In monkey dorsolateral prefrontal cortex (PFC), long-distance, horizontally oriented intrinsic axon collaterals interconnect clusters of pyramidal neurons in the supragranular layers. In order to study the electrophysiological responses mediated by these long-distance projections, an in vitro slice preparation of monkey PFC was used to obtain whole-cell patch clamp recordings from layer 3 pyramidal neurons. Using in vivo tracer injections, we found that long-distance projections were well preserved in PFC slices cut in the coronal plane. Postsynaptic currents were evoked by low-intensity electrical extracellular stimulation applied successively to 20–30 discrete sites located up to 2200 μm lateral to the recorded cell. Several criteria were applied to discriminate between monoand polysynaptic responses. Long-distance monosynaptic connections were mediated by fibers with relatively slow conduction velocity (0.14 m/s). Excitatory postsynaptic currents (EPSCs) evoked by stimulation of shortor long-distance horizontal connections did not differ in kinetic properties. The majority (77%) of the 35 layer 3 PFC neurons studied were monosynaptic targets of long-distance connections. EPSCs mediated by long-distance connections had amplitudes that were similar or even larger than short-distance EPSCs, suggesting that excitatory input provided by the former was relatively robust. For most neurons (87.5%) in which a full complement of monosynaptic EPSCs was evoked by multisite stimulation, the EPSC amplitude as a function of stimulation distance from the recorded cells exhibited statistically significant peaks. The spacing between peaks was similar to the spacing between interconnected clusters of neurons observed in previous anatomical studies. The results show that long-distance excitatory connections constitute a significant intrinsic pathway of synaptic communication in layer 3 of monkey PFC.
As is the case for other regions of the macaque monkey neocortex, pyramidal neurons in the supragranular layers of the dorsolateral prefrontal cortex (PFC) furnish intrinsic axon collaterals that travel for substantial distances, up to several millimeters, tangential to the pial surface (Levitt et al., 1993). The transport of anterograde and retrograde tracers from focal injections in monkey PFC (areas 9 and 46) has revealed that clusters of pyramidal neurons in layers 2–3 give rise to intrinsic, horizontal projections that terminate in discrete clusters of labeled axon terminals in layers 1–3 (Levitt et al., 1993; Kritzer and Goldman-Rakic, 1995; Pucak et al., 1996; Woo et al., 1997). In primary sensory regions, the axons of these long-distance horizontal connections terminate in patch-like clusters (Lund et al., 1993; Yoshioka et al., 1996) and appear to preferentially link columns of cells with similar functional properties (Gilbert and Wiesel, 1989; Malach et al., 1993; Weliky and Katz, 1994; Bosking et al., 1997; Yabuta and Callaway, 1998). In contrast, in the monkey PFC, the clusters of labeled axon terminals and neurons are elongated and form a series of regularly spaced stripes (Levitt et al., 1993; Pucak et al., 1996). Although it seems likely that the PFC neurons clustered in a given stripe share certain functional characteristics (Goldman-Rakic, 1995), this remains to be experimentally tested. Furthermore, in monkey primary visual cortex, the axon terminals of horizontal connections contact pyramidal and nonpyramidal neurons in a proportion similar to the frequency of each cell type in the cortex, suggesting that these connections do not have target cell selectivity (McGuire et al., 1991). In contrast, although the proportions of pyramidal to nonpyramidal neurons are similar in the monkey visual and prefrontal cortices, >95% of the horizontal connections in the PFC contact dendritic spines and thus appear to preferentially target pyramidal neurons (Melchitzky et al., 1998). Since these contacts form exclusively asymmetric synapses, these findings suggest that excitatory, long-distance, horizontal connections in the PFC are biased to innervate other pyramidal neurons.
Given that the stripes revealed by a given tracer injection in the PFC appear to be reciprocally connected (Pucak et al., 1996), it is reasonable to hypothesize that other pyramidal neurons in the superficial layers are the principal synaptic targets of these connections (Melchitzky et al., 1998). However, a proportion of the dendritic spines found in the superficial layers are located on the dendrites of neurons whose somata lie in deeper cortical layers. Indeed, in rat somatosensory cortex, the axon collaterals of layer 3 pyramidal neurons establish horizontal connections with layer 3 and layer 5 pyramidal neurons with similar probabilities (Thomson and Deuchars, 1997; Markram, 1997; Thomson and Bannister, 1998). In addition, the functional strength of the long-distance horizontal connections presumed to exist between supragranular pyramidal neurons in monkey PFC is not known.
Consequently, in the present study, we recorded visually identified layer 3 pyramidal neurons using whole-cell patch clamp methods applied in a living brain slice preparation from monkey PFC. The improved signal-to-noise ratio of in vitro tight seal whole-cell recordings allowed very sensitive detection of small amplitude, fast synaptic events. Hence, it was possible to employ low stimulation intensities, restricting the spread of the stimulation current and thus improving the spatial resolution of the stimulation procedure. Horizontal connections were activated by low-intensity stimulation applied at multiple sites in the superficial layers via a multielectrode stimulation array, similar to that employed previously for functional mapping of horizontal connectivity in primary visual cortex (Weliky and Katz, 1994; Weliky et al., 1995). This experimental approach was employed to address the following questions: (i) Do pyramidal neurons in layer 3 of monkey PFC receive monosynaptic, excitatory input from horizontal, long-distance connections? (ii) If so, how robust is this input as compared with input provided by short-distance connections? (iii) What proportion of layer 3 cells receive this type of synaptic input? (iv) Does the spatial pattern of excitatory synaptic responses evoked by multisite stimulation reflect the clustered organization of horizontal connectivity in the supragranular layers of monkey PFC?
