Journal of Neurophysiology

Isolation of the Kernel for Respiratory Rhythm Generation in a Novel Preparation: The Pre-Bötzinger Complex “Island”

Shereé M. Johnson, Naohiro Koshiya, Jeffrey C. Smith


The pre-Bötzinger complex (pre-BötC), a bilaterally distributed network of rhythmogenic neurons within the ventrolateral medulla, has been proposed to be the critical locus for respiratory rhythm generation in mammals. To date, thin transverse medullary slice preparations that capture the pre-BötC have served as the optimal experimental model to study the region's inherent cellular and network properties. We have reduced the thin slices to isolated pre-BötC “islands” to further establish whether the pre-BötC has intrinsic rhythmicity and is the kernel for rhythmogenesis in the slice. We recorded neuron population activity locally in the pre-BötC with macroelectrodes and fluorescent imaging of Ca2+ activities with Calcium Green-1AM dye before and after excising the island. The isolated island remained rhythmically active with a population burst profile similar to the inspiratory burst in the slice. Rhythmic population activity persisted in islands after block of GABAAergic and glycinergic synaptic inhibition. The loci of pre-BötC Ca2+ activity imaged in thin slices and islands were similar, and imaged pre-BötC neurons exhibited synchronized flashing after blocking synaptic inhibition. Population burst frequency increased monotonically as extracellular potassium concentration was elevated, consistent with mathematical models consisting entirely of an excitatory network of synaptically coupled pacemaker neurons with heterogeneous, voltage-dependent bursting properties. Our results provide further evidence for a rhythmogenic kernel in the pre-BötC in vitro and demonstrate that the islands are ideal preparations for studying the kernel's intrinsic properties.


The rhythm of breathing in mammals has been proposed to originate in the pre-Bötzinger complex (pre-BötC), a functionally and structurally defined subregion of the ventrolateral medulla (Rekling and Feldman 1998; Smith et al. 1991, 2000). This region is hypothesized to contain a rhythmogenic kernel, consisting of a bilaterally distributed network of synaptically coupled excitatory neurons with voltage-dependent bursting pacemaker properties (Butera et al. 1999a,b; Koshiya and Smith 1999; Smith et al. 1991). Thin in vitro slice preparations (Koshiya and Smith 1999; Smith et al. 1991), which optimally expose the pre-BötC and retain a rhythmically active respiratory network have been developed to study cellular and network mechanisms of rhythm generation. Evidence for an excitatory pre-BötC pacemaker network was recently obtained by optical imaging (Koshiya and Smith 1999), and models (Butera et al. 1999a,b) have been proposed to explain cellular pacemaker and network mechanisms of rhythmogenesis.

A fundamental hypothesis of these previous studies is that the local pre-BötC network is intrinsically rhythmic due to endogenous bursting properties of the pacemaker cells in the network. Models of this pacemaker network postulate that the oscillation frequency can be controlled by tonic excitation due to the pacemaker cells' voltage-dependent oscillatory bursting properties (Butera et al. 1999a,b; Koshiya and Smith 1999). In this study, we have further addressed these hypotheses. We excised the pre-BötC from one side of thin slices to produce a pre-BötC “island” from which we could record local population activity and determine intrinsic rhythmicity of the network in the isolated kernel. We blocked synaptic inhibition in the island to verify that rhythmogenesis is maintained as predicted for an excitatory pacemaker-network, and we imaged population activity to verify that the locus of activity was identical in thin slices and islands. We also examined frequency control of pre-BötC population activity by varying tonic excitation with changes in extracellular K+ concentration ([K+]o) and obtained results consistent with model predications. Our results provide further evidence for a pre-BötC pacemaker-network and show that islands are ideal preparations for analyzing the kernel's intrinsic rhythmogenic properties. A preliminary report has been presented in abstract form (Johnson et al. 2000).


