Spatiotemporal organization of frog respiratory neurons visualized on the ventral medullary surface

Spatiotemporal organization of frog respiratory neurons visualized on the ventral medullary surface

Respiratory Physiology & Neurobiology 161 (2008) 281–290 Contents lists available at ScienceDirect Respiratory Physiology & Neurobiology journal hom...

2MB Sizes 7 Downloads 97 Views

Respiratory Physiology & Neurobiology 161 (2008) 281–290

Contents lists available at ScienceDirect

Respiratory Physiology & Neurobiology journal homepage: www.elsevier.com/locate/resphysiol

Spatiotemporal organization of frog respiratory neurons visualized on the ventral medullary surface Yoshitaka Oku a,∗ , Naofumi Kimura b , Haruko Masumiya a , Yasumasa Okada c a

Department of Physiology, Hyogo College of Medicine, Nishinomiya, Hyogo 663-8501, Japan Department of Pharmacology, Jikei University School of Medicine, Tokyo 105-8461, Japan c Department of Medicine, Keio University Tsukigase Rehabilitation Center, Shizuoka 410-3215, Japan b

a r t i c l e

i n f o

Article history: Accepted 6 March 2008 Keywords: Control of breathing Brainstem Respiratory neurons Oscillators Evolution of breathing Frog Buccal ventilation Lung ventilation Voltage imaging Cross-correlation Strychnine Reciprocal inhibitory networks

a b s t r a c t We visualized the spatiotemporal activity of respiratory-related neurons in the frog using the isolated brainstem spinal cord preparation. We recorded optical signals from the ventral surface of the medulla using a voltage-sensitive dye, and calculated cross-correlations with the integrated respiratory activity of the trigeminal nerve. Lung burst-related depolarizing optical signals were observed bilaterally as longitudinal columns in the ventrolateral medulla between the levels of trigeminal and hypoglossal rootlets, mostly caudal to the vagal rootlet. However, we could not differentiate between neurons involved in rhythm generation and motoneurons. The dye weakened the buccal rhythm and slowed the lung rhythm, which might have influenced the results. Extracellular recording of respiratory neurons verified the optically identified area. Strychnine disrupted the spatiotemporal organization of optical signals, although trigeminal periodic bursts persisted. We conclude that the pattern generator but not the rhythm generator of lung burst in the frog involves glycinergic mechanisms and lies as longitudinal columns in the reticular formation of the ventrolateral medulla. © 2008 Elsevier B.V. All rights reserved.

1. Introduction Common frogs have two ventilation modes, buccal ventilation and lung ventilation. The buccal ventilation alternately repeats dilation and constriction of the buccal floor. The lung ventilation of the frog (Rana catesbeiana) is accomplished by the following four steps: first, the augmented activity of buccal dilator muscle inhales air into the buccal cavity via opened nares; second, the glottis opens and air is exhaled from the lung and goes through the nares; third, closing the nares, air is stuffed into the lung by the elevation of the buccal floor; fourth, the glottis closes and the nares is opened again (de Jongh and Gans, 1969). It is confirmed that these coordinated alternative motor patterns are preserved in the in vitro preparation that has no pulmonary afferent feedback (Kimura et al., 1997). In this preparation, the activation sequence occurs in the following order: the buccal dilating activity in the sternohyoid branch of the hypoglossal nerve (step 1), the glottal opening activity in the laryngeal branch of the cranial nerve (CN) X (step 2), the buccal compressing activity in the main branch of the CN

∗ Corresponding author. Tel.: +81 798 45 6387; fax: +81 798 48 9643. E-mail address: [email protected] (Y. Oku). 1569-9048/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.resp.2008.03.002

XII (hypoglossal nerve; i.e., second spinal nerve in the frog) (step 3), and the glottal closing activity (step 4), though the ratio of deflation and inflation might not be balanced in the fictive respiration. The rhythms of lung ventilation and buccal ventilation can be manipulated independently (Wilson et al., 2002), and can be uncoupled (Vasilakos et al., 2006). Thus it is postulated that paired coupled oscillators generate the respiratory rhythm in the frog. The locations of the two oscillators in the frog, responsible for the buccal and lung ventilations, were explored by responsiveness of each ventilation mode to microinjection of excitatory or inhibitory agonists (Wilson et al., 2002). According to this study, putative locations of these oscillators were separately identified, and termed as ‘buccal area’ and ‘lung area’, respectively. They reported that the lung area is located between the levels of CN VIII and IX, and the buccal area is located at the level of CN X. Neuronal mechanisms of these complex ventilatory motor patterns have not yet been elucidated. To understand the mechanisms of central motor pattern generation, we must know all types of neurons involved, their locations, and their neuronal connections. There are reports that described characteristics of several types of respiratory-related neurons (Kogo and Remmers, 1994; McLean and Remmers, 1997), however to date, there is even no information

