INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Timing is perhaps one of the most critical features of signal processing and integration in neural communication (Ferster and Spruston 1995). The distribution in time of action potentials and synaptic events critically determines interactions between groups of neurons in both the peripheral and CNS. In synaptic transmission, the time between activation of a pathway by a stimulus until its arrival as a synaptic response, the latency, is thought to be directly related to the nature and complexity of that pathway. This timing of events dictates the impact of the convergence of multiple synaptic inputs on single neurons and thus contributes to regulation of neuron excitability and the overall performance characteristics of a network or circuit of neurons. Understanding the mechanisms that shape synaptic timing is then a basic building block to understanding the cellular basis of function in any brain region.

The nucleus of the solitary tract (NTS) provides an interesting model for investigating synaptic performance and interactions. NTS is a major integrative site where visceral sensory information enters the brain and projections originate to a variety of brain stem and supra medullary nuclei with reciprocal projections back to NTS (Loewy 1990). The sensory axons from cranial nerves traverse the brain stem through the bundle of the solitary tract (ST) to their synapses within NTS. This anatomical feature of ST offers direct access to afferent axons for electrical stimulation within brain stem slices (Andresen and Yang 1990). ST stimulation evokes several synaptic responses within NTS neurons reflecting both activation of visceral sensory synapses as well as responses via indirect polysynaptic paths initiated in local circuits by ST activation (Andresen and Yang 1990, 1995; Miles 1986; Zhang and Mifflin 1998).

The most common approach to distinguishing mono-synaptic from polysynaptic pathways relies on the shortness of the absolute latency and whether evoked synaptic events faithfully follow pairs of closely timed shocks (Andresen and Yang 1995; Aylwin et al. 1997; Berger and Averill 1983; Deuchars et al. 2000; Gil et al. 1999; Miles 1986; Yoshimura and Jessell 1989; Zhang and Mifflin 2000). In the present study, we assessed whether these commonly used electrophysiological criteria clearly distinguish mono- from polysynaptic responses. To do this, we examined synaptic responses to bursts of ST activation in NTS neurons displaying a wide range of absolute latencies and assessed their response variability. We also determined the synaptic failure rates in these neurons. Last, we compared these synaptic performance characteristics to those of NTS neurons that were anatomically identified as second-order neurons by the presence of tracer-labeled sensory boutons on their cell bodies. Together, the results suggest that, contrary to conventional expectations, only synaptic jitter, the intraneuronal variability of the synaptic latency, reliably indicated monosynaptic contacts.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Slice preparation

Brain stem slices were prepared from adult (100-300 g) male Sprague-Dawley rats (Charles River). Rats were deeply anesthetized with ether and quickly killed by cervical dislocation. The hindbrain was removed and placed for 1 min in freezing (3 to 0°C) artificial cerebrospinal fluid (ACSF) containing (in mM) 125 NaCl, 3 KCl, 1.2 KH₂PO₄, 1.2 MgSO₄, 25 NaHCO₃, 10 dextrose, and 2 CaCl₂, bubbled with 95% O₂-5% CO₂. The medulla was trimmed rostrally and caudally to yield a 1-cm block centered on the obex. The cerebellum was removed, and a wedge of tissue was cut from the ventral surface to orient the brain stem such that 250-µm slices contained lengthy segments of sensory axons in the ST and their terminations on neurons within NTS (Mendelowitz et al. 1992). Slices were cut with a sapphire knife (Delaware Diamond Knives) mounted in a vibrating microtome (Leica VT-1000). Slices were placed in a custom perfusion chamber and secured with a nylon mesh on a platinum frame so that no strands crossed the ST or medial portions of NTS. This prevented compression of the sensory axons and minimized fluorescent background light emitted from the nylon fibers. The slices were perfused with ACSF at 34-37°C; pH 7.4; 300 mOsm. Recordings were made from NTS neurons located just medial to the ST and within 200 µm rostral or caudal from obex, the area of densest aortic baroreceptor afferent terminations (Mendelowitz et al. 1992).

Visualized patch recordings

Neurons were visualized using a Zeiss Axioskop microscope equipped with fluorescence, differential interference contrast (DIC) optics (×40 water immersion lens), and an infrared (IR) sensitive camera. Patch electrodes were guided using IR DIC microscopy to neurons that were voltage clamped in the whole cell configuration (Axopatch 200A). Recording electrodes (1.8-3.5 M) were filled with an intracellular solution containing (in mM) 10 NaCl, 130 K Gluconate, 11 EGTA, 1 CaCl₂, 2 MgCl₂, and 10 HEPES; pH 7.3; 295 mOsm. In some experiments (noted in the text), 130 K Gluconate in the internal solution was replaced with 125 KCl (high internal Cl solution) to displace the chloride reversal potential away from the resting membrane potential. Data were filtered at 5 kHz and sampled at 10-20 kHz using p-Clamp7 software (Axon Instruments). Neurons were accepted for recording if they had an initial seal resistance >1 G and over the initial 10 min required no more than 100 pA to hold the neuron at 70 mV.