Thin slice and island preparations

We obtained thin transverse slice preparations (300–350 μm) from Sprague-Dawley neonatal rats (P0–P3) as previously described (Koshiya and Smith 1999; Smith et al. 1991). Slices were pinned down and perfused in a recording chamber with artificial cerebrospinal fluid containing (in mM) 124 NaCl, 25 NaHCO3, 3 KCl, 1.5 CaCl2, 1.0 MgSO4, 0.5 NaH2PO4, and 30 d-glucose equilibrated with 95% O2-5% CO2 at 27°C (pH = 7.4). Simultaneous extracellular recordings of hypoglossal (XII) nerve and local pre-BötC population activity (Thoby-Brisson et al. 2000) were obtained with suction electrodes (Fig.1; following text). [K+]o was elevated (8–9 mM) to maintain rhythmic respiratory network activity. After mapping the locus of pre-BötC population activity, we excised the “island” from one side of the slice (Figs. 1 and 3). In some islands, we varied [K+]o(3–18 mM) to modulate the spontaneous discharge frequency. We blocked GABAAergic and glycinergic inhibition in some islands via bath application of bicuculline methobromide (20 μM; Sigma) and strychnine (5 μM; Sigma) and in some cases, obtained population frequency versus [K+]o relationships. In other experiments, we also imaged Ca2+ activity (see following text) of inspiratory neurons in both the slice and island to compare loci of population activity.

Fig. 1.

A, bottom: schematic drawing of dual XII-pre Bötzinger complex (pre-BötC) recording configuration showing island structural borders (shaded region) and raw data from thin slice. Labeled medullary structures: NTS, nucleus tractus solitarius; SP 5, spinal trigeminal tract; XII, XIIn, hypoglossal motor nucleus and nerve; IO, inferior olive. Top: superimposed cycle-triggered averaged traces (60 cycles, arbitrary units) of inspiratory discharge from simultaneous XII and pre-BötC recordings. Inset: onset of local pre-BötC activity preceding XII discharge.B: schematic of excised pre-BötC island with integrated traces showing examples of population discharge from 2 islands [20 μM bicuculline (Bic) and 5 μM strychnine (Strych) at 10 mM [K+]o]. Top trace: spontaneous activity with amplitude fluctuations;bottom trace: example with more stable activity.C: superimposed traces (average of 20 cycles, arbitrary units) illustrating similar temporal profiles of pre-BötC activity in thin slice and island.

Electrophysiological recording and signal analysis

We recorded inspiratory XII motor discharge with suction electrodes applied to ventral roots. Pre-BötC population recordings were obtained with a glass pipette (100 μm ID) applied to the slice surface that exposed the pre-BötC. The suction electrode produced a small deformation in the slice, marking the recording site for recording in the island. Signals were amplified (50,000–100,000 times, CyberAmp 380, Axon Instruments), band-pass filtered (0.3–2 kHz), and digitized (4 kHz). Signals were rectified, smoothed by an analogue leaky integrator (τ = 20 ms), and analyzed off-line with automated algorithms (PowerLab with Chart software, AD Instruments). Statistical significance was determined by a Student's paired t-test on mean data. Data are presented as means ± SE unless indicated otherwise.

Calcium imaging of pre-BötC activity

Methods for Ca2+ imaging have been described in detail elsewhere (Koshiya and Smith 1999). Briefly, Calcium Green-1 AM (CaG; Molecular Probes) was microinjected into the slice near the midline (see Fig. 3) to retrogradely label pre-BötC neurons overnight (8–12 h). Tissue structure and CaG fluorescence were visualized with an upright videomicroscope (Zeiss, Axioskop FS1) system incorporating infrared differential interference contrast optics (IR-DIC), extended IR Newvicon camera (Hamamatsu), and a fluorescence image intensifying camera (ICCD-1000F, Videoscope International). Activity-related changes in CaG fluorescence intensity (F) were expressed as %ΔF/F. Calcium activity images were obtained by subtracting baseline images from images acquired during peak activity. Time courses of ΔF/F were obtained by image processing (Koshiya and Smith 1999).


Pre-BötC population activity

We used simultaneous XII and local pre-BötC recordings to examine inspiratory population activity prior to excising the island. Under all conditions, pre-BötC activity preceded XII inspiratory motor output by ∼100–300 ms (Fig. 1 A), consistent with the proposal that inspiratory drive originates in the pre-BötC. All islands (n = 21) continued to spontaneously generate rhythmic activity under control conditions ([K+]o = 9 mM), although there were differences between slices and islands in the population burst frequency and amplitude. The mean population burst frequency increased by 38% [mean frequency of 0.21 ± 0.02 Hz (slice) and 0.29 ± 0.04 Hz (islands), P < 0.05,n = 5 pairs]. Island population burst amplitude decreased (39.4 ± 13%, n = 8) with variable changes in burst duration. The temporal profile of the pre-BötC population burst was similar in slices and islands (Fig. 1 C) despite changes in burst amplitude. The rhythm also became less regular; i.e., the coefficient of variation of burst frequency increased from 0.236 (slices) to 0.458 (islands; n = 5 pairs, P < 0.05).