282

Y. Oku et al. / Respiratory Physiology & Neurobiology 161 (2008) 281–290

as to the whole distribution of respiratory-related neurons and their activation sequences. Optical imaging has been successfully applied to visualize the respiratory-related neuronal activity of the ventral medulla in rats (Onimaru and Homma, 2003). Respiratory-related neuronal areas and their temporal activation patterns in optical imaging experiments can be defined by the cross-correlation analysis (Okada et al., 2007; Oku et al., 2007). Therefore, the present study primarily aimed to identify the distribution of frog respiratory neurons that could be visualized from the ventral medulla and to characterize their spatiotemporal activation patterns. In the interpretation of our results, we make a distinction between rhythm generation and pattern generation. We use the term ‘activity of buccal or lung oscillator’ to refer to the activity of rhythm generation, whereas the term ‘pattern formation’ is used to refer to down stream events associated with the production of motor output patterns. Simultaneous recording of the alternatively active nerves innervating respiratory muscles from the in vitro preparation of the frog brainstem has revealed that glycinergic synaptic inhibition is essential for the formation of the augmenting and reciprocal patterns in the respiratory motor activities (Kimura et al., 1997). We then hypothesized that the blockade of glycinergic inhibitory neurotransmission leads to a disruption of spatiotemporal neuronal network activity during lung ventilation, provided that the formation of the normal ventilatory motor pattern during lung ventilation depends on glycinergic inhibitory network mechanisms in the frog. To test this hypothesis, we further visually analyzed the effect of elimination of glycinergic postsynaptic inhibition on the spatiotemporal organization of the respiratory neuronal networks. 2. Methods 2.1. Frog brainstem preparation Experimental protocols were approved by the Animal Research Committee of Hyogo College of Medicine, and were in accordance with the Guiding Principles for the Care and Use of Animals of the Physiological Society of Japan. Experiments were performed on juvenile and adult Rana catesbeiana metamorphosed within 4 months (11–41 g, n = 16). All animals were acquired from a commercial supplier (Kazuo Ohuchi, Misato, Saitama, Japan). All the frogs used in the study were more than 1 week in age after metamorphosis. We classified frogs whose body weight were less than 20 g as juvenile, and others as adult. According to this criterion, nine were juvenile and seven were adult. The frog brainstem was isolated according to the methods described previously (Kimura et al., 1997). In the preliminary study, we measured the size of the brainstem. The lengths between rootlets of CN V and XII were about 4.0 mm in the 30 g frog (the body length was 65 mm) and 3.7 mm in 15 g frog (the body length was 51 mm). The length between rootlets of CN V and XII of 5 g frog was 3.5 mm (the body length was 38 mm). Therefore, sizes of the brainstem used in the present study were almost the same, in spite of the large range of body weight. Each animal was anesthetized in ice-cold water containing tricaine methane sulphonate (1:10,000) until unresponsive to toe pinch. The cranium was opened, and then the animal was quickly decerebrated. The cranial and spinal nerves were cut, and the brainstem was transected at the levels rostral to the CN V rootlet and caudal to the CN XII rootlet. The choroid plexus, dura and arachnoid maters were carefully removed from the dorsal and ventral surfaces of the medulla. The basilar artery was also removed from the ventral surfaces of the medulla, but the pia mater was left intact. In the preliminary study, we confirmed that this procedure does not affect either the buccal or lung motor activity. Throughout the dissection and control recording, the brainstem was superfused with an artificial cere-

brospinal fluid (aCSF), with the composition (in mM) NaCl 104, KCl 4.0, MgCl2 1.4, d-glucose 10, NaHCO3 25 and CaCl2 2.4, that was equilibrated with 2% CO2 and 98% O2 and controlled at 22 ± 1 ◦ C (Torgerson et al., 1997). Measured pH of the aCSF was 7.8 in the experimental condition. Prior to the optical recording, the preparation was incubated in the aCSF containing a voltage-sensitive dye, di-2-ANEPEQ (0.1–0.2 mM, Invitrogen, Carlsbad, California, USA) and dyed for 40 min bubbling with 98% O2 and 2% CO2 . After staining, the preparation was rinsed in aCSF for 10–15 min to eliminate an excessive dye, transferred to a 2 ml recording chamber, and pinned with the ventral surface up. The recording chamber was continuously superfused with aCSF at a rate of 3 ml/min. To test the effect of glycinergic synaptic blockade, strychnine (1–5 ␮M, Sigma) was given to the superfusate. Fictive lung ventilation can be recorded as lung bursts from nerve rootlets of CN V, VII, X and XII in the isolated brainstem of adult frogs (McLean et al., 1995a). The activities recorded from these nerve rootlets of the in vitro brainstem preparations are synchronous during the active phase because these nerve rootlets include nerve fibers innervating both buccal dilating and compressing muscles or both glottal opening and closing muscles. We monitored respiratory-related nerve activity from the CN V rootlet with a suction electrode (Fig. 1). The signals recorded from the CN V rootlet also contained the buccal dilating and compressing activities (McLean et al., 1995a). However, because the respiratory activity of CN V can be recorded with highest voltages in these cranial nerves (McLean et al., 1995a), it is possible to distinguish between the buccal and lung ventilatory cycles (Torgerson et al., 1998). The raw nerve signal was amplified using a bioelectric amplifier AB651J (Nihon Kohden, Tokyo, Japan), bandpass filtered from 15 Hz to 3 kHz, and fed into a time-amplitude window discriminator EN611J (Nihon Kohden, Tokyo, Japan) to trigger the onset of lung or buccal burst activity. The window discriminator generated a TTLlevel pulse signal when the trigeminal nerve activity exceeded the minimally preset threshold level, and then the TTL pulse was fed into the optical recording system. In addition, the filtered signal was full wave rectified and integrated using a ‘leaky’ integrator EI601G (Nihon Kohden, Tokyo, Japan) with a time constant of 100 ms. 2.2. Optical recording The recording chamber was mounted on a fluorescence macro zoom microscope (MVX-10, Olympus Optical, Tokyo, Japan). Activity of respiratory-related neurons in the ventral medulla was analyzed using an optical recording system (MiCAM Ultima, BrainVision, Tokyo, Japan). Preparations were illuminated with a tungsten-halogen lamp (150 W) through a band-pass excitation filter ( = 480–550 nm). Epifluorescence through a long-pass barrier filter ( > 590 nm) was detected with a CMOS sensor array (MiCAM Ultima L-camera, BrainVision, Tokyo; 100 ␮m × 100 ␮m pixel size, 100 × 100 pixel array) (Okada et al., 2007; Oku et al., 2007). Magnification of the microscope was adjusted to 2.8×–3.3× depending on the size of the brainstem. One pixel corresponded to 30 × 30 ␮m–35 × 35 ␮m, and the image sensor covered a total of 3 × 3 mm–3.5 × 3.5 mm. Optical signals were sampled at 50 Hz (20 ms/frame). Raw and integrated CN V activities were recorded at 1 kHz, and stored in a hard disk together with optical signals for later analyses. A total of 256 frames were recorded starting at 64 frames (1.28 s) before the onset of lung burst activity, and averaged 10–30 times. 2.3. Extracellular recording Activity of medullary respiratory neurons was recorded extracellularly with a glass microelectrode filled with 3 M NaCl (DC

