Generation and Propagation of Subthreshold Waves in a ...

Generation and Propagation of Subthreshold Waves in a Network ofInferior Olivary Neurons

ANNA DEVOR AND YOSEF YAROMDepartment of Neurobiology, Institute of Life Sciences, Hebrew University, Jerusalem 91904, Israel

Received 30 August 2001; accepted in final form 7 February 2002

Devor, Anna and Yosef Yarom. Generation and propagation of sub-threshold waves in a network of inferior olivary neurons. J Neurophysiol87: 3059–3069, 2002; 10.1152/jn.00736.2001. The cells of the inferiorolivary (IO) nucleus generate a large repertoire of electrical signals,among them subthreshold oscillations of the membrane potential (STO).To date, subthreshold oscillations have been studied at the level ofsingle-cell recordings, from which network properties were inferred. Inthis study we used whole cell patch recordings and optical imaging toaddress the following issues: 1) synchrony of STO in neighboring neu-rons; 2) stability of the oscillatory activity in the temporal and spatialdomain; and 3) the size of the oscillating network. Recordings were madefrom 126 pairs of IO neurons in 13- to 30-day-old rats. An additional 262neurons were recorded individually. The frequency of STO varied from0.8 to 8.6 Hz. The frequency distribution revealed two subpopulationswith peaks at about 3 and 6 Hz. The maximum amplitude among the cellsvaried from 2 to 25 mV. Oscillations in most neurons showed ongoingmodulations in both frequency and amplitude. These modulations werelargely abolished following bath application of 40 �M 6-cyano-7-nitro-quinoxaline-2,3-dione (CNQX), a competitive non–N-methyl-D-aspartate(non-NMDA) receptor antagonist, suggesting that they were caused byglutamatergic action. In 35 of 61 recorded pairs at least one neuronexhibited STO permitting us to compare frequency and phase relations. In22 pairs there was coherent activity with zero phase difference betweenoscillations in the 2 cells. In these pairs, frequency and amplitude mod-ulation occurred simultaneously in both neurons. Electrotonic couplingwas tested in 13 pairs, that had coherent STO, and it was detected in 12.An additional seven pairs showed coherent oscillations but with a phasedifference of 20–50 ms. Electrotonic coupling was observed in three ofthese pairs. Electrotonic coupling was also observed in two of five pairsin which only one neuron oscillated. No coupling was detected in one pairwhere both neurons oscillated but at different frequencies. Optical imag-ing using a voltage-sensitive dye (RH 414) was performed on 40 IO slicesusing an array of 128 photodiodes. Patches of oscillatory activity wereobserved in 10 slices. Among them six showed spontaneous oscillations,and four exhibited oscillations following extracellular stimulation. Inagreement with cell pair recording, optical imaging demonstrated phase-shifted activity in the form of propagating waves of activity within anoscillating patch. We conclude that 1) STO exhibit ongoing modulationsof frequency and amplitude that are probably caused by extrinsic inputsto the IO nucleus; 2) electrotonically coupled neurons show a high levelof STO synchrony; and 3) the oscillatory activity can propagate within anetwork of coupled olivary neurons.

I N T R O D U C T I O N

Synchronous and rhythmic behavior of complex spikes inPurkinje cells of the cerebellar cortex has been well docu-

mented in both anesthetized and awake animals (Bell andKawasaki 1972; Lang 2001; Lang et al. 1996, 1999; Llinas andSasaki 1989; Sasaki et al. 1989; Sugihara et al. 1993; Wylie etal. 1995; Yamamoto et al. 2001). This synchronous and rhyth-mic behavior must originate in the inferior olivary (IO) nu-cleus, the only source of the cerebellar climbing fibers thatdrive complex spikes. Moreover, since the average intervalbetween complex spikes (about 1 s) is greater than the leastinterval of the rhythm revealed by auto- or cross-correlationanalysis (about 100 ms), one must assume that there is asubthreshold, synchronized rhythmic mechanism that operatesat the IO level and triggers a complex spike only on every 10thcycle, on average. Indeed, subthreshold membrane potentialoscillations (STO) in the IO nucleus have been thoroughlydescribed (Bal and McCormick 1997; Benardo and Foster1986; Bleasel and Pettigrew 1992; Lampl and Yarom 1997;Llinas and Yarom 1981a,b, 1986), and a number of possiblemodels have been suggested to account for this phenomenon(Loewenstein et al. 2001; Makarenko and Llinas 1998; Manoret al. 1997; Schweighofer et al. 1999; Yarom 1991).

Both the synchrony and the rhythmicity of complex spikesare modulated either pharmacologically, by applying drugsdirectly to the IO nucleus (Lang 2001; Lang et al. 1996), orduring motor behavior (Smith 1998; Welsh et al. 1995). Block-ing GABAergic neurotransmission by intraolivary injection ofpicrotoxin increases the average complex spike firing rate inPurkinje cells and decreases the frequency of the rhythm asrevealed by auto- or cross-correlation analysis of complexspikes (Lang et al. 1996). In contrast, blocking glutamatergicinputs by intraolivary injection of 6-cyano-7-nitroquinoxaline-2,3-dione [CNQX; an �-amino-3-hydroxy-5-methyl-4-isox-azolepropionic acid (AMPA) blocker] decreases the averagecomplex spike firing rate and increases the frequency revealedby auto- or cross-correlation analysis (Lang 2001). Both treat-ments enhance the spatial extent of synchrony of complexspikes in Purkinje cells. The fact that GABAergic and gluta-matergic blocks manipulate the firing rate of complex spikesand the frequency of their rhythmic activity suggests that STOof IO neurons are controlled by synaptic inputs to the nucleus(De Zeeuw et al. 1998). Dynamic nature of olivary subthresh-old activity is further supported by the recent study ofYamamoto et al. (2001), demonstrating that Purkinje cells inthe medial portion of the cerebellar cortex fire before those

Address for reprint requests: Y. Yarom, Dept. of Neurobiology, Institute ofLife Sciences, The Hebrew University, Jerusalem 91904, Israel (E-mail:[email protected]).

