1.7 (14—18)32 ± 6 (26—38)
  VisualLayer III (1)0.311192
 Layer V RS (0)
 Layer V BF (4)1.0 ± 0.6 (0.5—2.1)2.4 ± 1.0 (1.4—4.0)18.5 ± 9.1 (9—29)39 ± 26 (20—84)
 All (29)1.4 ± 1.1 (0.23—5.5)1.9 ± 0.7 (0.8—4.0)13.8 ± 6.0 (5—25)46 ± 21 (20—80)
Simple PPF
  SomatosensoryLayer III (5)0.9 ± 1.3 (0.1—3.5)2.2 ± 1.2 (1.2—4.2)12.0 ± 4.0 (5—16)138 ± 16 (120—147)
 Layer V RS (2)0.8, 0.51.2, 3.09.3, 5116, 197
 Layer V BF (0)
  MotorLayer III (0)
 Layer V RS (1)0.42.816.0200
 Layer V BF (1)0.21.513.0115
  VisualLayer III (1)
 Layer V RS (0)
 Layer V BF (1)0.42.520.0186
 All (11)0.7 ± 0.9 (0.1—3.5)2.3 ± 0.9 (1.2—4.2)13.0 ± 4.0 (5—20)149 ± 32 (115—200)
Boosting PPD
  SomatosensoryLayer III (2)0.7, 0.83.4, 2.06, 1753, 85
 Layer V RS (0)
 Layer V BF (0)
  MotorLayer III (0)
 Layer V RS (0)
 Layer V BF (1)0.31.91845
  VisualLayer III (0)
 Layer V RS (1)0.73.62098
 Layer V BF (1)0.72.03672
 All (5)0.6 ± 0.2 (0.3—0.8)2.6 ± 0.7 (1.9—3.6)21.4 ± 7.4 (16—36)71 ± 20 (45—98)
Boosting PPF
  SomatosensoryLayer III (2)0.6, 0.31.4, 2.517, 7.0117, 158
 Layer V RS (1)0.25.017133
 Layer V BF (1)0.53.424180
  MotorLayer III (1)0.72.224160
 Layer V RS (3)1.2 ± 0.5 (0.4—1.7)1.7 ± 0.6 (1—2.5)20 ± 4.5 (15—26)152 ± 21 (130—180)
 Layer V BF (1)0.81.411138
  VisualLayer III (0)
 Layer V RS (1)0.32.413142
 Layer V BF (3)0.6 ± 0.2 (0.4—0.9)2.7 ± 1.7 (1—5)15.8 ± 5.8 (11—24)168 ± 41 (115—214)
 All (13)0.7 ± 0.4 (0.2—1.7)2.4 ± 1.3 (1—5)17.0 ± 5.9 (11—26)153 ± 27 (130—214)

Simple PPD at pyramid–pyramid connections

The analysis of paired pulse effects on EPSP integrals allowed the sixty-nine pyramid–pyramid EPSPs to be subdivided (see Table 1). In forty-three pairs the shape of the 2nd EPSP was similar to that of the 1st (e.g. Fig. 1). These forty-three pairs were further divided into those exhibiting simple PPD (32) (see Fig. 1) and those exhibiting simple PPF. To obtain an indication of the time course of PPD that was relatively uncontaminated by postsynaptic voltage-dependent events, only those layer V EPSPs in which no change in shape between the 1st and 2nd EPSP was apparent were plotted (n= 22 pairs, filled circles in Fig. 2). A linear fit to these points (r= 0.77, n= 53) crosses the line of equality (2nd EPSP = 1st EPSP) between 60 and 70ms after the 1st spike. The points obtained from postsynaptic layer III neurones showed a similar trend but were more widely scattered than those obtained from layer V neurones. Only comparative data points from layer III are included in this figure (see D-AP7 and 0 Mg2+ data below and open symbols in Fig. 2). A consistent finding was that between 70 and 100 ms, when the first phase of depression had apparently subsided, 2nd EPSPs remained smaller than 1st EPSPs (0.86 ± 0.074, n= 8).

Figure 1.