Materials and Methods
PFC Slice Preparation
Tissue slices were obtained from 10 young adult male cynomolgus monkeys (Macaca fascicularis). All animals were treated according to the guidelines outlined in the NIH Guide to the Care and Use of Animals. Following injections of ketamine hydrochloride (25 mg/kg), dexamethasone phosphate (0.5 mg/kg) and atropine sulfate (0.05 mg/kg), an endotracheal tube was inserted and the animal was placed in a stereotaxic frame. Anesthesia was maintained with 1% halothane in 28% O2/air. A craniectomy was performed over the dorsal PFC and a small block of tissue (~4 × 6 × 4 mm) containing dorsal area 9 and, in some cases, the medial bank of the principal sulcus (area 46) was removed. The tissue block was immediately immersed in ice-cold modified artificial cerebral spinal fluid (modified ACSF) (composition, in mM: sucrose 230, KCl 1.9, Na2HPO4 1.2, NaHCO3 33, MgCl2 6, CaCl2 0.5, glucose 10, kynurenic acid 2; pH 7.3–7.4 when bubbled with 95%/5% O2/CO2 gas mixture). After hemostasis was achieved, the skin was closed and the animal was treated postoperatively with analgesics and antibiotics as previously described (Pucak et al., 1996). Animals recovered quickly and no behavioral deficits were observed. Two to four weeks later, the animals underwent the same procedure with the following modifications to obtain tissue from the opposite hemisphere. After the craniectomy was performed, the animal was given an overdose of pentobarbital (50 mg/kg) and was perfused intracardially with ice-cold modified-ACSF. A tissue block containing the portions of PFC areas 9 and 46 nonhomotopic to the first biopsy was quickly excised and then the remainder of the brain was removed for other studies.
The tissue blocks were glued to the stage of a vibratome and 400 or 500 μm thick slices were cut in the coronal plane while the block was submerged in ice-cold, modified ACSF. After slicing, the tissue was maintained at room temperature for at least 2 h in an incubation chamber containing low Ca2+/high Mg2+ ACSF (low Ca2+ ACSF, composition, in mM: NaCl 125, KCl 2, Na2HPO4 1.25, NaHCO3 26, CaCl2 1.0, MgCl2 6.0, glucose 10; pH 7.3–7.4 when bubbled with 95%/5% O2/CO2 gas mixture) before being transferred to the recording chamber.
Slices transferred to a submersion recording chamber were superfused with ACSF (composition, in mM: NaCl 125, KCl 2, Na2HPO4 1.25, NaHCO3 26, CaCl2 3.0, MgCl2 3.0, glucose 10; pH 7.3–7.4 when bubbled with 95%/5% O2/CO2 gas mixture) at room temperature (20–24°C). The relatively low K+ concentration in the external solution hyperpolarizes the cells in the slice, and the high Ca2+ and Mg2+ concentrations produce an increase in the spike threshold. Both factors decrease excitability and reduce the probability of firing intercalated neurons, thus minimizing polysynaptic responses (Berry and Pentreath, 1976).
Electrical stimulation was applied by means of a multiple electrode array similar to that described by Weliky and Katz (Weliky and Katz, 1994). The stimulation array was placed parallel to the pia (see Fig. 1) in superficial–middle layer 3 (300–600 μm from the pial surface). The electrode array consisted of 24–30 fine nichrome wires coated with Formvar except at the tips (bare diameter 50 μm, coated diameter 65 μm; A-M Systems Inc., Carlsberg, WA), which were glued together with cyanoacrylate glue. The wire tips to be in contact with the tissue had similar shapes and the mean (± SE) inter-tip distance for three different arrays was 70 ± 5, 89 ± 10 and 60 ± 5 μm. Bipolar electrical stimulation was applied between pairs of adjacent wires with the cathode always closer to the recorded cell (Fig. 1). Bipolar square current pulses (20–100 μs, 20–200 μA) were applied using a constant current stimulus isolation unit (Model SC-100, Winston Electronic Co., Millbrae, CA).
Electrical Characteristics of the Stimulation Array
For one stimulation array, the mean resistance between the input to the stimulator and the tip of 20 of the wires was 36 Ω, and the mean resistance between adjacent pairs of electrodes placed in the recording solution or in contact with the tissue was 1.68 ± 0.17 kΩ. Since the stimulus isolation unit delivers currents up to 10 mA with the output connected to a load resistance of up to 10 kΩ, in our experiments the current output was constant across the sections of tissue stimulated by each bipolar position in the array. However, a factor that could introduce variability in current output from the isolation unit is the electrode capacitance. The total capacitance between adjacent pairs of electrodes in one of our stimulation arrays was 1.4 ± 0.4 nF (range: 0.86–2.71 nF), and time constants (R × C) for charging the electrodes were on the order of microseconds (maximal value 2.8 μs). As a result, only a small fraction of the total current delivered during the stimulation pulse was diverted to charge the electrode capacitance. Thus, it is unlikely that there were significant differences in electrical current stimulation flowing between the electrodes in the array due to differences in electrode capacitance.
