INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Synchronous and rhythmic behavior of complex spikes in Purkinje cells of the cerebellar cortex has been well documented in both anesthetized and awake animals (Bell and Kawasaki 1972; Lang 2001; Lang et al. 1996, 1999; Llinas and Sasaki 1989; Sasaki et al. 1989; Sugihara et al. 1993; Wylie et al. 1995; Yamamoto et al. 2001). This synchronous and rhythmic behavior must originate in the inferior olivary (IO) nucleus, the only source of the cerebellar climbing fibers that drive complex spikes. Moreover, since the average interval between complex spikes (about 1 s) is greater than the least interval of the rhythm revealed by auto- or cross-correlation analysis (about 100 ms), one must assume that there is a subthreshold, synchronized rhythmic mechanism that operates at the IO level and triggers a complex spike only on every 10th cycle, on average. Indeed, subthreshold membrane potential oscillations (STO) in the IO nucleus have been thoroughly described (Bal and McCormick 1997; Benardo and Foster 1986; Bleasel and Pettigrew 1992; Lampl and Yarom 1997; Llinas and Yarom 1981a,b, 1986), and a number of possible models have been suggested to account for this phenomenon (Loewenstein et al. 2001; Makarenko and Llinas 1998; Manor et al. 1997; Schweighofer et al. 1999; Yarom 1991).

Both the synchrony and the rhythmicity of complex spikes are modulated either pharmacologically, by applying drugs directly to the IO nucleus (Lang 2001; Lang et al. 1996), or during motor behavior (Smith 1998; Welsh et al. 1995). Blocking GABAergic neurotransmission by intraolivary injection of picrotoxin increases the average complex spike firing rate in Purkinje cells and decreases the frequency of the rhythm as revealed by auto- or cross-correlation analysis of complex spikes (Lang et al. 1996). In contrast, blocking glutamatergic inputs by intraolivary injection of 6-cyano-7-nitroquinoxaline-2,3-dione [CNQX; an -amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) blocker] decreases the average complex spike firing rate and increases the frequency revealed by auto- or cross-correlation analysis (Lang 2001). Both treatments enhance the spatial extent of synchrony of complex spikes in Purkinje cells. The fact that GABAergic and glutamatergic blocks manipulate the firing rate of complex spikes and the frequency of their rhythmic activity suggests that STO of IO neurons are controlled by synaptic inputs to the nucleus (De Zeeuw et al. 1998). Dynamic nature of olivary subthreshold activity is further supported by the recent study of Yamamoto et al. (2001), demonstrating that Purkinje cells in the medial portion of the cerebellar cortex fire before those located more laterally. As a result, complex spike activity "propagated" in the mediolateral direction.

In the present study we examined the stability of STO using whole cell patch recordings from pairs of electrotonically coupled IO neurons, and optical imaging in brain slice preparations. We demonstrate that olivary STO exhibit unstable temporal patterns. This instability was significantly reduced following a block of synaptic transmission. Both pair recordings and optical imaging demonstrated phase-shifted oscillatory activity along the olivary nucleus.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Slice preparation

Parasagittal slices (300 µm) were prepared from the brain stem of Sabra strain rats 13-30 days of age. Animals were anesthetized by intraperitoneal injection of pentobarbital sodium (60 mg/kg) and perfused through the heart with 100 ml of cold (01°C) physiological solution (solution A, Table 1). Following decapitation, the brain stem was quickly removed and sliced (Campden Instruments LTD 752 M vibroslice) in cold sucrose solution (solution C, Table 1).

The slices were incubated in the sucrose solution at room temperature for 60 min. During this time the sucrose solution was slowly replaced by solution A. Sections were kept at room temperature in solution A until they were transferred into the recording chamber. This treatment was found to be critical for increasing the viability of IO slices manifested as higher occurrence of STO.

