Neurons in the monkey dorsolateral prefrontal cortex (DLPFC) fire persistently during the delay period of working memory tasks. To determine how repetitive firing affects the efficacy of synaptic inputs to DLPFC layer 3 neurons, we examined the effects of repetitive presynaptic stimulation on the amplitude and temporal summation of EPSPs. Recordings were obtained in monkey DLPFC brain slices from regular spiking (RS) pyramidal cells and two types of interneurons, fast spiking (FS) and adapting non-pyramidal (ANP) cells. Repetitive stimulation of presynaptic axons in layer 3 caused EPSP depression in RS and FS neurons, but EPSP facilitation in ANP cells. A shorter EPSP duration produced weaker temporal summation in FS neurons compared to the other cell classes. Thus, due to the combined effects of dynamic changes in EPSP amplitude and differences in temporal summation, the effect of a presynaptic spike train differed according to the postsynaptic cell class. Similar results were obtained when recording unitary EPSPs evoked in connected pairs of presynaptic RS pyramidal cells and postsynaptic RS, FS or ANP neurons. In addition, similar differences in the efficacy of sustained inputs among cell classes were observed when delay-related firing was reproduced in vitro by stimulating inputs with the timing of spike trains recorded from the DLPFC of monkeys performing a delayed-response task. We suggest that the transition from baseline firing rates to higher frequency delay-related firing may lead to the differential activation of distinct cell populations, with corresponding significant effects on the patterns of activity in local prefrontal circuits.
The primate dorsolateral prefrontal cortex (DLPFC) is involved in the planning and execution of complex behaviors, many of which require working memory, the ability to transiently store and manipulate information in order to guide subsequent behavior (Baddeley, 1986). In vivo recordings from the DLPFC of monkeys during delayed-response tasks, which include a delay between the presentation of a cue and the execution of a response guided by cue-related information, have provided insight into the neural basis of working memory. In these tasks, some DLPFC neurons show an increase in firing rate that is sustained throughout the delay period (Fuster and Alexander, 1971; Kubota and Niki, 1971). Sustained firing also occurs in other areas of the brain, but in the primate DLPFC it is more prominent, correlates with behavioral performance and persists in the presence of distractors that disrupt delay-period activity in other cortical regions (Miller and Cohen, 2001). These observations lead to the hypothesis that sustained firing of primate DLPFC neurons is the cellular basis for the active maintenance of information during working memory (Goldman-Rakic, 1995).
Sustained firing in DLPFC has been hypothesized to be generated and maintained locally by reverberant synaptic interactions in recurrent excitatory circuits (Goldman-Rakic, 1995; Lewis and Anderson, 1995; Durstewitz et al., 2000; Gutkin et al., 2001; Wang, 2001). Because a purely excitatory recurrent network is prone to runaway excitation (Nelson and Turrigiano, 1998), sustained firing must involve recruitment of GABA neurons. Consistent with this idea, both pyramidal cells and putative GABAergic interneurons exhibit sustained delay period activity in monkey DLPFC (Wilson et al., 1994; Rao et al., 1999, 2000). Moreover, data from single unit recordings in vivo suggest that delay period activity of both pyramidal cells and interneurons is driven by local excitatory synaptic inputs (Rao et al., 1999).
The efficacy of an excitatory synaptic input can be defined as the ability to depolarize the action potential initiation site, and thus to increase the probability of cell firing. The peak amplitude of the excitatory postsynaptic potential (EPSP) recorded at the soma is a good estimate of synaptic efficacy since the soma is electrically well-coupled with the action potential initiation site, which appears to be localized in proximal segments of the axon, in both pyramidal cells (Colbert and Johnston, 1996) and interneurons (Martina et al., 2000). Previous studies suggest that EPSP amplitude and thus the efficacy of local excitatory synaptic inputs may change significantly during periods of repetitive firing, such as that observed during the delay period of delayed-response tasks. In most types of synapses, repetitive presynaptic firing elicits either significant short-term depression (decrease) or facilitation (increase) of the EPSP amplitude (Zucker and Regehr, 2002). The postsynaptic impact of a presynaptic spike train differs at depressing versus facilitating synapses because they may act like low- or high-pass filters, respectively, and thus have different preferred frequencies at which synaptic transmission is optimized (Fuhrmann et al., 2002). In rat neocortex and hippocampus, facilitation or depression are observed at excitatory synapses in a target cell-specific manner. EPSPs typically show depression in pyramidal cells (Markram, 1997; Thomson and Deuchars, 1997), whereas EPSPs depress in certain subclasses of interneurons and facilitate in others (Thomson, 1997; Ali et al., 1998; Ali and Thomson, 1998; Markram et al., 1998b; Reyes et al., 1998; Kozloski et al., 2001; Rozov et al., 2001; Losonczy et al., 2002; Wang et al., 2002).
In addition to transient changes in EPSP amplitude, the efficacy of sustained synaptic inputs also depends on the extent of temporal summation of EPSPs, which in turn depends on the relation between presynaptic interspike intervals and EPSP duration. Thus, a presynaptic spike train could yield significantly different levels of postsynaptic depolarization if depression or facilitation are counteracted or enhanced by temporal summation. In rat cortical neurons, duration of EPSPs was found to differ between pyramidal cells and interneurons (Geiger et al., 1997; Angulo et al., 1999; Fricker and Miles, 2000; Maccaferri and Dingledine, 2002). However, in spite of the observation that EPSP duration may differ between certain interneurons and pyramidal cells, no studies have examined whether the effects of temporal summation interact with facilitation and depression in a target- or input-specific manner in the cortex.
If the efficacy of synaptic inputs is influenced by sustained stimulation in a cell class-specific manner, then sustained firing could lead to preferential activation of distinct cell populations, and have a significant impact on the cellular interactions within DLPFC local circuits. For example, short-term facilitation at excitatory connections could contribute to synaptic reverberation in recurrent circuits. Conversely, depression of excitatory synapses onto pyramidal cells or facilitation of excitatory synapses onto GABA neurons might act as negative feedback mechanisms. Theoretical studies have indeed proposed that short-term changes in synaptic efficacy may both drive the DLPFC network into attractor states that result in sustained firing and control the network firing rate in these states (Gutkin et al., 2001; Wang, 2001).