Stimuli were delivered by placing a 200-µm-diam concentric bipolar stimulating electrode (F. Haer) on the visible ST 1-3 mm from the site of the recorded neuron soma and recorded ST evoked postsynaptic currents (PSCs). Bursts of five ST stimuli every 4 or 5 s at frequencies of 50, 100, or 200 Hz were generated with a Master-8 isolated programmable stimulator (A.M.P.I. Jerusalem, Israel). At the cycle rate of 4-5 s, consistent synaptic responses were sustained for many tens of minutes with no burst-to-burst changes. Stimulus duration was set at 0.1 ms based both on experience activating myelinated and unmyelinated afferent volleys in the aortic depressor nerve (Fan and Andresen 1998) and on preliminary tests of the ST stimulus duration to evoke maximum synaptic responses. Neurons were rejected from further study if the evoked synaptic currents initiated a fast sodium current at clamped potentials more negative than 40 mV indicating inadequate voltage control. All drugs were dissolved in 100 µL DMSO, diluted with external solution and bath applied. 2,3-Dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulphonamide (NBQX) was purchased from Tocris-Cookson (Ballwin, MO), and all other drugs were purchased from Sigma (St. Louis, MO). Application of the highest concentration of vehicle used for dissolving drugs had no effect on neuron properties or synaptic responses (data not shown).

Labeling of the aortic depressor nerve

In 20-day-old, pentobarbital-anesthetized (50 mg/kg) rats, the aortic depressor nerve (ADN) was located and separated from the surrounding tissue 1 cm peripheral to joining the superior laryngeal nerve and entering the nodose ganglion (Mendelowitz et al. 1992). The nerve trunk was then placed within a preformed molded shell of dental impression compound (Coltene, Mahwah, NJ). The lipophilic fluorescent dye, DiA (Molecular Probes, Eugene, OR), was carefully placed on ADN, making sure that dye did not contact adjacent nerves and structures. To prevent dye migration to adjacent nerves, the ADN was then sealed in place by application of additional dental impression compound and allowed to cure in place. Animals were killed for experiments 2-3 mo following dye implantation. Previous studies suggest that this procedure fills synaptic boutons belonging to aortic arch baroreceptors, and these are concentrated over cell bodies in medial portions of caudal NTS (Mendelowitz et al. 1992). For electrophysiological experiments, fluorescent ADN boutons within NTS were located visually under brief exposures to excitation wavelengths (Chroma, PN31024, DiA) and then images captured digitally using camera integration time intervals of 2-5 s. Captured fluorescent bouton images were then overlaid with real time IR DIC images using an ARGUS 20 image processor (Hammamatsu Bridgewater, NJ) to establish co-localization of fluorescence with the underlying neural structure. The presence of fluorescent ADN sensory boutons coincident with the neuron soma identified second-order neurons within NTS with anatomically mono-synaptic connections, and the synaptic responses to ST activation were then studied in these neurons under voltage clamp (see Fig. 8).

Synaptic performance was assessed by measuring the synaptic latency as the time between the onset of the stimulus artifact and the onset of synaptic current. For improved precision, the membrane current trace was first differentiated, and these two onset times were taken as crossing points of a set threshold level (see RESULTS for details). For clarity, only the synaptic responses to the first ST shock in each burst was analyzed for the jitter calculation. Jitter was calculated as the standard deviation of the shock-to-shock variation in latency within each neuron averaged over multiple trials (25-50 trials). The absence of a synaptic current deflection following a stimulus shock to the ST was taken as a synaptic failure and the rate of failure calculated as the number of failed synaptic responses to bursts of stimuli (>50 Hz) across mulitple trials. Failure rate was calculated as the number of absent synaptic responses divided by the number of stimulus shocks delivered and expressed as a percentage. Derived synaptic parameters (latency, jitter, and failure rate) were compared across groups using two sample t-tests (Snedecor and Cochran 1980), and groups were considered significantly different at P values <0.05. In addition, correlations were tested as described in RESULTS using linear models with least-squares fitting (Origin 6.1, OriginLab).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Trains of electrical shocks to the ST (Fig. 1A) most often evoked large excitatory postsynaptic currents (EPSCs) in medial NTS neurons. Increases in the shock intensity of ST stimulation did not further increase the EPSC amplitude, suggesting that recruitment of additional fibers was minimal. If the ST shocks were repeated at short intervals (<2 s), the amplitudes of successive EPSCs within a burst of stimuli declined substantially. The decrease in amplitude was greater at higher frequencies, and this frequency-dependent depression (FDD) was similar in all neurons tested.

View larger version (17K): [in this window] [in a new window]   Fig. 1. Solitary tract (ST) stimulation in different medial nucleus tractus solitarius (NTS) neurons evoked excitatory postsynaptic currents (EPSCs) with different latencies. Successive shocks () induced a similar frequency-dependent depression of EPSC amplitude in these 2 examples of different neurons. Traces are averages of 5 consecutive sweeps. A: short-latency EPSC (latency, 2.2 ms; jitter, 46 µs) had no observable failures in individual traces (not shown). B: long-latency EPSC (latency, 7.5 ms; jitter, 152 µs) had several observable failures particularly on the 5th pulse of the train so that the average trace has little net deflection. In both cases, ST stimulus trains consisted of 5 pulses at 100 Hz, and neurons were voltage clamped at 70 mV.

To improve the precision in assessment of synaptic latency, the individual synaptic current traces were mathematically differentiated with respect to time (Fig. 2A). Differentiation accentuated the initial phase of the PSC onset. Latencies were calculated as the time between initiation of the ST shock until the time at which the differentiated signal trace crossed a horizontal noise threshold level. This threshold was set at a level equal to the greatest transient value of the fluctuations in the differentiated signal trace for 1-50 ms preceding the PSC. The first point of the PSC differential to exceed a noise threshold was designated as the beginning of the synaptic event (Fig. 2A, left small arrow). This technique provides a simple, precise measurement of the latency. Absolute latency values for a given neuron were calculated over a 30- to 60-min period of recording as one measure of synaptic performance.

