Journal of Neurophysiology

Neurokinin Receptor-Expressing Pre-Bötzinger Complex Neurons in Neonatal Mice Studied In Vitro

John A. Hayes, Christopher A. Del Negro


Breathing in mammals depends on inspiratory-related neural activity generated in the pre-Bötzinger complex (preBötC), where neurokinin receptor-expressing neurons (NKR+) have been hypothesized to play a critical rhythmogenic role. Currently, the extent to which the preBötC is populated by rhythmogenic NKR+ neurons and whether neurons without neurokinin receptor expression (NKR) share similar electrical properties with NKR+ neurons are not well understood. These interrelated problems must be resolved to understand the widespread excitatory effects of neuropeptides and the mechanism of respiratory rhythmogenesis. We recorded and imaged inspiratory neurons in neonatal mouse slices that isolate the preBötC and generate respiratory motor output in vitro. Using tetramethylrhodamine conjugated to the endogenous NKR agonist substance P (TMR-SP) to tag neurons that express NKRs, we show that early inspiratory neurons with small whole cell capacitance (CM) are 36% TMR-SP+ and 64% TMR-SP. Also, late inspiratory neurons with large CM are 67% TMR-SP+ and 33% are TMR-SP. Thus NKR+ and NKR neurons exhibit the same phenotypic properties, which suggests that they may share functional roles also. Substance P (SP) alone evoked a voltage-insensitive inward current (ISP) that reversed at –19 mV and was associated with an increase in membrane conductance in both NKR+ and NKR neurons. Gap junctions may be needed to confer SP sensitivity to neurons that appear to lack NKR expression. We propose that cell death in NKR+ preBötC neurons, by targeted lesion or neurodegeneration, may impair breathing behavior by killing less than one half of the rhythmogenic preBötC neurons and a large number of respiratory premotoneurons.


Breathing behavior in mammals is generated by respiratory neurons in the medullary brain stem (Ballanyi et al. 1999; Bianchi et al. 1995; Blessing 1997). A critical issue is the role of neurons that both express neurokinin-1 receptors (NK1Rs) and reside in the critical site for inspiratory breathing behavior, the pre-Bötzinger complex (preBötC) (Gray et al. 1999, 2001; Guyenet and Wang 2001; Janczewski and Feldman 2006; Smith et al. 1991; Stornetta et al. 2003a; Wang et al. 2001).

Substance P (SP) is an endogenous agonist for NKRs, which is most potent at NK1Rs (Medhurst and Hay 2002), but may also activate other tachykinin receptors. SP accelerates respiratory rhythm in vitro and has been shown to depolarize every preBötC neuron recorded intracellularly after pharmacologically silencing network activity (Gray et al. 1999; Murakoshi et al. 1985; Pena and Ramirez 2004). Neurons with early inspiratory activity discharge with an ascending ramp-like voltage trajectory, before the inspiratory motor output phase of the respiratory cycle and are widely considered important for rhythmogenesis (Bianchi et al. 1995; Feldman and Del Negro 2006; Onimaru and Homma 1992; Onimaru et al. 1989; Ramirez et al. 2002; Richter and Spyer 2001; Smith et al. 1991). Neurons with early inspiratory activity in vivo, which were dubbed preinspiratory (Guyenet and Wang 2001), express NK1Rs detectable with immunohistochemistry, whereas noninspiratory neurons that may have other respiratory-related functions are NK1R (Guyenet and Wang 2001). A large number of bulbospinal premotor neurons in the preBötC, which discharge later in the respiratory cycle compared with early inspiratory neurons, also express NKRs (Guyenet et al. 2002; Stornetta et al. 2003b).

The ribosomal toxin saporin (SAP) conjugated to SP, i.e., SP-SAP, leads to the abolition of normal breathing in otherwise intact and awake rats when injected into the preBötC (Gray et al. 2001; McKay et al. 2005). These observations are consistent with the hypothesis that the preBötC contains NKR+ interneurons that respond vigorously to SP and share phenotypic inspiratory discharge properties specialized for rhythm generation (Feldman and Del Negro 2006; Gray et al. 1999). However, the relative number of NKR+ and NKR preBötC neurons that serve rhythmogenic and/or premotor functions remains unknown, and the mechanism underlying the widespread excitatory effects of SP is not well understood.