PPD is dependent on interspike interval at a pyramid–pyramid connection

A, upper traces are 4 typical superimposed presynaptic sweeps each consisting of a pair of spikes (1st spikes superimposed). Four different interspike intervals are shown (thick and thin traces alternating). The lower traces are composite averages of EPSPs triggered by single presynaptic spikes and EPSPs elicited by spike pairs (at the same 4 interspike intervals, thick and thin traces alternating), the average triggered from the rising phase of the 2nd spike. In this example, there is little change in EPSP shape when 1st and 2nd EPSPs are compared. PPD that recovers with increasing interspike interval is apparent. At the longest interspike interval illustrated (52 ms), the 2nd EPSP is as large as the 1st. B, upper traces in each case illustrate typical single presynaptic sweeps each containing a spike pair. Two different interspike intervals are shown (5 ms thick and 7 ms thin traces). The middle traces are composite averages of EPSPs triggered by single presynaptic spikes and EPSPs elicited by spike pairs (thick and thin traces are 5 and 7 ms, respectively) triggered from the rising phase of the 2nd spike. In this example, there is a modest reduction in EPSP duration at 7 ms (see subtractions illustrated below, where the single-spike EPSP and the subtracted 2nd EPSPs are superimposed for comparison). Paired pulse depression that declines with increasing interspike interval is apparent even with these very similar brief interspike intervals. Cell pair number indicated at bottom left of A and B.

As reported previously, the single-spike EPSPs of these connections were consistent in shape from sweep to sweep at postsynaptic membrane potentials more negative than −65 to −70 mV. In all recordings in which 2nd EPSPs were similar in shape to 1st EPSPs, 2nd EPSPs were also found to be consistent in shape from sweep to sweep. Where 2nd EPSPs had the same time course as 1st EPSPs, blockade of NMDA receptors resulted in no significant change in paired pulse effects despite reductions in amplitude and/or time course (not illustrated).

PPF at pyramid-pyramid connections

In eleven of sixty-nine pyramid–pyramid pairs studied, PPF was apparent even when the 2nd EPSP appeared the same shape as the 1st. In these, the 2nd EPSP integral was between 1.1 and 2.0 times the 1st EPSP integral at interspike intervals between 5 and 20 ms (see Table 1). This PPF, although weaker, appeared to decay with the same time course as that seen in postsynaptic interneurones (see below and Fig. 8). However, unlike pyramid-interneurone connections, this PPF was followed by slight PPD at interspike intervals between 70 and 100 ms and in pyramid–pyramid pairs 3rd EPSPs in brief trains were consistently depressed after 2nd EPSP facilitation (see below).

Involvement of NMDA receptors in prolongation of 2nd EPSPs

In another twenty-five pyramid–pyramid pairs the 2nd EPSP was of longer duration than the 1st. In sixteen of these, the two-spike EPSP integral was more than twice the single-spike EPSP integral (see Fig. 3). To determine the extent to which voltage-dependent and particularly NMDA receptor-mediated events contributed to this prolongation, these EPSPs were challenged with postsynaptic hyper-polarization (3 layer III and 1 layer V pyramid), NMDA antagonists (D-AP5 in 1 layer V and D-AP7 in 2 layer III pyramids) and by removing extracellular Mg2+ (1 layer III pyramid). All these procedures removed the apparent PPF and resulted in a 2nd EPSP whose shape was now similar to that of the 1st (see Fig. 3) and whose amplitude and integral were smaller than those of the 1st (see Fig. 2). The open symbols in Fig. 2 illustrate the effect on some of these EPSPs of removing this component. The EPSPs recorded under these test conditions now fall within the wider population that had exhibited simple PPD under control conditions. This form of facilitation is therefore mediated postsynaptically and is here termed 2nd EPSP ‘boosting’ to distinguish it from presynaptically mediated PPF.

Figure 3.

Second EPSP ‘boosting’ at a pyramid–pyramid connection is removed by NMDA receptor blockade which reveals paired pulse and brief train depression

A, the upper trace shows a typical single presynaptic sweep. Each of the lower records is a composite of 4 averages. EPSPs elicited by single presynaptic spikes (dotted lines indicate their decay) in control and in D-AP5 and EPSPs initiated by pairs of presynaptic spikes (averages triggered by the rising phase of the 2nd spike) in control and in D-AP5. A comparison of control and D-AP5 is shown for two postsynaptic membrane potentials, −60 and −70 mV. D-AP5 removes the 2nd EPSP ‘boosting’ and reveals the underlying PPD. B, in each case the upper trace is a typical single presynaptic sweep and the lower traces are composite averages of single-, 2-, 3- and in the lower example, 4-spike EPSPs triggered by the 1st, 2nd, 3rd and 4th spike, respectively.