All previous studies of the mechanisms that influence the efficacy of sustained synaptic inputs were performed on neurons from rat cortex, mostly at immature stages of development. In addition, it has been suggested that non-primate mammals lack a neocortical region homologous or analogous to the DLPFC of primates (Preuss, 1995). Furthermore, the role of GABAergic neurons in the regulation of activity in prefrontal local circuits could differ between primates and rodents, since substantial species differences exist in the proportions and developmental origin of different interneuron subclasses (Conde et al., 1994; Gabbot and Bacon, 1996a,b; Anderson et al., 1997; Gabbot et al., 1997; Letinic et al., 2002). Thus, understanding the effects of sustained activation on synaptic efficacy in the adult monkey DLPFC in vitro would provide important insights into the local circuit dynamics accompanying the temporal patterns of activity observed in the adult primate DLPFC in vivo.
Consequently, we employed brain slice techniques to obtain tight-seal whole-cell recordings from visually identified neurons in layer 3 of DLPFC (areas 46 and 9) of adult macaque monkeys in order to address the following questions: Does sustained presynaptic firing differentially induce short-term EPSP facilitation or depression in pyramidal cells and specific classes of interneurons? Do differences in EPSP duration, and therefore temporal summation, influence sustained inputs in a cell-class specific manner? How does stimulation with natural patterns of delay-related activity affect the efficacy of excitatory inputs?
Materials and Methods
DLPFC slices (n = 89) were obtained from 15 young adult (3.5–6 kg; 4–5 years old) male cynomolgus monkeys (Macaca fascicularis) treated according to the guidelines outlined in the National Institutes of Health Guide for the Care and Use of Laboratory Animals, as approved by the University of Pittsburgh Institutional Animal Care and Use Committee. Tissue slices from these animals also were used in other studies (González-Burgos et al., 2000; Henze et al., 2000; Melchitzky et al., 2001; Urban et al., 2002). Animals were treated with ketamine hydrochloride (25 mg/kg, i.m.), dexamethasone phosphate (0.5 mg/kg, i.m.), and atropine sulfate (0.05 mg/kg, subcutaneous); an endotracheal tube was inserted and anesthesia was maintained with 1% halothane in 28% O2/air. Monkeys were placed in a stereotaxic apparatus and a craniectomy was performed over the DLPFC. The dura was removed in a location determined by stereotaxic coordinates and by the position of relevant sulcal landmarks, and a small block of tissue was excised containing a portion of dorsal area 9 and both the medial and lateral banks of the principal sulcus (area 46). After the surgery, the animals were treated with an antibiotic (chloramphenicol, 15 mg/kg, i.m.) and an analgesic (hydromorphone, 0.02 mg/kg, i.m.) three times a day for 3 days. All animals recovered quickly with no impairments in eating or drinking nor overt behavioral deficits. In most cases, the animals underwent the same procedure 2–4 weeks later to obtain tissue from the opposite hemisphere. During the second procedure, after the craniectomy, the animal was given an overdose of pentobarbital (30 mg/kg) and was perfused through the heart with ice-cold modified ACSF. A tissue block containing portions of areas 9 and 46 nonhomotopic to the first biopsy was quickly excised. Subsequent treatment of the tissue was the same for both days.
The tissue blocks were placed in an ice-cold solution composed of (in mM): 230 sucrose, 1.9 KCl, 1.2 Na2HPO4, 33 NaHCO3, 6 MgCl2, 1 CaCl2, 10 glucose and 2 kynurenic acid; pH 7.3–7.4 bubbled with 95% O2-5% CO2. Slices of 350 or 400 µm thickness were cut in the coronal plane and incubated at room temperature for at least 2 h in a solution consisting of (in mM): 126 NaCl, 2 KCl, 1.2 Na2HPO4, 10 glucose, 25 NaHCO3, 6.0 MgCl2 and 1.0 CaCl2. For recordings, slices were transferred to a submersion chamber and superfused with oxygenated ACSF (in mM: 126 NaCl, 2.5 KCl, 1.2 Na2HPO4, 25 NaHCO3, 2.0 CaCl2, 1.0 MgCl2, 10 glucose) at 32–33°C. Some recordings were done in the presence of the anti-oxidant sodium metabisulfite (75 µM). At this concentration, the anti-oxidant does not produce significant effects on cellular excitability or synaptic transmission (G. González-Burgos, unpublished observations; Sutor and ten Bruggencate, 1990).
Neurons were identified visually in layer 3 using infrared illumination and differential interference contrast optics (Stuart et al., 1993). Patch pipettes (4–7 MΩ) were pulled from borosilicate capillary glass and filled with the following internal solution (in mM): 120 methylsulfate, 10 KCl, 10 HEPES, 0.2 EGTA, 4.5 ATP, 0.3 GTP and 14 phosphocreatine, pH 7.2–7.3. Tight seal (seal resistance > 5 GΩ) whole-cell voltage recordings were obtained with Axoclamp-2A (Axon Instruments, Union City, CA) or BVC-700A (Dagan Corporation, Minneapolis, MN) amplifiers operating in bridge mode and employing capacitance compensation. Signals were low-pass filtered at 3 kHz, digitized at 10 or 20 kHz, and stored on disk for off-line analysis. Data acquisition and analysis were performed using customer-made programs written in LabView (National Instruments, Austin, TX).
Intrinsic membrane properties were determined from the voltage responses elicited by injection of series of current steps of 500 ms duration and amplitudes starting at –100 pA and reaching values between 400 and 1000 pA in 10 pA increments. Input resistance was determined from the slope of a linear regression fit to the linear portion of the relation between injected current (usually between –50 and –10 pA) and the voltage deflection near the end of the 500 ms step. Membrane time constant was determined by fit of a single exponential to the on or off voltage response to hyperpolarizing current steps of –10 pA. Even within these small current steps there was evidence of contribution of hyperpolarization activated currents to the voltage transients (see Fig. 1), suggesting that active membrane properties participate in determining the decay of voltage transients elicited from the cells’ resting membrane potential. Properties of single action potentials were measured using current steps close to threshold for each individual cell, which usually elicited either one or a few action potentials.
Simulated EPSPs were generated by injecting, into the soma, current with a time course similar to an excitatory synaptic current. Injection of synaptic-like current waveforms was done as described previously (González-Burgos and Barrionuevo, 2001). The kinetics of the current waveforms was adjusted to reproduce fast EPSPs recorded from one fast-spiking (FS) cell. Thereafter, parameters were kept constant and injected into other FS cells and neurons of the other classes.