The costs of publication of this article were defrayed in part by the paymentof page charges. The article must therefore be hereby marked ‘‘advertisement’’in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

J Neurophysiol87: 3059–3069, 2002; 10.1152/jn.00736.2001.

30590022-3077/02 $5.00 Copyright © 2002 The American Physiological Societywww.jn.org

located more laterally. As a result, complex spike activity“propagated” in the mediolateral direction.

In the present study we examined the stability of STO usingwhole cell patch recordings from pairs of electrotonically cou-pled IO neurons, and optical imaging in brain slice prepara-tions. We demonstrate that olivary STO exhibit unstable tem-poral patterns. This instability was significantly reducedfollowing a block of synaptic transmission. Both pair record-ings and optical imaging demonstrated phase-shifted oscilla-tory activity along the olivary nucleus.

M E T H O D S

Slice preparation

Parasagittal slices (300 �m) were prepared from the brain stem ofSabra strain rats 13–30 days of age. Animals were anesthetized byintraperitoneal injection of pentobarbital sodium (60 mg/kg) andperfused through the heart with 100 ml of cold (0�1°C) physiologicalsolution (solution A, Table 1). Following decapitation, the brain stemwas quickly removed and sliced (Campden Instruments LTD 752 Mvibroslice) in cold sucrose solution (solution C, Table 1).

The slices were incubated in the sucrose solution at room temper-ature for 60 min. During this time the sucrose solution was slowly

replaced by solution A. Sections were kept at room temperature insolution A until they were transferred into the recording chamber.This treatment was found to be critical for increasing the viability ofIO 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 (Table1). A constant temperature of 35°C was maintained by a temperaturefeedback controlled stabilization unit. In some experiments CNQX(Research Biochemicals International) was added to solution B to afinal concentration of 40 �M. Whole cell patch recordings wereperformed under visual control with infrared differential interferencecontrast (DIC) optics. Recordings were made throughout the IOnucleus from neurons whose cell bodies were located below thesurface of the slice. The pipettes, pulled on a Narishige pp-83 puller,were filled with the intracellular solution (Table 1) and had a DCresistance of 10–15 M�. A seal resistance of at least 1 G� wascreated before the membrane was ruptured. Cell capacitance andseries resistance were not compensated. Recordings were made fromsingle cells or from cell pairs using Axoclamp 2B amplifiers (AxonInstruments) in current-clamp mode. To avoid cross talk two ampli-fiers were used in cell pair recordings. The separation distance be-tween the cells in a pair was measured as the distance between somacenters.

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

Analysis

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

Electrotonic coupling between two simultaneously recorded neuronswas measured by injecting hyperpolarizing (negative) current pulses(150–250 ms) of various intensities into one of the recorded cells, andmeasuring voltage responses in both cells at the end of the pulse. A pairwas defined as coupled if a voltage deflection of more than 0.02 mVcould be observed in the postjunctional cell after averaging 15 responsesto negative 100-pA current pulse in the prejunctional cell. The couplingcoefficient was measured as the ratio between voltage responses of the

TABLE 1. Solutions used

Solution A*(6.2 mM

K�)Solution B*(3 mM K�)

Solution C*(Sucrose)

IntracellularSolution†

NaCl 124 127.2 4KCl 5 1.8 5MgSO4 1.3 1.3 1.3KH2PO4 1.2 1.2 1.2NaHCO3 26 26 26Sucrose 124Glucose 10 10 10CaCl2 2.4 2.4 2.4 10�3

K-gluconate 140EGTA 10�2

Mg-ATP 4HEPES 10

Concentrations are given in mM. * Solutions were aerated with 95% O2-5%CO2, pH 7.4. † pH 7.2.

FIG. 1. Different patterns of the spontaneoussubthreshold oscillations recorded from olivaryneurons. A: 3 examples of subthreshold oscillationsof the membrane potential (STO) recorded fromdifferent neurons in different slices. The variationsrange from continuous oscillations (top) to shortbursts of activity (bottom). B: distribution of aver-age STO frequency in 94 neurons. Note the parti-tion into 2 groups of cells: low- and high-frequencygroups separated at about 3.5 Hz. C–E: averagefrequency of STO, plotted as a function of STOmaximal amplitude (C), and the age of the animal(E); the STO maximum amplitude as a function ofanimal age is plotted in D. Significant positivecorrelations were found in C and E.

3060 A. DEVOR AND Y. YAROM

J Neurophysiol • VOL 87 • JUNE 2002 • www.jn.org

post- and the prejuctional cell to prolonged (150–250 ms), negativecurrent pulses of various intensities.