We examined these issues in vitro using fluorescent labeling to identify SP-sensitive neurons that presumably express NKRs, as assessed by their ability to internalize the fluorescent marker tetramethylrhodamine (TMR) conjugated to SP, i.e., TMR-SP (Pagliardini et al. 2005). We recorded TMR-SP+ and TMR-SP neurons and characterized their membrane properties and inspiratory discharge patterns. We examined the ionic mechanism for SP-evoked excitation and may now be able to explain how SP ubiquitously depolarizes preBötC neurons even though only a limited subset seems to express NKRs.


The Institutional Animal Care and Use Committee at the College of William and Mary approved all protocols. Transverse slices (550 μm thick) from neonatal (P0–P7) C57BL/6 mice were dissected as described previously (Del Negro et al. 2005; Pace et al. 2007). With the neuraxis pinned to a paraffin-coated block, oriented rostral side up with its ventral surface facing forward, we aggressively cut into the preBötC at the rostral surface to expose the putatively rhythmogenic neuron population. Based on objective criteria now verified by the “on-line histology” atlas published by the Ballanyi group (Ruangkittisakul et al. 2006), we made the first cut above the rostral-most XII nerve roots at the level of the dorsomedial cell column and principal lateral loop of the inferior olivary nucleus; thus the preBötC is located at or near the rostral surface (Ruangkittisakul et al. 2006). The caudal cut always captureed the obex. Therefore we recorded from the rostral side of the slice where the preBötC neurons were exposed and never from the caudal side.

Slices were perfused at 26–28°C with artificial cerebrospinal fluid (ACSF) containing (in mM) 124 NaCl, 9 KCl, 0.5 NaH2PO4, 25 NaHC03, 30 d-glucose, 1.5 CaCl2*H2O, and 1 MgSO4. We used 21 slices for the epifluorescence electrophysiology/imaging data (Figs. 1, 2, and 7 and Supplementary Figs. S1–S3),1 and report data from 29 slices for the two-photon/confocal imaging experiments (Figs. 6 and 7), and 33 slices for the voltage-clamp experiments (Figs. 35 and 7). To avoid tachyphylaxis and other consequences of multiple drug applications, each slice was used for one type of experiment: electrophysiological recordings of preBötC neurons after TMR-SP labeling, voltage-clamp experiments to characterize SP-evoked membrane current, or acquisition from a two-photon/confocal imaging experiment.

FIG. 1.

Intrinsic properties of inspiratory neurons in pre-Bötzinger complex (preBötC). A: inspiratory drive latency plotted vs. small whole cell capacitance (CM). Each symbol shows drive latency in a preBötC neuron for several consecutive cycles. B: representative early inspiratory neuron is shown on left, with broken lines showing measured latency. Arrow shows ascending voltage trajectory. A representative late inspiratory neuron is shown on right, also showing drive latency and an arrow showing decrementing voltage trajectory.

FIG. 2.

Tetramethylrhodamine conjugated to substance P (TMR-SP) labeling in preBötC neurons with different phenotypic properties. IR-DIC and epifluorescence images are shown in left columns, with corresponding intracellular traces to the right. A: TMR-SP early inspiratory neuron with silent interburst intervals. B: TMR-SP+ early inspiratory neuron with tonic low-frequency spiking properties. C: TMR-SP+ early inspiratory neuron with voltage-dependent pacemaker properties. D: TMR-SP+ late inspiratory neuron. E: TMR-SP+ expiratory neuron. Scale bar (25 μm) applies to all images in A–E.

FIG. 3.