Pyramidal neurons in layer 3 were identified visually using differential interference contrast (DIC) and infrared video-microscopy with Leitz Laborlux 12 or Zeiss Axioskop FS microscopes equipped with CCD (CCD-72S, Dage-MTI) or Newvicon image tube (NC-70, Dage-MTI) cameras, respectively. Recorded cells were located 380–900 μm from the pial surface (middle–deep layer 3) and were identified as pyramidal neurons by the shape and size of the soma and the presence of an apical dendrite (Fig. 1B). Pipettes were pulled from borosilicate glass and had resistances between 3 and 7 MΩ when placed in the bath and filled with the internal solution (composition, in mM: CsF, 120; CsCl, 10; HEPES, 10; EGTA, 5; DIDS, 1.0 mM, pH 7.2–7.4; osmolality: 270–290 mOsm). Giga-Ohm seals (resistance >2 GΩ) were obtained using the ‘blow and seal’ technique, as described elsewhere (Stuart et al., 1993). The access resistance (Racc) was not compensated and had values between 8 and 25 MΩ. Racc was continuously monitored and recordings were rejected for analysis when Racc changed by >20%. EPSCs were recorded at a nominal holding potential of –80 to –70 mV (unless indicated) with an Axopatch1C amplifier (Axon Instruments Inc., Foster City, CA), filtered at 2–5 kHz, digitized at 10 kHz and stored on disk for off-line analysis.
The parameters of the stimulation current delivered via the electrode array (see above) were adjusted at the closest stimulation position (position 1) to evoke large EPSCs without complex waveforms (Fig. 2A). After stimulation from positions 1–2, the stimulus current was kept constant and delivered successively to each of the other pairs in the multielectrode array. At least 10 EPSCs were evoked from every electrode pair in the array, intermittently returning to positions 1–2 to control for stability of the recording conditions (see Fig. 2B). The stimulation frequency was set at 0.1 Hz, because in preliminary experiments higher frequencies produced variable degrees of depression of the response amplitude.
Acquisition and analysis were performed using LabView (National Instruments Corp., Austin, TX) running customized software. The baseline noise was determined in each trace by measuring the peak current in a time window with the same duration as that used to measure peak EPSC amplitude, but positioned before the stimulation artifact.
In Vivo Extracellular Tracer Injection
Iontophoretic injections of 10% biotinylated dextran amine (BDA, mol. wt 10 000) in 0.01 M phosphate buffer, pH 7.4, were made in layer 3 of area 9 by passing positive current (5 μA, 7 s cycles) through glass pipettes (tip diameter 20–50 μm) for 20 min (Pucak et al., 1996). These injections were made in a subset of animals in order to determine the presence of horizontal axons and clustered connections in thin (100–200 μm) coronal slices that were adjacent to slices used for electrophysiological recording.
To determine whether mean EPSC amplitude changed with stimulation distance, one-way analysis of variance (ANOVA) was employed. When the results of the ANOVA were significant (P < 0.05), peaks and troughs were identified in the pattern of responses obtained for each cell following an approach similar to that previously used by others (Weliky and Katz, 1994; Weliky et al., 1995). Peaks and troughs were defined as those locations where EPSC amplitude was larger or smaller, respectively, than those observed at the two adjacent locations. If the EPSC amplitude of a peak was significantly different from both of the two closest flanking troughs using post-hoc HSD Neumann–Keuls comparisons between means (P < 0.05), then it was considered a significant peak. Unless otherwise stated, values are reported as mean ± SD.
Bicuculline methiodide, 4,4′-diisothiocyanostilbene-2,2′-disulfonic acid (DIDS), biocytin and kynurenic acid were purchased from Sigma (Saint Louis, MO). BDA and 4,4′-dinitrostilbene-2,2′-disulfonic acid (DNDS) were purchased from Molecular Probes (Eugene, OR).
Isolation of EPSCs Evoked by Stimulation of Horizontally Oriented Connections
Preservation of horizontally oriented axons in the PFC slice preparation was demonstrated in animals in which extracellular injections of anterograde tracers were placed in the superficial cortical layers of area 9 in vivo. As shown in Figure 3, labeled axons running parallel to the pia could be followed for up to 2000 μm, even in slices that were thinner (100 μm) than those used for the electrophysiological recordings (400–500 μm). In addition, clusters of labeled arborizing axons and terminals, similar to those observed in previous anatomical studies (Levitt et al., 1993; Pucak et al., 1996), were clearly evident in the slice preparation.
To examine the electrophysiological properties of the synaptic input furnished by horizontally oriented axons, wholecell recordings were obtained from visually identified layer 3 pyramidal neurons. Postsynaptic currents (PSCs) were evoked by electrical stimulation applied in superficial layer 3, lateral to the recording site (Fig. 1A). As seen in Figure 4A, initial experiments showed that the PSCs typically had mixed excitatory and inhibitory components (EPSCs and IPSCs, respectively), with the relative weights of the EPSCs and IPSCs varying significantly between evoked responses. The IPSCs were more frequently evoked by stimulation at the locations closest to the cell (300–500 μm away) and were probably dior polysynaptic, because they invariably had longer latencies than the EPSCs. In addition, the IPSCs had fast time courses and reversed at a membrane potential close to the equilibrium potential for chloride ions (~–65 mV in our experimental conditions, data not shown) suggesting that they were mediated by activation of GABAA receptor-gated chloride channels.