Recordings

The recording chamber, mounted on an upright microscope stage (Zeiss Axioskop), was continuously perfused with solution B (Table 1). A constant temperature of 35°C was maintained by a temperature feedback controlled stabilization unit. In some experiments CNQX (Research Biochemicals International) was added to solution B to a final concentration of 40 µM. Whole cell patch recordings were performed under visual control with infrared differential interference contrast (DIC) optics. Recordings were made throughout the IO nucleus from neurons whose cell bodies were located below the surface of the slice. The pipettes, pulled on a Narishige pp-83 puller, were filled with the intracellular solution (Table 1) and had a DC resistance of 10-15 M. A seal resistance of at least 1 G was created before the membrane was ruptured. Cell capacitance and series resistance were not compensated. Recordings were made from single cells or from cell pairs using Axoclamp 2B amplifiers (Axon Instruments) in current-clamp mode. To avoid cross talk two amplifiers were used in cell pair recordings. The separation distance between the cells in a pair was measured as the distance between soma centers.

Electrical signals were stored on VHS videotape (Neurocorder DR-484) for off-line analysis.

Analysis

All data acquisition and analysis was performed in the LabVIEW (National Instruments) environment. Time-frequency (TF) analysis software was kindly provided by Dr. M. Palva (Palva et al. 2000). TF analysis was performed on traces normalized by subtracting the mean and dividing by the SD. The amplitude of cross-correlograms is expressed in units of the correlation coefficient (rho).

Electrotonic coupling between two simultaneously recorded neurons was measured by injecting hyperpolarizing (negative) current pulses (150-250 ms) of various intensities into one of the recorded cells, and measuring voltage responses in both cells at the end of the pulse. A pair was defined as coupled if a voltage deflection of more than 0.02 mV could be observed in the postjunctional cell after averaging 15 responses to negative 100-pA current pulse in the prejunctional cell. The coupling coefficient was measured as the ratio between voltage responses of the post- and the prejuctional cell to prolonged (150-250 ms), negative current pulses of various intensities.

Optical imaging

Slices were incubated for 10-15 min in a solution that contained the voltage-sensitive styryl dye RH-414 (Molecular Probes). The dye was first dissolved in distilled water and then diluted in physiological solution to a concentration of 0.3 mM. A detailed description of the optical measurement system is given in Cohen and Yarom (1999). Briefly, optical signals were recorded from 128 photodiodes in a 2-dimentional array placed at the focal plane of a Nikon microscope equipped with an epi-illumination attachment using ×40 water immersion objective (plan 40/0.75; Zeiss). Light was transmitted via a G-2A filter block (Nikon) with excitation filter of 510-560 nm, emission filter of 590 nm and a dichroic mirror of 580 nm. An area of 600 × 600 µm² of the brain slice was imaged. The photodiode array was centered over the IO nucleus, with at least 90% of the imaged area within the nucleus. The time course of signals, digitized at 1 kHz, was compared at different locations within the nucleus. Data were usually displayed as traces of absolute change in fluorescence as a function of time at each location. Concentric metal electrodes were used to stimulate the surface of the slice outside of the olivary nucleus (using 0.05-0.1 ms, 1- to 2-V stimulation pulse). Unstained preparations did not generate any signal, suggesting that the amplitude of "intrinsic signals" was insufficient to be detected in our system.

Each recording epoch, determined by the opening of an electronic shutter in the path of the incident illuminator, lasted from 1 to 1.5 s. The amplitude of oscillatory activity observed decreased in consecutive epochs indicating phototoxicity. This phototoxicity limited the number of recording epochs to <10 per location. Since we did not compare the amplitude of signals recorded, we present the absolute (F) and not the relative (F/F) change in fluorescence. The relative change in fluorescence was about 0.01.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We performed double whole cell patch recordings from 126 pairs of neurons. The separation distance between the neurons in a pair varied between 15 and 110 µm. An additional 262 neurons were recorded individually (not as members of a pair) and were used to supplement the analysis of occurrence and pattern of the subthreshold oscillations. The occurrence of STO was largely contingent on the K⁺ concentration in the physiological solution (see METHODS). Among the 514 neurons recorded in total, 178 were recorded using 6.2 mM K⁺ (solution A, Table 1); 5.6% of these neurons showed STO (10 cells). The remaining 336 neurons were recorded using 3.0 mM K⁺ (solution B, Table 1), and 54.5% of these cells had STO (183 cells). Neurons recorded from the same slice showed similar patterns of STO, while neurons recorded from different slices, even if taken from the same animal, sometimes differed dramatically in their oscillatory pattern.