Effects of 1 μM SP on inspiratory neurons. Steady-state I-V relationship is shown for a typical experiment using K-gluconate patch solution and standard artificial cerebrospinal fluid (ACSF) with 200 μM Cd2+ and 1 μM TTX.


Voltage-clamp and current-clamp experiments were performed with a HEKA EPC-10 patch-clamp amplifier (Lambrecht, Germany). Network activity was monitored from XII nerves with extracellular suction electrodes and a high-gain differential amplifier with band-pass filtering (0.3–1 kHz). The root-mean-square (RMS) of voltage input to the differential amplifier (Dagan Instruments, Minneapolis, MN) was conditioned using a true RMS-to-DC converter (Analog Devices, Norwood, MA) to provide a full-wave rectified and smoothed XII waveform. Data were acquired digitally and analyzed using Igor Pro 5 (WaveMetrics), Chart 5 (AD Instruments, Colorado Springs, CO), Excel (Microsoft, Redmond, WA), and custom software. An 8-mV liquid junction potential was corrected off-line in current-clamp recordings and on-line in voltage-clamp recordings.

Whole cell capacitance (CM) was measured using 50-ms voltage steps from –60 mV to command potentials from –75 to –65 mV in a 10-step sequence. Charge (Q) was computed by integrating leak-subtracted capacitative current (ΔQ = ∫IC) and CM was calculated from CM = ΔQV. Series (access) resistance (RS) was monitored throughout voltage-clamp recordings according to the Thevenin equivalent circuit, which allows RS to be calculated from the decay time constant (τm) in response to small voltage steps with RS = τm/CM as long as RSRN. We monitored input resistance (RN) through P/N on-line leak protocols. To avoid voltage-clamp errors, we discarded experiments in which RS > 0.1 × RN. We compensated for RS as much as possible without loss of stability. We also rechecked RS and RN before each I-V protocol to assess voltage-clamp viability. The average uncompensated RS was 20.2 ± 2.0 MΩ with an average of 37 ± 3% RS compensation, and the average RN was 355.3 ± 84.2 MΩ (n = 24). Firing patterns of recorded neurons were consistent in on-cell and whole cell and remained constant for the duration of the experiments, which in current clamp could last 40–60 min.

Current-clamp protocols (Figs. 1 and 2) and some voltage-clamp recordings (Fig. 3) used the following patch solution containing (in mM) 140 K-gluconate, 5 NaCl, 0.1 EGTA, 10 HEPES, 2 Mg-ATP, and 0.3 Na(3)-GTP. KOH was used to equilibrate pH at 7.2.

Voltage-clamp experiments in Figs. 4 and 5 used ACSF containing (in mM) 84 NaCl, 3 KCl, 20–40 TEA-Cl, 25 NaHCO3, 5 4-AP, 30 d-glucose, 0.5 CaCl2, 2 MgSO4, 2 CsCl, 0.2 CdCl2, 20–40 sucrose (for equimolar balancing with TEA-Cl), and 0.001 TTX. Patch electrodes (4–6 MΩ) contained the following solution (in mM): 140 d-gluconic acid, 140 CsOH monohydrate, 10 TEA-Cl, 10 NaCl, 10 HEPES, 2.5 EGTA, 1.2 CaCl2 dihydrate, 2 Mg-ATP, and 0.3 Na(3)-GTP, with pH adjusted to 7.2 using HCl.

FIG. 4.

Effects of 1 μM SP on inspiratory neurons with additional blockade of K+ channels; 200 μM Cd2+ and 1 μM TTX were present in the bath (as in Fig. 3), but patch solution was based on Cs-gluconate, and ACSF contained a cocktail of K+ channel antagonists. A: on-cell and whole cell recordings of a typical inspiratory neuron. B: steady-state I-V protocols showing reversibility of SP-evoked response. C: steady-state I-V curve and effects of SP. Inset: ISP obtained by subtraction (SP –control). D: tail current analysis, protocol shown below I-V curve. ISP obtained by subtraction using tail protocol is also plotted in inset of C.

FIG. 5.