View larger version (31K): [in this window] [in a new window]   Fig. 2. Analysis of synaptic jitter. A: raw digitized current records were differentiated (middle trace) to facilitate accurate identification of the onset of the PSC. Time points for the onset of the stimulus and PSC were determined by differentiating current traces with respect to time (dl/dt). Short arrows (left and right) indicate data points in the differentiated signal that 1st exceed a noise threshold level (dashed line) indicated by the beginning time points for stimulus (left arrow) and PSC (right arrow). The total latency between the ST shock and the PSC (double headed, horizontal arrow) was determined by subtracting the time of the stimulus onset from the onset time of the PSC. B: the unimodal distribution of the histogram of absolute latency for 36 NTS neurons reveals no distinction for mono- or polysynaptic events. Absolute latency is expressed as the mean of the latency of 20-50 individual PSCs measured over a 30- to 60-min period.

The range of absolute PSC latencies to ST stimulation across NTS neurons was quite broad, and the distribution was fairly unimodal with no clear divisions across our sample of 49 neurons (Fig. 2B). All of the shortest latency PSCs (<3 ms) were excitatory. The longer latency (>3 ms) PSC group, however, included both excitatory and inhibitory PSCs (IPSCs). Given the distribution and mixture of excitatory and inhibitory events, the pathway underlying moderate latency events is ambiguous based on the absolute latency. Thus we examined the shock to shock variation in latency (jitter) for repeated ST-evoked PSCs within neurons as an additional measure of synaptic performance.

EPSCs in neurons with the shortest absolute latencies (Fig. 3A) had remarkably low jitter (<100 µs). This quantitatively high degree of temporal stability in synaptic performance is consistent with transmission via a mono-synaptic pathway. Most highly consistent synaptic responses reliably evoked EPSCs with fewer than 5% failures. These low-jitter synaptic responses were fast, excitatory inward currents that reversed near 0 mV, and their pharmacological profile was similar to previous reports from medial NTS neurons (Andresen and Yang 1995). The non-N-methyl-D-aspartate (non-NMDA) selective glutamate receptor antagonist NBQX completely blocked such EPSCs. Recording such responses at depolarized membrane potentials or in extracellular solutions nominally free of Mg²⁺ (Fig. 3B) failed to reveal any contribution of NMDA receptor activation by synaptically released glutamate.

View larger version (25K): [in this window] [in a new window]   Fig. 3. Short-latency EPSCs with minimal jitter were mediated by glutamate acting at non-N-methyl-D-aspartate (non-NMDA) receptors. A: short absolute latency EPSCs recorded from a medial NTS neuron voltage clamped at 70 mV. Five individual EPSC traces are overlaid to show shock-by-shock variation in latency to ST stimulation. Each trace is the 1st EPSC of a burst. Every ST shock () evoked an EPSC with the high reliability of a monosynaptic connection (latency, 2.6 ms; jitter, 68 µs and no observable failures, train of 5 pulses at 200 Hz). B: successive, high-frequency shocks reveal a frequency-dependent depression of EPSC amplitude. EPSCs were completely blocked with 50 µM 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulphonamide (NBQX) without extracellular Mg2+, suggesting that no NMDA component was activated by ST shocks.

In contrast, the synaptic performance of longer latency EPSCs often displayed substantial temporal variation. Synaptic jitter of long latency responses increased as much as 10-fold over those with short latencies (Fig. 4A). Interestingly, despite these highly variable synaptic responses, failures to bursts of ST stimuli were infrequent or absent in these neurons. In some neurons with these intermediate latency EPSCs (3 of 5), pharmacological blockade with NBQX in solutions with zero extracellular Mg²⁺ revealed a small, slow inward synaptic response (Fig. 4, B and C). Note that such NBQX-insensitive inward synaptic currents always had a longer absolute latency (Fig. 4C) than the NBQX-sensitive EPSC and failed to follow 50-Hz trains of ST stimuli. The NBQX insensitive currents were blocked with the NMDA receptor antagonist 2-amino-5-phosphonopentanoic acid (AP-5; Fig. 4, B and C). These observations (short-latency non-NMDA event and longer latency, NMDA event) are consistent with two events of separate origin activated at different synapses on the same neuron. Thus in neurons that appear to be polysynaptically linked to the ST (long latency with high jitter), a mix of synaptic responses was found that could include participation of non-NMDA and NMDA receptors.

View larger version (22K): [in this window] [in a new window]   Fig. 4. Moderate latency EPSCs evoked by ST stimulation have greater temporal variability and a different pharmacological profile. A: moderate latency EPSC evoked by ST stimulation (Vm = 70 mV; 0 mM extracellular Mg2+). There is greater sweep-to-sweep variation in latency (jitter, 120 µs). Five EPSCs are overlaid to show differences in latency between stimuli (left panel). Each trace is the 1st EPSC of a burst of stimuli. B: there are no observable failures to 100-Hz ST stimulation. Note the frequency-dependent depression. C: in a time-expanded view of the initial portion of this same case as B, a small, slow NBQX- (50 µM) insensitive response remains that was blocked by the NMDA receptor antagonist 2-amino-5-phosphonopentanoic acid (AP-5; 100 µM). The NBQX-sensitive response had a shorter latency than the NMDA response (3.3 vs. 6 ms), suggesting origin from different synapses.