Optical imaging was performed using 40 slices. Among them spontaneous oscillatory activity was observed in seven slices. In three of 33 slices tested that had no spontaneous oscillatory activity, oscillations were induced by extracellular stimulation of the slice rostrodorsal to the olivary nucleus.

Prevalence and stability of STO

Olivary neurons exhibited a variety of subthreshold activity patterns: from continuous nearly sinusoidal oscillations (Fig. 1A, top trace), to intermittent activity consisting of epochs of oscillations separated by silent periods, lasting from seconds to minutes (Fig. 1A, bottom trace). The average frequency of STO, measured as the number of waves within a time window of 20 s, varied from 0.8 to 8.6 Hz (Fig. 1B). In agreement with Placantonakis et al. (2000), the frequency distribution revealed two peaks at about 3 and 6 Hz.

View larger version (29K): [in this window] [in a new window]   Fig. 1. Different patterns of the spontaneous subthreshold oscillations recorded from olivary neurons. A: 3 examples of subthreshold oscillations of the membrane potential (STO) recorded from different neurons in different slices. The variations range from continuous oscillations (top) to short bursts of activity (bottom). B: distribution of average STO frequency in 94 neurons. Note the partition into 2 groups of cells: low- and high-frequency groups separated at about 3.5 Hz. C-E: average frequency of STO, plotted as a function of STO maximal amplitude (C), and the age of the animal (E); the STO maximum amplitude as a function of animal age is plotted in D. Significant positive correlations were found in C and E.

During epochs of oscillations, modulation of amplitude was observed. Figure 1A (middle trace) shows an example of regular modulation of the STO amplitude, manifested as beating oscillations. The maximum amplitude observed in different cells varied from 2 to 25 mV (Fig. 1, C and D). In the sample of 94 neurons, there was a significant positive correlation between STO frequency and STO amplitude (Fig. 1C; P < 0.01), and between STO frequency and age of the animal (Fig. 1E; P = 0.01). There was no correlation between STO amplitude and age of the animal (Fig. 1D).

The stability of STO frequency was examined by analyzing long (20 s) periods of activity (Fig. 2, A-D). Using TF analysis (see METHODS), we examined the temporal pattern of frequency change. In about one-half of the neurons the frequency was stable for prolonged periods of time (Fig. 2A). In the rest of the neurons, spontaneous modulations in frequency were observed. These modulations took a variety of forms. Figure 2 shows examples of sudden (B) and regular (C) frequency modulations. Spontaneous frequency changes occurred in either continuous or intermittently oscillating cells. Thus frequency during active periods of intermittent oscillations also showed spontaneous shifts. Within a given neuron's repertoire, changes in frequency were associated with changes in STO amplitude. Higher frequency was accompanied by an increase in amplitude.

View larger version (63K): [in this window] [in a new window]   Fig. 2. The frequency and amplitude of the STO undergo significant alterations. A-D: intracellular recordings from 4 different neurons in 4 different slices (left) and their corresponding time-frequency (TF) analysis (right). An example of a neuron with constant frequency is shown in A, and 2 examples of neurons with changing frequencies are shown in B and C. Changes in frequency were associated with changes in the amplitude (compare the trace on the left and its TF representation on the right). The amplitude is coded in color so that bright yellow corresponds to higher amplitude (see the color scale on the right). D: two-states oscillatory behavior in an inferior olive (IO) neuron. A sudden shift from a high to a low frequency is evident. The underlined time segments a and b are shown on a faster time scale at the bottom of the figure. Fourier analysis of each one of the segments (on the right) demonstrates the shift from about 4 to about 2.5 Hz.

