Interactions Between Multiple Rhythm Generators Produce Complex Patterns of Oscillation in the Developing Rat Spinal Cord

Rezan Demir, Bao-Xi Gao, Meyer B. Jackson, Lea Ziskind-Conhaim


Neural networks capable of generating coordinated rhythmic activity form at early stages of development in the spinal cord. In this study, voltage-imaging techniques were used to examine the spatiotemporal pattern of rhythmic activity in transverse slices of lumbar spinal cord from embryonic and neonatal rats. Real-time images were recorded in slices stained with the voltage-sensitive fluorescent dye RH414 using a 464-element photodiode array. Fluorescence signals showed spontaneous voltage oscillations with a frequency of 3 Hz. Simultaneous recordings of fluorescence and extracellular field potential demonstrated that the two signals oscillated with the same frequency and had a distinct phase relationship, indicating that the fluorescence changes represented changes in transmembrane potentials. The oscillations were reversibly blocked by cobalt (1 mM), indicating a dependence on Ca2+ influx through voltage-gated Ca2+ channels. Extracellular field potentials revealed oscillations with the same frequency in both stained and unstained slices. Oscillations were apparent throughout a slice, although their amplitudes varied in different regions. The largest amplitude oscillations were produced in the lateral regions. To examine the spatial organization of rhythm-generating networks, slices were cut into halves and quarters. Each fragment continued to oscillate with the same frequency as intact slices but with smaller amplitudes. This finding suggested that rhythm-generating networks were widely distributed and did not depend on long-range projections. In slices from neonatal rats, the oscillations exhibited a complex spatiotemporal pattern, with depolarizations alternating between mirror locations in the right and left sides of the cord. Furthermore, within each side depolarizations alternated between the lateral and medial regions. This medial-lateral pattern was preserved in hemisected slices, indicating that pathways intrinsic to each side coordinated this activity. A different pattern of oscillation was observed in slices from embryos with synchronous 3-Hz oscillations occurring in limited regions. Our study demonstrated that rhythm generators were distributed throughout transverse sections of the lumbar spinal cord and exhibited a complex spatiotemporal pattern of coordinated rhythmic activity.


Neural networks in the spinal cord are capable of generating rhythmic motor outputs independently of synaptic inputs from the brain and periphery (Grillner and Wallen 1985; Kiehn et al. 1997;O'Donovan 1999). These autonomous networks are referred to as central pattern generators (CPGs) (Gossard and Hultborn 1991; Grillner 1981) and are thought to regulate the rhythmic locomotor movements (Cohen et al. 1988;Rossignol 1996), which are evident during embryonic development of vertebrates (Hamburger and Balaban 1963;Narayanan et al. 1971; Windle and Griffin 1931). Rhythmic voltage oscillations with a pattern that resembles coordinated locomotion can be induced by synaptic receptor agonists in isolated vertebrate spinal cords, indicating that these in vitro models provide a useful tool for studying the organization of CPGs and the mechanisms by which they generate locomotor-like rhythmic activity.

Spontaneous bilateral voltage oscillations are apparent in isolated spinal cords of lamprey (Brodin and Grillner 1985),Xenopus (Van Mier et al. 1989), and chick embryo (Ho and O'Donovan 1993) but occur less frequently in spinal cords of neonatal rodents (Bonnot et al. 1998; Kremer and Lev-Tov 1998; Whelan et al. 2000). Therefore locomotor-like activity has been induced in the immature rat spinal cord using neurotransmitters such as 5-hydroxytryptamine (5-HT), and the glutamate receptor agonistN-methyl-d-aspartate (NMDA) (Beato et al. 1997; Cazalets et al. 1992; Kiehn and Kjaerulff 1996; Kudo and Yamada 1987;Schmidt et al. 1998; Smith and Feldman 1987). Ventral root recordings demonstrated that these chemically induced oscillations alternate between the two sides of the spinal cord in a pattern resembling the alternating movements of the left and right hindlimbs (Cowley and Schmidt 1994;Kremer and Lev-Tov 1997; MacLean et al. 1995; Raastad et al. 1997). This coordinated activity is controlled by networks that include inhibitory synapses between the rhythm-generating circuits in contralateral sides of the cord (Hochman and Schmidt 1998; Kjaerulff and Kiehn 1997; Kudo et al. 1998).

The location of the CPGs in the longitudinal and transverse axes of the rat spinal cord has been examined using various recording techniques. Most studies have demonstrated that rhythm-generating networks that control hindlimb movements are distributed throughout the lumbar and lower thoracic segments, with neurons in rostral regions exhibiting the highest rhythmogenic potential (Kiehn and Kjaerulff 1998). Intracellular recordings to identify the location of rhythmogenic neurons in the transverse axis of the cord have shown that neurons in the intermediate gray matter and close to the central canal are capable of generating rhythmic oscillations in response to chemical stimulation (Hochman et al. 1994; Kiehn and Kjaerulff 1996; MacLean et al. 1995). CPGs in the chick spinal cord may be organized differently, as Ca2+ imaging revealed that about 80% of the active cells are located in lateral regions of the cord, with a higher proportion in the ventral region (O'Donovan et al. 1994).