Optical imaging

Slices were incubated for 10–15 min in a solution that containedthe voltage-sensitive styryl dye RH-414 (Molecular Probes). The dyewas first dissolved in distilled water and then diluted in physiologicalsolution to a concentration of 0.3 mM. A detailed description of theoptical measurement system is given in Cohen and Yarom (1999).Briefly, optical signals were recorded from 128 photodiodes in a2-dimentional array placed at the focal plane of a Nikon microscopeequipped with an epi-illumination attachment using �40 water im-mersion 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 of600 � 600 �m2 of the brain slice was imaged. The photodiode arraywas centered over the IO nucleus, with at least 90% of the imaged areawithin the nucleus. The time course of signals, digitized at 1 kHz, wascompared at different locations within the nucleus. Data were usuallydisplayed as traces of absolute change in fluorescence as a function oftime at each location. Concentric metal electrodes were used tostimulate the surface of the slice outside of the olivary nucleus (using0.05–0.1 ms, 1- to 2-V stimulation pulse). Unstained preparations didnot generate any signal, suggesting that the amplitude of “intrinsicsignals” was insufficient to be detected in our system.

Each recording epoch, determined by the opening of an electronicshutter in the path of the incident illuminator, lasted from 1 to 1.5 s.

FIG. 2. The frequency and amplitude of theSTO undergo significant alterations. A–D: in-tracellular recordings from 4 different neuronsin 4 different slices (left) and their correspond-ing time-frequency (TF) analysis (right). Anexample of a neuron with constant frequency isshown in A, and 2 examples of neurons withchanging frequencies are shown in B and C.Changes in frequency were associated withchanges in the amplitude (compare the trace onthe left and its TF representation on the right).The amplitude is coded in color so that brightyellow corresponds to higher amplitude (seethe color scale on the right). D: two-statesoscillatory behavior in an inferior olive (IO)neuron. A sudden shift from a high to a lowfrequency is evident. The underlined time seg-ments a and b are shown on a faster time scaleat the bottom of the figure. Fourier analysis ofeach one of the segments (on the right) dem-onstrates the shift from about 4 to about 2.5 Hz.

FIG. 3. TTX stabilizes fre-quency and amplitude of STO.Intracellular recordings of STOfrom an olivary neuron (left) andtheir corresponding time-fre-quency (TF) analysis (right). A:STO recorded under controlconditions exhibited significantfrequency and amplitude modu-lation. B: bath application of 0.5�m of TTX largely abolishedspontaneous shifts. C: wash outof the drug restored the controlbehavior of STO. TF analysisreveals a complex pattern of fre-quency modulation before TTXapplication and after the washout, and a regular frequency ofabout 5 Hz following TTX.

3061SUBTHRESHOLD WAVES IN THE INFERIOR OLIVE


The amplitude of oscillatory activity observed decreased in consecu-tive epochs indicating phototoxicity. This phototoxicity limited thenumber of recording epochs to �10 per location. Since we did notcompare the amplitude of signals recorded, we present the absolute(�F) and not the relative (�F/F) change in fluorescence. The relativechange in fluorescence was about 0.01.

R E S U L T S

We performed double whole cell patch recordings from 126pairs of neurons. The separation distance between the neuronsin a pair varied between 15 and 110 �m. An additional 262neurons were recorded individually (not as members of a pair)and were used to supplement the analysis of occurrence andpattern of the subthreshold oscillations. The occurrence of STOwas largely contingent on the K� concentration in the physi-ological solution (see METHODS). Among the 514 neurons re-corded in total, 178 were recorded using 6.2 mM K� (solution

A, Table 1); 5.6% of these neurons showed STO (10 cells). Theremaining 336 neurons were recorded using 3.0 mM K� (so-lution B, Table 1), and 54.5% of these cells had STO (183cells). Neurons recorded from the same slice showed similarpatterns of STO, while neurons recorded from different slices,even if taken from the same animal, sometimes differed dra-matically in their oscillatory pattern.

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

Prevalence and stability of STO

Olivary neurons exhibited a variety of subthreshold activitypatterns: from continuous nearly sinusoidal oscillations (Fig.

FIG. 4. 6-Cyano-7-nitroquinoxaline-2,3-dione (CNQX) stabilizes fre-quency and amplitude of STO. Intra-cellular recordings of STO from 2 ol-ivary neurons (left) and theircorresponding time-frequency (TF)analysis (right). A: STO recorded be-fore (1) and 20 min after (2) bath ap-plication of 40 �m of CNQX. TF rep-resentations of the traces on the rightshow a complex pattern of frequencymodulation before CNQX application,and a regular frequency of about 4 Hzfollowing CNQX. B: 2nd example ofthe effect of CNQX, in a neuron re-corded from a different slice, demon-strating that the intermittent nature ofthe STO was not affected by CNQX.

FIG. 5. The synchronicity of the olivary subthreshold oscilla-tions. Simultaneous patch recordings from 4 pairs of olivary neuronsare shown in column a. In each pair, one of the neurons is shown inblue and another one in red. Columns b and c represent the corre-sponding cross-correlation analysis and normalized power spec-trum. 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 synchronousSTO where 20 ms of phase difference is evident in the cross-correlogram (b). The neurons differed in their secondary frequencycomponents, but had the same main frequency (c). C: an example ofa 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 revealedfrequency components similar to its pair.