STO activity at either of two specific frequencies was observed in four neurons recorded from different slices (Fig. 2D). The oscillations in these neurons switched between two frequencies that differed by more than 1 Hz (up to 3 Hz). Interestingly, the two frequencies always fell within the upper and the lower of the two ranges of frequency in the overall distribution (Fig. 1B). Time periods marked by "a" and "b" in Fig. 2D are presented at a faster time scale below the trace. Fourier analyses on the right demonstrate almost two-fold change in frequency accompanied by an almost a two-fold change in amplitude.

TTX and CNQX stabilize the temporal pattern of STO

Spontaneous changes in the frequency of STO could result either from intrinsic dynamics within the olivary network, or from synaptic inputs originating outside of the olivary nucleus. The latter can be examined by either a global blockade of synaptic input by TTX, or by a blockade of specific types of synaptic receptors. Indeed bath application of 0.5 µM TTX or 40 µM CNQX, a competitive non-N-methyl-D-aspartate (non-NMDA) receptor antagonist, abolished spontaneous shifts in frequency of STO, yielding prolonged periods of stable oscillatory activity.

Figure 3 shows an example of the effect of TTX. Under control conditions, irregular frequency shifts were accompanied by changes in amplitude (A). Application of TTX abolished the spontaneous shifts (B), which were restored following wash out of the drug (C). Figure 4 shows two examples of the effect of CNQX application in two different neurons recorded from two different slice preparations. Similar effects of CNQX application were observed in another 10 neurons recorded from different slices. In three neurons that had an exceptionally stable pattern of STO, application of CNQX had no effect. Although both TTX and CNQX treatments prevented the spontaneous shifts in frequency, they failed to produce continuous oscillations in previously intermittently oscillating neurons. Therefore the intermittent pattern of oscillations may represent intrinsic olivary network dynamics.

View larger version (48K): [in this window] [in a new window]   Fig. 3. TTX stabilizes frequency and amplitude of STO. Intracellular recordings of STO from an olivary neuron (left) and their corresponding time-frequency (TF) analysis (right). A: STO recorded under control conditions exhibited significant frequency and amplitude modulation. B: bath application of 0.5 µm of TTX largely abolished spontaneous shifts. C: wash out of the drug restored the control behavior of STO. TF analysis reveals a complex pattern of frequency modulation before TTX application and after the wash out, and a regular frequency of about 5 Hz following TTX.

View larger version (48K): [in this window] [in a new window]   Fig. 4. 6-Cyano-7-nitroquinoxaline-2,3-dione (CNQX) stabilizes frequency and amplitude of STO. Intracellular recordings of STO from 2 olivary neurons (left) and their corresponding time-frequency (TF) analysis (right). A: STO recorded before (1) and 20 min after (2) bath application of 40 µm of CNQX. TF representations of the traces on the right show a complex pattern of frequency modulation before CNQX application, and a regular frequency of about 4 Hz following CNQX. B: 2nd example of the effect of CNQX, in a neuron recorded from a different slice, demonstrating that the intermittent nature of the STO was not affected by CNQX.

Coherence of STO in pairs of olivary neurons

In all cell pairs in which both of the cells showed oscillatory behavior, both displayed similar activity pattern. Synchronous oscillatory activity with no phase difference was observed in 22 of 61 pairs. In nearly all of these pairs (90%), the distance between the cells was <60 µm. Electrotonic coupling was tested in 13 of these 22 pairs. In 12 pairs coupling was detected, with the coupling coefficient ranging from 0.17 to 0.01 (see METHODS). An example of synchronized behavior is shown in Fig. 5A. This cell pair exhibited intermittent oscillations in synchrony in both cells. Cross-correlation analysis (Fig. 5Ab) revealed synchronous oscillatory activity. The distinct peak at zero time in the cross-correlogram indicates zero phase difference. Both cells oscillated at exactly the same frequency, about 7 Hz, as shown by Fourier analysis (Fig. 5Ac).