As before, the 1st EPSP was consistent in shape from trial to trial, the standard deviation time course mirrored the shape of the EPSP and averages of the smallest single-spike EPSPs recorded had the same time course as averages of the largest in all but one unusual example. Second EPSPs in these twenty-five pairs were, however, less consistent from sweep to sweep and their time course appeared dependent on the amplitude of the 1st EPSP. As illustrated in Fig. 4A stronger ‘boosting’ or prolongation of the 2nd EPSP occurred after the larger 1st EPSPs. This effect was blocked by D-AP5 in the one pair so tested (Fig. 4B).

Figure 4.

Prolonged time course of 2nd EPSPs can be dependent on the amplitude of the 1st EPSP at a pyramid–pyramid connection, as well as on NMDA receptor activation

A, upper trace illustrates a typical single presynaptic sweep containing a spike pair. The lower traces are composite averages of EPSPs triggered by single presynaptic spikes and EPSPs elicited by spike pairs, the averages triggered from the rising phase of the 2nd spike. Before these averages were computed, 2 subsets of single sweeps were selected in which the 1st EPSP was smaller than average (thin trace) and in which the 1st EPSP was larger than average (thick trace). The shape of the 2nd EPSP (obtained by subtracting the averaged single-spike EPSP from the averaged 2-spike EPSP) was found to depend on the amplitude of the 1st EPSP (see comparison below), the larger the 1st EPSP, the more prolonged was the 2nd. B, this effect was lost when NMDA receptors were blocked. In the presence of d-AP5, the 2nd EPSPs that followed small 1st EPSPs were, on average, larger than those that followed large 1st EPSPs and there was no longer any difference apparent in their shape.

Prolonged 2nd EPSPs with PPD

Second EPSP prolongation associated with PPD (both of amplitude and integral) was observed in nine of the sixty-nine pairs and was also sensitive to membrane potential. The shape of the 2nd EPSP was strongly influenced by interspike interval in both groups exhibiting 2nd EPSP prolongation and the ‘boosting effect’ appeared strongest when the 2nd EPSP summed with the early falling phase of the 1st (see Fig. 5).

Figure 5.

Prolonged time course of 2nd EPSPs is dependent on interspike interval at a pyramid–pyramid connection

A, upper traces in each case illustrate typical single presynaptic sweeps containing spike pairs at 3 interspike intervals (thick and thin traces alternating). The lower traces are composite averages of EPSPs triggered by single presynaptic spikes and EPSPs elicited by spike pairs (at the 3 interspike intervals, thick and thin traces alternating), the averages triggered from the rising phase of the 2nd spike. Similar composites are shown for two postsynaptic membrane potentials, −64 and −68 mV. ‘Boosting’ of the EPSP duration can be seen to depend on both interspike interval and membrane potential. B, depression of the 3rd EPSP in the pair illustrated in A.

Saturating EPSPs

In four of the EPSPs in which similar interspike intervals could be studied at more than one postsynaptic membrane potential, an apparent voltage saturation effect was observed. That is, when the postsynaptic membrane potential was more negative than −70 mV the 2nd EPSP added significantly to the normalized EPSP integral (0.39–1.08 the integral of the 1st EPSP) while at potentials positive to −60 mV, the 2nd EPSP added much less to the integral (0.05–0.33). Figure 6 illustrates the strongest example recorded. The 2nd EPSP was so small at depolarized membrane potentials that this pair could not be assigned to any of the preceding categories as the shape of the 2nd EPSP could not be analysed.

Figure 6.

At pyramid–pyramid connections EPSPs can saturate

Averaged postsynaptic responses to single spikes and spike pairs, recorded at two membrane potentials (−70 and −54 mV) are illustrated. At −70 mV, the 2nd spike clearly elicits an EPSP. At −54 mV, despite significant voltage-dependent enhancement of the 1st EPSP, almost no additional voltage response was elicited by the 2nd spike.