Y. Oku et al. / Respiratory Physiology & Neurobiology 161 (2008) 281–290

283

Fig. 1. Schematic drawing of frog brainstem spinal cord preparation. (A) The respiratory motor activity was monitored from the CN V rootlet by a suction electrode. (B) A representative recording of the CN V respiratory activity (upper panel) and the full wave rectified, integrated CN V signal (lower panel). A lung burst (indicated as Lung) with a bigger amplitude is followed by a buccal burst (indicated as Buccal) with a smaller amplitude. Ld and Lc represent the buccal dilating and constricting phases of the lung activity, respectively. V, VI, VII/VIII, IX/X and XII: cranial nerves.

impedance 1.2–3 M). Neurons were recorded in the ventrolateral medulla (Okada et al., 1993b). Although we initially surveyed respiratory neurons widely, we encountered respiratory neurons mostly in the optically identified respiratory-related areas. We then focused our recording in the optically identified areas. The recording site was identified empirically using cranial nerve rootlets as landmarks. Tissue layers from the surface down to 600 ␮m were explored. Depths of recorded neurons from the ventral medullary surface were estimated by reading the penetrating distance of the micromanipulator that controlled the microelectrode advancement as described elsewhere (Okada et al., 1993b). Neuronal activity was amplified, band-pass filtered ( = 50 Hz–10 kHz), and stored on a hard disk together with the CN V activity for later analyses. Spikes were passed through a window-discriminator software (Chart 5 spike histogram module, ADInstruments, Castle Hill, Australia) to obtain unitary pulses from a single unit using spike amplitude and duration as discriminating criteria.

0.25, then F/F was set to be zero. A negative F/F corresponds to membrane depolarization. Cross-correlation coefficients between the integrated CN V activity and each pixel were calculated after removing linear drift components associated with photo-bleaching (Okada et al., 2007; Oku et al., 2007). The cross-correlation coefficient function was defined as: Rxy() =

x(t)y(t + )

  x2

y2

Preparations (n = 3) were transferred into a fixative solution (4% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4) after recording of neuronal activities, stored at 4 ◦ C for 2–3 weeks, paraffin-embedded, transversely sectioned (6 ␮m thick) and stained with hematoxylin–eosin. Identification of the level of each section followed Kemali and Braitenberg (1969) and Pogosyan et al. (2002). The anatomical location of respiratory-related neurons in the brainstem was estimated by optical and extracellular recordings, i.e., both optical and extracellular recording data were used to estimate the two-dimensional location on the ventral brainstem and extracellular recording data were used to estimate the depths from the ventral brainstem surface.

where x was the integrated CN V activity and y was optical signal at each pixel. Cross-correlation coefficients were estimated for the lag  ranging between −1 s and 1 s. Then the maximum of cross-correlation coefficients (CCmax) and the lag at which the cross-correlation became maximal (LAGmax) were estimated. CCmax represents the similarity between the two signals. LAGmax represents the activation timing relationship, and a negative LAGmax means that the mean activity of the pixel precedes CN V activity, although it does not necessarily mean that the onset of the activation of the pixel is earlier than that of CN V activity. As a test of significance of cross-correlation coefficient estimates, the upper 99.9% confidence value was calculated with p < 0.001 being considered significant. The confidence value was 0.19 for 256 frames data. Pixels having CCmax greater than the half of the maximal value of CCmax were plotted on the reference image. This threshold value was greater than the upper 99.9% confidence value and eliminated most noises associated with tissue vibration and scattering light. We defined these pixels as respiratory-related pixels. Effects of strychnine were statistically analyzed using a paired t-test. A p-values < 0.05 was considered significant. Values were shown as mean ± S.E.M.

2.5. Data analysis

3. Results

The change in fluorescence intensity (F) relative to the initial intensity of the fluorescence (F0 ) in each pixel was calculated. To normalize the difference in the amount of membrane-bound dye and illumination within the preparation, background fluorescence intensity at each pixel was divided by the maximal background fluorescence, then the ratio of F to the normalized background fluorescence intensity (F), i.e. the fractional change in fluorescence intensity (F/F), was calculated at each pixel in each frame (Fukuda et al., 2006; Okada et al., 2007; Oku et al., 2007). If F was less than

3.1. Voltage imaging of respiratory activity

2.4. Histological examination

Respiratory-related optical signals associated with lung bursts were appreciable even without averaging in all stained preparations (n = 10) by the ventral approach. To obtain better optical images, signals were averaged 10–30 times triggered by the lung burst activity. We successfully obtained respiratory-related optical signals associated with lung bursts. We applied a cross-correlation technique to visualize the distribution and spatiotemporal activa-

284

Y. Oku et al. / Respiratory Physiology & Neurobiology 161 (2008) 281–290

Fig. 2. Pseudocolor images of optical signals reveal that frog respiratory neurons are located bilaterally in the ventrolateral medulla as longitudinal columns between the levels of CN V and XII rootlets. (A) Cross-correlations between the integrated CN V activity and optical signals are shown as a pseudocolor image (a representative case obtained from one preparation). (B) Pixels whose CCmax exceeds the half of the maximal value of CCmax are plotted on the reference frame photograph, and are colored depending on LAGmax. The pseudocolor map shows that neurons within warm hue pixels depolarize earlier than those within cold hue pixels. Note that the onset of respiratory-related optical signals of warm hue pixels is the same as that of cold hue pixels, but the time courses of depolarization are different. (C) The raw optical signal traces depicted in red and blue correspond to areas circled in red and dark blue in panel B. Note that the lung burst duration of CN V activity is longer than that of Fig. 1B (unstained preparation) due to dye staining. VII/VIII, IX/X and XII: cranial nerves.