1A, top trace), to intermittent activity consisting of epochs ofoscillations separated by silent periods, lasting from seconds tominutes (Fig. 1A, bottom trace). The average frequency ofSTO, measured as the number of waves within a time windowof 20 s, varied from 0.8 to 8.6 Hz (Fig. 1B). In agreement withPlacantonakis et al. (2000), the frequency distribution revealedtwo peaks at about 3 and 6 Hz.

During epochs of oscillations, modulation of amplitude wasobserved. Figure 1A (middle trace) shows an example ofregular modulation of the STO amplitude, manifested as beat-ing oscillations. The maximum amplitude observed in differentcells varied from 2 to 25 mV (Fig. 1, C and D). In the sampleof 94 neurons, there was a significant positive correlationbetween 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 ampli-tude and age of the animal (Fig. 1D).

The stability of STO frequency was examined by analyzinglong (20 s) periods of activity (Fig. 2, A–D). Using TF analysis(see METHODS), we examined the temporal pattern of frequencychange. In about one-half of the neurons the frequency wasstable for prolonged periods of time (Fig. 2A). In the rest of theneurons, spontaneous modulations in frequency were observed.These modulations took a variety of forms. Figure 2 shows

examples of sudden (B) and regular (C) frequency modula-tions. Spontaneous frequency changes occurred in either con-tinuous or intermittently oscillating cells. Thus frequency dur-ing active periods of intermittent oscillations also showedspontaneous shifts. Within a given neuron’s repertoire, changesin frequency were associated with changes in STO amplitude.Higher frequency was accompanied by an increase in ampli-tude.

STO activity at either of two specific frequencies was ob-served in four neurons recorded from different slices (Fig. 2D).The oscillations in these neurons switched between two fre-quencies that differed by more than 1 Hz (up to 3 Hz). Inter-estingly, the two frequencies always fell within the upper andthe lower of the two ranges of frequency in the overall distri-bution (Fig. 1B). Time periods marked by “a” and “b” in Fig.2D are presented at a faster time scale below the trace. Fourieranalyses on the right demonstrate almost two-fold change infrequency accompanied by an almost a two-fold change inamplitude.

TTX and CNQX stabilize the temporal pattern of STO

Spontaneous changes in the frequency of STO could resulteither from intrinsic dynamics within the olivary network, or

FIG. 6. The variations in amplitude and frequency of the sub-threshold oscillations occurred synchronously in a pair of simul-taneously recorded neurons. A: simultaneous recordings from 2intermittently oscillating neurons. Note synchronous modulationin amplitude. B: Fourier analysis shows almost identical spectra.C: a 20-ms phase difference was revealed by the cross-correlationanalysis. D: the results of TF analysis of the traces shown in A.Note synchronous modulations in frequency in both cells.



from synaptic inputs originating outside of the olivary nucleus.The latter can be examined by either a global blockade ofsynaptic input by TTX, or by a blockade of specific types ofsynaptic receptors. Indeed bath application of 0.5 �M TTX or40 �M CNQX, a competitive non–N-methyl-D-aspartate (non-NMDA) receptor antagonist, abolished spontaneous shifts infrequency of STO, yielding prolonged periods of stable oscil-latory activity.

Figure 3 shows an example of the effect of TTX. Undercontrol conditions, irregular frequency shifts were accompa-nied by changes in amplitude (A). Application of TTX abol-ished the spontaneous shifts (B), which were restored follow-ing wash out of the drug (C). Figure 4 shows two examples ofthe effect of CNQX application in two different neurons re-corded from two different slice preparations. Similar effects ofCNQX application were observed in another 10 neurons re-corded from different slices. In three neurons that had anexceptionally stable pattern of STO, application of CNQX hadno effect. Although both TTX and CNQX treatments preventedthe spontaneous shifts in frequency, they failed to producecontinuous oscillations in previously intermittently oscillatingneurons. Therefore the intermittent pattern of oscillations mayrepresent intrinsic olivary network dynamics.

Coherence of STO in pairs of olivary neurons

In all cell pairs in which both of the cells showed oscillatorybehavior, both displayed similar activity pattern. Synchronousoscillatory activity with no phase difference was observed in22 of 61 pairs. In nearly all of these pairs (90%), the distancebetween the cells was �60 �m. Electrotonic coupling wastested in 13 of these 22 pairs. In 12 pairs coupling wasdetected, with the coupling coefficient ranging from 0.17 to0.01 (see METHODS). An example of synchronized behavior isshown in Fig. 5A. This cell pair exhibited intermittent oscilla-tions in synchrony in both cells. Cross-correlation analysis(Fig. 5Ab) revealed synchronous oscillatory activity. The dis-tinct peak at zero time in the cross-correlogram indicateszero phase difference. Both cells oscillated at exactly thesame frequency, about 7 Hz, as shown by Fourier analysis(Fig. 5Ac).