Because it is critical to study EPSCs in isolation, the presence of the fast IPSC complicated the electrophysiological analysis of the excitatory input. However, when the GABAA antagonist bicuculline was bath-applied to the PFC slices, it induced epileptiform activity even at concentrations (0.5 μM) that are too low to block the IPSCs (data not shown), as previously reported (Gutnick et al., 1982; Chagnac-Amitai and Connors, 1989). Recent studies have shown that, when applied intracellularly to single cells in neocortical slices, cyanostilbene compounds like DIDS block GABAA receptor channels without inducing such hyperexcitability (Nelson et al., 1994; Dudek and Friedlander, 1996; Shao and Burkhalter, 1996). Indeed, after dialysis of PFC pyramidal cells for at least 20 min with a CsF-based pipette solution containing 0.5–1.0 mM DIDS, no fast IPSCs were observed (Fig. 4B). Slow IPSCs, such as those expected from the activation of potassium channels by GABAB receptors (McCormick, 1989), also were not observed when recording with the cesium-based internal solution. Together, these data indicate that intracellular blockade of fast and slow synaptic inhibition allowed us to study isolated EPSCs in PFC pyramidal neurons.
Identification of Monosynaptic EPSCs Evoked by Horizontally Oriented Connections
To determine if layer 3 pyramidal neurons were targets of long-distance horizontal excitatory monosynaptic inputs, EPSCs were evoked from different locations in the slice via the multielectrode array. Whole-cell recordings were obtained from a total of 47 layer 3 pyramidal neurons. In 12 of these 47 neurons, either the recordings were lost during the period of dialysis with DIDS or a continuous and rapid rundown of the EPSCs precluded an adequate interpretation of the results. In the remaining 35 neurons, stable EPSCs were evoked from 10 or more pairs of stimulation positions in the slice (16.9 ± 5.1 positions per cell). Figure 5 shows responses in one layer 3 pyramidal neuron evoked from each of 24 electrode stimulation positions.
As stimulation was applied at positions increasingly distant from the recorded neuron, EPSC amplitude diminished, and fewer stimulations sites evoked detectable EPSCs. However, in 30 of the 35 neurons, EPSCs were detected when stimulation was applied at long-distance (>600 μm from the recorded cell) horizontal positions in the slice. As expected, low-intensity stimulation did not evoke fast inward currents due to action potentials in any of the 35 cells, indicating that EPSCs were subthreshold, and that the probability of antidromic stimulation of the axon collaterals was low.
As shown in Figure 6, four types of responses were evoked via the stimulation array. One type of response (Fig. 6A) showed no detectable EPSCs. Three other types of evoked responses exhibited detectable EPSCs that were classified according to waveform shape, incidence of observed failures, and variability of latency between stimulus and EPSC onset. As shown in Figure 6B, one type of EPSC was characterized by monophasic waveforms which had no failures and a low trial-to-trial variability in latency. Another type of EPSC had multiphasic waveforms (Fig. 6C) despite stimulation parameters that were adjusted to evoke only monophasic EPSCs from array positions 1–2 (see Fig. 2A). These complex EPSCs had an early component with constant latency and no failures, and one or more late components with frequent failures and latency fluctuations of several milliseconds. Finally, other EPSCs had monophasic waveforms, but exhibited frequent failures and highly variable latencies (Fig. 6D). In the 35 neurons examined in this study, the mean (± SD) percentage of stimulation sites evoking each type of recording trace was as follows: 56.5 ± 16.2% of the sites evoked no detectable EPSCs; 27.7 ± 18.6% of the sites evoked monophasic EPSCs; 13.4 ± 12.6% of the sites evoked complex EPSCs; and 2.3 ± 5.0% of the sites evoked EPSCs with variable latency and failures. We interpret the presence of failures and variable latency as resulting from the involvement of intercalated neurons firing intermittently and at variable latencies (Berry and Pentreath, 1976). Therefore, the late components of the complex EPSC waveforms (Fig. 6C) and the EPSCs with variable latency and failures (Fig. 6D) were regarded as polysynaptic and excluded from further data analysis, whereas the monophasic EPSCs (Fig. 6B) and the early components of the complex EPSC waveforms (Fig. 6C) were considered potentially monosynaptic and examined further.
As a final test for monosynapticity, we plotted EPSC latency as a function of horizontal distance between the stimulation and recording sites for each neuron. Assuming that the synaptic delay is nearly identical at all active synapses, then latency should increase linearly with horizontal distance as a consequence of the distance that the presynaptic action potential travels. For 18 of the 30 layer 3 pyramidal neurons in which long-distance responses were observed, the plot of mean EPSC latency versus horizontal distance was well fit by a linear regression line (Fig. 7A), confirming that all the EPSCs in those neurons were evoked monosynaptically. For the remaining 12 neurons, the latency–distance plot was not well fit by a linear relation (Fig. 7B), suggesting that, despite having low variability in latency and lack of failures, the EPSCs evoked from some of the stimulation sites probably were not monosynaptic.