Effects of 1 μM SP on preBötC neurons with blockade of K+ channels and gap junctions; 100 μM carbenoxolone (CBX) was added in control to attenuate gap junction–mediated transmission. A: steady-state I-V protocols. B: steady-state I-V curves. C: tail-current analysis. SP had no effect on membrane properties in 100 μM CBX.

Epifluorescence microscopy

Slices were incubated in 1 μM TMR-SP (Invitrogen, Carlsbad, CA) for 8–12 min at 32°C and moved to the perfusion chamber for intracellular recording. We visualized preBötC neurons with Koehler illumination and differential interference contrast (DIC) videomicroscopy, which facilitated patch-clamp recordings, and switched to epifluorescence (X-cite-120, EXFO, Mississauga, Ontario, Canada) and a rhodamine filter to capture TMR-SP. Positively labeled neurons (TMR-SP+) were distinguished by labeling around the somatic border, in a perinuclear area, and occasionally along dendrites (e.g., Figs. 2 and S1). TMR-SP labeling became more diffuse over time (>1–2 h) (Grady et al. 1995).

Control experiments were performed (Fig. S2) by preincubating a slice in 10 μM unconjugated SP for 5 min at 32°C and applying 1 μM TMR-SP following the protocol described above. After exposure to both forms of SP, only sparse TMR-SP labeling could be detected in regions that included the preBötC; the extent of the labeling was dramatically less than when TMR-SP was applied without prior exposure to unconjugated SP. This is most apparent in the nucleus ambiguous (NA; Bieger and Hopkins 1987), which is heavily populated by NK1R+ neurons and is adjacent to the preBötC dorsally (Gray et al. 1999; Pagliardini et al. 2005). In control experiments, the NA showed substantially less TMR-SP labeling (Fig. S2A) than using the standard TMR-SP loading protocol (Fig. S2B).

Epifluorescence images were acquired with a 12-bit charge-coupled device (CCD) monochromatic camera, the QImaging Retiga 1300i (Surrey, British Columbia, Canada) using a long-working distance water immersion ×40 objective with a 0.80 numerical aperture. Before image acquisition, the pipette tip or somatic border was focused with IR-DIC videomicroscopy; IR-DIC images were typically exposed for 150–250 ms with 1 × 1 binning. Epifluorescence images were exposed for 10 s with 1 × 1 binning at maximum fluorescence intensity. The images were pseudocolored with a black-to-red look-up table in iVision software (Biovision Technologies, Exton, PA). Background subtraction was performed by plotting a histogram of pixel intensities and truncating all values less than the lowest peak. For publication figures, we copied images to Photoshop (Adobe Systems, San Jose, CA) and enhanced contrast and applied a 1-pixel radius Gaussian blur.

Confocal and two-photon microscopy

Fifty micrograms of fluo-4 acetoxymethyl ester (fluo-4 AM; Invitrogen) was dissolved in 50 μl of pluronic (20%) + DMSO (Invitrogen) and vortexed for 10 min. After that, 750 μl of 30°C 9 mM [K+] ACSF was added to the dye solution and vortexed until the dye was evenly distributed. The solution was divided into two tubes (∼375 μl each), and an additional 375 μl of 30°C 9 mM [K+] ACSF was added to each tube for a final concentration of 29.4 μM fluo-4-AM. Slices were incubated 40–50 min in 29.4 μM fluo-4 AM at 32°C and incubated in a separate chamber of 1 μM TMR-SP at 32°C for 10–14 min. We imaged cellular Ca2+ fluctuations using an inverted Nikon Radiance microscope (Nikon, Melville, NY) and a Mai-Tai Ti:sapphire femtosecond laser (Spectraphysics, Mountain View, CA) tuned to 800-nm excitation wavelength. Data were acquired digitally and saved to disk on a PC running Windows NT and LaserSharp software by Zeiss Microimaging (Thornwood, NY).