Thirteen pairs in which at least one neuron exhibited STOhad noncoherent activity. The majority of them (62%) werelocated at distances of more than 60 �m. In 7 of these 13 pairsthe STO were phase shifted by 20–50 ms. In three of theseseven pairs, all oscillating with a phase shift of 20 ms, elec-trotonic coupling was detected. Figure 5B shows an example ofa pair that oscillated with a phase shift of 50 ms as revealed bycross-correlation analysis (Fig. 5Bb). Both cells had their mainfrequency component at 4.2 Hz (Fig. 5Bc). The neuron shownin the top trace had additional components at lower frequen-cies. In two additional pairs the neurons oscillated at differentfrequencies. For example, in the pair, illustrated in Fig. 5C, theneurons oscillated at 4.6 and 5.4 Hz. No coupling was detectedin these pairs. In the remaining four pairs, only one neuronoscillated, while the other one was apparently quiescent. Elec-trotonic coupling was detected in two of them. Fourier analysisof the apparently quiescent neuron in these two pairs revealedfrequency 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 be-havior exhibited an STO characterized by modulations in boththe frequency and amplitude. One example of an electrotoni-cally coupled pair is shown in Fig. 6. In this case, both neuronshad intermittent STO, where during periods of oscillationsprominent amplitude (Fig. 6A) and frequency (Fig. 6B) mod-ulations were observed. The identical frequency spectrum,revealed by Fourier analysis (Fig. 6B), demonstrates that bothcells oscillated in the same frequency range. Indeed, the TFrepresentation (Fig. 6D) shows that the modulation of thefrequency 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 accuratesynchronous modulation in frequency must be a consequenceof network activity.

Optical imaging of oscillatory IO activity

The synchronized STO observed during double patch re-cordings imply that coherent oscillations occur in a large

FIG. 7. Spontaneous subthreshold oscillations in the inferior olive nucleusrevealed by optical imaging. A: spatiotemporal presentation of the subthresholdoscillations. Each trace represents the change in fluorescence recorded by asingle photodiode in its relative location as a function of time. Fourier analysiswas performed on all diodes. All diodes, where the amplitude of the mainfrequency component was �30% of that of the best diode, are marked in red.All diodes, where the amplitude of the main frequency component wasbetween 10 and 30% of that of the best diode, are marked in yellow. Note thecentral patch of high-amplitude oscillations. B: red and blue traces on the leftrepresent 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 synchronouslywith zero phase difference.



population of olivary neurons. To examine the populationbehavior of olivary STO we used optical imaging in slicepreparations (see METHODS). Spontaneous or evoked oscillatoryactivity was observed in 25% of the slices examined (10 of 40slices). Oscillation frequency ranged from 2 to 8 Hz, similar tothat observed in patch recordings. Moreover, also in agreementwith patch recordings, this activity was not always observed atevery recording session (see METHODS), but followed an inter-mittent pattern.

The oscillatory activity usually occupied a large portion (butnot all) of the imaged area that was centered on the IO nucleus(see METHODS). An example of spontaneous oscillatory activityis shown in Fig. 7. Each trace represents the change in fluo-rescence, recorded by a single photodiode, as a function oftime. Oscillatory activity at 8 Hz is clearly visible in the middleof the field (red). A rim of low-amplitude oscillations, whichgradually merge with the background noise, surrounds thiscentral core (yellow). Figure 7B represents analysis of signalsaveraged from areas designated in Fig. 7A. Fourier analysisshows 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 indicateszero phase difference. Similar synchronous activity with nophase difference was observed in another three slices withspontaneous oscillations.

The spatial distribution of the optical signal is shown as afunction of time in Fig. 8. Each frame represents the activitythroughout the imaged area at the designated time (top leftcorner). Two cycles of oscillatory activity are shown. Theupswing of the first wave started at about 148 ms, reached its

peak at 172 ms, and declined to minimum at 220 ms. Note thatin accordance with Fig. 7 the STO are confined to a certain areaand occurred simultaneously in the entire active area.

Synchronous oscillations were also evoked by stimulatingthe slice rostrodorsal to the olivary nucleus in three olivaryslices that did not have spontaneous oscillatory activity. Thestimulus evoked a brief biphasic response (first peak in Fig.9B) that propagated across the imaged area at a velocity ofabout 7 cm/s (Fig. 10A). This brief response, which representspostsynaptic responses of olivary neurons, was followed by aprolonged period of oscillations characterized by gradual de-creasing amplitude. As with the spontaneous oscillations, thespatial distribution (Fig. 10B) revealed a patchy organization ofsynchronized activity. The temporal analysis showed 7-Hzoscillations (Fig. 9B, middle panel) with zero phase differencebetween signals recorded from the red and blue boxes markedon Fig. 9A (Fig. 9B, right panel).

In three slices with spontaneous oscillations, cross-correla-tion analysis between diodes at different locations revealedsynchronous but phase-shifted oscillatory activity. An exampleis shown in Fig. 11. Like the previous examples, the oscilla-tions occurred around a central core (Fig. 11A). However, closeexamination of signals averaged from diodes marked by redand blue boxes on Fig. 11A revealed a 50-ms phase shift of theoscillatory activity (Fig. 11B, right), although the oscillationsoccurred at the same frequency (Fig. 11B, middle). Analyzingthe optical signal frame-by-frame revealed that within eachoscillatory cycle the depolarization started at the top of theimaged area (Fig. 12, frame 396) and gradually propagated tothe bottom (Fig. 12, frame 516). The trough of the wave

FIG. 8. Spatial distribution of the oscillatory activityshown in Fig. 7, presented in color code over the wholearray (600 � 600 �m2). The largest peak-to-peak amplitudeof the signal was fitted with rainbow colors from dark purpleto red, where red is the depolarizing peak of the oscillations.The time, denoted in the top left corner of each frame, ismeasured in ms starting from the beginning of the recordingshown in Fig. 7. Only 2 cycles of oscillations are presented.In this example, the entire active area oscillatessimultaneously.



showed a similar pattern of propagation. In this example theoscillations propagated across the slice with a velocity of about0.7 cm/s.