The Majority of Layer 3 Neurons Are Postsynaptic Targets of Long-distance Horizontal Monosynaptic Connections
A major goal of the present study was to establish the proportion of PFC layer 3 pyramidal neurons that are targets of long-distance horizontal monosynaptic connections. If monosynaptic EPSCs were evoked in a neuron by at least one stimulation site located >600 μm from the cell, then that neuron was assumed to be a target of long-distance, horizontal connections. As mentioned above, 18 cells exhibited EPSCs in response to long-distance stimulation, with latency–distance relations indicating that the evoked EPSCs were indeed monosynaptic. To determine if some of the long-distance EPSCs evoked in the other neurons were also monosynaptic, the experimentally observed EPSC latencies with predicted latencies calculated for each stimulation distance were correlated using a model linear relation. This linear model had parameters equal to the average values obtained from the linear latency–distance plots in the first group of 18 neurons. The conduction velocity of the horizontal axon collaterals, estimated from the slope of the linear regressions, was 0.14 ± 0.04 m/s and the synaptic delay, estimated from the y-axis intercepts of the linear regressions, was 2.29 ± 0.49 ms. As expected, when the plot of observed versus predicted latency was calculated for EPSCs in the first group of 18 cells, the points were always close to the theoretical line of identity between observed and predicted latency (Fig. 8A). For nine other cells (Fig. 8B), at least one of the long-distance EPSCs had a deviation in latency that was less than or equal to the maximal deviation from the identity line observed for EPSCs in Figure 8A. Therefore, we concluded that these nine pyramidal cells were also targets of long-distance excitatory monosynaptic connections. Thus, 27 of the 35 (77%) recorded layer 3 pyramidal neurons were targets of long-distance monosynaptic inputs.
Although a complete characterization of the biophysical properties of these monosynaptic EPSCs was beyond the scope of this work, we compared some of these properties between EPSCs evoked from shortvs. long-distance stimulation. The mean (± SD) extrapolated reversal potential for monosynaptic EPSCs evoked by long-distance stimulation was 9.9 ± 12.0 mV (n = 8). In five of the 18 cells in which all evoked EPSCs were monosynaptic we selected monophasic EPSC waveforms evoked by stimulation at a short (510 ± 30 μm) or long (1340 ± 174 μm) distance from the same cell. For these EPSCs, recorded at somatic holding potentials of –80 to –70 mV, the observed kinetic parameters were similar for responses evoked by longand short-distance stimulation. The EPSC 10–90% rise time was 4.89 ± 1.97 ms for short-distance and 3.30 ± 1.74 ms for long-distance stimulation, whereas the EPSC duration at half peak amplitude was 17.7 ± 7.7 ms for short-distance and 11.4 ± 3.4 ms for long-distance stimulation. In contrast to monosynaptic responses evoked by long-distance stimulation (which must be mediated by long-distance axonal projections), EPSCs evoked by short -distance extracellular stimulation may not represent exclusively short-distance inputs. Consequently, it is not clear to what extent the kinetic properties described above represent two distinct populations of synaptic contacts. The excitatory postsynaptic potential (EPSP) observed between a pair of simultaneously recorded layer 3 monkey PFC pyramidal neurons (with somata separated by <100 μm) had a 10–90% rise time of 3.25 ms and a width at half amplitude of 15.25 ms (unpublished results). Therefore, these data suggest that the kinetics of monosynaptic responses evoked by short-distance stimulation or by an actual short-distance input are not significantly different. Moreover, no significant differences in kinetics were observed between shortand long-distance responses.
The Strength of Evoked Monosynaptic EPSCs Varies as a Function of Stimulation Distance
To activate horizontally oriented connections in the present experiments, a constant low-intensity stimulation current was applied throughout the multielectrode stimulation array. Therefore, differences between the strengths of responses evoked by different stimulation sites are likely due to variations in the number of horizontal connections available for activation at each stimulation site. Indeed, a periodic variation in the evoked EPSC amplitude as a function of stimulation distance is what one would predict if the multisite electrical stimulation reflects the clustered organization of intrinsic, horizontal connections in the PFC (Levitt et al., 1993; Pucak et al., 1996). Therefore, we determined if the amplitude of the evoked EPSCs varied as a function of distance from the recorded neuron for the 18 neurons in which all the observed EPSCs were monosynaptic. In two of the neurons, the EPSC amplitude decayed with distance in a monotonic fashion (Fig. 9A). In the other 16 cells (Fig. 9B–D), although the EPSC strength tended to decay with distance (see also Fig. 5), the distribution of monosynaptic EPSC amplitudes as a function of stimulation distance showed peaks that were statistically significant. However, not all of the peaks were experimentally equivalent, because some troughs represented baseline resting conductance (e.g. no detectable EPSC; see Fig. 9B,D), whereas other troughs were actually small amplitude EPSCs (see Fig. 9C). Interestingly, in two neurons (data not shown) in which it was possible to test two different levels of stimulation current throughout the array, the response size evoked from each multielectrode position was greater with the higher applied current (e.g. small EPSCs were observed for positions that evoked no response at the lower current) without altering the overall pattern of responses.
For 15 neurons, >1 peak was observed and the distances separating consecutive peaks ranged from 135 to 810 μm (n = 29). For most neurons, the amplitude of the peak EPSCs tended to be smaller for more distant stimulation sites. However, in several examples (e.g. Fig. 9C) EPSC strength was similar for all spatial peaks, independent of the distance between the cell and stimulation site, and in some cases EPSC strength was even greater for stimulation sites that were more distant (Fig. 9C).
This study, in concert with an earlier report (Krimer and Goldman-Rakic, 1997), demonstrates the feasibility and value of a primate PFC slice preparation for electrophysiological, anatomical and biochemical in vitro studies. Given the marked structural and functional differences between the PFC of rodents and primates (Preuss, 1995), this slice preparation makes it possible to study cellular neurophysiological features that may be unique to the primate PFC. In particular, this preparation has enabled us, in the present report, to address the following questions.
Do Pyramidal Neurons in Layer 3 of Monkey PFC Receive Monosynaptic, Excitatory Input from Horizontal, Long-distance Connections?