tion profiles of respiratory neurons. A pseudocolor image of CCmax of each pixel represents the distribution of respiratory-related neurons (Fig. 2A). In all preparations, respiratory-related areas were identified bilaterally as longitudinal columnar regions in the ventrolateral medulla between CN V and XII. There was no difference in the location of respiratory-related areas depending on the weight of frogs. We then plotted pixels whose CCmax exceeded the half of the maximal value of CCmax, and colored them depending on LAGmax (Fig. 2B). Warm hue pixels indicate that the neurons within these pixels depolarized faster than neurons within cold hue pixels. As depicted in Fig. 2C, the onset of respiratory-related optical signals of warm hue pixels was the same as that of cold hue pixels, but the time courses of depolarization were different. In other words,

these pixels finished depolarizing earlier than neurons within cold hue pixels. The area depolarized at the earliest timing (red pixels) was located in an area about 700–1000 ␮m lateral to the midline and 500–800 ␮m rostral to the CN XII rootlet, caudal to the ‘buccal area’ defined by previous workers (Wilson et al., 2002). Subgroups of pixels with delayed peak activation were distributed rostrally and medially within the respiratory-related area. Di-2-ANEPEQ affected both lung and buccal ventilations. Although we did not measure the control values before staining, when we compared the values of 6 stained preparations with those of 6 unstained preparations, the frequency of lung bursts was lower (unstained 6.2 ± 2.9 burst/min vs. stained 2.9 ± 0.7 burst/min) and the burst duration was longer (unstained 666 ± 218 ms, range

Fig. 3. The distribution of recorded neurons corresponds well to the optically identified respiratory-related area. (A and B) On the left, the locations of recorded neurons (from 6 preparations) are shown on a schematic drawing, and on the right, optically identified respiratory-related areas are colored depending on the activation timing as same way as Fig. 2 (from a representative single preparation). V, VII, VIII, IX/X and XII: cranial nerves.

Y. Oku et al. / Respiratory Physiology & Neurobiology 161 (2008) 281–290

386–952 ms vs. stained 1067 ± 120 ms, range 760–1448 ms, statistical tests were not performed because of insufficient sample numbers) in stained preparations. The buccal dilator activities preceding the lung burst were recorded in all preparations, however, fast buccal bursts (buccal ventilation) were not detected or weakened after staining except for only one preparation. The amplitude of buccal ventilation was too small as a trigger signal for optical recordings. In the preparation with obvious buccal ventilation, in spite of averaging 30 times triggered by the buccal burst, no significant optical signal related to the buccal bursts was detected on the ventral surface of the medulla. 3.2. Extracellular recording of neuronal activity To verify the optically identified distribution of respiratoryrelated neurons, we next conducted extracellular recordings mainly in the identified areas. We recorded a total of 64 respiratory neurons with various firing patterns from 6 unstained preparations. The distributions of respiratory neurons are shown in Fig. 3. There was no difference in the location of respiratory neurons depending on the weight of frogs. It should be noted, however, that the surveyed area was biased, since we surveyed respiratory neurons in the vicinity of the optically identified respiratory-related areas. These neurons were recorded at the depth of approximately 100–500 ␮m from the ventral surface. There were characteristic distributions depending on types of respiratory neurons. Frog respiratory phases are very complicated and the correspondence to respiratory phases in mammals is uncertain. Therefore, we classified recorded respiratory neurons into three categories in relation to the two ventilation modes: buccal (B) neurons fired during buccal ventilation, lung (L) neurons fired during lung ventilation, and non-lung (non-L) neurons did not fire during lung ventilation. Among 64 respiratory neurons recorded, only four B neurons were recorded from the midlevel between CN X and XII. All the B neurons also fired during the lung burst phase (Fig. 4). Further, L neurons were categorized into three types based on their firing characteristics: early-L neurons

285

started firing at the beginning of lung ventilation but stopped firing in the mid-phase of CN V activity, throughout-L neurons fired during most of the lung burst phase, and late-L neurons fired during the latter half of lung burst (Fig. 5). There were 13 early-L neurons, 34 throughout-L neurons and 10 late-L neurons. Early-L neurons were distributed between the levels of CN X and XII rootlets densely in two separated areas, and one area was caudal to the CN X rootlet and the other was rostral to the CN XII rootlet. Throughout-L neurons were densely distributed between the levels of CN X and XII rootlets. Late-L neurons were widely distributed between the levels of CN VII and XII rootlets. The distributions of early-L, throughoutL and late-L neurons were intermingled in the area between the levels of CN X and XII rootlets. However, the distribution of respiratory neurons in the area rostral to the CN X rootlet was sparse, and all the neurons recorded were exclusively late-L. Non-L neurons were further categorized into following two types. Biphasic non-L neurons fired during the pre- and post-lung phases but never fired during the mid-lung burst phase, and thus resembled to ‘N-type neurons’ as previously reported (Kogo and Remmers, 1994; McLean and Remmers, 1997). A tonic non-L neuron fired throughout the interval between lung ventilation but was suppressed during lung burst, and this type was not previously reported in frogs. Two biphasic N neurons were recorded from the mid-level between CN X and XII. The tonic non-L neuron was recorded from just rostral to the CN X exit. 3.3. Histological examination To further clarify the localization of respiratory neurons, we demarcated respiratory-related areas estimated by the combination of optical and extracellular recordings on transverse sections of the frog brainstem (Fig. 6). Similarly to mammals (Ezure, 2004), the location of respiratory neurons did not correspond to any known nucleus; respiratory neurons were distributed bilaterally in the reticular formation of the ventrolateral medulla as longitudinal columns in the frog.

Fig. 4. Extracellular recordings of a lung neuron (A) and a buccal neuron (B) together with trigeminal (V) nerve recordings. Buccal (B) neurons fire during both lung and buccal ventilations, whereas lung (L) neurons fire only during lung ventilation. In each panel, the upper trace shows window-discriminated unitary pulses, the middle trace is raw extracellular recording, and the lower trace is CN V activity. Note that in the panel B, spikes with large amplitude are activities of L neurons, which are excluded from unitary pulses by the window discriminator software.

286

Y. Oku et al. / Respiratory Physiology & Neurobiology 161 (2008) 281–290

Fig. 5. Various types of frog respiratory neurons are exemplified. Early-L neurons start firing at the beginning of lung ventilation but stop firing in the mid-phase of CN V activity, throughout-L neurons fire during most of the lung burst phase, and late-L neurons fire during the latter half of lung burst. Biphasic non-L neurons fire during preand post-lung phases but never fire during mid-lung burst phase, and tonic non-L neurons fire throughout the interval between lung ventilations but are suppressed during lung burst. In each panel, the upper trace shows window-discriminated unitary pulses, the middle trace is raw extracellular recording, and the lower trace is CN V activity.