Synaptically connected pairs were identified during simultaneous recording from two to four neurons (Urban et al., 2002). Action potentials were evoked in the presynaptic pyramidal neurons by injecting short (3 ms) suprathreshold current steps that elicited spikes with little trial-to-trial variability. Once the presence of a synaptic connection was established, 50–200 single EPSPs were collected at a stimulation frequency of 0.1 or 0.05 Hz. Subsequently, trains of presynaptic spikes were evoked following protocols identical to those described in the next section for the focal stimulation experiments (see below).
Presynaptic fibers were stimulated using borosilicate theta-glass pipettes (1.5 mm outside diameter, Warner Instruments Corporation, Hamden, CT) pulled to a tip diameter of 2–3 µm and half-filled with freshly oxygenated extracellular solution. Chlorided silver wires placed inside each compartment of the theta glass and connected to a stimulus isolation unit (Model A350D-A, World Precision Instruments, Sarasota, FL) were used to apply bipolar focal stimulation. Timing and duration of the stimulation pulses were digitally controlled with a personal computer running customer-written Labview programs. The stimulation electrode was placed in layer 3 at ∼100 µm above (or below) and ∼50–100 µm lateral to the soma of the recorded neurons. Its position was finely adjusted by means of a motorized micromanipulator (Model MP-285, Sutter Instruments Co., Novato, CA). Stimulation current parameters (duration between 50 and 500 µs; intensity between 20 and 300 µA) and final position were adjusted to elicit small amplitude responses resembling monosynaptic EPSPs, as described in results. As in our previous studies (González-Burgos et al., 2000; Melchitzky et al., 2001) we found no evidence of spontaneous epileptiform activity during GABAA receptor blockade by 10 µM bicuculline (n = 16 slices). In contrast with our previous studies using metal stimulation electrodes (González-Burgos et al., 2000; Melchitzky et al., 2001) here we found no hyperexcitability upon stimulation in the presence of bicuculline. Because the EPSPs evoked by focal stimulation had amplitudes similar to those of unitary EPSPs, it is likely that focal extracellular stimulation as employed here activated only a few presynaptic axons. Therefore, the absence of paroxysmal responses upon stimulation is probably due to the fact that focal stimulation activated a very limited portion of the excitatory network in the slice.
In most experiments, biocytin (0.2–0.5%) was included in the internal solution. After recordings were finished, slices were incubated for 5–20 min at 32–33°C and then fixed in 4% paraformaldehyde in 0.1 M phosphate-buffered saline. The fixed slices were then transferred to 0.1 M Na-phosphate buffer, serially resectioned at 50 µm, and processed for visualization of the biotin label using the Vectastain Elite ABC kit (Vector Laboratories Inc., Burlingame, CA) and diaminobenzidine. Neurons were reconstructed employing the Neurolucida tracing system (MicroBrightField, Inc., Williston, VT).
EPSP Data Analysis
To estimate the EPSP amplitude during repetitive stimulation excluding EPSP summation, we measured the peak EPSP as the membrane potential at the peak of each EPSP minus the potential measured immediately prior to the EPSP onset. For EPSPs after the first in a train, this measure underestimates true EPSP amplitude, therefore overestimates depression and underestimates facilitation. The magnitude of this error depends on the relation between EPSP rise time and decay: the faster the EPSP rise relative to decay, the smaller the error introduced. To estimate the error introduced by the peak EPSP measure, we selected recordings in which single EPSPs had 20–80% rise times that were among the fastest and slowest in our data sample (1.5 and 3.1 ms, respectively). In these recordings, we elicited first single EPSPs and then two EPSPs using paired-pulse stimulation at 20 Hz. The amplitude of the second EPSP elicited by the paired-pulses was then estimated in two ways. First, we measured the peak EPSP as described above. Second, we subtracted the average of the traces with a single EPSP from the average of the traces with EPSP pairs. This subtraction yields the second EPSP in isolation thus estimating the amplitude excluding under- or over-estimation of the true EPSP amplitude. The peak EPSP was 96% and 83.5% of the EPSP amplitude estimated by subtraction of traces, for the fast and slow rising EPSPs, respectively. Therefore, the peak EPSP provides a reasonable estimation of the EPSP amplitude. Identical conclusions were made in a previous study in which peak EPSP (measured as described above) was compared with estimations of the EPSP amplitude obtained after extrapolation of the decay of the preceding EPSPs in the train (Markram, 1997).
To quantify the actual depolarization elicited by each EPSP in a train, including the effects of temporal summation, we measured the peak depolarization as the membrane potential at the peak of each EPSP minus the resting potential measured immediately prior to the onset of the first EPSP in the train. Both peak EPSP and peak depolarization were measured from average traces obtained from at least 10 but in most cases 20 repetitions of each stimulus protocol. Consecutive traces were averaged independently of the presence or absence of response failures, because failure rate cannot be estimated accurately for responses evoked with extracellular stimulation of afferent fibers (see Results). Thus, the contribution of changes in failure rate during repetitive stimulation to depression and facilitation could not be estimated. Duration of single EPSPs was determined on average traces, by measuring the width of the EPSPs at half the maximal amplitude. Measurements of EPSP amplitude an peak depolarization were done on average traces most experiments
Results are expressed as means ± SEM unless otherwise indicated. Statistical significance of differences between group means was tested using t-test, paired t-test and one- or two-way ANOVA followed by contrasts, as stated in each case.