View larger version (40K): [in this window] [in a new window]   Fig. 5. The synchronicity of the olivary subthreshold oscillations. Simultaneous patch recordings from 4 pairs of olivary neurons are shown in column a. In each pair, one of the neurons is shown in blue and another one in red. Columns b and c represent the corresponding cross-correlation analysis and normalized power spectrum. A: an example of synchronous STO with zero phase difference (b). The neurons displayed synchronous modulations in amplitude (a), and identical power spectra (c). B: an example of synchronous STO where 20 ms of phase difference is evident in the cross-correlogram (b). The neurons differed in their secondary frequency components, but had the same main frequency (c). C: an example of a pair of olivary neurons that oscillated at different frequencies. D: an example of a pair of olivary neurons where only 1 cell oscillated. However, Fourier analysis of the apparently silent neuron revealed frequency components similar to its pair.

Thirteen pairs in which at least one neuron exhibited STO had noncoherent activity. The majority of them (62%) were located at distances of more than 60 µm. In 7 of these 13 pairs the STO were phase shifted by 20-50 ms. In three of these seven pairs, all oscillating with a phase shift of 20 ms, electrotonic coupling was detected. Figure 5B shows an example of a pair that oscillated with a phase shift of 50 ms as revealed by cross-correlation analysis (Fig. 5Bb). Both cells had their main frequency component at 4.2 Hz (Fig. 5Bc). The neuron shown in the top trace had additional components at lower frequencies. In two additional pairs the neurons oscillated at different frequencies. For example, in the pair, illustrated in Fig. 5C, the neurons oscillated at 4.6 and 5.4 Hz. No coupling was detected in these pairs. In the remaining four pairs, only one neuron oscillated, while the other one was apparently quiescent. Electrotonic coupling was detected in two of them. Fourier analysis of the apparently quiescent neuron in these two pairs revealed frequency components matching the active partner (Fig. 5Cc). As a result, some correlation was present (Fig. 5Cb).

All pairs that showed a phase shift in their oscillatory behavior exhibited an STO characterized by modulations in both the frequency and amplitude. One example of an electrotonically coupled pair is shown in Fig. 6. In this case, both neurons had intermittent STO, where during periods of oscillations prominent amplitude (Fig. 6A) and frequency (Fig. 6B) modulations were observed. The identical frequency spectrum, revealed by Fourier analysis (Fig. 6B), demonstrates that both cells oscillated in the same frequency range. Indeed, the TF representation (Fig. 6D) shows that the modulation of the frequency and amplitude occurred simultaneously in both cells. In this particular cell pair the range of frequency shifts (Fig. 6D) was unusually large, from 5 to 10 Hz. Such an accurate synchronous modulation in frequency must be a consequence of network activity.

View larger version (60K): [in this window] [in a new window]   Fig. 6. The variations in amplitude and frequency of the subthreshold oscillations occurred synchronously in a pair of simultaneously recorded neurons. A: simultaneous recordings from 2 intermittently oscillating neurons. Note synchronous modulation in amplitude. B: Fourier analysis shows almost identical spectra. C: a 20-ms phase difference was revealed by the cross-correlation analysis. D: the results of TF analysis of the traces shown in A. Note synchronous modulations in frequency in both cells.

Optical imaging of oscillatory IO activity

The synchronized STO observed during double patch recordings imply that coherent oscillations occur in a large population of olivary neurons. To examine the population behavior of olivary STO we used optical imaging in slice preparations (see METHODS). Spontaneous or evoked oscillatory activity was observed in 25% of the slices examined (10 of 40 slices). Oscillation frequency ranged from 2 to 8 Hz, similar to that observed in patch recordings. Moreover, also in agreement with patch recordings, this activity was not always observed at every recording session (see METHODS), but followed an intermittent pattern.

The oscillatory activity usually occupied a large portion (but not all) of the imaged area that was centered on the IO nucleus (see METHODS). An example of spontaneous oscillatory activity is shown in Fig. 7. Each trace represents the change in fluorescence, recorded by a single photodiode, as a function of time. Oscillatory activity at 8 Hz is clearly visible in the middle of the field (red). A rim of low-amplitude oscillations, which gradually merge with the background noise, surrounds this central core (yellow). Figure 7B represents analysis of signals averaged from areas designated in Fig. 7A. Fourier analysis shows that both areas oscillated at virtually identical frequency (Fig. 7B, middle). The distinct peak at zero time in the cross-correlogram (Fig. 7B, right) of these averaged signals indicates zero phase difference. Similar synchronous activity with no phase difference was observed in another three slices with spontaneous oscillations.