D I S C U S S I O N

Two significant issues were presented. First, we used mod-ern techniques to substantiate the hypothesis that subthresholdoscillatory activity in olivary neurons is generated by a com-plex interplay between neuronal properties and network orga-nization. Second, we demonstrated that STO are dynamic inspace and time as manifest in frequency and amplitude insta-bility, and in the ability to propagate across the nucleus.

Network behavior of the subthreshold oscillations

The remarkable similarity of subthreshold activity recordedsimultaneously from two olivary neurons (Fig. 5) cannot beexplained unless one assumes that the entire network partici-pates in its generation. If we assume a hypothetical pacemakerneuron that generates STO with peak-to-peak amplitude of 25mV and a coupling coefficient of 0.05 (Devor and Yarom2002), the average postjunctional responses will be less than1.25 mV. Such a low amplitude cannot trigger oscillatoryactivity in the postjunctional cell. Thus the rather low couplingcoefficient between olivary neurons argues against the possi-bility of a central pacemaker neuron that drives the entire

system. Optical imaging further supports the network hypoth-esis. Oscillatory activity recorded optically had frequency andwaveform parameters similar to the STO recorded intracellu-larly. The percentage of slices in which oscillations weredetected optically (25%) was significantly lower than the per-centage of slices in which oscillations were recorded using thewhole cell patch method (about 50%). This discrepancy is notsurprising when one takes into account that only in-phaseoscillations are likely to be detected by the imaging system.Each diode in the array integrated voltage signals scatteredacross 50 � 50 �m2 in x-y plane and, possibly, through the fullthickness of the slice. Phase shifts in oscillatory activity amongneurons in this volume are expected to decrease, or even cancelthe signal, depending on the degree of synchrony. Consistentwith this possibility, in some preparations with no spontaneousoscillations, oscillations were induced by extracellular stimulusthat temporally synchronized activity in olivary neurons. Weconclude that subthreshold oscillations are generated by net-work interactions, and are not driven by pacemaker elements.

Spatiotemporal modifiability of subthreshold oscillations

The temporal modifiability of STO is manifest as spontane-ous shifts in oscillation frequency and amplitude. These spon-taneous changes might represent an intrinsic feature of theolivary network, or be imposed on the network by extrinsicsynaptic inputs. Consistent with the latter possibility are re-ports that a number of neurotransmitters are able to modulatethe oscillatory activity of olivary neurons as reflected in cere-bellar complex spikes (Lang 2001; Lang et al. 1996; Llinas andSasaki 1989; Placantonakis et al. 2000). Among them, intra-olivary injection of non-NMDA glutamate receptor antagonistsin vivo were shown to increase the rhythmicity of complexspike activity (Lang 2001). Consistent with this study, we haveshown that CNQX stabilizes STO by abolishing spontaneousfrequency modulations. Assuming that STO is the source of thecomplex spike rhythmicity, such an effect is expected to in-crease the number of peaks in the complex spike autocorrela-tion as observed by Lang (2001). The possibility that chemicalsynapses play an important role in subthreshold oscillations isfurther supported by the finding that some neurons in ourpreparations displayed two-states behavior, shifting suddenlybetween two alternative frequencies (Fig. 2). In these cases, thetwo frequencies belonged to the two subpopulations of thepopulation frequency distribution (Fig. 1). Interestingly, Pla-cantonakis and Welsh (2001) in a recent study of olivary STOin slice preparations, described two types of oscillatory activitythat occurred in similar frequency ranges. The higher fre-quency oscillations depended on the low-threshold calciumconductance. The lower frequency oscillations depended onNMDA receptor activation. It is possible, therefore that thesudden switches in frequency observed in our two-states neu-rons reflect activation of NMDA receptors.

Since no chemical synapses exist between olivary neurons,and no interneurons have been found in the nucleus, the mod-ulatory glutamate neurons must be located outside of the oli-vary nucleus, perhaps in the dorsal column nuclei or thesurrounding reticular formation (De Zeeuw et al. 1989, 1990).The remarkable synchrony in the frequency and amplitude

FIG. 9. Stimulus evoked subthreshold oscillations in the inferior olivenucleus revealed by optical imaging. The conventions are the same as in Fig.7. The stimulating electrode was positioned at the bottom left corner, outsideof the imaged area. Note in B the fast initial response that preceded theoscillatory activity.



modulation in pairs of olivary neurons (Fig. 5) strongly sug-gests that these modulations occur on the level of populationsof coupled cells. In other words, sporadic synaptic inputs to asingle neuron would not be powerful enough to cause thechange. Therefore the glutamatergic effect observed in ourexperiments can be explained either by a synchronous synapticinput to a population of olivary neurons, or by nonsynapticrelease of glutamate, for example, from glial cells. The stabi-lizing effect of TTX supports the former possibility and sug-gests that well-controlled neuronal inputs, limited to a sub-population of olivary neurons, would create clusters of neuronsoscillating at different frequencies.