TMR-SP labeling was measured at the same workstation (without moving the slice) using a 543-nm Green HeNe laser at full intensity (1.5 mW), a pinhole size of 2.2 Airy units, at 512 × 512 pixels scanned with 25 lines per second (lps) using an accumulate feature to optimize the signal-to-noise ratio. This involved 30 scans per image where the pixel intensities of each scan were divided by two and added to the previous scans (∼10-min acquisition time). We used a Nikon ×40 Plan Fluor objective with a numerical aperture of 0.75, which resulted in a 5-μm depth of field. We acquired one green (500–530 nm) and one red (600–650 nm) channel for TMR-SP images. For each image, background correction was accomplished by subtracting the RMS of the pixel intensity across the image from every pixel value. We renormalized each image and subtracted 8% of the green signal (the fluo-4 channel) from the red channel to correct for the overlap of the emission spectra, i.e., the portion of fluo-4 emission expected in the 600- to 650-nm band. Images were analyzed using iVision and ImageJ (National Institutes of Health, Bethesda, MD). Like the epifluorescence data, for publication figures, we copied images to Photoshop and enhanced contrast and applied a 1-pixel radius Gaussian blur.

Peak acquisitions (i.e., Fig. 6A, fluo-4) were achieved by scanning the focal plane repeatedly at 25 lps (1,024 × 1,024 pixels) until at least one pixel saturated. Time-series recordings of Ca2+ activity (Fig. 6B) were scanned at 256 × 256 pixels and 500 lps (∼2 Hz) for ∼125 frames. Summed Ca2+ activity of time-series two-photon experiments (Fig. 6C, small inset panels) was plotted by taking the minimum pixel value for the whole time-course of the acquisition bout and subtracting this baseline from the entire time series. We summed all the fluorescence measurements for the whole time series collapsed onto one aggregate image. This produces an image that convolves expiratory and inspiratory neurons (and other transiently active cells), highlighted in warm colors.

FIG. 6.

Simultaneous measurements of inspiratory activity and TMR-SP labeling in preBötC neurons. A: fluo-4 image shows a peak acquisition of Ca2+ labeling; TMR-SP image shows TMR-SP+ cells in the same region. Scale bar is 50 μm. B: changes in fluorescence intensity from regions of interest (ROIs) indicated by numerals in A, plotted with synchronized XII activity. Maximum ΔF/F range for each ROI was used for illustrative purposes (1: 50%, 2: 26%, 3: 39%, 4: 49%, 5: 50%, 6: 32%, 7: 54%, 8: 57%). C: magnified images shows background subtracted and summed Ca2+ activity of a TMR-SP+ inspiratory neuron (7), a TMR-SP inspiratory neuron (4), a TMR-SP+ expiratory neuron (8), and peak acquired view of 3 TMR-SP+ nonrhythmic neurons (9–11).

We scanned one plane per slice preparation for experiments that contained both TMR-SP and fluo-4; the imaging plane was <5 μm in thickness. Using the two-photon (Ti:Sapphire) laser to detect Ca2+ transients through fluo-4 fluorescence changes, we first probed for inspiratory neurons. If we detected inspiratory neurons in a given plane, we applied a long-duration confocal scan to detect TMR-SP labeling using the 543-nm laser (HeNe). This long-lasting exposure at high intensity bleached TMR-SP; thus we terminated the experiment after acquiring these data in the selected plane. If in a given plane we failed to detect inspiratory neurons, we did not apply the long-duration confocal scan but instead we incremented our z-axis by 5 μm (≤30–40 μm into the tissue) to probe for inspiratory neurons in the adjacent tissue layer.

Some pilot experiments with just TMR-SP were performed using the Ti:Sapphire laser to study the viability of the labeling technique (Fig. S1C).


We compared neurons on the basis of drive potential latency (Fig. 1B), CM, and TMR-SP labeling. We tested for normality and applied Student's t-test or Wilcoxon signed ranks tests as appropriate to detect statistically significant differences. Mean values are reported as mean ± SE.