Many PSCs with the greatest jitter were inhibitory, and some regularly failed in response to high-frequency trains of stimuli (Fig. 5). Interestingly, the long-latency, high jitter IPSCs could be blocked with either the GABA_A receptor antagonist bicuculline or the excitatory non-NMDA receptor antagonist NBQX (Fig. 5, B and C). This direct pharmacological evidence is consistent with a polysynaptic connecting path from the ST. For these types of PSCs, one step in the path from ST was mediated by non-NMDA glutamate receptors, and this event was responsible for evoking the recorded GABA_A IPSC. Because of their relatively high temporal variability, long latencies, and common failures, these polysynaptic responses were the least reliable.

View larger version (18K): [in this window] [in a new window]   Fig. 5. Polysynaptic inhibitory postsynaptic current (IPSC) evoked by ST activation. A: traces show synaptic responses to trains of 4 shocks delivered at 50 Hz ( with vertical lines indicate each shock). The latency was quite long (absolute latency, 5.1 ms) and highly variable (jitter, 259 µs). Failures were evident in each of the bottom 3 traces in A but occurred at different pulses within each of these repeated trains. Note as well that frequency-dependent depression of this IPSC was substantial. This response was recorded using the high internal chloride solution and was an inward IPSC under the recording conditions (Vm = 80 mV). This IPSC was reversibly blocked with either the non-NMDA antagonist NBQX (B) or the GABAA receptor antagonist bicuculline (C). All traces in these panels are single records and not averaged. Overall data are consistent with a serial polysynaptic pathway in which the observed PSC was blocked at separate points by 2 different selective antagonists. NBQX blockade implies that ST stimulation evoked an initial, unobserved glutamatergic EPSC that was responsible for evoking the GABAa IPSC (bicuculline block) observed in the recorded cell. Such a polysynaptic pathway was consistent with synaptic performance (high jitter) as well as the pharmacological profile.

Our pool of 49 NTS synaptic responses included both EPSCs and IPSCs. Across this sample, synaptic jitter generally increased as the absolute latency increased with a strong and significant positive correlation (R = 0.573, Fig. 6A). Given the nature of the characterization of individual synaptic responses described above, this pool represents a mixture of second and higher order responses. Subregions of this jitter relationship may reflect differences in this mixture of mono- and polysynaptic response pathways. At the very shortest latencies (<2.8 ms), values fell within a quite restricted range of jitter values with no jitter values >100 µs (Fig. 6A). Beyond this short-latency range, jitter values ranged widely. Note that in the range of 3-9 ms, jitter could be as low as the lowest values found at the shortest latencies or over 700 µs. In the simplest interpretation, some long-latency events were highly reliable and likely were mediated by simple, monosynaptic pathways despite their late arrival after ST activation. The high jitter synaptic group included the pharmacologically, disynaptic GABA_A IPSCs described above. Beginning at approximately double the shortest latency that we recorded, intermediate latency EPSCs (>2.8 ms) had jitters ranging from among the lowest of the entire sample to substantial (>100 µs). These intermediate latencies could not be unequivocally classified as mono- or polysynaptic based on their latency or jitter characteristics to ST burst stimulation.

View larger version (16K): [in this window] [in a new window]   Fig. 6. A: summary relationship between jitter and the absolute latency. Jitter was positively and significantly correlated to increasing latency (correlation coefficient = 0.573). Points represent values from 49 separate medial NTS synaptic responses. Jitter was expressed as the standard deviation of the absolute latency. Dotted line is a least-squares linear regression fit to the data (y = 52.6X  55.9; R = 0.577; n = 49; P < 0.0001). At latencies >3.0 ms, many synaptic responses displayed very high jitter or very low jitter so that values fell far from the regression fit. Circled points indicate neurons that were anatomically identified as being 2nd order to sensory contacts with DiA labeling. Data points derived from related figures are indicated for several individual values by arrows.  indicates average jitter point for EPSC with expanded display in B. B: individual histograms showing the distribution of individual latency values to a series of ST shocks for 2 individual neurons with similar overall latency (about 5 ms) but very different synaptic jitter.  and filled bars display latency values for an EPSC from a DiA aortic depressor nerve (ADN)-labeled, anatomically second-order NTS neuron. Note the narrow range of values (<63 µs) for this monosynaptic response. In contrast, the open bars display the range of latency values for the pharmacologically polysynaptic response of the neuron from Fig. 5. In such cases, the range of latency values is quite large (nearly 1.3 ms) and the distribution less symmetrical.

Conventionally, polysynaptic events have been distinguished by their inability to sustain synaptic responses to two closely timed stimuli. This is based on the premise that increasing the number of synapses increases the likelihood that an element will fail in the pathway required for transmission. If failures can distinguish mono- from polysynaptic responses, then the frequency of failures to bursts of stimulation should be directly correlated to events with the shortest latencies and the smallest jitters. Our data do not support this concept. Responses with high failure rates (Fig. 7A) occurred over such a broad range of latencies that no relation was detected by regression analysis (P = 0.540). Likewise, the magnitude of synaptic jitter (Fig. 7B) was unrelated to failure rate (P = 0.139). Also, responses that were clearly disynaptic (e.g., Fig. 4) could have zero failures at high frequencies of activation. Thus the absence of overt failures poorly predicts synaptic order. In this preparation, monosynaptic PSCs appear to have jitter of <100 µs. As a group (n = 27), such EPSCs had an average latency of 2.8 ± 1.0 ms with a jitter of 54 ± 21 µs and an average failure rate of 5.9 ± 11.5% (e.g., Fig. 3). These electrophysiologically based reliability criteria for mono-synaptic events are consistent with the pharmacological data. High jitter (>100 µsec) synaptic PSCs (n = 22) had an average latency of 4.6 ± 1.6 ms, an average jitter of 237.4 ± 170.1 µs, and average failure rate of 6.8 ± 7.9%. High jitter (>100 µs) EPSCs (Fig. 4) and all IPSCs (Fig. 5) were considered polysynaptic. To test further the relationship between jitter and synaptic pathway order, we recorded ST evoked EPSCs from NTS neurons with anatomically verified, direct connections to ST sensory afferents.