Brief trains of EPSPs in pyramid–pyramid connections

Brief, three-spike trains were examined in twenty-five pyramid–pyramid pairs including examples of all the subtypes discussed above. A 3rd spike occurring within 30 ms of the 1st spike elicited an EPSP that added between 0.04 and 1.50 to the normalized EPSP integral (0.57 ± 0.48). Between 30 and 60 ms the 3rd EPSP added between 0.43 and 1.60 (0.75 ± 37.2) and between 60 and 110 ms, between 0.56 and 1.34 (0.96 ± 0.27). The result of this was that three-spike EPSP integrals were rarely as large as three times the single-spike EPSP integral (2.49 ± 0.38 at < 30 ms) (2.57 ± 0.50 at 30.60 ms) unless the train lasted > 100 ms. Even in the pairs that exhibited PPP, the 3rd EPSP of a brief train was smaller than the 1st (0.71 ± 0.19, n= 4) when it occurred within 60 ms of the 1st spike. Thus although some pyramid-pyramid EPSPs displayed PPP, the profound and incrementing brief train facilitation seen in pyramid–interneurone pairs (see below) was absent at pyramid-pyramid connections in 2.5 mm Ca2+ (see Figs 3B) and 4th and 5th spike EPSPs were smaller than 1st spike EPSPs when they occurred within 100 ms of the 1st spike (see Fig. 3B).

Pyramid–interneurone connections

To provide a comparison with the pyramid–pyramid pairs, nineteen pyramid–interneurone pairs recorded under similar conditions are included in this study. With 50–200 sweeps contributing to each average, larger differences in average amplitude (e.g. > 25%) are required in pyramid-inter-neurone EPSPs to reach significance, than in pyramid–pyramid EPSPs, since the c.v. is typically > 0.8 and EPSP amplitude distributions are often skewed in fast spiking interneurones.

PPF and brief train facilitation at pyramid-interneurone connections

As reported previously and in marked contrast to pyramid–pyramid EPSPs, paired pulse and frequency-dependent brief train facilitation are profound at connections from pyramids to two broad classes of interneurones, classical fast spiking interneurones with smooth, beaded dendrites (n= 10 pairs) and low threshold spiking, or burst-firing interneurones with sparsely spiny, non-beaded dendrites (n= 5) (see Kawaguchi, 1993, 1995; Kawaguchi & Kubota, 1993, for interneurone classification; Thomson et al. 1993b; Thomson, West & Deuchars, 1995).

In these fifteen pyramid–interneurone connections, paired pulse and brief train facilitation were dependent on inter-spike interval/train duration (Figs 7 and 8). No difference was apparent when EPSPs recorded in classical fast spiking and burst-firing interneurones were compared in this regard and the data were therefore pooled in Fig. 8 (see also Table 2). The plots illustrated in Fig. 8 give an indication of the time course of this facilitation. The dashed lines on the figure show the double-log fits obtained for two-, three- and four-spike PPSPs. The line fitted to the points obtained for two-spike PPSP integrals (r = 0.79, n= 31) reaches equality (2nd PPSP = 1st PPSP) between 50 and 60 ms after the 1st spike. When three-spike PPSPs were plotted (r= 0.68, n= 34) facilitation lasted longer, between 90 and 100 ms after the 1st spike when the three-spike PPSP integral had declined to 3 times the 1st PPSP integral. With four-spike PPSPs (r= 0.71, n= 12) facilitation lasted > 2 s. The points included in Fig. 8 indicate that the effects of trains of five and six spikes may also last for several seconds. This was further indicated when single sweeps were selected according to the number of spikes in the train preceding that sweep (by either 1 or 3 s). First EPSPs that followed trains of five or more spikes were, on average, larger than those that followed two- to three-spike trains and included a smaller proportion of transmission failures. For example, in one interleaved data set, 1st EPSPs that were elicited 3 s after trains of 2.3 spikes were 0.77 ± 0.91 mV in average amplitude and included forty-nine failures of transmission in 138 trials (35-5% failures). Three seconds after a five- to six-spike train, the average 1st EPSP was larger, 1.17 ± 1.01 mV and included fewer failures (48 in 230 trials; 21% failures; Fig. 9). These very large standard deviations and high proportions of apparent failures of transmission are typical of these pyramid-interneurone EPSPs (Thomson et al. 1993b, 1995).

Figure 7.

PPF is dependent on interspike interval at a pyramid–interneurone connection

Upper traces in each case illustrate typical single presynaptic sweeps each consisting of a pair of spikes. Three different interspike intervals are shown (8, 11 and 17 ms, thin and thick traces alternating). The lower traces are composite averages of EPSPs triggered by single presynaptic spikes and EPSPs elicited by spike pairs (at these 3 interspike intervals, thin and thick traces alternating), the average triggered from the rising phase of the 2nd spike. Paired pulse facilitation that declines with increasing interspike interval is apparent.