Electrophysiological and Morphological Characteristics of Pyramidal Cells and Two Classes of Interneurons in Layer 3 of Monkey DLPFC
In slices prepared from DLPFC of young adult macaque monkeys, we visually identified layer 3 pyramidal cells and putative interneurons. The latter had a small soma, rounded or fusiform in shape, and no evidence of a thick apical dendrite. Neurons identified as pyramidal exhibited a firing pattern (Fig. 1A) typical of regular spiking (RS) pyramidal cells (Connors and Gutnick, 1990). Most cells identified visually as interneurons exhibited spikes with short duration, prominent afterhyperpolarization (AHP) and little or no spike frequency adaptation (Fig. 1B) regardless of the current intensity injected (data not shown). These electrophysiological features are consistent with those of fast spiking (FS) interneurons (Connors and Gutnick, 1990). In a second electrophysiological type of interneurons, termed adapting non-pyramidal (ANP), the spike frequency decreased substantially during injection of depolarizing current (Fig. 1C). The intrinsic membrane properties of FS and ANP neurons were distinct from each other and from RS pyramidal cells (Table 1). Many of the intrinsic membrane properties described for FS and ANP cells (Table 1) differ significantly from those described for interneurons with or without spike frequency adaptation in rat cortex (Gupta et al., 2000; Wang et al., 2002). In particular, the absence of a prominent low threshold spiking population of interneurons contrasts markedly with findings in the cortex of immature rats (Kawaguchi and Kubota, 1997; Gupta et al., 2000; Beierlein et al., 2000; Galarreta and Hestrin, 2001; Xiang et al., 2002). Whether these contrasts reflect differences in developmental stage, cortical area or species remains to be investigated.
Morphological analysis of the recorded neurons (Fig. 1D–F) confirmed that RS neurons were pyramidal (Fig. 1D), whereas FS and ANP cells had the morphological features of interneurons (Fig. 1E,F). Interneuron dendrites were mostly smooth and clearly distinguishable from those of pyramidal cells, however dendritic structure did not clearly distinguish FS from ANP cells. The axonal arbor in FS neurons tended to spread horizontally within layers 2/3, similarly to the medium and wide arbor basket cells described previously (Lund and Lewis, 1993). The axons of some of the ANP cells (7/16 analyzed for morphology) tended to be narrow and to spread vertically across layers (Fig. 1F), as described previously for double bouquet cells (Lund and Lewis, 1993). However, in other cases (8/16) the axons of ANP cells were difficult to distinguish from those of FS neurons. A further characterization of the morphology and intrinsic electrophysiology of interneurons in layers 2/3 of DLPFC will be reported elsewhere (L. S. Krimer et al., unpublished results).
Differential Properties of EPSPs Evoked in Pyramidal Cells and Two Types of Interneurons
To examine the efficacy of sustained synaptic inputs, in most experiments EPSPs were elicited with focal extracellular stimulation. The evoked responses had small amplitudes (<4 mV peak) and had short and constant latency (Fig. 2), consistent with that of monosynaptic EPSPs (see González-Burgos et al., 2000). No evidence of synaptic inhibition was found in these responses when depolarizing the cells above the equilibrium potential for chloride (∼–65 mV in our conditions). In addition, application of 10 µM bicuculline did not affect the responses (data not shown) in any of the cases tested (n = 12 RS, 6 FS and 4 ANP cells).
The time course of single EPSPs evoked with focal stimulation differed across cell classes (Fig. 2, lower panels). Specifically, EPSP duration was significantly shorter in FS cells (15.0 ± 1.8 ms, n = 10; ANOVA followed by Tukey’s contrast, P < 0.05) than in ANP (35.0 ± 6.7 ms, n = 10) or RS neurons (31.4 ± 4.8 ms, n = 10). Because we could not determine if the responses evoked with focal stimulation resulted from activation of more than one presynaptic axon, neither the amplitude nor failure rate could be accurately compared between cell groups. Nevertheless, it is worth noting that failures were typically more frequent in EPSPs recorded from ANP neurons (Fig. 2).
The high quality of our DLPFC slice preparation permitted to record simultaneously from pyramidal cells and interneurons in order to determine the physiological properties of unitary EPSPs, i.e. EPSPs evoked by stimulation of a single presynaptic pyramidal neuron, in synaptically connected pairs. We recorded from a total of nine excitatory synaptic connections between presynaptic RS cells and neighboring postsynaptic neurons of each class (three RS–FS, five RS–RS; one RS–ANP). The unitary EPSPs evoked by single spikes in the presynaptic RS cells, had a short latency (less than 4 ms) in all cases. The unitary EPSPs had low failure rates in both FS and RS neurons, but high failure rate in the RS–ANP pair (Fig. 3A). In addition, the duration of unitary EPSPs tended to be shorter in FS cells compared with either ANP or RS neurons (Fig. 3B), similar to the observations for EPSPs evoked with extracellular stimulation (Fig. 2).
Intrinsic Membrane Properties Differentially Shape EPSP Duration in RS, FS and ANP Neurons
The different duration of EPSPs observed in FS versus ANP or RS cells raises the question of how these differences are generated. Previous studies showed that the duration of EPSPs in pyramidal cells is shaped by membrane conductances that are active at rest, or that are activated or inactivated by depolarization. Because membrane properties differ significantly between RS, FS and ANP cells, we hypothesized that the slower EPSP decay in RS and ANP neurons could be determined at least in part by active and/or passive membrane properties different from those of FS cells. To test this idea, we generated EPSP-like depolarizations (EPSP-LD) by injecting current waveforms similar to actual excitatory postsynaptic currents (EPSCs) into the soma of individual FS, ANP or RS neurons. The kinetics of the EPSC-like waveform was adjusted during recording from a FS neuron, in order to generate an EPSP-LD that mimicked the fast kinetics of the synaptically evoked EPSPs recorded from that FS cell. Thereafter, the kinetics of the EPSC-like waveform was kept constant across experiments and was injected into other FS cells as well as into neurons of the other classes. If intrinsic membrane properties of RS and ANP cells indeed determine a slower decay of the EPSPs, then this effect should be revealed after injecting a fast decaying EPSC-like waveform. As shown in Figure 4, the fast EPSC-like waveform generated EPSP-LDs with significantly shorter duration in FS cells than in either ANP or RS neurons.
If the effect of membrane properties on the duration of EPSP-LDs involves voltage-dependent conductances, then EPSP duration may change with membrane potential. We showed previously that changes in EPSP duration with depolarization strongly influence the efficacy of repetitive excitatory inputs in PFC pyramidal neurons (González-Burgos and Barrionuevo, 2001). Therefore, we examined the voltage-dependence of the duration of the EPSP-LDs. As shown in Figure 5A,B, in RS neurons EPSP-LDs were increased in amplitude and duration at depolarized potentials, as in pyramidal cells from rat cortex (González-Burgos and Barrionuevo, 2001; Stuart and Sakmann, 1995). In marked contrast, in FS cells the amplitude of EPSP-LDs increased, but the responses were shortened by depolarization (Fig. 5A,B), whereas in ANP cells the EPSP-LDs did not change significantly with membrane potential (Fig. 5A,B).