In this study, we examined the spatiotemporal pattern of spontaneous voltage activities in transverse slices of lumbar spinal cord from neonatal and embryonic rat using voltage-sensitive fluorescent dye (Grinvald et al. 1988; Wu and Cohen 1993). Real-time voltage imaging revealed spontaneous oscillations with a frequency of approximately 3 Hz in intact and sectioned slices. The pattern of rhythmic activity suggested that synaptic connections played a principle role in the coordination of multiple rhythm-generating circuits.

A preliminary report of this work has been presented in abstract form (Demir et al. 1999).


Spinal cord slices

Lumbar segments were isolated from Sprague-Dawley rat embryos at 15–16 days of gestation (E15–16, birth is at E21–22), and from 1- to 4-day old postnatal rats (P1–4). The spinal cord dissection procedure was similar to that described previously (Gao and Ziskind-Conhaim 1998). Briefly, a pregnant rat was lightly anesthetized with ether and decapitated. Embryos were removed, transferred to cold dissection solution, and decapitated. Postnatal rats were anesthetized by hypothermia and decapitated. The isolated spinal cord was placed in oxygenated cold dissection solution containing (in mM) 140 NaCl, 5 KCl, 4 CaCl2, 1.1 MgCl2, 4.2 HEPES, and 11 glucose (pH 7.2–7.4). The spinal cord was embedded in agar (2%) and glued to a chamber containing cold dissection solution. Transverse slices (400 μm) were cut at the level of L1–L5using a Vibratome.

Fluorescence imaging

Slices were stained by incubating in the voltage-sensitive fluorescent dye RH414 (200 μM, Molecular Probes, Eugene, OR) for 1 h at room temperature. Stained slices were then transferred to a recording chamber mounted on the stage of a Reichert Jung Diastar fluorescence microscope (Leica, Deerfield, IL). Slices were held in a glass-bottomed chamber with a grid of nylon threads. The slices were superfused with extracellular solution at a rate of 1–2 ml/min. Experiments were carried out at a temperature of 29–31°C (Demir et al. 1998). The extracellular solution was composed of (in mM) 113 NaCl, 3 KCl, 2 CaCl2, 1 MgCl2, 25 NaHCO3, 1 NaH2PO4, and 11 glucose. This solution was aerated with 95% O2-5% CO2 and had a pH of 7.2. In experiments in which CoCl2 (1 mM) was added to the extracellular solution, NaHCO3 and NaH2PO4 were omitted from the solution to prevent precipitation; the NaCl concentration was raised to 130 mM and 10 mM HEPES was added as a buffer.

Imaging techniques employed in this study were based on the approach ofWu and Cohen (1993) and have been described in detail previously (Demir et al. 1998; Jackson and Scharfman 1996). The preparation was illuminated with a 100-W tungsten-halogen bulb powered by a regulated power supply (Kepco, Flushing, NY). The microscope was equipped with a 475- to 565-nm band-pass excitation filter, a 570-nm dichroic mirror, and a 610-nm long-pass emission filter. Fluorescent light was focused onto a photodiode fiber optic device with 464 hexagonally arranged detectors. Each optical fiber was coupled to a photodiode, and each photodiode signal was amplified to a final level of 0.2 V/pA of photocurrent. Fluorescence records were obtained with either ×5 (NA 0.25) or ×10 (NA 0.5) Zeiss fluar objectives. The spatial resolution of the optical images was determined by the distance between the centers of the fields of adjacent photodiodes, which was 144 and 61 μm for the ×5 and ×10 objectives, respectively.

Amplified photodiode signals were low-pass filtered at 500 Hz, high-pass filtered with a 0.5-s time constant, and read into a Pentium computer using multiplexers and an A/D conversion board from Microstar (Bellevue, WA). The program Neuroplex (RedShirtImaging, Fairfield, CT) was used to read in the data at full-frame acquisition intervals of 0.94 ms for periods of up to 39 s. Analysis was performed with Neuroplex and with additional programs written in IDL (Research Systems, Boulder, CO). For display, fluorescence traces were digitally low-pass filtered at 100 Hz. To improve the signal-to-noise ratio, averages were often made of signals from two to four neighboring detectors. Because signals showed essentially no variation over the short distances between adjacent detector fields, the averaged signals did not alter the shape of the fluorescence traces.

Electrophysiological recordings

Extracellular field potentials were recorded with an Axopatch 1A amplifier (Axon Instruments, Foster City, CA) using glass electrodes (approximately 7–10 μm diameter) filled with 1 M NaCl. These signals were read into the computer simultaneously with imaging data.


Mean values were tested for statistically significant differences with Student's t-test. The hypothesis that a normalized mean was significantly different from one was tested with the Z test.