We compared NKR expression in six different experimental approaches using a resampling method (Manly 2007). We used the fraction of NKR+ neurons detected in a given experiment as our benchmark and counted the number of times a uniformly distributed randomly generated number in the interval [0,1] fell below that fraction when drawing the same number of samples. We repeated this algorithm in 10,000 simulated experiments and tallied the outcomes to generate a histogram that reports the likelihood of each experimental sample drawn by chance. Removing the highest and lowest 250 counts yields 95% credible intervals for the experiment.

We tested whether the fraction of SP-sensitive neurons detected in voltage clamp (Fig. 4; 86.7%) was significantly higher than the TMR-SP+ fraction detected with other techniques. This was performed by comparing the SP-sensitive fraction detected in voltage clamp (control) to the fraction of TMR-SP+ neurons found in other conditions. These include 1) epifluorescence imaging (Fig. 2), 2) confocal imaging (Fig. 6), 3) the SP-sensitive fraction in voltage clamp with carbenoxolone (CBX, Sigma-Aldrich, St. Louis, MO; Fig. 5), 4) the meta-analyzed NK1R immunoreactive (NK1R-ir) data set from adult rats (Guyenet and Wang 2001), and finally 5) single-cell RT-PCR evidence for NK1Rs in inspiratory neurons of neonatal rats (Manzke et al. 2003). Altogether, the pooled fraction of neurons that show evidence of NKR/NK1R expression was 40% (Fig. 7). To calculate the likelihood of the voltage-clamp experiment being different from the other experiments by chance, we used the above resampling technique and calculated a P value by dividing the number of samples drawn that equaled or exceeded 86.7% (the SP-sensitive fraction in this experiment) by 10,000 simulated experiments.

FIG. 7.

Analysis of NKR+ preBötC neuron distribution. Box plot shows fraction of TMR-SP+ neurons measured with conventional epifluorescence (epifluor.) displayed for the early inspiratory neurons (early) and for the entire sample of all inspiratory neurons (all); fraction of TMR-SP+ inspiratory neurons measured with confocal laser-scanning microscopy (confocal LSM); fraction of neurons with measurable ISP in control (vclamp control) and in the presence of 100 μM CBX (vclamp + CBX). Fraction of NK1R-ir neurons detected in adult rats are also shown (adult in vivo) meta-analyzed from Guyenet and Wang (2001). Meta-analyzed data from Manzke et al. (2003) using single-cell RT-PCR to determine NK1R+ inspiratory neurons is also included. Bold horizontal lines in each category show mean, and thin lines show 95% credible intervals. Overlapping regions of epifluor., confocal LSM, vclamp + CBX, adult in vivo, and RT-PCR data are bounded by gray rectangle. **Statistical significance at P < 0.01.


Electrical properties and TMR-SP labeling in inspiratory neurons

Inspiratory neurons were separable on the basis of multiple phenotypic properties (Figs. 1 and 2). We measured CM in preBötC neurons satisfying reliable voltage-clamp conditions (see methods). We also characterized inspiratory drive latency in all neurons with reliable current-clamp recordings, defined as the difference between the onset of inspiratory excitatory postsynaptic potentials (EPSPs) and the beginning of the XII motor output (Rekling et al. 1996). Figure 1A plots inspiratory drive latency versus CM in neurons that had both reliable voltage- and current-clamp recordings. We found a subset of early inspiratory neurons with significantly longer latency (241.2 ± 8.4 ms, n = 8) and lower CM (45.6 ± 1.5 pF, n = 8) compared with a subset of late inspiratory neurons that had significantly larger CM (85.9 ± 6.5 pF, n = 5; t-test: P < 0.01) and shorter latency (103.7 ± 8.2 ms, n = 5; Wilcoxon: P < 0.01). Early inspiratory neurons showed an incremental discharge pattern (Fig. 1B, left, arrow), whereas late inspiratory neurons exhibited a rapid onset with decrementing discharge pattern (Fig. 1B, right, arrow).