Table 2. Properties of pyramid—interneurone EPSPs
Paired pulse effectLayer and firing characteristics (n cell pairs)EPSP amplitude (mv)EPSP rise time (ms)EPSP width at half-amplitude (ms)2nd EPSP integral as %1st EPSP integral*
  1. Values are means ±s.d. with the range shown in parentheses. Pyramid—interneurone pairs are subdivided according to the paired pulse effects elicited and by the cortical region, layer and firing characteristics of the postsynaptic interneurone: fast spiking (FS) or low threshold spiking (LTS) interneurones. The amplitude of the 1st EPSP (after a 3 s interval), 10—90% rise time and width at half-amplitude were measured at membrane potentials between −65 and −75 mV. Where the normalized 2nd EPSP integral appears in square brackets, only an interspike interval between 30 and 50 ms was available for analysis and these values are not included in the means given. The pyramid—interneurone EPSPs that exhibited PPD are also described by the arrangement of the connection, i.e. these EPSPs either involved cells that were reciprocally (Rec.) connected with their pyramidal partner, or were layer V interneurones receiving descending input from a layer III pyramidal cell. Neither of these characteristics applied to any of the EPSPs exhibiting PPF. * Interspike interval, 5—20 ms.

Simple PPF
  SomatosensoryLayer III FS (4)0.64 ± 0.37 (0.2—1.2)0.8 ± 0.5 (0.2—1.6)3.9 ± 1.4 (2.3—6.0)339 ± 24 (413—447)
 Layer III LTS (0)
 Layer V FS (2)0.06, 0.651.0, 2.06.5, 6.7700, 1600
 Layer V LTS (2)0.15, 0.970.4, 0.44.0, 7.0385, 386
  MotorLayer III FS (0)
 Layer III LTS (1)0.410.84.9333
 Layer V FS (2)0.1, 0.3, 0.30.6, 1.0, 0.87.5, 6.5, 6.0[286] 1092, 1650
 Layer V LTS (1)
  VisualLayer III FS (1)[161]
 Layer III LTS (1)
 Layer V FS (0)
 Layer V LTS (0)
 All FS cells (10)0.55 ± 0.46 (0.1—1.5)0.91 ± 0.50 (0.2—1.6)5.09 ± 1.91 (1.9—7.0)840 ± 514 (413—1650)
 All LTS cells (5)0.45 ± 0.29 (0.15—0.97)0.68 ± 0.30 (0.4—1.2)6.18 ± 1.49 (4.0—8.0)528 ± 160 (333—838)
Simple PPD
  SomatosensoryIII to V FS (1)2.40.3320
 Rec. III to III FS (1)1.23.21547
  MotorRec. III to III FS (1)
  VisualIII to V FS (1)
 All (4)1.07 ± 0.84 (0.3—2.4)1.32 ± 1.11 (0.3—3.2)7.22 ± 4.64 (3—15)58 ± 27 (20—90)
Figure 9.

Incrementing brief train facilitation at a pyramid–interneurone connection can last for 3 s

In each case the upper trace is a typical single presynaptic sweep and the lower traces are composite averages of single-, 2-, 3-, 4-, 5- and, in B, 6-spike EPSPs triggered by the 1st, 2nd, 3rd, 4th, 5th and 6th spikes, respectively. Note the additional enhancement of the 3rd EPSP compared with the 2nd EPSP and the maintenance of this facilitation throughout the train, despite lengthening interspike intervals. The responses illustrated in A followed 2- to 3-spike trains and those in B followed 5- to 6-spike trains at intertrain intervals of 3 s. Note the enhancement of the 1st EPSP in B compared with that in A.

NMDA receptors are not responsible for facilitation at pyramid–interneurone connections

Also illustrated in the plot in Fig. 8 and the EPSPs in Fig. 10 are the data obtained from one layer III pyramidal cell-burst-firing interneurone pair challenged with the NMDA receptor antagonist D-AP7 (30 μm, 15–25 min). Although blockade of NMDA receptors reduced the average EPSP amplitude and integral, it only slightly reduced PPF and brief train facilitation. In addition, PPF and brief train facilitation were not significantly affected by postsynaptic membrane potential at these connections (compare Figs 9 and 10), despite changes in average amplitude.

Figure 10.

PPF at pyramid–interneurone connections is not dependent on NMDA receptor activation

The upper trace is a typical single presynaptic sweep showing a 3-spike train. Below are composite postsynaptic averages comprising EPSPs elicited by single presynaptic spikes, EPSPs elicited by double spikes (averages triggered from the rising phase of the 2nd spike) and EPSPs elicited by spike triplets (triggered from the 3rd spike), recorded under control conditions and in the presence of the NMDA antagonist D-AP7 (30 μm, 15–25 min). This drug significantly reduces EPSP amplitude but has little effect on paired pulse, or brief train facilitation.