Our previous anatomical studies demonstrated that the horizontal, long-distance axon collaterals furnished by supragranular pyramidal neurons in monkey PFC synapse, almost exclusively, on the dendritic spines of other PFC pyramidal neurons (Melchitzky et al., 1998). In the present study, we demonstrate electrophysiologically that these long-distance axon collaterals (originating up to ~2.0 mm from the recorded layer 3 pyramidal neuron) provide monosynaptic excitatory inputs to layer 3 pyramidal neurons. The monosynapticity of these connections was verified using several criteria. First, the studied EPSCs had to have a low failure rate, in contrast to the high failure rate expected for polysynaptic inputs (Berry and Pentreath, 1976). Second, the EPSCs had to have latencies that varied by <1 ms between successive trials, because the latency of monosynaptic connections between pairs of pyramidal neurons has been reported to fluctuate by ~1 ms (Miles and Wong, 1986; Markram et al., 1997; Hardingham and Larkman, 1998). Third, we required that the increase in latency as a function of horizontal distance be linear, and that the conduction velocity (calculated from the slope of the linear regression) be compatible with that of thin and unmyelinated axons (Andersen et al., 1978).
To improve the duration and stability of whole-cell recordings, and the overall viability of the slices, the present experiments were performed at room temperature, a condition known to affect synaptic transmission and axonal conduction velocity (Hardingham and Larkman, 1998). For example, in neocortical slices maintained at 34–35°C (Thomson et al., 1988; Lohmann and Rorig, 1994) conduction velocity was 0.1 and 0.3 m/s for pyramidal cell axon collaterals. In contrast, in the present study we found that at room temperature (20–24°C) conduction velocity was slower (mean: 0.14 m/s). This is consistent with previous findings that low temperature markedly slows action potential conduction velocity (Andersen et al., 1978; Berg-Johnsen and Langmoen, 1992). Our estimations of synaptic delay also are consistent with monosynaptic EPSCs recorded at low temperatures. We estimated (from the intercepts of the linear regressions) a synaptic delay of 2.29 ± 0.49 ms for layer 3 horizontal connections at room temperature, a value consistent with the report that the synaptic delay at mammalian glutamate synapses is temperature-sensitive, increasing from 0.2 ms at 35°C to 3 ms at 20°C (Sabatini and Regehr, 1996). In the present experiments, a long synaptic delay helped to discriminate between monoand polysynaptic responses because the latter should exhibit an additional latency of at least one synaptic delay (see Fig. 6C,D).
In light of previously published results (Sabatini and Regehr, 1996; Hardingham and Larkman, 1998), two other effects of low temperature to be considered are slow kinetics of the EPSC and decreased probability of release. Although slow kinetics would change the temporal window for synaptic integration, it should not affect the present results because we did not study summation of synaptic events. A decrease in probability of release would result in a high probability of observed failures, even if the synaptic connections have multiple synaptic contacts (see the following section).
In the present study, we did not examine the pharmacology of the receptors mediating the EPSCs evoked by activation of horizontal connections. Previous electron microscopy studies showed that most, if not all, of the synapses established by long-distance horizontal connections in PFC are asymmetric, and thus very likely excitatory (Melchitzky et al., 1998). Consistent with these results, in every cell tested, monosynaptic EPSCs evoked by activation of long-distance horizontal connections had an apparent reversal potential close to 0 mV, as expected if synaptic excitatory glutamate receptor-gated channels mediated the response (Hestrin et al., 1990).
We also measured the time course (rise time and duration) of small, monosynaptic and monophasic EPSCs evoked by stimulation at different distances from a given neuron. We did not find significant differences in the kinetics of EPSCs as a function of this distance, suggesting that localization of synaptic contacts on the dendritic arbor of layer 3 pyramidal neurons was similar for shortand long-distance horizontal connections [however, see the findings of Markram et al. (Markram et al., 1997)]. The time course of the EPSCs also was similar to the time course of short-distance EPSPs evoked by a single presynaptic pyramidal neuron during simultaneous current-clamp recordings from a pair of synaptically connected layer 3 pyramidal neurons in monkey PFC. The similarity in waveform kinetics between EPSCs and EPSPs also indicates that the synapses mediating the EPSCs in the present study were located in a membrane region in which voltage was poorly controlled by the somatic voltageclamp (Spruston et al., 1993). Finally, if the location of synaptic contacts is in fact related to the effectiveness of a connection to drive the postsynaptic cell, then the similarity in the time course of EPSCs evoked from shortvs. long-distance stimulation sites would indicate that differences in strength between shortand long-distance connections are dependent only on the number of synaptic contacts that each target cell receives from each source (see below).
How Robust is the Input from the Long-distance as Compared with the Short-distance Connections?
A large proportion of the excitatory input to cortical pyramidal neurons is generally assumed to be provided by short-distance, local axon collaterals of neighboring pyramidal cells (Douglas et al., 1995; Markram, 1997). However, in many of our experiments, monosynaptic EPSCs elicited from distal stimulation sites had a similar or larger amplitude than EPSCs elicited from more proximal sites (Fig. 9). We assumed that the evoked EPSCs were not unitary and that therefore the EPSC amplitude was directly proportional to the number of stimulated synaptic connections. If this assumption is correct (see below), then the presence of large EPSCs evoked from distant sites is contrary to the expectation that monosynaptic input would decline progressively with distance, due to the severing of horizontal axons during the slicing procedure. Therefore, this finding suggests that long-distance, horizontal, intrinsic projections are a relatively strong source of excitatory input to layer 3 pyramidal neurons.