View larger version (18K): [in this window] [in a new window]   Fig. 7. Summary of relationships between failure rates (responses missing expressed as % of all stimuli) of individual synaptic responses and either the absolute latency (A) or the jitter (B). The failure rate was independent of the absolute latency or the magnitude of synaptic jitter. Dotted lines are least-squares linear regression fits to the respective datasets (A, latency: Y = 0.55X + 4.30; R = 0.09; n = 49; P = 0.540; B, jitter: A = 0.015X + 4.32; R = 0.21; n = 49; P = 0.139). Circled points depict neurons with anatomically identified sensory contacts.

The ADN in the rat contains primary afferent fibers that enter the brain stem travel through the ST and terminate in NTS. The ADN forms axo-somatic synapses with NTS second-order neurons (Mendelowitz et al. 1992). In a subset of experiments, we studied brain stems of animals with fluorescently tagged ADNs. We searched for and recorded from neurons with DiA-labeled primary terminals within NTS (see METHODS). Time-integrated images revealed the presence of punctate fluorescently labeled boutons (Fig. 8, green dots). Using IR illumination, real-time DIC images of the underlying neuron structure were compared with the locations of fluorescent boutons. The real-time IR DIC images were used to guide patch electrodes to the anatomically identified, second-order neurons and record synaptic responses. Train stimulation of the ST evoked fast EPSCs that rapidly depressed in these identified neurons (Fig. 9). These EPSCs ranged in latency from 1.9 to 4.8 ms (Fig. 6, circled points), and failure rates (Fig. 7, circled points) ranged from no observable failures to 39%. Interestingly, jitter was remarkably invariant across these synaptic responses ranging from 31 to 61 µs and, overall, the values fell along the minimum jitter range across our whole sample of NTS synaptic responses (Fig. 6, circle points). NBQX completely blocked all these synaptic responses. In this study, we had characterized five bicuculline-sensitive, ST-evoked IPSCs and concluded that these were clearly polysynaptic, both pharmacologically and according to their synaptic performance. We then compared these two groups (proposed monosynaptic and polysynaptic IPSC responses) for their relative synaptic performance parameters. EPSCs from identified second-order neurons had shorter latencies and much lower jitters than the GABA_A IPSCs (latency, 2.95 ± 0.71 vs. 5.56 ± 0.74 ms, mean ± SE, P = 0.027; jitter, 42.3 ± 6.5 vs. 416.3 ± 94.4 µs, P = 0.013, n = 4 and 6, respectively). Surprisingly, however, the failure rates across these two groups were similar (27.8 ± 9.0 vs. 9.7 ± 4.4% failures; n = 4 and 6, P = 0.08, respectively).

View larger version (123K): [in this window] [in a new window]   Fig. 8. Example of location of fluorescent labeling from aortic baroreceptor afferent boutons on medial NTS neuron. Punctate fluorescent label was found primarily directly over the soma of NTS neurons. Overlay of a time-integrated fluorescent DiA signal with a real-time infrared (IR) illuminated differential interference contrast (DIC) image allowed a patch electrode to be guided to the labeled neuron. Fluorescent ADN terminals formed green patches co-incident with cell body identifying a 2nd-order NTS neuron within the slice.

View larger version (25K): [in this window] [in a new window]   Fig. 9. Synaptic responses to ST stimulation of an anatomically identified neuron (see Fig. 8 for image of recorded neuron). A: single unaveraged traces of fast EPSCs evoked by ST when voltage clamped at 70 mV. Note the consistent and short latency (latency, 2.3 ms) and the low jitter (jitter, 61 µs). Each ST shock evoked an EPSC (train of 5 pulses at 100 Hz; , stimulus). No failures occurred in this example despite robust frequency-dependent depression. B: overlay of 5 successive EPSCs indicative of only small jitter in the latency across stimuli. Responses were completely blocked with the non-NMDA glutamate receptor antagonist NBQX. The NBQX trace is an average of 5 individual sweeps. Such low jitter glutamatergic EPSCs were found in all labeled neurons (see additional data in Figs. 6 and 7). Thus responses of anatomically identified 2nd-order neurons suggest that while neurons monosynaptically activated by ST have low jitter, they can have remarkably high rates of failure and long latencies (see Fig. 6).

Our studies used three strategies to identify neurons in our slices: synaptic electrophysiology, antagonist pharmacology, and dye labeling of second-order neurons. Comparing anatomically monosynaptic neurons to pharmacologically polysynaptic responses display characteristically different variation patterns to a series of ST shocks. The individual latency values for ST EPSCs in anatomically monosynaptic neurons were very narrowly and fairly symmetrically distributed around the mean latency value (Fig. 6B). This was typical of low jitter, presumed monosynaptic EPSCs. In contrast, individual latency values for polysynaptic ST responses varied across a large range of latencies (several ms), and the distributions could be quite asymmetrical (Fig. 6B, open bars).