View larger version (44K): [in this window] [in a new window]   Fig. 7. Spontaneous subthreshold oscillations in the inferior olive nucleus revealed by optical imaging. A: spatiotemporal presentation of the subthreshold oscillations. Each trace represents the change in fluorescence recorded by a single photodiode in its relative location as a function of time. Fourier analysis was performed on all diodes. All diodes, where the amplitude of the main frequency component was 30% of that of the best diode, are marked in red. All diodes, where the amplitude of the main frequency component was between 10 and 30% of that of the best diode, are marked in yellow. Note the central patch of high-amplitude oscillations. B: red and blue traces on the left represent averaged signals from diodes inside red and blue frames in A, respectively. Their almost identical power spectra are shown in the middle. Cross-correlation reveals that the oscillatory activity occurred synchronously with zero phase difference.

The spatial distribution of the optical signal is shown as a function of time in Fig. 8. Each frame represents the activity throughout the imaged area at the designated time (top left corner). Two cycles of oscillatory activity are shown. The upswing of the first wave started at about 148 ms, reached its peak at 172 ms, and declined to minimum at 220 ms. Note that in accordance with Fig. 7 the STO are confined to a certain area and occurred simultaneously in the entire active area.

View larger version (170K): [in this window] [in a new window]   Fig. 8. Spatial distribution of the oscillatory activity shown in Fig. 7, presented in color code over the whole array (600 × 600 µm2). The largest peak-to-peak amplitude of the signal was fitted with rainbow colors from dark purple to red, where red is the depolarizing peak of the oscillations. The time, denoted in the top left corner of each frame, is measured in ms starting from the beginning of the recording shown in Fig. 7. Only 2 cycles of oscillations are presented. In this example, the entire active area oscillates simultaneously.

Synchronous oscillations were also evoked by stimulating the slice rostrodorsal to the olivary nucleus in three olivary slices that did not have spontaneous oscillatory activity. The stimulus evoked a brief biphasic response (first peak in Fig. 9B) that propagated across the imaged area at a velocity of about 7 cm/s (Fig. 10A). This brief response, which represents postsynaptic responses of olivary neurons, was followed by a prolonged period of oscillations characterized by gradual decreasing amplitude. As with the spontaneous oscillations, the spatial distribution (Fig. 10B) revealed a patchy organization of synchronized activity. The temporal analysis showed 7-Hz oscillations (Fig. 9B, middle panel) with zero phase difference between signals recorded from the red and blue boxes marked on Fig. 9A (Fig. 9B, right panel).

View larger version (44K): [in this window] [in a new window]   Fig. 9. Stimulus evoked subthreshold oscillations in the inferior olive nucleus revealed by optical imaging. The conventions are the same as in Fig. 7. The stimulating electrode was positioned at the bottom left corner, outside of the imaged area. Note in B the fast initial response that preceded the oscillatory activity.

View larger version (127K): [in this window] [in a new window]   Fig. 10. Spatial distribution of the oscillatory activity shown in Fig. 9. The conventions are the same as in Fig. 8. A: propagation of postsynaptic activity triggered by the stimulus. The activation started at the bottom left corner, closest to the location of the stimulating electrode. Time resolution from frame to frame is 2 ms. The stimulus was given 10 ms after starting data acquisition. B: frame-by-frame illustration of the oscillatory activity that followed the stimulus. Time resolution was 12 ms between the presented frames. Note that as opposed to the postsynaptic response in A, the oscillatory activity occurred simultaneously at the center of the imaged area.

In three slices with spontaneous oscillations, cross-correlation analysis between diodes at different locations revealed synchronous but phase-shifted oscillatory activity. An example is shown in Fig. 11. Like the previous examples, the oscillations occurred around a central core (Fig. 11A). However, close examination of signals averaged from diodes marked by red and blue boxes on Fig. 11A revealed a 50-ms phase shift of the oscillatory activity (Fig. 11B, right), although the oscillations occurred at the same frequency (Fig. 11B, middle). Analyzing the optical signal frame-by-frame revealed that within each oscillatory cycle the depolarization started at the top of the imaged area (Fig. 12, frame 396) and gradually propagated to the bottom (Fig. 12, frame 516). The trough of the wave showed a similar pattern of propagation. In this example the oscillations propagated across the slice with a velocity of about 0.7 cm/s.