The monosynaptic EPSCs recorded in the present study were likely to be mediated by more than one synaptic connection because (i) activation of unitary connections requires minimal stimulation, which we did not use; and (ii) in neocortical slices, unitary EPSCs evoked by minimal stimulation have been reported to show high failure rates (Rumpel et al., 1998; Gil et al., 1999), which we did not observe. In our experiments, failures were detected only in EPSCs that, in concert, had a highly variable latency and thus were polysynaptic. On the other hand, the conclusion that unitary responses must exhibit significant failure rates may not be correct. For example, in neocortical slices at 32–34°C, unitary EPSPs had extremely low failure rates (0–5%) during simultaneous recordings from synaptically connected pyramidal cells (Markram et al., 1997; Galarreta and Hestrin, 1998). This observation was explained by the fact that pyramid-to-pyramid connections in neocortex involve an average of six synaptic contacts (Markram et al., 1997). Also, the sampling procedure in the present study was biased to detect connections with a high probability of release since only 10–20 traces were collected for each stimulation site. Finally, the amplitudes of the EPSCs recorded in this study were generally small and similar to that of unitary EPSCs observed in paired recordings (Galarreta and Hestrin, 1998).
However, compared with paired recordings, a high failure rate is to be expected with extracellular stimulation, due to sources other than failures of release, e.g. stimulation failures due to threshold stimulation intensity (Gil et al., 1999) and conduction failures, which are more likely for cut axons than for intact ones (Stratford et al., 1996). Moreover, unitary EPSPs in paired recordings display almost no failures at 35°C but very high failure rates when recorded at low temperatures, because of a pronounced decrease in probability of release that seems to overcome synapse multiplicity in pyramid-to-pyramid connections (Hardingham and Larkman, 1998). Therefore, it seems reasonable to conclude that in our experimental conditions (which include room temperature and non-minimal extracellular electrical stimulation) the absence of failures indicates a multi-connection EPSC.
What Proportion of Layer 3 Pyramidal Cells Receive Long-distance, Excitatory, Monosynaptic Inputs?
Our findings also suggest that most pyramidal neurons in layer 3 are targets of long-distance, horizontal projections. Specifically, low-intensity stimulation at long distances from the recorded layer 3 pyramidal cell evoked monosynaptic EPSCs in the majority (77%) of these neurons. However, this proportion is likely to be an underestimate since some long-distance axon collaterals were probably severed by slicing of the tissue blocks. Although the present study does not indicate whether horizontal projections synapse selectively onto layer 3 pyramidal neurons, our results show that these cells frequently receive this type of synaptic input.
Alternatively, if our estimate is accurate, the fact that not all the recorded layer 3 neurons appear to receive long-distance input raises the possibility that a subpopulation (23%) of layer 3 pyramidal neurons does not participate in long-distance interactions. Interestingly, in monkey visual cortex, ~25% of layer 2/3 pyramidal neurons give rise only to short-distance connections (Yabuta and Callaway, 1998).
Does the Spatial Pattern of Excitatory Monosynaptic Responses Evoked by Multisite Stimulation Reflect the Clustered Organization of Horizontal Connectivity in Monkey PFC?
Previous anatomical studies showed that PFC supragranular pyramidal neurons are organized into discrete clusters separated by gaps, and that these clusters are reciprocally connected by intrinsic, horizontally oriented axon collaterals (Levitt et al., 1993; Kritzer and Goldman-Rakic, 1995; Pucak et al., 1996). In addition, the labeled horizontal axons present in the gaps between clusters are nonbranching and nonvaricose, suggesting that these axons have a very low probability of connecting with neurons located in the intervening gaps. Based on the spatial arrangement of these horizontal connections, one would expect a periodic variation in the strength of functional connectivity, with peaks and troughs in connectivity strength occurring at the locations of the clusters and gaps, respectively.
The functional significance of these findings depends, in part, on the neural elements activated by the extracellular electrical stimulation employed in this study. Experiments in rat visual cortex slices suggest that axon branches, but not cell bodies, dendrites or axon initial segments, are activated by electrical stimulation (Nowak and Bullier, 1998a,b). In contrast, other studies show that electrical or glutamate stimulation produce similar spatial patterns of activation in slices from neocortex (Dalva et al., 1997; Yuste et al., 1997), arguing against a predominance of electrical stimulation of axons versus cell somata. Since it is possible that the stimulation we used activated both these types of presynaptic elements, it is important to consider how they would contribute to the observed profiles of horizontal monosynaptic EPSCs. Specifically, if the stimulation results solely in the activation of axons, one would expect a simple decay of response strength with distance, because the number of intact horizontal axons (and thus of inputs available for stimulation) would progressively decrease as a function of distance from the recorded cell, due to the slicing of the tissue blocks. On the other hand, if the stimulation activates only cell bodies (or axon initial segments), then stimulation at the locations of the cell clusters would produce peaks in EPSC amplitude, whereas stimulation at the gaps between clusters would evoke no response.