3.4. Blockade of glycinergic transmission We then tested whether such spatiotemporal organization of respiratory neuronal activities was disturbed by a glycine receptor antagonist strychnine. Strychnine (1–5 ␮M) did not change the frequency of lung bursts (2.9 ± 0.7 burst/min vs. 3.2 ± 1.0 burst/min, n = 6), but augmented the amplitude, shortened the duration of CN V bursts (1067 ± 120 ms vs. 527 ± 45 ms, p = 0.0035), and altered the pattern from augmenting to decrementing (Fig. 7). As exemplified in Fig. 8, strychnine disrupted the spatiotemporal organization of the respiratory neuronal network in three of six preparations, although rhythmic CN V bursts persisted after the application of strychnine. Longitudinal columnar arrangements of respiratory-related areas disappeared, and the burst-related pixels were widely distributed in the ventral medulla by strychnine. Furthermore, the time difference in the activation timings almost disappeared. In the other three preparations, longitudinal columnar arrangements were preserved after the application of strychnine. 4. Discussion We identified the whole distribution of lung ventilation-related neurons in the vicinity of the ventral surface of the frog brainstem and evaluated the spatiotemporal activation profiles of these neurons by voltage imaging and cross-correlation analysis. Optical imaging did not reveal distinct rhythm generating areas with different activation timings, but succeeded to identify respiratoryrelated areas involved in the motor pattern formation associated with lung ventilation. Similarly to mammals, respiratory neurons were distributed bilaterally in the reticular formation of the ventrolateral medulla as longitudinal columns in the frog. Further, strychnine disrupted the spatiotemporal organization of

respiratory-related optical activities, supporting the idea that the motor pattern formation during lung ventilation depends on glycinergic synaptic transmission. 4.1. Limitations of optical imaging In neonatal rats, we estimate that our method can detect neuronal activities up to 500 ␮m deep from the surface based on dye penetration and oxygen partial pressure gradient (Okada et al., 1993a, 2007). We consider that the failure to detect any optical signals associated with rhythm generation in the present study may be attributed to this limitation of the optical technique. Namely, the activity of either buccal or lung rhythm generator was not detected, possibly because the areas responsible for producing them are too deep to be seen from the surface. Staining substantially affected the buccal and lung ventilations. Dye staining lowered the frequency of the lung bursts, prolonged the duration of the lung bursts, and often abolished the buccal bursts. This suggests a kind of ‘toxic’ effect of the dye. Toxic effects of di-2-ANEPEQ on respiratory rhythm generation have been described in rats (Onimaru and Homma, 2003). The buccal activity seen in the optical recording may be an enhanced activity associated specifically with the onset of a lung breath. Although the mechanism of the toxic effect has not been clarified, there is a possibility that the lung oscillator was uncoupled from the buccal oscillator, which was suppressed after staining with a dye. This suggests that lung ventilation (both the buccal dilation and compression) does not require the buccal oscillator although in more intact preparations the two may act in a coordinated fashion. Interestingly, a caudal area in the ventral medulla, more caudal than the ‘buccal area’ described by others (Wilson et al., 2002; Vasilakos et al., 2005), remained strongly active during the fictive lung ventilation in the present study,

Y. Oku et al. / Respiratory Physiology & Neurobiology 161 (2008) 281–290

287

Fig. 6. Locations of respiratory-related neurons, estimated by optical and extracellular recordings, are indicated on transverse sections of the frog brainstem as dotted circles. nVI: Nucleus nervi abducentis, Ols: Oliva superior, nVIIId: Nucleus dorsalis nervi octavi, nIXm: Nucleus motorius nervi glossopharyngei, nXm: Nucleus motorius nervi vagi, nXII: Nucleus nervi hypoglossi, VI: Nervus abducens, IX: Nervus glossopharyngeus, X: Nervus vagus, fas sol: Fasciculus solitarius. Lines A–D correspond to the levels of transverse sections A–D, respectively. V, VII/VIII, IX/X and XII: cranial nerves.

irrespective of the presence or the absence of the buccal ventilation. There are additional important limitations of the optical method. First, it does not distinguish neuronal types among propriobulbar, premotor and motor neurons. Second, optical signals

do not reflect activity of a single neuron but summated activity of multiple neurons of possibly heterogeneous populations. Opti¨ cal signals could even originate from astrocytes (Hulsmann et al., 2003; Oku et al., 2007). Third, since the fluorescent signal (F/F) associated with spontaneous neuronal activity is very weak, the

Fig. 7. Strychnine altered the pattern of lung ventilation from augmenting to decrementing. (A) Control. (B) After 5 ␮M strychnine application. Raw (upper panel) and integrated (lower panel) CN V motor activities during lung ventilation are shown.

288

Y. Oku et al. / Respiratory Physiology & Neurobiology 161 (2008) 281–290

Fig. 8. Strychnine radically disrupted spatiotemporal organization of these respiratory neurons. (A and B) Correlation images of the control. See the legend of Fig. 2A and B for detailed explanation. (C and D) Correlation images after the application of 5 ␮M strychnine. VII/VIII, IX/X and XII: cranial nerves.