View larger version (44K): [in this window] [in a new window]   Fig. 11. Optical imaging of noncoherent oscillatory activity. The conventions are the same as in Figs. 7 and 9. Note that although the 2 designated areas have the same frequency (B, middle), the cross-correlation analysis shows a phase difference of 50 ms.

View larger version (177K): [in this window] [in a new window]   Fig. 12. Spatial distribution of the oscillatory activity shown in Fig. 11. The conventions are the same as in Figs. 8 and 10. Time resolution from frame to frame is 24 ms. Note that the activity started at the top and propagated to the bottom of the imaged field.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Two significant issues were presented. First, we used modern techniques to substantiate the hypothesis that subthreshold oscillatory activity in olivary neurons is generated by a complex interplay between neuronal properties and network organization. Second, we demonstrated that STO are dynamic in space and time as manifest in frequency and amplitude instability, and in the ability to propagate across the nucleus.

Network behavior of the subthreshold oscillations

The remarkable similarity of subthreshold activity recorded simultaneously from two olivary neurons (Fig. 5) cannot be explained unless one assumes that the entire network participates in its generation. If we assume a hypothetical pacemaker neuron that generates STO with peak-to-peak amplitude of 25 mV and a coupling coefficient of 0.05 (Devor and Yarom 2002), the average postjunctional responses will be less than 1.25 mV. Such a low amplitude cannot trigger oscillatory activity in the postjunctional cell. Thus the rather low coupling coefficient between olivary neurons argues against the possibility of a central pacemaker neuron that drives the entire system. Optical imaging further supports the network hypothesis. Oscillatory activity recorded optically had frequency and waveform parameters similar to the STO recorded intracellularly. The percentage of slices in which oscillations were detected optically (25%) was significantly lower than the percentage of slices in which oscillations were recorded using the whole cell patch method (about 50%). This discrepancy is not surprising when one takes into account that only in-phase oscillations are likely to be detected by the imaging system. Each diode in the array integrated voltage signals scattered across 50 × 50 µm² in x-y plane and, possibly, through the full thickness of the slice. Phase shifts in oscillatory activity among neurons in this volume are expected to decrease, or even cancel the signal, depending on the degree of synchrony. Consistent with this possibility, in some preparations with no spontaneous oscillations, oscillations were induced by extracellular stimulus that temporally synchronized activity in olivary neurons. We conclude that subthreshold oscillations are generated by network interactions, and are not driven by pacemaker elements.

Spatiotemporal modifiability of subthreshold oscillations

The temporal modifiability of STO is manifest as spontaneous shifts in oscillation frequency and amplitude. These spontaneous changes might represent an intrinsic feature of the olivary network, or be imposed on the network by extrinsic synaptic inputs. Consistent with the latter possibility are reports that a number of neurotransmitters are able to modulate the oscillatory activity of olivary neurons as reflected in cerebellar complex spikes (Lang 2001; Lang et al. 1996; Llinas and Sasaki 1989; Placantonakis et al. 2000). Among them, intraolivary injection of non-NMDA glutamate receptor antagonists in vivo were shown to increase the rhythmicity of complex spike activity (Lang 2001). Consistent with this study, we have shown that CNQX stabilizes STO by abolishing spontaneous frequency modulations. Assuming that STO is the source of the complex spike rhythmicity, such an effect is expected to increase the number of peaks in the complex spike autocorrelation as observed by Lang (2001). The possibility that chemical synapses play an important role in subthreshold oscillations is further supported by the finding that some neurons in our preparations displayed two-states behavior, shifting suddenly between two alternative frequencies (Fig. 2). In these cases, the two frequencies belonged to the two subpopulations of the population frequency distribution (Fig. 1). Interestingly, Placantonakis and Welsh (2001) in a recent study of olivary STO in slice preparations, described two types of oscillatory activity that occurred in similar frequency ranges. The higher frequency oscillations depended on the low-threshold calcium conductance. The lower frequency oscillations depended on NMDA receptor activation. It is possible, therefore that the sudden switches in frequency observed in our two-states neurons reflect activation of NMDA receptors.