Frequency of spontaneous oscillations

Fluorescent dye signals were recorded simultaneously from 464 locations in transverse slices of spinal cords using photodiodes arranged in a hexagonal array (Fig. 1). In 23% of the slices (n = 37/160) oscillations were widely distributed, and visible in most regions of the slice (Fig. 1). Voltage oscillations had the same frequency at every location where they were seen. These oscillations continued at a relatively constant frequency in experiments lasting more than 20 min and were observed at times ranging from 1 to 6 h after slice preparation. In an additional 13% of slices, oscillations were confined to the lateral regions, along the dorsal and ventral edges, and around the central canal. In those slices, the largest amplitude oscillations were recorded in the lateral regions (see following text). In approximately 15% of slices, relatively large-amplitude oscillations were recorded at the beginning of the experiment, but the amplitudes declined quickly within a few minutes. If oscillations were apparent in a given spinal cord, they were recorded in at least two slices/cord. The reasons for the seemingly random and infrequent occurrence of spontaneous oscillations or their absence over periods of weeks are unknown. Preliminary experiments indicated that there was no correlation between the generation of oscillations in a given slice and its location in the rostral and caudal lumbar cord (slices from two animals), but more data will be necessary to address this question definitively. Data presented in this study were only from the 37 slices in which spatially widespread oscillations lasted at least 20 min.

Fig. 1.

Fluorescence traces from 464 locations in a transverse slice of lumbar spinal cord of a P3 rat. Traces were overlaid on a sketch of the slice from which the recording was made. Each trace is 0.91 s in duration, with about 2.5 oscillation cycles (3 Hz). Fluorescence oscillations were clearly visible in more than half the sites, but most other sites had oscillations that were too small to be detected with this gain. Each trace was normalized to its own resting light intensity. Data were acquired with a ×10 objective. Traces for dysfunctional detectors or detectors near the edge receiving too little light were left blank.

Oscillations had a relatively similar frequency of 2.91 ± 0.09 (SD) Hz (range, 2.75–3.1 Hz; n = 16), regardless of the areas in a slice where they were recorded. However, the amplitude of the oscillations varied considerably, and in the example shown in Fig. 1, the strongest oscillations were recorded in lateral regions of the slice and close to the midline. Quantitative analysis of recordings from several slices showed that the largest amplitude oscillations were generated in the lateral regions (mean ΔF/F = 0.0041 ± 0.0005; mean ± SE, n = 8). Normalization of the amplitudes in other regions to the value in the lateral region of a slice provided a relative estimation of spatial variations in the amplitude of the fluorescence signals (Fig. 2). These data showed that the amplitude of oscillations in ventral regions were generally larger than oscillations in dorsal regions of the slice. The larger amplitudes might suggest that lateral and ventral regions contributed more to the rhythmogenicity than other regions of the slice. However, amplitudes of fluorescence changes must be interpreted cautiously (Grinvald et al. 1988), and although smaller amplitudes were recorded in the dorsal region, lesion experiments showed that that region was capable of generating spontaneous oscillations when it was synaptically isolated from the ventral region (see following text). In most slices, the amplitude of the oscillations did not change during each recording period (up to 25 s). In a small number of slices (n = 5/37), the amplitude waxed and waned in an irregular manner with a time scale of roughly 5–10 s; such changes occurred synchronously in all regions of a slice. These findings suggested that rhythm-generating networks were preserved in 400-μm transverse sections of the lumbar spinal cord, which, at P1–4, encompass approximately one half of a single lumbar segment.

Fig. 2.

Amplitudes of oscillations in different regions of the cord. Peak-to-peak changes in fluorescence were normalized to the resting light from each detector (ΔF/F), and traces from 4 neighboring detectors at the indicated locations were averaged. These averages for a given experiment were then normalized to the amplitude of the oscillation recorded from the lateral location, which generally showed the largest amplitude. These ratios are plotted as means ± SE (n = 8), and the ratio at lateral locations was taken as one. The mean ΔF/F at the lateral site was 0.0041 ± 0.0005. The difference between dorsal and ventral is statistically significant by the Student's t-test (P < 0.05). The difference between the lateral and medial amplitudes is statistically significant by the ztest (P < 0.05).

To test the hypothesis that fluorescence changes represent changes in transmembrane potentials of cells and processes within the optical field of a given detector, the temporal correlation between fluorescence and extracellular field potential was examined. Extracellular recording electrodes were placed in regions of the spinal cord slice that showed strong fluorescence oscillations. Simultaneous recordings of fluorescence and extracellular field potential showed that the two signals oscillated with the same frequency and had a distinctive phase relationship (Fig.3 A; n = 3). The falling phase of the fluorescence change occurred at approximately the same time as the peak positivity in the field potential, indicating that outward currents produced the hyperpolarizing phase of the voltage oscillation and inward currents produced the depolarizing phase of the voltage oscillations.

Fig. 3.

A: fluorescence and field potentials in a slice from a P3 rat. Fluorescence signals (A1) and field potentials (A2) were recorded simultaneously from the same site in the dorsal horn. The fluorescence trace was averages from 2 neighboring detectors and normalized to resting light (ΔF/F). The midpoint of the falling phase of the fluorescence occurred at approximately the same time as the peak positivity in field potential (- - -). Thus inward currents coincided with the depolarizing phase, and outward currents coincided with the hyperpolarizing phase of the voltage oscillation.B: an example of field potentials recorded continuously in an unstained P3 slice, showing a 3-Hz oscillation.