Modulation of island activity with [K+]o

Models of the pacemaker kernel predict that population burst frequency will increase monotonically with a decrease in population burst amplitude as tonic excitation is elevated by parameters such as [K+]o (Butera et al. 1999b; Del Negro 2001). To examine this, we incremented [K+]o (7–18 mM) and found a monotonic increase in burst frequency and decrease in amplitude (Fig. 2, A and C). Burst frequency increased from 0.18 ± 0.05 Hz at 8 mM [K+]o to 0.78 ± 0.06 Hz at 17 mM [K+]o(Fig. 2 C, n = 9).

Fig. 2.

Modulation of island population activity with [K+]o. A andB: cycle-triggered averaged traces (20 cycles) of pre-BötC population burst from 2 representative islands over a range of [K+]o (7–15 mM in A; 5–15 mM in B) without (A) and with (B) block of GABAAergic and glycinergic inhibition [(20 μM bicuculline (Bic), 5 μM strychnine (Strych)]. Average burst amplitude decreased progressively with increasing [K+]o.C: monotonic [K+]o vs. burst frequency relationships for islands without (n = 9) and with (n = 2) block of inhibition.

Island activity after block of synaptic inhibition

Pre-BötC rhythmic activity persisted in islands following block of GABAAergic and glycinergic inhibition (20 μM bicuculline, 5μM strychnine; n = 5, Figs.1 B and 2, B and C). Mean burst amplitude increased by 41% (n = 5 islands;P < 0.05) and burst frequency decreased by 32% (from 0.30 ± 0.05 to 0.20 ± 0.03 Hz, at 10 mM [K+]o; P< 0.05, n = 5). In two islands [K+]o was varied (3–15 mM); population burst frequency increased monotonically (Fig.2 C) and burst amplitude decreased (Fig. 2 B), similar to islands without inhibitory block. After block of inhibition, the range of [K+]o over which the islands exhibited rhythmic population activity was displaced to lower [K+]o and stable bursting occurred at the lower end at 4–5 mM [K+]o. This suggests that there is ongoing endogenous tonic inhibition in the pre-BötC regulating excitability of the kernel that accounts for the higher [K+]o required to maintain rhythmic activity in the island and slice without inhibitory block. Burst frequency would theoretically be controlled by the balance of tonic excitatory and inhibitory synaptic inputs to the kernel (Butera et al. 1999b).

Rhythmic calcium activity in pre-BötC neurons

Pre-BötC neurons were retrogradely labeled with CaG (Koshiya and Smith 1999) and pre-BötC Ca2+ activity was imaged in slices and islands (n = 7). CaG fluorescence intensity increased during inspiratory activity in slices (Fig. 3,A and C) and remained rhythmic in islands (Fig.3, B and D). Superposition of Ca2+ activity images demonstrated that the locus of pre-BötC activity was essentially identical in slices and islands (∼300 μm diam; n = 7/7). The magnitude of ΔF/F in the island however was typically smaller than in the slice at the same [K+]o (7–8 mM; not shown) and was restored to near control level after block of synaptic inhibition. Similar to electrophysiological recordings (ref. Fig.1 B), there was greater cycle-to-cycle variability in amplitude of population activity in the island, with small-amplitude Ca2+ oscillations (Fig. 3 D, ▴) between larger amplitude transients. Individual neurons in the pre-BötC exhibited synchronized rhythmic Ca2+ transients before (not shown) and after (Fig. 3, E and F) blocking inhibition.

Fig. 3.