EPSC/IPSC sequences are common in NTS (Andresen and Yang 1990) and are often observed in response to ST stimulation in this slice preparation (Fig. 10). Such IPSCs follow the EPSC component (Fig. 10A), and the IPSCs have an apparent latency of 4-6 ms from the ST stimulus. With this timing, a single shock to the ST evoked the IPSC beginning in the falling phase of the EPSC and as a result truncating the EPSC amplitude and time course (Fig. 10A). Blocking the IPSC with bicuculline isolated the full characteristics of the EPSC (Fig. 10B). In this example, the initial EPSC in the complex response was very reliable (1.52-ms latency; 28-µs jitter) and had no observable failures at 200 Hz (V_m = 80 mV). However, during trains of ST shocks, at short interstimulus intervals, the long-latency IPSC evoked by the first ST shock coincided with the second, fast EPSC in the train (Fig. 10C). This process continued through the train. Such a response appears as a failure of the EPSC and would be considered evidence for a polysynaptic pathway using the common criteria of paired pulses separated by 5 ms. In addition, note that even the polysynaptic IPSC in this example followed 200-Hz stimulus trains (Fig. 10C, arrows). Thus simple two-pulse paradigms separated by 5 ms were insufficient to consistently separate mono- from polysynaptic NTS responses.

View larger version (15K): [in this window] [in a new window]   Fig. 10. In some medial NTS neurons, ST stimulation evoked complex synaptic responses. A: ST stimulation evoked a sequential EPSC/IPSC complex. The EPSC was of short latency (1.52-ms latency) and low jitter (28 µs) consistent with reliability criteria for a monosynaptic response. No PSC failures were observed even at 200-Hz ST stimulation. The long-latency IPSC began before the full decay of the EPSC and therefore truncated its amplitude and time course due to the temporal overlap of the 2 synaptic events. The combined synaptic interactions (A) were most evident at a holding potential of 40 mV, which was displaced from the reversal potential for Cl. B: in the presence of the GABAA receptor antagonist bicuculline (100 µM), the full time course of the EPSC was evident at 70 mV. C: in control conditions near normal resting membrane potential (60 mV), the long-latency IPSC collided with subsequent evoked EPSCs in the train. Arrows have been placed at intervals of predicted onset of the IPSCs based on A. The result obliterated the net EPSC and mimicked a synaptic failure (no evident EPSC). In this example, the polysynaptic IPSC followed successive stimuli at 200-Hz ST stimulation. Each trace is an average of 5 individual sweeps.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

To understand the mechanisms of information processing within a brain region, the nature of the pathway followed, i.e., the synaptic order, is critically important. In our study of medial NTS neurons, we sought to distinguish by electrophysiological means the responses mediated through monosynaptic pathways from those arising from polysynaptic pathways. We recorded synaptic currents evoked by ST activation and used three measures of synaptic reliability including latency, jitter, and failure rate to determine whether these neurons received monosynaptic or higher order inputs from ST. We made the conventional assumption that as synaptic order increased, latencies would become longer for more complex, polysynaptic pathways. In addition, we assumed that the reproducibility of the precise timing of the synaptic responses would also decline with an increasingly complex pathway and result in an increase in synaptic jitter. Third, we anticipated that monosynaptic responses would rarely fail, while more complex pathways (3rd order and higher) would fail more frequently.

Our fundamental measure of synaptic performance was latency. In medial NTS, the latency between the stimulus and the onset of the PSC varied considerably from neuron to neuron. The excitation pathway from stimulus to response involves multiple steps (Sabatini and Regehr 1999), and, among the mechanisms contributing to latency of the measured PSCs, time is required for 1) activation of the afferent nerve fiber, 2) conduction to the terminal site, 3) synaptic transmitter release, and 4) development of the postsynaptic response. Thus in the simplest case, a monosynaptic response requires a single iteration of each of these processes and should require completion in minimum time for the shortest latencies. In the case of polysynaptic responses, all of these processes are repeated at other, higher order neurons in the pathway and should thus proportionately add to the observed absolute latency and should make such complex pathways inherently more variable. Thus the PSCs arising from ST activation through monosynaptic and polysynaptic pathways should display different frequency-dependent characteristics related to pathway complexity.

To test the relation between latency and synaptic performance in NTS, we utilized a horizontal slice preparation of the brain stem. This slice orientation exposed the rostral-caudal course of the ST input pathway, and an electrode was placed under visual guidance directly on ST relatively distant (1-3 mm) from the recording site in medial NTS. Such remote activation of the afferent pathway minimized confounding local activation of interneurons. Finally for comparison, we used a newly developed technique to record selectively from a group of anatomically identified second-order neurons. These neurons were identified by anterogradely transported DiA as having anatomically monosynaptic contacts from aortic baroreceptor afferents. Such identified neurons were considered positively identified, second-order NTS neurons of the baroreflex sensory pathway (Loewy 1990). Only one of our conventional assumptions was consistently upheld across the data: jitter was minimal (<61 µs) in all ST responses of anatomically identified second-order neurons, and thus we conclude that variation in the synaptic latency from stimulus to stimulus within neurons is the best measure of pathway complexity. In contrast, the absolute latency values as well as the rates of failure ranged widely across all recorded neurons and were uncorrelated to the degree of jitter including that of neurons anatomically identified as second order. Such results raise serious concerns about strategies based on dividing NTS neurons into putative mono- and polysynaptic groups by absolute latency alone (e.g., Andresen and Yang 1990; Kasparov and Paton 1999).