To rule out the possibility that the 3-Hz oscillations resulted from dye exposure, extracellular recordings were carried out in unstained slices. Such recordings revealed 3.18 ± 0.12-Hz oscillations at multiple regions in three of six slices (Fig. 3 B). These findings indicated that the oscillations in stained slices did not result from effects of the dye on electrical activity.

To determine whether the fluorescence oscillations were dependent on membrane currents, Ca2+ channels were blocked by bath-applied CoCl2 (1 mM). The rhythmic oscillations were reversibly blocked (Fig.4, n = 3), indicating that Ca2+ entry via voltage-dependent Ca2+ channels was a necessary condition for voltage oscillations.

Fig. 4.

Inhibition of spontaneous oscillations by Co2+ (1 mM). Fluorescence was averaged from 4 neighboring detectors in the ventral horn of a P2 slice. Top: the signal before bath application of Co2+ (control). Middle: the fluorescence recorded at the same site, 7 min after CoCl2application. Bottom: the fluorescence recorded 20 min after washing out Co2+ (wash).

Pattern of oscillations

Fluorescence images of dye-stained spinal cord slices revealed an alternating pattern of oscillation in opposite sides of the cord. The peak fluorescence in one side corresponded to a trough in fluorescence in a mirror location on the contralateral side (Fig.5 B). This pattern was especially clear in a sequence of maps in which fluorescence intensity was encoded as color (Fig. 5 C). Red indicates high fluorescence, corresponding to a positive change in membrane potential, while blue indicates low fluorescence, corresponding to a negative change in membrane potential. Color maps at 20 ms intervals showed bursts of red, appearing alternately on the right and left sides of the slice with a frequency of about 3 Hz. This pattern of coordinated bilateral oscillation between the two halves of the spinal cord resembled the pattern of neurochemically induced alternating rhythmicity recorded in ventral roots and motor nerves of the isolated rat spinal cord (reviewed by Cazalets et al. 1998;Kiehn and Kjaerulff 1998). These findings suggested that the neural pathways coordinating bilateral rhythmic oscillations between the two sides of the cord were preserved in 400-μm-thick sections of the lumbar spinal cord.

Fig. 5.

Alternating patterns of 3-Hz oscillations in a spinal cord slice from a P4 rat. A: an overlay of fluorescence traces on a video picture of the slice (dorsal is up). The diagonal lines in this and other video pictures are the nylon threads of the grid that held the slice in the recording chamber. B: signals from the indicated sites show the temporal relations more clearly. The activity recorded at the left-lateral site (trace 1) is in-phase with activity from a right-medial site near the central canal on the contralateral side (trace 3), and is out-of-phase with activity from the right-lateral (contralateral) site (trace 4). Each trace is an average from 4 neighboring detectors. The dashed vertical lines help visualize timing differences. C: with fluorescence encoded as color, the spatiotemporal pattern of the oscillation can be seen as a left-lateral/right-medial peak (frame 6) alternating with a right-lateral/left-medial peak (frame 14). Frames are shown at 19.8-ms intervals; the sequence starts with the upper left frame, continuing to the right, and down by rows. Every 3rd frame is numbered to aid in finding specific events. The dark vertical line in each frame is the midline along the central canal (A), where no activity is seen.

Detailed analysis of the imaging data revealed that the pattern of oscillation was more complex than a simple right-left alternation between the two sides of the cord. Oscillations in a broad lateral region on one side of the spinal cord were synchronous with oscillations in a narrow medial region on the opposite side (Fig.5 B: compare trace 1 with 3, and trace 2 with 4). Thus the pattern was lateral-right/medial-left alternating, and it was apparent in all slices examined at this level of detail (n = 8). This pattern spanned the entire dorsal-ventral axis as can be seen from the vertical ribbons of color in the sequence of maps in Fig.5 C. For example, frame 6 shows a broad red zone laterally on the left side and a narrow red zone medially on the right side. Likewise, one half-cycle later, frame 14 shows a broad red zone laterally on the right side and a narrow red zone medially on the left side. Oscillations in the medial zone were restricted to a narrow region that was difficult to clearly see in color maps with the lower spatial resolution provided by the ×5 objective. Visualization in a color map such as Fig. 5 C required the ×10 objective, and the medial-lateral pattern was not apparent in lower magnification views (Figs. 6 and 8).

Fig. 6.

Oscillations after a midline lesion in a slice of a P2 rat.A: an image of the slices shows the 2 halves of the slice with the locations of the photodiode fields indicated. Note the lower magnification compared with Fig. 5. Before separating the 2 sides, alternating rhythm was apparent between the left and right sides of the slice, as demonstrated in Fig. 5. Oscillations continued independently in the 2 sides of the hemisected slice, either synchronously (B1) or alternately (B2). Five minutes elapsed between the recordings shown in B, 1 and 2, and this was sufficient time for a complete change in the relative phase of oscillation in the opposite sides of the slice.