sampling rate is limited to a relatively low range (20 ms/frame in the present study), and cycle-triggered averaging is often necessary to increase signal-to-noise ratio. These limitations could obscure a brief pre-inspiratory signal or a signal related to buccal ventilation. 4.2. Distribution and activation sequence of lung ventilation-related neurons In our previous study we reported that the combination of voltage imaging and cross-correlation analysis is a powerful tool to map the distribution of respiratory neurons in neonatal rat brainstem spinal cord preparations (Okada et al., 2007; Oku et al., 2007). Applying the same technique to the brainstem spinal cord preparation of Rana catesbeiana, we succeeded for the first time to elucidate the distribution of frog respiratory neurons visible from the ventral surface of the brainstem. We have previously shown that in P0–P1 rats, two distinct areas with different activation profiles that represent the activities of the two respiratory rhythm generators, the para-facial respiratory group (pFRG) (Onimaru et al., 1988; Onimaru and Homma, 2003) and the pre¨ ¨ Botzinger complex (preBotC) (Smith et al., 1991), can be identified by a clustering method (Oku et al., 2007). However, activation profiles of respiratory neurons in the adult frog did not consist of such distinct clusters. Further, we did not detect any optical activity that preceded the lung burst. We therefore assume that optical signals do not reflect the activity of rhythm generators, but detect that of down stream events associated with the pattern formation of motor outputs. We do not know why the same technique reveals rhythm-generating sites in rats but not in frogs. It could be simply because the rhythm generator sites are located close to the surface in the rat but too

deep in the frog to detect by optical imaging (also see Section 4.1). Frog respiratory neurons associated with lung ventilation were localized bilaterally in the ventrolateral medulla as longitudinal columns between the levels of CN V and XII rootlets. Previously, frog respiratory neurons have been recorded from the mid-rostral brainstem between CN VII and X rootlets (Kogo and Remmers, 1994; McLean and Remmers, 1997). However, the aggregate of frog respiratory neurons optically identified in the present study were much more densely distributed in the caudal area of the ventrolateral medulla at the levels between CN X and XII rootlets. Indeed, we found a large population of respiratory neurons in the region 500–800 ␮m rostral to the CN XII rootlet and approximately 100–500 ␮m deep by extracellular recording study (Fig. 3). Thus the future exploration of respiratory-related neurons should include this area. In addition, we note that optically identified rostral respiratory-related area is located more laterally to the proposed ‘lung area’, and is distributed as a thinner column continuing to the caudal respiratory-related column (compare Figs. 2 and 3 of the present report and Fig. 9 of Vasilakos et al., 2005). Previous studies had already located groups of respiratory neurons either by microinjections of neuromediators (McLean et al., 1995b; Wilson et al., 2002) or transections of the brainstem (Torgerson et al., 2001; Wilson et al., 2002). All these studies found a region that seemed essential for lung ventilation, which was located between the levels of CN VIII and X. Furthermore, the results of the transection studies showed that this region could itself produce the lung rhythm. Therefore, the present results showing that most of the respiratory neurons associated with lung ventilation were located more caudally, between the levels of CN X and XII, seem a priori contradictory. Further, the area where the optical signal first appears is caudal to the ‘buccal area’ defined by previous

Y. Oku et al. / Respiratory Physiology & Neurobiology 161 (2008) 281–290

workers (Wilson et al., 2002); it is closer to the CN XII rootlet than to the CN X rootlet. Since the present results were obtained from frogs of a wide spectrum of body weights and thus a wide spectrum of developmental stages, the size of brainstem or the stage of development of the animals could explain these discrepancies. However, we think that these are not the case because there was no difference in the location of respiratory neurons depending on the weight of frogs in either optical measurements or extracellular recordings. Further, we confirmed in the preliminary study that the sizes of the brainstem used in the present study were almost the same, in spite of the large range of body weight (see Section 2). We suggest that these apparent discrepancies are reconciled if optical signals represent motoneuronal activities that are associated with pattern formation rather than activity of rhythm generation. The initial optical activity does not correspond to the buccal dilation phase (step 1), but mostly compatible with the buccal compression phase, because the neural activity associated with buccal dilation begins before and peaks just as the optical signal emerges. However, we think that the earliest depolarization in this area is a reflection of a dense distribution of early-L neurons that are active during the buccal dilation phase. As we discussed in the previous section, optical signals reflect summated neuronal activities of multiple neurons with possibly different activity patterns. Besides early-L neurons, there are throughout-L neurons and late-L neurons in the area where the optical signal first appears (Fig. 3), and the firing frequency of early-L neurons is lower than that of throughout-L neurons or late-L neurons (Fig. 5). Therefore, optical signals from early-L neurons would be smaller than those from throughout-L neurons and late-L neurons, and the onset of early-L neuron activity might not be detected by optical measurement. The more rostral cluster of early-L neurons near CN IX/X corresponds to a cold hue area (Fig. 3), suggesting that neuronal populations as a whole in this region depolarize later. It is possible that we missed recording of late-L neurons due to insufficient number of recorded neurons. 4.3. Neuronal correlates of buccal ventilation By triggering with the fast buccal bursts, optically active area related to the buccal bursts was not detected on the ventral surface of the medulla. The reason why we were not able to detect optical signals related to the buccal burst is unknown. As discussed in Section 4.1, the area responsible for the buccal fast oscillations may be too deep or may emit too weak signals to detect by optical imaging. However, we found an area, slightly caudal to the ‘buccal area’, where a small population of B neurons was concentrated by extracellular recordings with the aid of optical identification of the respiratory neuronal localization. Those B neurons may be responsible for buccal rhythm generation or buccal motor outputs. Only four of 64 neurons fired during buccal ventilation. However, in the extracellular recordings, it is unknown whether the membrane potential of the neuron has the sub-threshold buccal modulation. With this regard, our classification is different from that used in the previous intracellular recording study (Kogo and Remmers, 1994). There is a possibility that the membrane potential of many L neurons is modulated by the buccal activity. 4.4. Comparison between respiratory rhythm generators of frogs and rats It is interesting that mammals also have two respiratory rhythm ¨ (Smith et al., 1991; Onimaru generators, the pFRG and the preBotC and Homma, 2003). These oscillators may have originated from