The present study revealed two patterns in the distribution of monosynaptic EPSC amplitudes evoked as a function of horizontal distance in superficial layers of monkey PFC. In a minority of cells (12.5%), EPSC amplitude decayed with distance in a monotonic fashion (Fig. 9A). By contrast, in 87.5% of the cells, the distribution of monosynaptic EPSC amplitudes (as a function of horizontal distance from the recorded cell) showed a series of amplitude peaks superimposed on an overall monotonic decay. As suggested elsewhere (Nowak and Bullier, 1998a,b), the decay in EPSC amplitude with stimulation distance indicates that activation of axons is a major factor in determining EPSC amplitude. However, the presence of peaks in the spatial patterns of EPSC amplitude observed for the majority of PFC layer 3 pyramidal cells would also be consistent with the activation of discrete clusters of cell bodies whose intrinsic axon collaterals innervate the recorded cell. Interestingly, the range of horizontal distances (135–810 μm) between the stimulation sites that evoked peaks in EPSC amplitude was very similar to the range of distances (200–1200 μm) between centers of stripe-like clusters in anatomical studies of monkey PFC (Levitt et al., 1993). In fact, previous studies in other cortical regions suggest that clustered connections are the basis of a peaked distribution in EPSC amplitudes. For example, using whole-cell recordings and multisite electrical stimulation in visual cortex slices, Weliky and colleagues (Weliky and Katz, 1994; Weliky et al., 1995) observed EPSC peaks when stimulation was applied at the locations of clusters of neurons that shared orientation selectivity with the recorded cell. These iso-orientation clusters of visual cortex neurons are known to be linked by intrinsic horizontal axons (Gilbert and Wiesel, 1989). However, the present study does not provide direct evidence for a correspondence between location of clusters of connected neurons and stimulation sites evoking peaks in EPSC amplitude. Therefore, whether the spatial peaks in EPSC amplitude result from clustered connectivity in the PFC slices remains to be demonstrated.
Functional Significance of Long-distance, Horizontal, Excitatory inputs to Layer 3 Pyramidal Cells in Monkey PFC
The present study demonstrates that long-distance horizontal connections constitute a significant intrinsic pathway of synaptic communication in superficial layers of monkey PFC. Our experimental approach does not, however, directly address the functional role of these connections during dynamic operations of the PFC in vivo. Our present in vitro results, in concert with previous anatomical and in vivo electrophysiological studies, may provide support for several possible functional roles for these connections.
It is possible that, as in other regions of neocortex, PFC neurons with similar functional properties are clustered and that the horizontal connections link these clusters of cells. In spatial delayed-response tasks, the delay period activity of PFC neurons is selective for the location of the cue in the visual field, suggesting that individual neurons have specific ‘memory fields’ (Funahashi et al., 1989; Rainer et al., 1998). By analogy to the links between iso-orientation columns in the visual cortex, the long-distance horizontal connections in the PFC have been proposed to link clusters of cells sharing memory fields (Goldman-Rakic, 1995). In addition, in visual cortex, horizontal connections have been proposed to mediate the long-distance, between-column synchronization of activity during visual stimulation (Mountcastle, 1998). Because cells of different columns usually have different receptive fields, horizontal connections are thought to mediate the effects of stimulation outside a given cell's visual receptive field (Gilbert, 1998). Therefore, by a strict analogy to the visual cortex, horizontal connections in the PFC would mediate lateral interactions between clusters of neurons with different memory fields but which share some other functional property, such as object selectivity (Rainer et al., 1998).
Although delay period activity of monkey PFC neurons is thought to be the cellular basis of working memory (GoldmanRakic, 1995), the mechanisms generating the sustained firing of PFC neurons during the delay period are still largely unknown. It has been suggested that sustained activity may be subserved, at least in part, by reverberating excitatory circuits (Funahashi and Kubota, 1994), but the organization of this circuitry is not known. The presence of reciprocal, long-distance, horizontal connections between clusters of supragranular pyramidal neurons (Pucak et al., 1996), the selectivity of these axon collaterals for pyramidal cell targets (Melchitzky et al., 1998) and the observation that the vast majority of layer 3 pyramidal neurons receive robust, long-distance, monosynaptic inputs (present study) suggest that, in the monkey PFC, these horizontal connections could serve an important role in promoting recurrent activation of supragranular pyramidal cells. That is, the mutual excitation of these cell clusters via reciprocal, horizontally oriented inputs could contribute to the sustained activity of PFC cells during the delay period of delayed-response tasks (Lewis and Anderson, 1995). However, sustained neuronal activity is also present in parietal and temporal cortices during the delay period of working memory tasks (Fuster and Jervey, 1981; Miller and Desimone, 1994; Zhou and Fuster, 1996). Interestingly, compared with the smaller patch-like arrangement of horizontally connected cell groups in posterior cortical regions (Lund et al., 1993), the stripe-like organization of cell clusters in the PFC suggests that a larger number of pyramidal cells per cluster participate in horizontal interconnections. If these reciprocal, horizontal connections subserve reverberating activity, then the larger the number of pyramidal cells recruited, the more robust their sustained firing is likely to be. Therefore, the stripe-like architecture of the PFC could explain why the delay-related activity of PFC cells is less likely to be interrupted by intervening stimuli than the delay-related firing of neurons in temporal or parietal cortices (Miller et al., 1996).
We thank Darrell Henze and Nathan Urban for providing the acquisition and analysis software and for the helpful discussions. We also thank Brad Booth, Mary Brady and Darlene Melchitzky for their excellent technical assistance. This work was supported by NIMH grants MH51234 and MH45156 and Independent Scientist Award MH00519.
Address correspondence to David A. Lewis, MD, University of Pittsburgh, 3811 O'Hara Street, W1650 BST, Pittsburgh, PA 15213, USA. Email: firstname.lastname@example.org.