The alternating pattern between medial and lateral regions was further examined by analyzing the correlation of fluorescence at different sites and estimating the R value from linear regression analysis. Two nearby sites from the lateral region were plotted against one another, giving an R value of approximately 0.9. By contrast, when a lateral site was plotted against a medial site, linear regression analysis yielded an R value of approximately −0.8. The plots appeared linear over the entire range of fluorescence values, and gave P values of less than 0.001.

Oscillatory pattern in hemisected slices

The complex pattern of oscillation revealed in Fig. 5 raises questions about the underlying circuitry. Contralateral projections presumably coordinate the alternating oscillations between the left and right sides of the slice, and such projections could also play roles in rhythmogenesis and in the coordination of the medial-lateral pattern. To investigate the function of contralateral projections in modulating coordinated oscillations, sagittal lesions were performed along the midline. Although oscillations continued in each of the separate halves after making the midline lesion (Fig. 6 B, 1 and2), they were no longer coordinated. Oscillations appeared either synchronous (Fig. 6 B1) or asynchronous (Fig.6 B2), as the phase relation slowly changed with time (approximately 5 min elapsed between Fig. 6 B, 1 and2). The average frequency of oscillations in these hemisected slices was 2.81 ± 0.05 (SD) Hz (n = 4 slices), similar to the frequency stated above for intact slices (2.91 ± 0.09 Hz, n = 16). The oscillations in hemisected slices showed the same widespread distribution as control slices, with activity spanning the dorsal-ventral axis. Thus each side of the cord possessed self-contained rhythmogenic activity. Furthermore, the circuitry intrinsic to each half of a spinal cord slice was sufficient to determine the frequency of these oscillations.

Contralateral projections might also play a role in coordinating the lateral-medial pattern of oscillations seen within each side of the cord. This hypothesis was tested by imaging oscillations in hemisected slices. Comparing the fluorescence signals from medial and lateral regions in the same side of a hemisected spinal cord slice showed that the alternating medial-lateral pattern evident in the intact slice (Fig. 5) was preserved in the absence of contralateral projections (Fig. 7, n = 6/6 slices). The oscillations in medial locations (Fig. 7, sites 2 and 3) were alternating with the oscillations in the corresponding lateral locations (sites 1 and 4, respectively). These findings suggested that the medial-lateral component of the oscillatory pattern was driven by circuitry intrinsic to each half of a spinal cord slice and was independent of contralateral projections.

Fig. 7.

Medial-lateral patterned oscillation in hemisected slices. Traces were taken from the numbered locations marked in the picture above. Oscillations in medial sites (traces 2 and 3) were out-of-phase with oscillations in the corresponding lateral locations (traces 1 and 4, respectively).

Spatial distribution of rhythm generators

Our finding that rhythmic voltage oscillations were visible throughout transverse slices of spinal cord raised the possibility that rhythm-generating networks were widely distributed through the entire slice. Alternatively, the oscillations could be triggered by a few unitary rhythmogenic networks localized in specific regions of the cord with spread to other regions depending on synaptic projections. To distinguish between these possibilities, spinal cord slices were sectioned along sagittal and horizontal axes and subsequently monitored with fluorescence imaging (n = 4). Oscillations were observed in all such spinal cord fragments, and their frequency was similar to that recorded in intact slices (Fig.8). Each of the four quarters showed oscillations at a frequency of 2.99 ± 0.16 Hz (n= 4 slices), similar to that seen in slices prior to sectioning. The preservation of oscillations in four fragments indicated that synaptic connections between the dorsal and ventral horns were not necessary to trigger the oscillations and did not play a role in regulating the frequency. These results implied that both dorsal and ventral regions of the slice contained rhythm-generating circuitry capable of producing voltage oscillations.

Fig. 8.

Oscillations in quarter sections of a slice from a P2 rat. Overlays of fluorescence traces on video images show activity before (A1), and after cutting the slice midsagittally and midhorizontally (B1). ⋆ inA1 indicates the site used for amplitudes in Fig.9 A. The same slice was also imaged after making only the midsagittal cut (Fig. 6). A2 and B2: sequences of color maps at 19.8-ms intervals show the oscillations as in Figs. 5 and 6. B2 shows that oscillations continued in all four quarters after making the cuts. C: fluorescence traces from each of the 4 quarter sections ofB1 show that the oscillation is maintained after cutting. Trace numbers correspond to the labels of quarters inB1.

Although the frequency of the oscillation was not altered as the slice was cut into successively smaller pieces (Figs. 6-8), the amplitudes of the oscillations decreased with increasing number of lesions (Fig.9). A possible explanation is that recurrent excitatory synapses contribute to the rhythmogenic circuit. Larger networks with more neurons would summate inputs from more synapses to increase the overall amplitude of the voltage changes during rhythmic oscillations. Reducing the number of synapses in spinal cord fragments could result in smaller amplitude oscillations. However, this result must be interpreted with caution because it is conceivable that the decrease in amplitude resulted from more general forms of tissue damage caused by the lesions. We attempted to carefully measure amplitudes in the centers of fragments away from the cut edges. However, because processes could extend to the cut edge, we cannot rule out the possibility of damage in the middle of a fragment.