289

those for the gill and lung of the earliest air breathers (Wilson et al., 2006). The locations of the two oscillators have been well identified in rats (for review see Feldman and Del Negro, 2006). In frogs, opioid agonist for ␮-receptor selectively reduces the frequency of lung ventilation (Vasilakos et al., 2006), whereas in rats, it depresses ¨ (Gray et al., 1999). On the basis of this similarity, it is conpreBotC sidered that the frog ‘buccal area’ corresponds to the rat pFRG, and ¨ the ‘lung area’ to the preBotC. Strangely, the topographic relations between the two oscillators in frogs and rats are reversed along the rostro-caudal axis. In the present study, we observed the ramp of activity occurs caudal to ‘buccal area’, which possibly correlated with both buccal dilation and buccal compression. This activity may be homologous to the inspiratory and expiratory pattern formation activities of the rostral and caudal ventral respiratory groups (Ezure, 2004). We could not identify activities of buccal and lung oscillators, and thus the apparent reversal of rhythm generator sites between frogs and rats remains to be elucidated. 4.5. Role of glycinergic neurotransmission on spatiotemporal organization of respiratory neuronal network Strychnine abolishes the gill bursting activity of tadpoles while lung-like bursting activity remains (Galante et al., 1996; Broch et al., 2002). The lung burst frequency of tadopoles increases in response to strychnine or bicuculline (Galante et al., 1996; Broch et al., 2002), whereas similar concentration of these agents decreases dose-dependently lung burst frequency in adult bullfrog preparations (Broch et al., 2002). These findings lead to the idea that the lung ventilation of tadpoles is driven by a pacemaker mechanism and the gill ventilation is driven by a neuronal network mechanism (Galante et al., 1996), but the lung ventilation of adult frogs may shift to a network-dependent mechanism (Broch et al., 2002). A combination of strychnine and bicuculline or chloridefree buffer does not abolish the lung bursts in tadpoles (Galante et al., 1996), but chloride-free buffer stops the lung bursts in adult bullfrogs and grass-frogs (Broch et al., 2002). Microinjection of GABA into the ventrolateral medulla of the juvenile and adult Rana catesbeiana in the rostral level between CN VIII and IX stops the lung rhythm and that in the caudal level of CN X suppresses the buccal rhythm, respectively (McLean et al., 1995b; Wilson et al., 2002). These results suggest that respiratory rhythmogenesis in adult frogs requires chloride-dependent inhibitory mechanisms. A common finding in these studies is that strychnine changes the augmenting pattern of the lung burst to the decrementing pattern (Kimura et al., 1997; Broch et al., 2002). Importantly, blockade of the glycinergic inhibitory synaptic mechanisms by strychnine eliminates the alternatively active patterns in the buccal dilating and compressing nerve activities in the in vitro brainstem preparation of adult frogs (Kimura et al., 1997). Under the blockade of glycinergic inhibitory synaptic mechanisms by strychnine, all motoneurons are synchronously activated, and thus motor nerves that work antagonistically in a normal condition fire synchronously, and eventually the frog is unable to ventilate (Kimura et al., 1997). These results indicate that the neural structure responsible for the pattern formation of the respiratory activity involves strychninesensitive inhibitory mechanisms. We thought if the optically identified respiratory-related neural structure is essential for generating the normal ventilatory motor pattern, then the spatiotemporal pattern of the neuronal activities must be strongly affected by elimination of glycinergic postsynaptic inhibition with strychnine. Strychnine dramatically disrupted the spatiotemporal organization of respiratory neuronal activities. Optically active area widely spread on the whole ventral brainstem, and the spatiotemporal activation sequence almost

290

Y. Oku et al. / Respiratory Physiology & Neurobiology 161 (2008) 281–290

disappeared. The latter phenomenon is consistent with the observation that strychnine changes the pattern of respiratory neuronal activation from reciprocal to synchronous (Kimura et al., 1997). These results suggest that the glycinergic inhibitory neurotransmission is essential for the respiratory motor pattern formation, but is not required for the rhythm generation. However, it is possible that the mechanism of rhythm generation is altered by blockade of glycinergic inputs, from a network-driven mechanism dependent on postsynaptic chloride-mediated inhibition to a pacemaker-driven mechanism that basically depends on endogenous membrane properties of respiratory neurons. The effects of glycine receptor blockades on the generation of respiratory rhythm and motor pattern in the frog are also similar to those in mammals. Blockade of glycine receptor does not abolish respiratory rhythm in brainstem spinal cord preparations (Onimaru et al., 1990) and ‘breathing’ slice preparations (Shao and Feldman, 1997) of neonatal rats, but suppresses respiratory rhythm in adult cats in vivo (Pierrefiche et al., 1998). Loss of glycinergic inhibition disrupts normal respiratory motor pattern but does not stop the rhythm in the working heart-brainstem preparation from adult ¨ mice (Busselberg et al., 2001). In conclusion, frog respiratory neurons are distributed as columnar aggregates bilaterally in the reticular formation of the ventrolateral medulla between the levels of CN V and XII rootlets. Strychnine radically disrupts spatiotemporal organization of the optically measured respiratory neuronal activities. These results suggest that the optically identified area involves glycinergic neurotransmission and plays an important role in generating the motor pattern during lung ventilation. The present study also indicates that the combination of optical imaging and cross-correlation analysis has a great advantage in analyzing dynamic configuration of complicated neuronal networks. Acknowledgement We thank M. Ito for technical assistance. References Broch, L., Morales, R.D., Sandoval, A.V., Hedrick, M.S., 2002. Regulation of the respiratory central pattern generator by chloride-dependent inhibition during development in the bullfrog (Rana catesbeiana). J. Exp. Biol. 205, 1161–1169. ¨ Busselberg, D., Bischoff, A.M., Paton, J.F., Richter, D.W., 2001. Reorganisation of res¨ piratory network activity after loss of glycinergic inhibition. Pflugers Arch. 441, 444–449. de Jongh, H.J., Gans, C., 1969. On the mechanisms of respiration in the bullfrog, Rana catesbeiana: a reassessment. J. Morphol. 127, 259–290. Ezure, K., 2004. Reflections on respiratory rhythm generation. Prog. Brain Res. 143, 67–74. Feldman, J.L., Del Negro, C.A., 2006. Looking for inspiration: new perspectives on respiratory rhythm. Nat. Rev. Neurosci. 7, 232–242. Fukuda, K., Okada, Y., Yoshida, H., Aoyama, R., Nakamura, M., Chiba, K., Toyama, Y., 2006. Ischemia-induced disturbance of neural network function in the rat spinal cord analyzed by voltage-imaging. Neuroscience 140, 1453–1465. Galante, R.J., Kubin, L., Fishman, A.P., Pack, A.I., 1996. Role of chloride-mediated inhibition in respiratory rhythmogenesis in an in vitro brainstem of tadpole, Rana catesbeiana. J. Physiol. 492, 545–558. Gray, P.A., Rekling, J.C., Bocchiaro, C.M., Feldman, J.L., 1999. Modulation of respiratory ¨ frequency by peptidergic input to rhythmogenic neurons in the preBotzinger complex. Science 286, 1566–1568.