The propagating waves of oscillatory activity observed byoptical imaging constitute a remarkable manifestation of spa-tial modifiability of STO. Simultaneous recordings from pairsof olivary neurons occasionally exhibited a phase difference ofup to 50 ms (Fig. 5). This phase difference might indicateeither that subpopulations of cells are operating in differentphases, or that the subthreshold activity propagates across thenucleus. The later possibility was confirmed by optical imag-ing. 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 olivaryneurons has been reported to exhibit a directional preference(Devor and Yarom 2002). Therefore in addition to its role ingenerating oscillations, asymmetric electrotonic couplingmight contribute to the spatial propagation of subthresholdactivity. Propagation of subthreshold waves along the IO nu-cleus may explain propagation of complex spikes in the cere-bellar 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 networkcan be deduced from our optical recordings. The observedoscillations were always confined to only a part of the imagedfield, in other words to a subpopulation of olivary cells and notthe whole nucleus. Assuming that a photodiode detects thesignal through the 300-�m thickness of the slice, and that thecell density in the olivary nucleus of 3 � 2 cells in a 50 � 50 �

FIG. 10. Spatial distribution of the oscillatory activityshown in Fig. 9. The conventions are the same as in Fig. 8.A: propagation of postsynaptic activity triggered by thestimulus. The activation started at the bottom left corner,closest to the location of the stimulating electrode. Timeresolution from frame to frame is 2 ms. The stimulus wasgiven 10 ms after starting data acquisition. B: frame-by-frame illustration of the oscillatory activity that followedthe stimulus. Time resolution was 12 ms between thepresented frames. Note that as opposed to the postsynapticresponse in A, the oscillatory activity occurred simulta-neously at the center of the imaged area.



25 �m cube (Devor and Yarom 2002), the area of oscillationsmarked by red and yellow in Fig. 6A corresponds to about2,000 neurons. The IO in the rat was estimated to contain about30,000 neurons (Ito 1984). Therefore the area marked by redand yellow in Fig. 6A constitutes approximately 7% of thenucleus. Consequently, axons of these olivary neurons wouldcause coherent complex spikes in 7% of Purkinje cells, orabout 20,000 cells. Coherent complex spike activity has beenreported in parasaggital bands about 500 �m wide. Assumingthat the distance between Purkinje cells in the rostrocaudaldirection is 83 �m, and in the mediolateral direction is 55 �m(Palay and Chan-Palay 1974), the length of the correspondingparasaggital band in the cerebellar cortex would be about 160mm. It is important to note, however, that the majority of thesynaptic inputs present in vivo are missing in our slice prepa-ration. Among these are the GABAergic inputs from the deepcerebellar nuclei that synapse in the immediate vicinity ofolivary gap junctions. These inputs can potentially uncoupleolivary neurons, fragmenting the network (Llinas 1974). Onthe other hand, only a part of the network is present in eachpreparation due to the slicing procedure.

The properties of olivary STO described in the present studycan account for several aspects of complex spike behaviorobserved in in vivo experiments. Therefore our observationsstrongly support the hypothesis that rhythmicity, as well assynchronicity, of complex spikes are due to subthreshold ac-tivity in the olive. The dynamics of the subthreshold behaviorshould therefore be regarded as an indication that the frequencyof the subthreshold oscillations carries information essentialfor cerebellar function.

FIG. 11. Optical imaging of noncoherent oscillatory activity. The conven-tions are the same as in Figs. 7 and 9. Note that although the 2 designated areashave the same frequency (B, middle), the cross-correlation analysis shows aphase difference of 50 ms.

FIG. 12. Spatial distribution of the oscillatory activityshown 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 thebottom of the imaged field.



This study was suppoted by the US-Israel Binational Science Foundationand the European Commission.

REFERENCES

BAL T AND MCCORMICK DA. Synchronized oscillations in the inferior olive arecontrolled by the hyperpolarization-activated cation current Ih. J Neuro-physiol 77: 3145–3156, 1997.

BELL CC AND KAWASAKI T. Relations among climbing fiber response of nearbyPurkinje cells. J Neurophysiol 35: 155–169, 1972.

BENARDO LS AND FOSTER RE. Oscillatory behavior in inferior olivary neurons:mechanism, modulation, cell aggregates. Brain Res Bull 17: 773–784, 1986.

BLEASEL AF AND PETTIGREW AG. Development and properties of spontaneousoscillations of the membrane potential in inferior olivary neurons in the rat.Dev Brain Res 65: 43–50, 1992.

COHEN D AND YAROM Y. Optical measurements of synchronized activity inisolated mammalian cerebellum. Neuroscience 94: 859–866, 1999.

DEVOR A AND YAROM Y. Electrotonic coupling in the inferior olivary nucleusrevealed by simultaneous double patch recordings. J Neurophysiol. 87:3048–3058, 2002.

DE ZEEUW CI, HOLSTEGE JC, RUIGROK TJ, AND VOOGD J. Ultrastructural studyof the GABAergic, cerebellar, and mesodiencephalic innervation of the catmedial accessory olive: anterograde tracing combined with immunocyto-chemistry. J Comp Neurol 284: 12–35, 1989.