To determine if the duration of synaptically evoked EPSPs changes with depolarization similarly to the duration of EPSP-LDs, and to further test the voltage-dependence of EPSP shape, we elicited EPSPs at a range of subthreshold potentials. Because the integral under the EPSP waveforms may be more sensitive to changes in overall EPSP shape than the duration at half EPSP amplitude, we determined the integral under the waveforms for both EPSPs and EPSP-LDs. As shown in Figure 5C, both EPSP and EPSP-LD integrals were amplified with depolarization in RS neurons, although the effect seemed stronger for synaptically evoked EPSPs. In FS cells, the integral of both EPSPs and EPSP-LDs decreased to a similar extent at depolarized potentials (Fig. 5C), whereas in ANP cells, the integrals did not vary in a consistent manner with depolarization (Fig. 5C). Together, these results show that in RS and FS neurons, voltage-dependent conductances play a significant role in shaping the duration of EPSPs, whereas in ANP cells the EPSP duration is independent of changes in somatic membrane potential.
Sustained Synaptic Inputs Have Different Efficacy in Two Types of Interneurons and Pyramidal Cells
To determine if sustained presynaptic stimulation affects EPSP amplitude in a target-cell specific manner, we applied stimulus trains of 10 pulses at frequencies of 5, 10, 20 or 50 Hz. During stimulation at the lower frequencies (5 and 10 Hz) the interspike intervals are substantially longer than EPSP duration and thus EPSPs elicit depolarization without temporal summation. In contrast, during 20 Hz and 50 Hz stimulation, EPSPs elicit depolarization through the combined effects of changes in EPSP amplitude and summation. Stimulation with 5 and 10 Hz trains elicited synaptic depression of the EPSP amplitude in RS and FS neurons, but facilitation in ANP cells (Fig. 6). During 20 and 50 Hz stimulation, EPSP amplitude also was depressed in both RS and FS neurons (Fig. 6) although the overall depolarization caused by the trains seemed larger for RS than FS neurons, particularly at 50 Hz. In ANP cells the depolarization induced by subsequent EPSPs in a train increased significantly, consistent with the facilitation observed at the lower frequencies (Fig. 6).
To estimate the magnitude of depression and facilitation and the contribution of temporal summation at all frequencies, we estimated separately the amplitude of each EPSP in the train (peak EPSP) and the depolarization elicited by each EPSP relative to the cell’s resting membrane potential (peak depolarization). Whereas the peak depolarization includes the effect of temporal summation, the peak EPSP excludes most of this effect (see Materials and Methods). Thus, if the EPSP duration is short enough to prevent temporal summation, then the two measures should be very similar. The changes in peak EPSP revealed that repetitive stimulation induced depression of the EPSP amplitude in a frequency-dependent manner in RS and FS cells. Furthermore, the peak EPSP (Fig. 7A) was similarly depressed at the end of the trains in RS and FS neurons at all frequencies. In ANP neurons, the peak EPSP showed facilitation at the end of 5, 10 and 20 Hz trains, but was on average unchanged by 50 Hz stimulus trains (Fig. 7A).
As expected based on the observed EPSP durations, we found that during 5- and 10 Hz-stimulation the changes in peak EPSP (Fig. 7A) and peak depolarization (Fig. 7B) were not significantly different, demonstrating a negligible contribution of temporal summation at these frequencies. In contrast, during 20 Hz and 50 Hz stimulation, in RS cells the peak depolarization (Fig. 7B) was significantly larger than the peak EPSP measured at the end of the same trains (Fig. 7B), suggesting an important contribution of temporal summation. In FS neurons, the peak depolarization differed significantly from the peak EPSP during 50 Hz, but not during 20 Hz, stimulation, suggesting that the effect of temporal summation was substantially weaker in FS than in RS neurons. In ANP cells, the effect of temporal summation appeared to be the strongest, showing the largest differences between peak depolarization and peak EPSP at the end of 20 and 50 Hz stimulus trains (Fig. 7).
We found that repetitions of the stimulus trains induced similar changes in EPSP amplitude that, in addition, were short-lasting, since the amplitude of the first EPSP recovered completely during the 10–20 s intervals between trains (data not shown). Moreover, when single test stimuli were delivered at different intervals following a 20 Hz train, EPSPs recovered from depression rapidly after the end of the trains in both RS and FS cells. Relative to that of the first EPSP in the train, the amplitude of test EPSPs evoked 0.2 or 1.0 s after the end of the trains was 76.5 ± 6.1% and 107 ± 22% in RS cells (n = 3); 55.8 ± 1.2% and 120 ± 24% in FS neurons (n = 3) and 179.9 ± 63% and 141 ± 12% in ANP cells (n = 3). These data indicate that in ANP neurons the EPSP amplitude remained slightly potentiated during the initial second post-train.
We also determined the effects of presynaptic stimulus trains on the efficacy of unitary EPSPs elicited in synaptically connected pairs that involved presynaptic RS pyramidal cells and different classes of postsynaptic neurons. In neocortical synapses, low failure rates correlate with synaptic depression, whereas facilitation is found in connections with a high rate of failures (Tsodyks and Markram, 1997; Atzori et al., 2001; Rozov et al., 2001). Thus, the low failure rate observed here for single unitary EPSPs in RS and FS cells (see Fig. 3A) suggests the presence of depression during repetitive presynaptic stimulation, whereas a high rate of failures (Fig. 3A) suggests the presence of unitary EPSP facilitation in ANP neurons. In addition, the longer EPSP duration (Fig. 3B) suggests that EPSP summation would be stronger in RS and ANP compared with FS cells. By delivering trains of short, suprathreshold current steps into the presynaptic RS pyramidal neurons (Fig. 8A), we tested these predictions for unitary EPSPs in FS (three pairs) and RS neurons (one pair). In all of the FS and RS cells, unitary EPSPs showed strong paired-pulse depression at all tested frequencies (Fig. 8B). In addition, unitary EPSPs exhibited strong depression of the peak EPSP during presynaptic spike trains in both RS–FS and RS–RS pairs, with weak temporal summation in FS cells, but significant summation at 20 or 50 Hz in RS cells, as revealed by determining the peak depolarization (Fig. 8B). Therefore, in a given cell class, sustained presynaptic stimulation affected the efficacy of unitary EPSPs and of focal extracellular stimulation-evoked EPSPs in a qualitatively similar manner.