Since no chemical synapses exist between olivary neurons, and no interneurons have been found in the nucleus, the modulatory glutamate neurons must be located outside of the olivary nucleus, perhaps in the dorsal column nuclei or the surrounding reticular formation (De Zeeuw et al. 1989, 1990). The remarkable synchrony in the frequency and amplitude modulation in pairs of olivary neurons (Fig. 5) strongly suggests that these modulations occur on the level of populations of coupled cells. In other words, sporadic synaptic inputs to a single neuron would not be powerful enough to cause the change. Therefore the glutamatergic effect observed in our experiments can be explained either by a synchronous synaptic input to a population of olivary neurons, or by nonsynaptic release of glutamate, for example, from glial cells. The stabilizing effect of TTX supports the former possibility and suggests that well-controlled neuronal inputs, limited to a subpopulation of olivary neurons, would create clusters of neurons oscillating at different frequencies.

The propagating waves of oscillatory activity observed by optical imaging constitute a remarkable manifestation of spatial modifiability of STO. Simultaneous recordings from pairs of olivary neurons occasionally exhibited a phase difference of up to 50 ms (Fig. 5). This phase difference might indicate either that subpopulations of cells are operating in different phases, or that the subthreshold activity propagates across the nucleus. The later possibility was confirmed by optical imaging. It is premature to speculate on the mechanism underlying the propagation of subthreshold waves. It is interesting to note, however, that electrotonic coupling between pairs of olivary neurons has been reported to exhibit a directional preference (Devor and Yarom 2002). Therefore in addition to its role in generating oscillations, asymmetric electrotonic coupling might contribute to the spatial propagation of subthreshold activity. Propagation of subthreshold waves along the IO nucleus may explain propagation of complex spikes in the cerebellar cortex as has been reported by Llinas and Sasaki (1989) and Yamamoto et al. (2001).

Size of synchronously oscillating populations

A preliminary estimate of the size of the oscillating network can be deduced from our optical recordings. The observed oscillations were always confined to only a part of the imaged field, in other words to a subpopulation of olivary cells and not the whole nucleus. Assuming that a photodiode detects the signal through the 300-µm thickness of the slice, and that the cell density in the olivary nucleus of 3 ± 2 cells in a 50 × 50 × 25 µm cube (Devor and Yarom 2002), the area of oscillations marked by red and yellow in Fig. 6A corresponds to about 2,000 neurons. The IO in the rat was estimated to contain about 30,000 neurons (Ito 1984). Therefore the area marked by red and yellow in Fig. 6A constitutes approximately 7% of the nucleus. Consequently, axons of these olivary neurons would cause coherent complex spikes in 7% of Purkinje cells, or about 20,000 cells. Coherent complex spike activity has been reported in parasaggital bands about 500 µm wide. Assuming that the distance between Purkinje cells in the rostrocaudal direction is 83 µm, and in the mediolateral direction is 55 µm (Palay and Chan-Palay 1974), the length of the corresponding parasaggital band in the cerebellar cortex would be about 160 mm. It is important to note, however, that the majority of the synaptic inputs present in vivo are missing in our slice preparation. Among these are the GABAergic inputs from the deep cerebellar nuclei that synapse in the immediate vicinity of olivary gap junctions. These inputs can potentially uncouple olivary neurons, fragmenting the network (Llinas 1974). On the other hand, only a part of the network is present in each preparation due to the slicing procedure.

The properties of olivary STO described in the present study can account for several aspects of complex spike behavior observed in in vivo experiments. Therefore our observations strongly support the hypothesis that rhythmicity, as well as synchronicity, of complex spikes are due to subthreshold activity in the olive. The dynamics of the subthreshold behavior should therefore be regarded as an indication that the frequency of the subthreshold oscillations carries information essential for cerebellar function.