Ca2+ activity of pre-BötC neurons.A: composite image showing slice structure (gray), overnight spread of CaG at midline injection site (semitransparent pseudocolor, left inset), and Ca2+activity localized in pre-BötC (pseudocolor rectangle) during XII inspiratory discharge. Pins holding tissue on silicone elastomer (Sylgard) floor of recording chamber are marked (stars). Blue contour lines indicate outline of excised island in B. B: image of island with Ca2+ activity (pseudocolor) at peak of population burst. C: time course of pre-BötC Ca2+ activity (ΔF/F, spatial average) in slice (A) synchronized with XII discharge (∫XII). D: pre-BötC Ca2+activity (ΔF/F) in island (B) showing rhythmic Ca2+ transients with amplitude fluctuations (smaller transients indicated by arrowheads). E: CaG-labeled neurons (contoured and labeled a–d) in another island preparation, shown with gray scale (boxes) optimized for each neuron. F: synchronized rhythmic Ca2+ activity in neurons inE. E and F recorded with 20 μM bicuculline, 5 μM strychnine, and 9 mM [K+]o. Signals were denoised with a wavelet transform-based algorithm (see Koshiya and Smith 1999 for details).


We have shown that the physically isolated pre-BötC island spontaneously generates rhythmic population and single-cell activity that persists after blocking fast inhibitory synaptic transmission. The locus and temporal profile of population activity corresponded to that in the slice, as confirmed by Ca2+ imaging and local recording. These results provide further evidence that the pre-BötC is the kernel for rhythmic inspiratory burst generation in the in vitro slice. The persistence of rhythm after block of inhibition also supports the proposal that the kernel has intrinsic rhythmic properties arising from an excitatory network of synchronized inspiratory neurons with bursting pacemaker properties (Butera et al. 1999a,b; Koshiya and Smith 1999;Smith et al. 1991). Similar conclusions have been previously reached from synaptic inhibition block experiments in more intact in vitro preparations (reviewed in Rekling and Feldman 1998; Smith et al. 2000) and now confirmed at the level of the isolated kernel.

The oscillatory behavior of the island was similar but not identical to the slice. The amplitude of population activity was reduced, and both burst amplitude and frequency exhibited greater variability than in situ. This may result primarily from the reduction in synaptic connectivity when the contralateral pre-BötC is disconnected in excising the island. In the slice, pre-BötC pacemaker cells are connected by axons crossing the midline (Koshiya and Smith 1999) that bilaterally synchronize pre-BötC activity (Butera et al. 1999b). Network models show that the pacemaker population rhythm becomes unstable when synaptic connectivity is made sparse (Butera et al. 1999b). These models also predict that burst frequency should increase as the coupling in the network is reduced as shown here for the islands and also recently shown for split slice preparations (Del Negro et al. 2001).

The pre-BötC island represents the most reduced preparation yet developed to isolate and study the inspiratory rhythm-generating kernel. Our results suggest that the island population activity may reflect more closely the intrinsic properties of the kernel. In the slice, synaptic inputs from active neurons extrinsic to the pre-BötC can modulate neuronal excitability in the kernel. Recent studies show that when excitability of thin slices is elevated with [K+]o, the population burst frequency can be increased over an order of magnitude as demonstrated for islands, but the burst frequency versus [K+]o relationship has a reduced slope and becomes nonmonotonic at high levels of tonic excitation ([K+]o > ∼15 mM) (Del Negro et al. 2001). The relationship is monotonic over the entire frequency range for the island both before and after blocking synaptic inhibition. A similar monotonic relationship has been predicted from models of the kernel consisting entirely of an excitatory network of synaptically coupled pacemaker neurons with heterogeneous, voltage-dependent bursting properties (Butera et al. 1999b). Thus the oscillatory behavior of the island appears to closely represent an excitatory population of voltage-dependent pacemaker cells. Our results provide further evidence for a pre-BötC excitatory pacemaker-network kernel and show that the island is an ideal preparation for analyzing the kernel's rhythmogenic properties.


The authors thank C. Del Negro for assisting with data analysis and providing data in Fig. 1 A and C. Wilson for reviewing the manuscript.

Present address of N. Koshiya: Blanchette Rockefeller Neurosciences Institute, Johns Hopkins University Academic and Research Building, 9601 Medical Center Dr., Rockville, MD 20850.


  • Address for reprint requests: J. C. Smith, Cellular and Systems Neurobiology Section, Laboratory of Neural Control, 49 Convent Dr., Rm. 3A50 MSC 4455, NINDS, NIH, Bethesda, MD 20892-4455 (E-mail:jsmith{at}


View Abstract