The Postsynaptic Effect of Presynaptic Delay-related Firing Differs Between Cell Classes
How the efficacy of excitatory inputs is affected during delay-related firing may not be well predicted based on the effect observed with short and constant frequency trains. Previous studies found large fluctuations of EPSP amplitude when synaptic inputs were stimulated with spike trains of variable rather than constant frequency (Varela et al., 1997; Dobrunz and Stevens, 1999). Unlike other natural firing patterns, delay-related activity contains a seconds-long episode of firing above baseline that may particularly affect the pre- and postsynaptic mechanisms underlying the efficacy of repetitive synaptic inputs. Thus, to examine the postsynaptic effect of sustained excitatory inputs during presynaptic delay-related firing, we set out to reproduce delay-related firing in the in vitro slice preparation. To this end, synaptic inputs were stimulated with the timing of spike trains recorded from the DLPFC of monkeys performing an oculomotor behavioral task that included a delayed response (M. Roesch and C. R. Olson, unpublished observations). Out of recordings obtained from a single unit that exhibited robust delay-related activity (Fig. 9A), we selected a single trial for which the instantaneous firing rates (mean ± SE) differed between the delay period (43 ± 7 Hz; range 2–250 Hz; median = 30.8 Hz) and the pre- and post-delay periods combined (16 ± 9 Hz; range: 1.7–111 Hz; median = 3.5 Hz). These values are representative of the typical delay-related firing observed during in vivo electrophysiological studies, i.e. sustained firing at relatively low rates (10–50 Hz) above baseline (∼5 Hz). Consistent with the similarity between individual trials (Fig. 9A), these instantaneous firing rates were very similar to those calculated for other trials and thus also similar to the rates calculated after averaging all trials (data not shown). This suggests that the firing rates observed during each of the individual trials were representative of the overall robust delay-related firing behavior of this unit. Focal extracellular stimulation was adjusted to elicit EPSPs with small mean amplitudes (1.4–3.5 mV) and 10-pulse 20 Hz trains were applied first, which elicited depression in RS and FS and facilitation in ANP neurons. Then, the same inputs were stimulated with the natural spike train.
We found that the natural spike train did not cause substantial changes in the peak EPSP prior to the delay period. The peak EPSP for the stimulus just before the onset of the delay, compared with the first EPSP in the spike train, was 97.8 ± 4.4% in RS neurons (n = 5) and 102.2 ± 10.8% in ANP cells (n = 4). In FS neurons (n = 4), the EPSP amplitude decreased between the first and second stimulus in the spike train (73.8 ± 6.3% of the first in the train) but did not change further until the last stimulus before the onset of the delay (peak EPSP 69.5 ± 4.7% of the first in the train). In contrast to these small changes in EPSP amplitude observed before the delay, recordings of EPSPs during the delay period showed that synaptic efficacy changed significantly during this phase (Fig. 9B). Although transmission was sustained during the delay period in neurons of the three classes, the peak EPSP was on average depressed throughout the delay, relative to the first stimulus in the spike train. Although depression dominated EPSP dynamics during the delay in all three cell types, EPSPs were more strongly depressed in FS and RS than in ANP cells (Fig. 9C), consistent with a stronger contribution of facilitation mechanisms in the synaptic inputs onto ANP cells. In addition, the average peak depolarization attained throughout the delay period showed that the postsynaptic depolarizing effect of the presynaptic spike train was strongest in ANP cells and stronger in RS than in FS neurons (Fig. 9C). Indeed, the much larger depolarization often elicited delay-related firing in ANP cells (Fig. 9D), and reduction of the stimulus intensity or injection of hyperpolarizing current was required to prevent firing. These results are consistent with the relative contributions of EPSP dynamics and summation in RS, FS and ANP cells observed in the experiments with 10-pulse stimulus trains.
In order to understand the impact of sustained neuronal activity on the dynamics of synaptic communication in local circuits during working memory tasks, we answered several critical questions regarding the properties of excitatory synaptic inputs onto different neuronal classes in adult monkey DLPFC. First, we found that sustained presynaptic firing induces short-term EPSP facilitation or depression in a cell class-specific manner in pyramidal cells and two populations of interneurons. Second, we found that differences in EPSP duration, and therefore temporal summation, also contribute to the cell-class specific efficacy of sustained inputs. Third, we found that cell class-specific differences in the efficacy of sustained synaptic inputs are also present during stimulation with the natural temporal patterns of activation observed during working memory tasks.
Potential Mechanisms Underlying Cell Class-specific EPSP Dynamics in Pyramidal Cells and Two Populations of Interneurons
Facilitation and depression are primarily determined by presynaptic factors (Zucker and Regehr, 2002), such as the transmitter release machinery or presynaptic calcium dynamics (Rozov et al., 2001). Typically, synapses with low probability of release display high failure rate and EPSP facilitation, whereas synapses with high release probability exhibit low failure rate and synaptic depression (Atzori et al., 2001; Rozov et al., 2001). Although we did not investigate the mechanisms underlying cell-class specific EPSP dynamics, our results are consistent with this scenario, since EPSP failure rate was lower in FS or RS cells than in ANP neurons. We found that repetitive extracellular stimulation of unidentified inputs induced depression or facilitation consistently in postsynaptic cells of a given class. In addition, several properties of the EPSPs evoked with extracellular stimulation were similar than those of unitary EPSPs. Previous studies have also shown that EPSPs or EPSCs evoked in cortical neurons by extracellular stimulation of presynaptic axons have dynamic properties similar to those of EPSPs evoked by stimulation of identified inputs onto neurons of the same class (Stratford et al., 1996; Tarczy-Hornoch et al., 1999; Hempel et al., 2000; Losonczy et al., 2002). The consistent properties of extracellularly evoked EPSPs observed in this and previous studies can be explained if the kind of EPSP dynamics is determined entirely by the class of postsynaptic neuron, independently of the presynaptic input source. Therefore, if the type of EPSP dynamics depends on presynaptic mechanisms, then the postsynaptic cell class must influence the physiology of the presynaptic terminals in a retrograde manner. Alternatively, it is possible that the type of dynamics differs among inputs from different sources and is determined entirely by the presynaptic input class, but that focal stimulation predominantly activates inputs of a certain kind, because of more numerous or more excitable fibers. Comparison of the total number of excitatory synapses per single layer 2/3 pyramidal cell (DeFelipe and Farinas, 1992) with the number of synaptic contacts received by each layer 2/3 cell from neighboring layer 2/3 pyramidal cells (Hellwig, 2000) suggests that neighboring pyramidal cells might be the major source of input onto layer 2/3 pyramidal cells. Whether nearby pyramidal cells also could be the main source of input onto layers 2/3 interneurons currently is not known. However, recent studies in the DLPFC of monkeys suggest that interneurons are targeted in a much greater proportion by the local axon collaterals of layer 2/3 pyramids than by any other known source of excitatory inputs to the PFC (Melchitzky et al., 1998, 2001; Melchitzky and Lewis, 2003). In any case, the EPSPs evoked by focal extracellular stimulation in our study had cell class-specific properties that were very similar to those of unitary EPSPs and thus served as a useful model for examining the effects of sustained stimulation on local excitatory synaptic inputs in the adult monkey DLPFC.