PPD at pyramid–interneurone connections

The present larger sample includes four EPSPs that were elicited in fast spiking interneurones by simultaneously recorded pyramids and unusually demonstrated PPD. It is at present unclear why these EPSPs are different, but all these connections were atypical within the present sample in that they represented either a rare reciprocal connection (2 of 48 one-way connected pairs, including interneurone-to-pyramid connections), or a descending input from layer III to layer V (see Table 2). All fifteen pyramid–interneurone connections described above that exhibited PPF were one-way connections within a single layer. A much larger sample is required before the neocortical interneurones whose pyramidal inputs exhibit PPD can be classified.

Changes in extracellular Ca2+ at pyramid–pyramid connections

In one pyramid–pyramid pair that had exhibited slight PPF in 2.5 mm Ca2+, extracellular Ca2+ was changed to 1 mm and then to 5 mm (Fig. 11). In 1 mm Ca2+ the EPSP exhibited PPF at an interspike interval of 9 ms (normalized two-spike EPSP integral 2.5), but not at 60 ms. The three-and four-spike EPSPs exhibited brief train facilitation. The normalized three-spike EPSP integral was 3.9 at 80–100 ms and the normalized four-spike EPSP integral was 5 at 100–120 ms, the 4th EPSP being similar in size to the 3rd. In 5 mm Ca2+ PPD was apparent at 10 and at 60 ms, i.e. the normalized two-spike EPSP integrals were 1.8 and 1.9, respectively. Normalized three-spike EPSP integrals were < 2.5 at < 60 ms, but at 100 ms 3rd EPSPs were only slightly depressed, resembling the other pyramid–pyramid connections displaying simple PPD described above. Therefore, by lowering extracellular Ca2+ the pyramid-pyramid connections can be brought to resemble pyramid–interneurone connections more closely. It should also be noted that despite paired pulse and brief train facilitation in 1 mm Ca2+, the postsynaptic response to brief trains as well as to single spikes was considerably smaller in 1 mm than in 5 mm Ca2+. The single-spike EPSP integral in 5 mm Ca2+ was 2.3 times that in 1 mm Ca2+ and even the three-spike EPSP integral was 1.7 times that in 1 mm Ca2+, indicating perhaps that only a proportion of the available release sites were utilized even during three- to four-spike trains in low Ca2+.


This study highlights some of the many factors that should be taken into account when the efficacy of synaptic transmission in a dynamic system is considered. Connections between pyramidal cells, in all layers and regions studied, displayed presynaptically mediated depression during brief spike trains, though this can be overcome early in the train by recruitment of voltage-dependent currents at some connections. In striking contrast, pyramidal inputs from pyramidal cells to certain classes of interneurones consistently displayed profound frequency-dependent, incrementing facilitation that appeared quite independent of postsynaptic mechanisms in the subthreshold range. The manner in which these two classes of connections respond to different patterns of presynaptic firing and to changes in firing pattern and rate are therefore likely to be very different.

Simple PPD and PPF at pyramid–pyramid connections

The present analysis confirms previous reports (Thomson & West, 1993; Thomson et al. 1993a; Markram & Tsodyks, 1996) that pyramid–pyramid connections (in 2.5 mm Ca2+, 1 mm Mg2+) typically display PPD (58 of 69 pairs studied). It extends the observations to visual cortex and quantifies and describes the time course of this depression further. In previous studies, this form of depression was deduced to be primarily presynaptic in origin from the results of three tests: proportion of failures, c.v. analysis and comparison of EPSP amplitude distributions (Thomson et al. 1993a; Thomson & Deuchars, 1995). This PPD lasts between 60 and 70 ms at 35–36 °C in adult neocortex and even in those pyramid–pyramid connections exhibiting PPF, 3rd spikes within 60 ms elicited a smaller EPSP than a single spike. Only when extracellular Ca2+, and thereby release probability, was lowered were 3rd and 4th EPSPs in brief trains larger than 1st EPSPs. This is consistent with previous suggestions that single release sites, having once released, are refractory for tens of milliseconds (Betz, 1970; Thomson et al. 1993a; Stevens & Wang, 1995). Whether refractoriness results from presynaptic autoreceptors or from some delay in a new vesicle binding to the active zone was not addressed directly. Approximately 10% depression was apparent between 70 and 100 ms in all pyramid–pyramid connections so studied and may represent postsynaptic receptor desensitization (Trussell et al. 1993) outlasting presynaptic refractoriness. The mechanism(s) required for presynaptically mediated PPF is, however, present at pyramid–pyramid connections, although it is often masked. Simple PPF was seen in 15% of such connections in 2-5 mm Ca2+ and in all pairs tested in low extracellular Ca2+. Presynaptic PPF is therefore likely to contribute to the recruitment of previously quiescent release sites, even where PPD dominates, probably blurring somewhat the measurement of simple PPD time course.