Fig. 9.

A: amplitudes of oscillations from intact spinal cord slices (1), from halves of slices (2, see Fig. 6), and from quarters (3, see Fig.8 B). Traces from whole and cut slices showed that the frequency of the oscillation remained the same, but the amplitude decreased in smaller sections. The traces were all from the same slice before and after making the indicated cuts. Averages were computed from 4 detectors at the site indicated by ⋆ in Fig.8 A1. B: mean amplitudes ± SE are plotted for oscillations in intact slices, halves, and quarters (n = 4). Amplitudes were taken from ventral locations.

Developmental changes in the pattern of oscillations

Rhythmic activity has been reported in the spinal cord of rat embryos at E15 (Greer et al. 1992; Nishimura et al. 1996; Ozaki et al. 1996), a few days after motoneurons are generated in the lumbar spinal cord (Nornes and Das 1974) and synaptic connections are formed (Saito 1979; Ziskind-Conhaim 1988,1990). Persistent spontaneous oscillations were recorded in 25% of spinal cord slices from E15–16 rats (n = 9/36). At this age, the 400-μm transverse slice encompassed more than one lumbar segment, but the probability of recording spontaneous oscillations was not higher than that recorded after birth, when slices constituted about half of a single lumbar segment.

The oscillatory pattern in embryonic slices was different from that recorded in slices of neonatal rats. Unlike the widely distributed oscillations at P1–4, in embryonic slices, rhythmic activity was only seen in limited areas such as those indicated by the red contours in Fig. 10 A. In those slices the frequency was 3.50 ± 0.95 Hz (n = 7), slightly higher and more variable than in neonatal rats. The amplitude of the oscillations (ΔF/F = 0.00060 ± 0.00059; mean ± SE; n = 7) was significantly smaller than in neonatal slices (Fig. 10 B). Moreover, in seven of the nine slices, when oscillations were visible on both sides of the cord, they were synchronous (traces 2 and 3, Fig.10 B) rather than alternating as was seen in slices from neonatal rats. The alternating lateral-medial oscillations that were generated in each half of the slice at P1–4 were not apparent at this age. An alternating pattern of oscillation in opposite halves of the cord was apparent in only 2/9 of the slices tested, and they were both from the same embryo. Similar to the synchronous oscillations, the oscillations alternating between the two sides were restricted to small areas. Those oscillations were not included in data analysis. Spontaneous, synchronous ventral root potentials have been recorded at intervals of 1–2 min in opposite sides of E15–16 intact spinal cords (Nishimura et al. 1996). Therefore the developmental transition from synchronous to alternating coordinated patterns of oscillations in the slice preparation might be related to the developmental changes observed in the isolated spinal cord.

Fig. 10.

Spontaneous oscillations in a spinal cord slice from an E15 embryo.A: overlay of fluorescence traces on a video image (dorsal is top left). The red contours surround the areas of visible oscillatory activity. B: traces averaged from four neighboring detectors in areas 1 and 3 on one side of the cord and 2 and 4 on the other side. The amplitudes of the fluorescence changes were significantly smaller than those recorded in postnatal rats, but their frequency was similar (see text). The oscillations at sites 2–4 were all in-phase, although the signal recorded at site 4 was small.C: color maps at 19.8-ms intervals show the temporal pattern of oscillations. Because of the small signal-to-noise ratio, images were spatially smoothed.


Properties of spontaneous oscillations

This is the first study of the spatiotemporal pattern ofspontaneous rhythmic oscillations in spinal cord slices of embryonic and neonatal rats. Imaging of membrane potential with a voltage-sensitive fluorescent dye showed oscillations with a frequency that was consistently close to 3 Hz and was recorded in the absence of neuromodulatory substances and electrical stimulation. Fluorescence signals oscillated with the same frequency as extracellular field potentials recorded simultaneously at the same site, suggesting that the rising and falling phases of fluorescence resulted from alternating inward and outward membrane currents.

The ionic mechanisms underlying these oscillations are unknown. Rhythmic, spontaneous action potentials such as those recorded in brain stem neurons (Koshiya and Smith 1999), as well as excitatory and inhibitory synaptic currents (Raastad et al. 1997), might contribute to the rhythmic activity. The fluorescence of RH414 reflects an average membrane potential of all somata and processes within the field imaged by a given detector, but the identity of the cells responsible for these changes is unknown. Action potentials in various neuronal types might account for some of the fluorescence change, and a contribution from glial cells is also possible. The primary objective of the present study was to characterize the spatiotemporal pattern of activity in transverse sections of the cord rather than determine the underlying cellular interactions. The roles of specific types of neurons were recently evaluated by imaging techniques in the simpler motor system of the leech (Cacciatore et al. 1999). Simultaneous intracellular recordings, or selective dye-staining procedures (Wenner et al. 1996) will be necessary to relate the spatiotemporal dynamics of oscillations to specific forms of cellular activity in the vertebrate spinal cord. Our finding that the amplitudes of oscillations were greatest in lateral and ventral regions (Fig. 2) supports a previous report demonstrating a higher density of cells with spontaneous oscillatory Ca2+ transients in the ventrolateral region of the spinal cord of chick embryos (O'Donovan et al. 1994).