The fact that the EPSP duration differed across cell classes in a similar manner for unitary EPSPs and EPSPs evoked by focal extracellular stimulation of unidentified inputs may be explained by the effects of membrane properties on the EPSP shape, as revealed by the experiments in which EPSP-LDs were elicited by injection of EPSC-like waveforms, independently of actual synaptic inputs. These differences in EPSP duration likely contributed to the differential influence of temporal summation to the cell class-specific efficacy of sustained inputs. Because the EPSC-like waveforms were injected into the soma, the conductances shaping EPSP duration in RS and FS neurons are probably located in perisomatic compartments, as are the voltage-dependent Na+ channels that shape EPSP duration in rat cortical pyramidal cells (González-Burgos and Barrionuevo, 2001; Stuart and Sakmann, 1995). The shortening of EPSPs by depolarization reported here for FS neurons was also found in FS cells from rat visual cortex (Galarreta and Hestrin, 2001), suggesting that the voltage-dependence of EPSP duration is generally opposite in cortical pyramidal cells compared with FS interneurons. This differential role of membrane properties is consistent with and may account for the recent finding that EPSPs and also IPSPs exhibit a faster decay time course in FS interneurons than in pyramidal cells (Thomson et al., 2002). In the ANP cell class, EPSP duration was relatively unaffected by changes in somatic membrane potential. Similarly, in some hippocampal interneurons, EPSP duration is relatively voltage-independent, due to an effect of 4-AP/TEA-sensitive K+ channels (Fricker and Miles, 2000). Alternatively to an effect of active conductances, the longer EPSP duration and lack of voltage-dependence of EPSPs in ANP cells could be due to different passive membrane properties, since membrane time constant was significantly larger in ANP cells compared with FS neurons (Table 1). In addition to the cells’ membrane properties, EPSP duration may depend on the EPSC kinetics determined by different subtypes of synaptic glutamate receptors. Investigation of the role of glutamate receptor subtypes was outside the scope of the present study and remains to be established.
Efficacy of Sustained Synaptic Inputs during Presynaptic Delay-related Firing
Whereas several studies have determined the effects of short stimulus trains on the efficacy of cortical synapses, the effect of prolonged and sustained presynaptic firing with natural temporal patterns has not been previously examined. The dissimilar physiological properties (i.e. release probability and mechanisms of vesicle pool depletion) of synapses that show depression or facilitation with short stimulus trains could differentially affect transmission of presynaptic spike trains with the temporal pattern of delay-related firing. By reproducing delay-related firing in vitro, we found that synapses onto RS and FS neurons, although depressed, sustain transmission throughout the delay period. The postsynaptic depolarization elicited during the delay was, however, greater in RS than in FS cells, probably because of the effects of summation. These differences may have important implications for the activation of the FS and RS cell populations during delay-related firing, as discussed in the following section. We also found that synapses onto ANP neurons are on average depressed during stimulation with the temporal pattern of delay-related firing. Nevertheless, delay-related EPSP depression was significantly less pronounced in ANP neurons than in the other cell classes. In synapses onto ANP neurons, facilitation was evident during constant-frequency stimulation at 5–20 Hz but not at 50 Hz. Thus, it is likely that both facilitation and depression mechanisms operate in these synapses, with the contribution of depression increasing in a frequency-dependent manner. Indeed, to account for the experimental data, theoretical models of depressing or facilitating synapses include a simultaneous contribution of both depression and facilitation mechanisms (Markram et al., 1998a; Matveev and Wang, 2000). Nevertheless, our results show that in spite that the contribution of synaptic depression seems to increase at increasing firing rates, the cell class-specific differences in input efficacy are still observed during the delay period.
Functional Implications for Local Circuit Function in DLPFC
During sustained activation of synaptic inputs, we found that the EPSP amplitude depressed similarly in FS and RS neurons, but that EPSP summation had a weaker effect in FS cells, likely because of a shorter EPSP duration. This suggests that compared with RS cells, recruiting FS cells would require coincident synaptic inputs such as those found during synchronous firing or burst firing. However, currently there is no evidence for the presence of synchrony or burstiness in delay-related activity of DLPFC neurons. Indeed, theoretical studies suggested that delay-related firing must be asynchronous, because synchronous firing turns off activity in a reverberating network (Gutkin et al., 2001; Tegner et al., 2002). A previous study suggested that the delay-related activity of both FS and RS neurons is probably driven by monosynaptic input from nearby RS pyramidal cells (Rao et al., 1999). Due to the differences in temporal summation, a larger number of asynchronously active presynaptic RS cells would be required to recruit FS cells than to recruit other RS cells, assuming that unitary EPSPs have similar average amplitude in RS compared with FS cells. Consequently, the strength of FS neuron-mediated inhibition would increase above baseline levels only once a critical number of pyramidal cells are recruited, acting basically as a feedback mechanism. If so, then there may be an initial condition in which synaptic reverberation between pyramidal cells starts without influence of strong feedback inhibition. In this state, recurrent excitation could produce local amplification of cue-related signals (Douglas et al., 1995), conveyed by brief and relatively weak inputs arriving from posterior cortical areas. During a stable reverberating state with asynchronous inputs, feedback inhibition may prevent runaway excitation. In contrast, during synchronous inputs the recruitment of FS neurons would increase sharply, such that feedback inhibition could now contribute to the termination of sustained firing. These predictions can be tested by recording in vivo from FS and RS units with delay-period activity and determining the timing of delay-related firing of RS and FS cells relative to the delay time window.