Once we recognized the correlation between CM and inspiratory drive latency, we divided preBötC neurons into two classes: 1) early inspiratory neurons had an average drive latency of 201.5 ± 8.1 ms (n = 22, with a total of 256 latency measurements) and 2) late inspiratory neurons had an average drive latency of 88.9 ± 6.5 ms (n = 6, with a total of 66 latency measurements). Inspiratory drive latency was significantly different between these two classes (t-test: P < 0.05).

Nine early inspiratory neurons showed a low baseline membrane potential and thus a silent interburst phase (Fig. 2A), whereas 13 spiked at low rates during the interinspiratory burst interval (Fig. 2B). One neuron exhibited ectopic bursts at depolarized baseline membrane potentials (Fig. 2C), suggesting voltage-dependent pacemaker properties (Del Negro et al. 2002, 2005; Ramirez et al. 2004; Smith et al. 1991; Thoby-Brisson and Ramirez 2001). These factors in early inspiratory neurons did not correlate with CM so we did not further subdivide the early inspiratory data set.

After patch-recording, we tested for TMR-SP labeling. TMR-SP labeled 8 of 22 (36.4%) early inspiratory neurons. Figure 2 shows one TMR-SP inspiratory neuron (Fig. 2A) and two TMR-SP+ inspiratory neurons (Fig. 2, B and C) with discharge properties described above. Four of six late inspiratory neurons exhibited TMR-SP labeling (66.6%) as shown in Fig. 2D. For the combined sample of 28 early and late inspiratory neurons, 42.8% were TMR-SP+.

We also recorded TMR-SP+ expiratory neurons, which discharge at high rates throughout the interburst interval but are actively inhibited during the inspiratory phase (Fig. 2E, n = 18). These data were surprising because adult rat expiratory neurons showed no NK1R immunoreactivity (Guyenet and Wang 2001). Finally, nonrhythmic TMR-SP+ cells were detected within the preBötC region but were not counted because their identity could not be verified in vitro (Fig. S3).

Excitation of inspiratory neurons by substance P

SP has been hypothesized to excite preBötC neurons by evoking a low-threshold, voltage-dependent, and TTX-insensitive Na+ current (Pena and Ramirez 2004). Because we were interested in SP-evoked excitation, we did not prelabel neurons with TMR-SP to avoid NKR desensitization. We measured the steady-state current-voltage (I-V) relationship with a K-gluconate patch solution and standard ACSF while blocking TTX-sensitive Na+ currents and Cd2+-sensitive voltage-gated Ca2+ currents (Fig. 3). The slope of the I-V curve increased in 0.5–1 μM SP and crossed the control I-V curve at approximately –20 mV, which suggests the opening of a mixed cation channel (n = 4).

To isolate the SP-induced current (ISP), we used patch solution containing Cs+ and TEA, with Cs+, TEA, and 4-AP in the ACSF to block K+ currents and hyperpolarization-activated cation current (Ih). We identified inspiratory neurons in the on-cell configuration (Fig. 4A ) by observing the onset latency of inspiratory discharge. It was impossible to determine early versus late inspiratory phenotypes because Cs+-patch solution elevates input resistance and depolarizes the reversal potential for chloride. Nevertheless, CM ranged from 23.3 to 78.9 pF, (mean CM was 49.2 ± 7.0 pF, n = 9), which suggests both early and late inspiratory phenotypes were sampled (Fig. 1). We measured the I-V relationship in control and 1 μM SP and obtained ISP by subtraction (Fig. 4, B and C). ISP was linear (n = 13) and reversed at ESP = –19.4 ± 0.02 mV (n = 9), similar to the ISP reversal potential measurement with the K-gluconate patch solution (Fig. 3).

Next, we examined whether ISP expressed any voltage-dependent properties. Because the I-V protocol could cause voltage-dependent inactivation of ISP during 500-ms-long voltage steps, we analyzed tail currents from –60 to +10 mV after a prepulse to +10 mV for 100 ms (Fig. 4D). ISP was computed by subtraction. We compared ISP tail currents to steady-state ISP in the range of –60 to 0 mV. In both cases, ISP was identical throughout the voltage range (Fig. 4C, inset), suggesting no voltage-dependent inactivation of ISP.