Voltage-dependent currents and synapse location may contribute to 2nd EPSP ‘boosting’

A large component of the voltage dependence of pyramid–pyramid EPSPs is blocked by NMDA receptor antagonists (Thomson, Girdlestone & West, 1988, 1989; Deuchars & Thomson, 1996). However, inward Na+ currents, more readily activated by the EPSP when the neurone is depolarized (Deisz, Fortin & Zieglgänsberger, 1991; Schwindt & Grill, 1995) and voltage-gated Ca2+ channels (Markram & Sakmann, 1994; Magee & Johnston, 1995; Magee et al. 1995) may also contribute, resulting in a larger, more prolonged local depolarization than is provided by the very brief AMPA/kainate receptor/channel current alone (Silver, Traynelis & Cull-Candy, 1992) and possibly also enhancing the NMDA receptor current.

These additive events result in the strongly non-linear voltage relations of pyramid–pyramid EPSPs which allow some tentative conclusions about interactions between active sites to be made. Even a few millivolts of depolarization close to an active synapse can dramatically increase the amplitude and duration of these EPSPs in the immediate subthreshold range. In the majority of connections, all 1st EPSPs are of similar duration and 2nd EPSPs are the same shape as 1st EPSPs at all subthreshold membrane potentials. It can therefore be concluded that the synaptic sites contributing to the 1st EPSP are electrotonically distant from each other and from those contributing to the 2nd. With widely separated synapses, even large voltage changes at one discrete site might produce only a small voltage change at another and events would sum more or less linearly at the soma, or in a common major dendrite. With release site refractoriness delaying a 2nd activation of each site, different synaptic sites and different dendritic compartments would be involved in subsequent EPSPs. Indeed, where the synaptic connections between single pyramidal cells have been reconstructed anatomically, many have been found to involve synapses distributed on several different second- or third-order dendritic branches (Deuchars, West & Thomson, 1994; Lübke, Markram, Frötscher & Sakmann, 1996).

Second EPSP ‘boosting’ was, however, seen in about one-third of pyramid–pyramid connections and appeared due to postsynaptic voltage-dependent events since it was removed by hyperpolarization or by removing the voltage-dependent NMDA receptor-mediated component. This ‘boosting’ could be large enough to obscure the underlying PPD, though it was not strong enough to overcome depression of 3rd and 4th EPSPs. Synapses concentrated in closely neighbouring compartments would interact more powerfully than those located further apart. The depolarization generated by one EPSP would depolarize closely neighbouring sites, facilitating activation of voltage-dependent currents and removal of Mg2+ blockade from NMDA receptor/channels during a 2nd EPSP.

‘Boosting’ was not, however, the only voltage-dependent effect observed; EPSPs could also saturate, presumably where the depolarization produced by one reduced the driving force for current flow during the next EPSP. It should also be remembered that the present data deal only with single axon inputs. Interactions between active sites are more likely to occur in situ when many presynaptic elements are active. The profoundly non-linear (both supra-and sublinear) summation that can occur in pyramidal dendrites will greatly influence the efficacy of coincident inputs. Widely dispersed inputs might sum near linearly, but concentrated inputs would follow a sigmoidal summation curve. Here, because pyramid–pyramid EPSPs can be very broad in the immediate subthreshold range, ‘boosting’ would enhance even loosely coincident events until voltage saturation was reached in that dendritic region. It is clear that voltage-dependent outward currents also shape EPSPs, but their contribution to paired pulse effects has not been determined.