Our electrophysiological analyses strongly depended on our measure of latency. Low resistance recording electrodes helped reduce electrical recording noise, and voltage clamp reduced membrane potential influences. To improve the latency and jitter measurement, the derivative of individual data traces was used to improve the resolution of the rising edge in indicating the initiation point of the synaptic current. By relying on the leading edge of the synaptic current, our approach minimized potential influences of variations in amplitude or kinetic differences across individual synaptic traces that might add the variation if alternative aspects of the waveform such as 10% peak amplitude point were used.

The latency of monosynaptic PSCs includes synaptic delay and conduction time. Presynaptic processes involved in transmitter release and diffusion are estimated to be as short as 90 µs at other fast mammalian synapses (Sabatini and Regehr 1996, 1999). In NTS, the latency between extracellularly recorded terminal action potentials and focal synaptic potentials was about 600 µs (Berger and Averill 1983; Champagnat et al. 1985). These values for synaptic delays are quite short compared with our measured synaptic latencies. Contributing factors that might differ along pathways to medial NTS neurons may include myelination of varying degrees together with the loss of myelin that occurs as afferent fibers leave the ST (Kalia and Richter 1988). Indirect paths with multiple branch points could greatly slow the conduction time of the evoked action potential, but these are poorly described for NTS. Dividing the latency by the distance between the stimulation and recording electrodes (roughly averaging 2 mm), a crude index of conduction was <2 m/s, a conduction rate that would correspond to an unmyelinated axon in the peripheral nervous system. A true central conduction velocity along the presynaptic axon would be greater if these were corrected by the conduction delay time. Despite the short portion of the afferent paths preserved in our brain slice preparation, the range of absolute latencies was severalfold even among anatomically identified second-order NTS neurons. Together, the data suggest that a major portion of the synaptic latency may reside in substantial differences in the afferent pathway character and length for monosynaptic responses.

The location and density of synaptic innervation varies across regions of the CNS and importantly impacts synaptic integration (Thomson 2000). The presence of synaptic boutons from ADN on the soma and proximal dendrites of the second-order NTS neuron (Chan et al. 2000; Mendelowitz et al. 1992) are likely responsible for the very powerful sensory synaptic responses. Additional DiA staining might be located more distally but was undetectable with our approach. Nonetheless, the striking prominence of afferent baroreceptor boutons at the soma suggests that these synaptic inputs are likely to dominate synaptic integration near the axon hillock at the soma. Increases in ST stimulus current intensity evoked a nongraded, all-or-nothing synaptic current in our horizontal slice preparation (Andresen and Yang 1995). The tip face of the concentric stimulating electrode is quite large (50-µm-diam inner core) compared with the dimensions of either single afferent axons, and, since it visibly covers a major portion of the ST, the tip should be in close proximity to multiple ST axons at the location of contact. Axon diversity in conduction velocity and myelination is pronounced in the vagus and ADN source nerve trunks (Andresen and Kunze 1994). We interpret this inability to grade either the amplitude or the number of synaptic events as evidence that, in general, single ST axons contact the recorded NTS neurons. Thus such monosynaptic responses might be considered "unitary" (Titz and Keller 1997). This electrophysiological finding together with the distribution of fluorescently labeled ADN terminals into visible, multiple clusters is consistent with multiple innervation by single axons onto single neurons. This might explain the complexity of some apparently monosynaptic EPSC waveforms. Some EPSCs appear to have multiple inflections or, in some cases, fully separated synaptic events that we and others have observed (Miles 1986) and yet that we cannot independently recruit with stimulus intensity changes. In addition, if this conclusion is correct, it suggests that individual NTS neurons may directly receive only relatively discrete sensory inputs, i.e., afferent convergence at second-order medial NTS neurons may be rare.

This axosomatic input pattern is common at other sensory regions within the brain stem and contrasts with the spatial distribution and low unitary synaptic strength typical of many cortical regions (Conti and Weinberg 1999). Such brain stem synapses are at the initial stage of sensory processing and appear to be specialized for high efficacy with single action potentials releasing multiple vesicles (Conti and Weinberg 1999). This contrasts to the axospinous synapse pattern commonly found in cortical structures where the probability of release may be more restricted (Conti and Weinberg 1999; Thomson 2000). The brain stem sensory synaptic pattern commonly exhibits substantial frequency-dependent depression of the degree that we observed in our NTS neurons, and such release characteristics would be consistent with the surprisingly low failure rates within NTS (Thomson 2000). The result for throughput within NTS is a large synaptic safety factor especially for the initial EPSC (Andresen and Yang 1995).