¨ Hulsmann, S., Straub, H., Richter, D.W., Speckmann, E.J., 2003. Blockade of astrocytes causes stimulation-induced depolarization as revealed by voltage sensitive dyes in mouse brainstem slices. Exp. Brain. Res. 150, 117–121. Kemali, M., Braitenberg, V., 1969. Atlas of the Frog’s Brain. Springer, Berlin. Kimura, N., Perry, S.F., Remmers, J.E., 1997. Strychnine eliminates reciprocation and augmentation of respiratory bursts of the in vitro frog brainstem. Neurosci. Lett. 225, 9–12. Kogo, N., Remmers, J.E., 1994. Neural organization of the ventilatory activity in the frog, Rana catesbeiana. II. J. Neurobiol. 25, 1080–1094. McLean, H.A., Kimura, N., Kogo, N., Perry, S.F., Remmers, J.E., 1995a. Fictive respiratory rhythm in the isolated brainstem of frogs. J. Comp. Physiol. [A] 176, 703–713. McLean, H.A., Perry, S.F., Remmers, J.E., 1995b. Two regions in the isolated brainstem of the frog that modulate respiratory-related activity. J. Comp. Physiol. [A] 177, 135–144. McLean, H.A., Remmers, J.E., 1997. Characterization of respiratory-related neurons in the isolated brainstem of the frog. J. Comp. Physiol. [A] 181, 153– 159. ¨ Okada, Y., Muckenhoff, K., Holtermann, G., Acker, H., Scheid, P., 1993a. Depth profiles of pH and PO2 in the isolated brain stem-spinal cord of the neonatal rat. Respir. Physiol. 93, 315–326. ¨ Okada, Y., Muckenhoff, K., Scheid, P., 1993b. Hypercapnia and medullary neurons in the isolated brain stem-spinal cord of the rat. Respir. Physiol. 93, 327– 336. Okada, Y., Masumiya, H., Tamura, Y., Oku, Y., 2007. Respiratory and metabolic acidosis differentially affect the respiratory neuronal network in the ventral medulla of neonatal rats. Eur. J. Neurosci. 26, 2834–2843. Oku, Y., Masumiya, H., Okada, Y., 2007. Postnatal developmental changes in activation profiles of the respiratory neuronal network in the rat ventral medulla. J. Physiol. 585, 175–186. Onimaru, H., Arata, A., Homma, I., 1988. Primary respiratory rhythm generator in the medulla of brainstem-spinal cord preparation from newborn rat. Brain Res. 445, 314–324. Onimaru, H., Arata, A., Homma, I., 1990. Inhibitory synaptic inputs to the respiratory ¨ rhythm generator in the medulla isolated from newborn rats. Pflugers Arch. 417, 425–432. Onimaru, H., Homma, I., 2003. A novel functional neuron group for respiratory rhythm generation in the ventral medulla. J. Neurosci. 23, 1478–1486. Pierrefiche, O., Schwarzacher, S.W., Bischoff, A.M., Richter, D.W., 1998. Blockade of ¨ synaptic inhibition within the pre-Botzinger complex in the cat suppresses respiratory rhythm generation in vivo. J. Physiol. 509, 245–254. Pogosyan, V.I., Arutyunyan, T.L., Aglintsyan, T.S., Danielyanand, M.A., Fanardjian, V.V., 2002. Morphological study of spatial distribution of vestibulospinal neurons in the frog Rana ridibunda. J. Evol. Biochem. Physiol. 38, 773–780. Shao, X.M., Feldman, J.L., 1997. Respiratory rhythm generation and synaptic inhi¨ bition of expiratory neurons in pre-Botzinger complex: differential roles of glycinergic and GABAergic neural transmission. J. Neurophysiol. 77, 1853– 1860. Smith, J.C., Ellenberger, H.H., Ballanyi, K., Richter, D.W., Feldman, J.L., 1991. Pre¨ Botzinger complex: a brainstem region that may generate respiratory rhythm in mammals. Science 254, 726–729. Torgerson, C.S., Gdovin, M.J., Remmers, J.E., 1997. Ontogeny of central chemoreception during fictive gill and lung ventilation in an in vitro braistem preparation of Rana catesbeiana. J. Exp. Biol. 200, 2063–2072. Torgerson, C.S., Gdovin, M.J., Remmers, J.E., 1998. Fictive gill and lung ventilation in the pre-and postmetamorphic tadopole brain stem. J. Neurophysiol. 80, 2015–2022. Torgerson, C.S., Gdovin, M.J., Remmers, J.E., 2001. Sites of respiratory rhythmogenesis during development in the tadpole. Am. J. Physiol. Regul. Integr. Comp. Physiol. 280, R913–R920. Vasilakos, K., Wilson, R.J., Kimura, N., Remmers, J.E., 2005. Ancient gill and lung oscillators may generate the respiratory rhythm of frogs and rats. J. Neurobiol. 62, 369–385. Vasilakos, K., Kimura, N., Wilson, R.J., Remmers, J.E., 2006. Lung and buccal ventilation in the frog: uncoupling coupled oscillators. Physiol. Biochem. Zool. 79, 1010–1018. Wilson, R.J., Vasilakos, K., Harris, M.B., Straus, C., Remmers, J.E., 2002. Evidence that ventilatory rhythmogenesis in the frog involves two distinct neuronal oscillators. J. Physiol. 540, 557–570. Wilson, R.J., Vasilakos, K., Remmers, J.E., 2006. Phylogeny of vertebrate respiratory rhythm generators: the oscillator homology hypothesis. Respir. Physiol. Neurobiol. 154, 47–60.