Our findings that oscillations can be generated in 400-μm transverse slices suggested that intersegmental projections were not required for rhythmogenesis. Spontaneous rhythmic bursts also persist in small isolated single lumbar and sacral hemisegments of newborn mice (Bonnot and Morin 1998; Whelan et al. 2000) and in slice cultures of embryonic rat spinal cord in which inhibitory synapses are not well established (Streit 1993). Moreover, neurochemically induced rhythmic activity was recorded in isolated spinal cords in which action potential-dependent chemical transmission was blocked (Tresch and Kiehn 2000). However, in contrast to our observation that intersegmental projections were not required for rhythm generation, other studies have suggested that such projections are important as indicated by the gradual reduction in effectiveness of neurochemical substances to induce rhythmic activity with progressive sectioning of the lumbar spinal cord (Kjaerulff and Kiehn 1996;Kremer and Lev-Tov 1997). Some of the differences between the findings of those studies might be attributed to the different preparations and procedures used to induce oscillatory activity.

The reasons for not detecting rhythmic activity in all slices are unknown, but it is possible that the low incidence of rhythmogenesis is related to the age-dependent decline in the potential of the postnatal cord to generate coordinated oscillations (Bonnot et al. 1998). The low incidence of spontaneous oscillations might also result from the fact that at P1–4 the slice encompasses only a portion of a single lumbar segment, which might not always contain functionally intact rhythm-generating networks. We cannot rule out the possibility that damage during the isolation procedure also contributed to the variable appearance of spontaneous oscillations.

Spinal cord oscillations have been reported with a wide range of frequencies. The frequency of spontaneous oscillations in the mouse lumbar cord was approximately 1 Hz (Bonnot et al. 1998;Whelan et al. 2000), while in spinal cord slices maintained in culture it was in the range of 4–5 Hz (Streit 1993). Neurochemically induced locomotor-like oscillations in intact spinal cords of neonatal rats ranged from 0.5 to 2 Hz (Kremer and Lev Tov 1997; Kudo and Yamada 1987; Raastad et al. 1997), but high-frequency ventral root potentials (5–10 Hz) could also be generated by NMDA, 5-HT, or norepinephrine (Cazalets et al. 1990). The latter oscillations were sometimes superimposed on low-frequency (0.2–0.5 Hz) coordinated locomotor-like oscillatory activity. NMDA-induced oscillations were also recorded intracellularly in interneurons (8.6 Hz) and motoneurons (4.4 Hz) of neonatal rats (Hochman et al. 1994; MacLean et al. 1997). The frequency of 3 Hz for spontaneous rhythmic activity in our study falls within the range of reported values. The variability in frequency could reflect the many experimental differences in these studies, such as age and rodent species, and the concentrations of neuromodulatory substances used to induce rhythmic activity (Hochman et al. 1994; Schmidt et al. 1998). It is possible that longitudinal inhibition plays a role in regulating oscillation in the intact spinal cord, and its absence in slices could increase the frequency compared with intact preparations. Membrane potential could also be a factor, as evident by the reports that high-K+-induced depolarization increases the frequency of oscillations in the spinal cord and brain stem (Bracci et al. 1996; Koshiya and Smith 1999; Kremer and Lev-Tov 1998). However, it is unlikely that the relatively high-frequency recorded in our study was related to extracellular K+ concentration because a relatively low concentration of K+ was used in our study.

Coordinated pattern of oscillations

The voltage oscillations recorded in this study exhibited a complex pattern, with depolarizations alternating between the right and left sides of the spinal cord and between the medial and lateral regions within each side of the cord. This alternating medial-lateral pattern, not described previously, was preserved in separate halves of a slice, indicating that neural connections intrinsic to each side can maintain the pattern and contralateral projections do not set the phase relation between the medial and lateral depolarizations on the same side. It is likely that inhibitory pathways between medial and lateral regions in each side of the cord mediate the alternating activity (Fig.11). It should be noted that although the coordinated, alternating pattern of oscillations observed in our study resembles the locomotor-like rhythmicity, we have no experimental evidence to show that those activities are functionally correlated.

Fig. 11.

Diagram of neural circuitry mediating patterned rhythmic activity in spinal cord slices. Rhythm generators (RGs) are indicated by the circles enclosing waves and are distributed over dorsal and ventral areas of the slice. Phase-setting pathways are depicted as arrows between RGs. Excitatory pathways (↔) maintain synchrony along the dorsal-ventral axis, to produce ribbons of simultaneous depolarization (Fig. 5 C). Excitatory bilateral projections synchronize the activity in the right-lateral and left-medial regions. Inhibitory pathways (>—<) project horizontally to coordinate alternating phases of excitation in lateral and medial regions.