Several factors, however, affect the observed amplitudes of synaptic responses and determine whether responses are propagated further within NTS. We observed both monosynaptic and polysynaptic currents in response to ST activation in single or separate neurons. In our studies, some polysynaptic responses faithfully tracked input stimulation throughout even high-frequency (200 Hz) trains. Some neurons received monosynaptic EPSCs followed by IPSCs with a longer latency (4-6 ms) suggestive of a disynaptic path for the IPSC. In such train responses, the late polysynaptic IPSC arrived in time to interrupt the subsequent EPSCs in these trains. Stimulating in the ST in horizontal slices, we never observed IPSCs that could not be blocked by non-NMDA glutamate receptor antagonists. This observation suggests that ST stimuli do not activate descending inhibitory fibers to NTS such as those from parabrachial nucleus (Felder and Mifflin 1988; Mifflin and Felder 1990). Combination EPSC/IPSC responses to sensory axon activation are commonly reported in NTS both in vivo as well as in vitro (Andresen and Mendelowitz 1996; Aylwin et al. 1997; Brooks et al. 1992; Champagnat et al. 1985; Felder 1986; Feldman and Felder 1991). Depending on the sensory input timing, an EPSC may be completely shunted by a following IPSC. These common, rapidly recurrent inhibitory inputs may thus confound detection of monosynaptic excitatory events especially in extracellular recordings since such local circuitry would powerfully mask afferent input responses.

The functional ramifications of frequency-dependent depression in NTS are substantial. Relatively modest stimulus frequencies substantially depress ST-evoked synaptic responses (Andresen and Yang 1995; Champagnat et al. 1986). In some cases, depression to high frequencies (200 Hz) in NTS was great enough to induce failures within our short trains of stimuli by the third or fourth stimulus despite the fact that such neurons met anatomical and/or very low jitter criteria for a monosynaptic pathway. Frequencies that maximally release neurotransmitter stores could tend toward failure by depletion or strong drive to recurrent inhibitory neurons. This may explain the prominent failure rates in several of our dye-identified second-order neurons. Such failures challenge the operating definition of monosynaptic events as distinguished by their ability to follow paired shocks separated by 5 ms (Aylwin et al. 1997; Deuchars et al. 2000; Gil et al. 1999; Miles 1986; Yoshimura and Jessell 1989; Zhang and Mifflin 2000). Our results suggest a potential pitfall of this paired-pulse protocol, particularly in NTS. The likelihood is substantial that monosynaptic sensory NTS responses may be mistakenly classified as polysynaptic due to apparent failures that actually arise from either monosynaptic depression or disynaptic recurrent inhibitory interactions.

The chain of events required for synaptic transmission via a monosynaptic pathway must be repeated for polysynaptic circuits following generation of an action potential in the postsynaptic neurons. Any intrinsic variation in the timing of these mechanisms should increase the temporal variability (jitter) of higher order synaptic latencies. The triggering of the action potential by postsynaptic current may be a major source of variation at the 50-µs time scale and reflect variation in the action potential threshold. Action potential threshold and synaptic amplitude are actively modulated by a variety of mechanisms including neurotransmitter actions. These add sources of variability to polysynaptic pathways and should substantially increase their response jitter. Conduction time in a single neuron is likely to be relatively invariant and thus should contribute little to variation in synaptic latency assessed within neurons by jitter. Also, unlike failures, jitter of PSCs evoked by infrequent stimuli should be less affected by inhibitory circuits. As predicted, jitter was greater in pharmacologically verified polysynaptic responses, and low jitter EPSCs were pharmacologically distinct from high jitter EPSCs or IPSCs. Most importantly, the jitter of ST-evoked EPSCs in identified second-order neurons was very low despite a wide range of latencies and failure rates. As such, jitter appears most sensitive to the processes thought to contribute to intrinsic differences in synaptic performance with increasing pathway complexity and thus is the best indicative measure of synaptic order.

Functional perspective

Emerging evidence from brain stem sensory pathways suggests that these circuits may follow somewhat different principles of integration that determine synaptic performance compared with cortical models (Conti and Weinberg 1999; Thomson 2000). For example, in auditory portions of the brain stem, sensory synapses are large and concentration on the soma and transmitter release probability at the calyx of Held is quite high and can approach 0.9 (Von Gersdorff et al. 1997). This contrasts to typical cortical models of integration in which synapses tend to be dendritic with relatively small individual contributions and lower release probability; e.g., at hippocampal spiny interneurons release probability drops to as low as 0.01 (Thomson et al. 1995). Depression and transmitter depletion are most often dominant performance characteristics of high release synapses (Thomson 2000). The NTS sensory synapse appears to follow the high release, somatic delivery model (Andresen and Yang 1995; Schild et al. 1995). The axosomatic location of baroreceptor sensory synapses (Mendelowitz et al. 1992) that allowed us to identify second-order neurons in NTS must impact the integrative properties of the sensory-NTS pathway differently than the dendritically weighted patterns found in cortical structures. This axosomatic model for NTS may represent a strategy for spatio-temporal integration that incorporates multiple specializations contributing to the high synaptic efficacy found at other brain stem sensory synapses (Trussell 1999). Functionally, this high probability of transmitter release may confer the high degree of rate-dependent depression observed at these NTS sensory synapses. Functionally for baroreceptor afferents, this rate-sensitive property could strongly contribute to the bidirectional rate sensitivity found with reflex responses to baroreceptor activation (Franz 1969; Jung and Katona 1990). Baroreceptors discharge substantially even under resting blood pressure conditions and in the discharge frequency range that produces substantial synaptic depression (Andresen and Kunze 1994). When blood pressure drops, the accompanying decrease in baroreceptor afferent discharge and thus in synaptic activation should release an ongoing synaptic depression, and this process may contribute to the well-known disproportionate reflex responses for increases compared with decreases in blood pressure. Such performance characteristics may make modulation of depression by presynaptic mechanisms a particularly powerful site for overall autonomic reflex modulation by accessory, nonglutamatergic transmitters so common in NTS (Andresen and Kunze 1994).