In contrast to FS interneurons, the potential role of ANP neurons during working memory tasks is less clear. We found that the axonal arborization of ANP neurons was frequently vertically oriented, differing from the axon of FS cells, which had an appearance resembling that of basket cells or chandelier neurons, as will be reported elsewhere (L. S. Krimer et al., unpublished results). Whereas the majority of cortical FS/basket cells target perisomatic compartments of postsynaptic neurons located within the same layer, interneurons with vertically oriented axons typically target dendritic postsynaptic cell compartments and may project across several layers (Kawaguchi and Kubota, 1997; Somogyi et al., 1998; Thomson and Bannister, 2003). Interestingly, a recent report confirmed and extended the results of previous studies indicating that dendritic responses to GABA are excitatory under conditions in which somatic GABA responses are inhibitory (Gulledge and Stuart, 2003). In addition to the morphology and intrinsic electrophysiology, interneurons are typically grouped based on expression of Ca2+-binding proteins. Most FS cells contain the Ca2+-binding protein parvalbumin and most parvalbumin-positive cells are FS (Thomson and Bannister, 2003; Thomson et al., 2002). In contrast, cells with spike frequency adaptation and vertical axons may contain either of the Ca2+-binding proteins calretinin or calbindin (Conde et al., 1994; Kawaguchi and Kubota, 1997). Whereas calretinin-containing interneurons target primarily dendrites of other calretinin interneurons, the major postsynaptic targets of calbindin-containing cells are not so clearly identified, appearing to include dendrites of both pyramidal cells and interneurons (Gonchar and Burkhalter, 1999; Meskenaite, 1997; Thomson and Bannister, 2003). Thus, understanding the functional relevance of the increased efficacy of excitatory inputs onto ANP neurons during sustained activation depends on future identification of their postsynaptic targets and on knowing whether ANP neurons produce an excitatory or inhibitory postsynaptic effect.
Altogether, our findings show that during sustained presynaptic firing the efficacy of excitatory inputs in local circuits of the primate DLPFC is not invariable, but changes dynamically and in a postsynaptic cell class-specific manner. Because sustained activity was produced by the experimenter, we did not directly address the mechanisms generating sustained firing in the DLPFC network. However, our results may help address these mechanisms in future studies. For example, incorporation of cell class-specific synaptic depression, facilitation and temporal summation should improve significantly the biophysically realistic network models that reproduce delay activity, in which synaptic weight at the connections between the model cells is typically constant in the short term, and include a single population of inhibitory model neurons. The effects of synaptic depression, facilitation and EPSP duration can be influenced by neuromodulators (Gil et al., 1997), which may increase or decrease the amplitude of single EPSPs but simultaneously attenuate or enhance depression or facilitation (Markram and Tsodyks, 1996). Thus, the combined effects of activity and neuromodulators could elicit different types of EPSP dynamics at distinct synaptic connections, creating diverse activity patterns in local circuits of the DLPFC during different behavioral states.
In conclusion, the present study provides the first experimental evidence of the presence and types of synaptic dynamics found in local circuits of the DLPFC in the adult primate brain. By describing how the efficacy of excitatory transmission onto different kinds of postsynaptic neurons is affected by physiologically relevant temporal patterns of activation, these results constitute a significant step towards understanding how activity may flow between elements of the prefrontal local circuits when the DLPFC network is engaged in working memory operations.
We thank Jeremy Seamans for comments on a previous version of this manuscript; Darrell Henze for providing the Labview programs used for data acquisition; Matthew Roesch and Carl Olson for kindly providing the spike trains recorded from monkey prefrontal cortex in vivo, and Ingelore Kroener and Olga Krimer for excellent assistance with histology and cell reconstructions. This work was supported by NIMH (grants MH51234, MH41456 and MH63561) and by a NARSAD Young Investigator Award (to G. González-Burgos).
Address correspondence to Guillermo González-Burgos, Translational Neuroscience Program, Department of Psychiatry, University of Pittsburgh School of Medicine, Room W1651 Biomedical Science Tower, 3811 O’Hara St., Pittsburgh, PA 15213–2593, USA. Email: firstname.lastname@example.org.
|Regular spiking||Fast spiking||Adapting non-pyramidal|
|Resting membrane potential||–74.4 ± 1.1 mV, n = 20||–67.7 ± 1.2 mV, n = 20||–68.2 ± 1.3 mV, n = 17|
|Input resistance||102.4 ± 22.0 MΩ, n = 14||165.8 ± 15.0 MΩ, n = 20||291.4 ± 31.0 MΩ, n = 17|
|Time constant||22.3 ± 3.6 ms, n = 14||9.5 ± 0.8 ms, n = 15||24.4 ± 1.82 ms, n = 14|
|AHP amplitude||11.4 ± 1.1 mV, n = 18||27.0 ± 2.2 mV, n = 19||21.3 ± 1.5 mV, n = 17|
|Spike width at half amplitude||1.05 ± 0.08 ms, n = 20||0.39 ± 0.03 ms, n = 20||0.68 ± 0.08 ms, n = 17|
|Regular spiking||Fast spiking||Adapting non-pyramidal|
|Resting membrane potential||–74.4 ± 1.1 mV, n = 20||–67.7 ± 1.2 mV, n = 20||–68.2 ± 1.3 mV, n = 17|
|Input resistance||102.4 ± 22.0 MΩ, n = 14||165.8 ± 15.0 MΩ, n = 20||291.4 ± 31.0 MΩ, n = 17|
|Time constant||22.3 ± 3.6 ms, n = 14||9.5 ± 0.8 ms, n = 15||24.4 ± 1.82 ms, n = 14|
|AHP amplitude||11.4 ± 1.1 mV, n = 18||27.0 ± 2.2 mV, n = 19||21.3 ± 1.5 mV, n = 17|
|Spike width at half amplitude||1.05 ± 0.08 ms, n = 20||0.39 ± 0.03 ms, n = 20||0.68 ± 0.08 ms, n = 17|
All of the differences between group means were significant (P < 0.02; one-way ANOVA followed by comparisons with Tukey’s test) except the difference between resting membrane potential of FS and ANP cells and between time constant of RS and ANP neurons. AHP: after hyperpolarizing potential.