The identity of the neurons that participate in the alternating lateral-medial oscillations is unknown, but it is unlikely that this pattern represents the activation of different pools of motoneurons that innervate muscles with opposite functions (e.g., extensor versus flexor). Although motoneurons with inverse functions are clustered in different areas in the ventral horn, none are thought to be located in a narrow zone close to the midline (McHanwell and Biscoe 1981; Nicolopoulos-Stournaris and Iles 1983). Moreover, motoneurons innervating different muscles fire at different times during rhythmic locomotion, but in the chick this reflects firing at different phases of the oscillation rather than phase differences between the oscillations of different motoneurons (O'Donovan 1989).

Spatial distribution of rhythmogenic circuitry

The alternating depolarization between the right and left sides is reminiscent of network oscillations generated by reciprocal inhibition. Reciprocal inhibition between populations of neurons has been shown to give rise to oscillations in various CPGs performing diverse functions (Marder and Calabrese 1996). In these models, activity in one half follows the termination of inhibition from the contralateral half. However, such models cannot explain the rhythmogenicity in spinal slices because each half continued to oscillate after it was synaptically isolated by midline lesion. Indeed, fragments as small as one-fourth of a slice oscillated spontaneously, indicating that rhythmogenicity is widely distributed over ventral and dorsal regions of the lumbar spinal cord (Fig. 11). The finding ofKremer and Lev-Tov (1998) that neurochemically induced rhythmicity could be seen in the dorsal horn after the surgical removal of the ventral horn is consistent with our findings that rhythmogenesis is not restricted to intermediate and ventral regions.

The fact that oscillations were observed in spinal cord fragments ruled out a role for long-range projections in the generation of spontaneous oscillations. Our findings support the hypothesis that rhythm generation by pacemaker cells, such as the neurons involved in respiratory oscillations, maintain their rhythmic activity independently of excitatory synaptic transmission (Koshiya and Smith 1999). It is conceivable that excitatory connections between high excitability rhythm generators play a role in the synchronization and coordination of activity in various regions of the lumbar spinal cord. Excitatory pathways could project parasagittally (Fig. 11) to produce the ribbons of synchronous activity (Fig. 5). Bilateral excitatory projections between lateral and contralateral medial regions could synchronize oscillations in the right and left sides of the cord. Amplification of activity by excitatory interactions might account for the larger amplitudes in larger slice fragments (Fig. 9).

The presence of spontaneous oscillations in fragments of spinal cord slices suggested that local circuitry and intrinsic neuronal properties are sufficient for rhythmogenesis. Mathematical models have indicated that gap junctions could play an important role in the synchronization of rhythmic activity (Kepler et al. 1990; Manor et al. 1997), and experimental evidence for this was recently reported in the neonatal rat spinal cord (Tresch and Kiehn 2000). Thus the age-dependent decrease in coupling (Kandler and Katz 1995; Walton and Navarrete 1991) could explain the developmental decline in rhythmogenesis in the mouse spinal cord (Bonnot et al. 1998).

In contrast to the presumed synchronizing role of excitatory synapses, inhibitory synapses are likely to perform phase-setting functions in coordinating the complex pattern of rhythmic activity. It has been suggested that inhibitory synapses regulate the alternating left-right oscillations in the ventral roots (Bracci et al. 1996;Cowley and Schmidt 1995; Kjaerulff and Kiehn 1997; Kremer and Lev-Tov 1997; Kudo and Nishimaru 1998). The findings that synchronous rather than alternating oscillations were generated in the opposite sides of slices from embryos may reflect the later development of inhibitory synapses relative to excitatory synapses in the rodent spinal cord (Gao et al. 1998; Jackson et al. 1982).

In summary, the alternating pattern of spontaneous rhythmic activity in opposite sides of transverse lumbar spinal cord slices resembles spontaneous locomotor-like activity in isolated spinal cords of neonatal rats. However, the relationship between the rhythmic activity reported in this study and the CPG activity responsible for regulating rhythmic locomotor movements remains as an interesting question for future study. Spontaneous oscillations exhibit a complex pattern of alternating depolarization not only between the left and right sides but also between the lateral and medial regions within each side of the spinal cord. Multiple rhythm generators are distributed through the transverse section of the developing lumbar spinal cord, which oscillate at a characteristic 3-Hz frequency independently of long-range synaptic pathways. The synchronization and coordination of activity between these rhythmogenic loci presumably depends on both excitatory and inhibitory synaptic projections.


This work was supported by National Institute of Neurological Disorders and Stroke Grants NS-37212 (M. B. Jackson) and NS-23808 (L. Ziskind-Conhaim).


  • Address for reprint requests: L. Ziskind-Conhaim, Dept. of Physiology, University of Wisconsin Medical School, 1300 University Ave., Madison, WI 53706 (E-mail: lconhaim{at}


View Abstract