Paired pulse and brief train facilitation at pyramid–interneurone connections

In striking contrast to the depression typical of pyramidal inputs to other pyramidal cells, some pyramid–interneurone connections display a dramatic frequency- and train-duration-dependent facilitation. As reported previously the probability of transmitter release at these synapses is extremely low. Total failures of transmission occur in response to a large proportion of 1st spikes (30 to > 80%; Thomson et al. 1993b, 1995), despite multiple synaptic contacts (Deuchars & Thomson, 1995). The plots illustrated in Fig. 8 indicate that PPP is dramatic, i.e. more than a 10-fold enhancement at the briefest interspike intervals, but declines rapidly, within 50 ms. The facilitatory effects of two and three preceding spikes (on 3rd and 4th EPSPs, respectively) were increasingly long lasting (up to 90–100 ms and > 2 s, respectively) and even at the longest intervals studied (3 s) a preceding five- to six-spike train facilitated subsequent EPSPs. In addition to these pre-synaptically mediated differences, pyramid-interneurone EPSPs differ from those recorded in pyramidal cells in their time course, typically half the duration of pyramidal EPSPs at membrane potentials around −70 mV. Moreover, since EPSPs in these interneurones display conventional voltage relations, their duration close to spike threshold is very much briefer than those occurring in pyramidal cells. Therefore, not only must a presynaptic pyramidal cell fire repetitively to excite these interneurones significantly, its discharge must be very nearly coincident with that of another repetitively firing pyramidal cell for summation to be effective.

PPD at pyramid-interneurone connections

Paired pulse depression at pyramid–interneurone connections in neocortex has not previously been reported. However, the existence of even a few such connections might call into question the hypothesis that the target neurone, via some retrograde signal, selects the presynaptic release probability and thereby the frequency-dependent characteristics of its inputs (see also Katz et al. 1993; Walrond et al. 1993; Davis & Murphey, 1993; Govind et al. 1994). However, of nineteen neocortical pyramid-inter-neurone EPSPs in which brief trains have been studied to date, only four exhibited PPD and these four connections were different also in other respects. Two were the only two layer III to layer V pyramid–interneurone connections so far studied. The others were the only two reciprocal connections so far encountered in forty-eight pairs displaying one-way connections (which includes 29 interneurone-to-pyramid pairs). Classification of interneurones in neocortex still presents some major ambiguities, but in hippocampus where axonal arbours are more readily classified, strong PPP has been consistently observed in one readily definable class of interneurones, while PPD is equally consistently observed in others (A. B. Ali & A. M. Thomson, unpublished observations). The postsynaptic determinant(s) of pre-synaptic release pattern remain to be determined; it is clearly more complex than simply whether glutamate or GABA is present postsynaptically.

Pyramidal discharge and release patterns

Numerically, pyramidal axon terminals represent a large component of the excitatory drive to other pyramidal cells and to interneurones, both via the very local connections described here and longer-range connections within and between cortical regions. After a period of quiescence, pyramidal cells typically discharge two to five spikes at high frequency before frequency adaptation slows discharge rates. Very high continuous firing rates would gain little at either of the major types of synapse discussed here. At pyramid–pyramid connections profound depression would result not only from the short term PPD discussed here, but from a more slowly developing form of presynaptic depression proposed to result from depletion of readily available transmitter (Thomson et al. 1993a; Thomson & Deuchars, 1995). At many pyramid–interneurone connections the facilitation that is first initiated by a brief high frequency spike train is then readily maintained at much lower frequencies. For maximum efficacy at both types of connection, therefore, brief, two- to three-spike bursts alternating with single spikes at intervals long enough to allow recovery from refractoriness (> 60–70 ms) but short enough for some facilitation to remain (< 90–2000 ms) would be ideal. On the other hand, for selective activation of pyramidal cells, tonic discharge at rates below 20 Hz, or for more selective activation of some interneurones, repetitive burst firing at mean rates above 30–40 Hz would be effective.

In vivo each pyramidal cell is likely to display a wide range of firing frequencies and patterns dependent on the inputs it receives and on, for example, arousal state. It remains to be determined how depressed pyramid–pyramid connections, or how facilitated pyramid–interneurone connections can be under these different conditions. This was the first example of very different release patterns at different terminals of a single axon to be described in the mammalian CNS. However, since there are now several invertebrate examples of phasic and tonic outputs from a single presynaptic axon (Katz et al. 1993; Walrond et al. 1993; Davis & Murphey, 1993; Govind et al. 1994), it appears that target-dependent modification of presynaptic release pattern has been well conserved through phylogeny and may be a widespread mechanism for tuning circuit activity throughout the CNS.


My thanks to The Wellcome Trust and the Medical Research Council for financial support. D-AP7 was a kind gift from Sandoz Pharma, Basel.