DE ZEEUW CI, HOLSTEGE JC, RUIGROK TJ, AND VOOGD J. Mesodiencephalicand cerebellar terminals terminate upon the same dendritic spines in theglomeruli of the cat and rat inferior olive: an ultrastructural study using acombination of [3H]leucine and wheat germ agglutinin coupled horseradishperoxidase anterograde tracing. Neuroscience 34: 645–655, 1990.

DE ZEEUW CI, SIMPSON JI, HOOGENRAAD CC, KOEKKOEK SK, AND RUIGROK TJ.Microcircuitry and function of the inferior olive. Trends Neurosci 21:

391–400, 1998.ITO M. The Cerebellum and Neural Control. New York: Raven, 1984.LAMPL I AND YAROM Y. Subthreshold oscillations and resonant behavior: two

manifestations of the same mechanism. Neuroscience 78: 325–341, 1997.LANG E. Organization of olivocerebellar activity in the absence of excitatory

glutamatergic input. J Neurosci 21: 1663–1675, 2001.LANG EJ, SUGIHARA I, AND LLINAS R. GABAergic modulation of complex

spike activity by the cerebellar nucleoolivary pathway in rat. J Neurophysiol76: 255–275, 1996.

LANG EJ, SUGIHARA I, WELSH JP, AND LLINAS R. Patterns of spontaneouspurkinje cell complex spike activity in the awake rat. J Neurosci 19:2728–2739, 1999.

LLINAS R. Eighteenth Bowdwich lecture. Motor aspects of cerebellar control.Physiologist 17: 19–46, 1974.

LLINAS R AND SASAKI K. The functional organization of the olivocerebellarsystem as examined by multiple Purkinje cell recordings. Eur J Neurosci 1:587–602, 1989.

LLINAS R AND YAROM Y. Electrophysiology of mammalian inferior olivary

neurons in vitro. Different types of voltage-dependent ionic conductances.J Physiol (Lond) 315: 549–567, 1981a.

LLINAS R AND YAROM Y. Properties and distribution of ionic conductancesgenerating electroresponsiveness of mammalian inferior olivary neurons invitro. J Physiol (Lond) 315: 569–584, 1981b.

LLINAS R AND YAROM Y. Oscillatory properties of guinea pig inferior olivaryneurons and their pharmacological modulation: an in vitro study. J Physiol(Lond) 376: 163–182, 1986.

LOEWENSTEIN Y, YAROM Y, AND SOMPOLINSKY H. The generation of oscilla-tions in networks of electrically coupled cells. Proc Natl Acad Sci USA 98:8095–8100, 2001.

MAKARENKO V AND LLINAS R. Experimentally determined chaotic phasesynchronization in a neuronal system. Proc Natl Acad Sci USA 95: 15747–15752, 1998.

MANOR Y, RINZEL J, SEGEV I, AND YAROM Y. Low-amplitude oscillations inthe inferior olive: a model based on electrical coupling with heterogeneouschannel densities. J Neurophysiol 77: 2736–2752, 1997.

PALAY SL AND CHAN-PALAY V. Cerebellar Cortex: Cytology and Organiza-tion. New York: Springer-Verlag, 1974.

PALVA JM, LAMSA K, LAURI SE, RAUVALA H, KAILA K, AND TAIRA T. Fastnetwork oscillations in the newborn rat hippocampus in vitro. J Neurosci 30:1170–1178, 2000.

PLACANTONAKIS DG, SCHWARZ C, AND WELSH JP. Serotonin suppresses sub-threshold and suprathreshold oscillatory activity of rat inferior olivaryneurones in vitro. J Physiol (Lond) 524: 833–851, 2000.

PLACANTONAKIS DG AND WELSH JP. Two distinct oscillatory states determinedby the NMDA receptor in rat inferior olive. J Physiol (Lond) 534: 123–140,2001.

SASAKI K, BOWER JM, AND LLINAS R. Multiple Purkinje cell recording inrodent cerebellar cortex. Eur J Neurosci 1: 572–586, 1989.

SCHWEIGHOFER N, DOYA K, AND KAWATO M. Electrophysiological propertiesof inferior olive neurons: a compartmental model. J Neurophysiol 82:804–817, 1999.

SMITH SS. Step cycle-related oscillatory properties of inferior olivary neuronsrecorded in ensembles. Neuroscience 82: 69–80, 1998.

SUGIHARA I, LANG EJ, AND LLINAS R. Uniform olivocerebellar conduction timeunderlies Purkinje cell complex spike synchrony in the rat cerebellum.J Physiol (Lond) 470: 243–271, 1993.

WELSH JP, LANG EJ, SUGIHARA I, AND LLINAS R. Dynamic organization ofmotor control within the olivocerebellar system. Nature 374: 453–456,1995.

WYLIE DR, DE ZEEUW CI, AND SIMPSON JI. Temporal relations of the complexspike activity of Purkinje cell pairs in the vestibulocerebellum of rabbits.J Neurosci 15: 2875–2887, 1995.

YAMAMOTO T, FUKUDA M, AND LLINAS R. Bilaterally synchronous complexspike Purkinje cell activity in the mammalian cerebellum. Eur J Neurosci13: 327–339, 2001.

YAROM Y. Rhythmogenesis in a hybrid system—interconnecting an olivaryneuron to an analog network of coupled oscillators. Neuroscience 44:263–275, 1991.



Generation and Propagation of Subthreshold Waves in a ...

Documents

Transcript of Generation and Propagation of Subthreshold Waves in a ...