We detected ISP in 13/15 neurons (87%). These data are consistent with current-clamp studies showing that SP depolarizes every inspiratory neuron tested in vitro (Gray et al. 1999; Pena and Ramirez 2004). However, TMR-SP labeling was only present in 42.8% of inspiratory neurons in our previous experiments. This disparity could reflect a failure to detect TMR-SP labeling in some NKR+ neurons or that SP-sensitive glial cells play some role in exciting inspiratory preBötC neurons that are otherwise SP-insensitive and NKR.

An alternative explanation for the disparity between the large number of preBötC neurons with measurable ISP and the smaller TMR-SP+ subset is that gap junctions (Rekling et al. 2000) might confer the effects of SP to NKR neurons. To test whether gap junctions were required to evoke ISP, we repeated the voltage-clamp protocols from Fig. 4 after >15-min exposure to 100 μM CBX to block gap junctions. Our intra- and extracellular solutions minimized undesired effects of CBX on intrinsic membrane properties such as leak currents (Rekling et al. 2000; Rouach et al. 2003). We evoked ISP in 7 of 14 inspiratory neurons with CBX present, whereas 7 inspiratory neurons did not respond to SP either in I-V or tail-current protocols (Fig. 5).

TMR-SP labels respiratory phasic and nonrhythmic neurons in the mouse preBötC

We measured the rhythmic Ca2+ activity of inspiratory neurons using two-photon laser-scanning microscopy (TPLSM), which enabled us to scan 5-μm-thick focal planes in a defined 293 × 293-μm area within the preBötC. In 49 planes from 29 slices, we observed 344 inspiratory neurons and detected a maximum of 11 and a minimum of 1 inspiratory neuron per plane (average 6), which is commensurate with inspiratory neuron counts recently reported using TPLSM in neonatal rats (Ruangkittisakul et al. 2006). Fluorescence changes could not differentiate early versus late inspiratory phenotypes; with drive latencies typically <500 ms (see Fig. 1), our maximum 4-Hz sampling rate was too low to make reliable distinctions.

Figure 6A (top) shows the peak acquisition of Ca2+ emission over several respiratory cycles in a typical experiment. Figure 6A (bottom) shows TMR-SP emission in the same region detected with confocal laser-scanning microscopy (CLSM). Cycle-to-cycle activity from neurons in Fig. 6A are plotted with XII activity in Fig. 6B: neurons 1–7 were inspiratory, whereas neuron 8 was expiratory. Neurons 4, 7, and 8 are shown at higher magnification in Fig. 6C. We detected 13/31 (41.9%) TMR-SP+ inspiratory neurons with TPLSM/CLSM and 3/3 (100%) TMR-SP+ expiratory neurons in a total of four imaging planes acquired in four slices. Several TMR-SP+ nonrhythmic cells were situated among the inspiratory neurons (such as 9–11, Fig. 6C).

Comparing the relative fraction of preBötC neurons with evidence for NKR expression in several experimental conditions

Altogether we used three methods to quantify the fraction of NKR+ inspiratory neurons in the preBötC: epifluorescence yielded 12/28 (42.9%) TMR-SP+, TPLSM/CLSM yielded 13/31 (41.9%) TMR-SP+, and ISP was measured in voltage clamp in 7/14 (50%) inspiratory neurons in the presence of 100 μM CBX. These measurements in the neonatal mouse preBötC are comparable with the fraction of early inspiratory neurons, dubbed preinspiratory (preI) by the authors, which were recorded in adult rats in vivo and subsequently found to be NK1R+ by immunohistochemistry: 11/32 (34.4%) (Guyenet and Wang 2001). Additionally, our results are consistent with the fraction of inspiratory neurons (4/13, 30.7%) with NK1R expression measured using single-cell RT-PCR methods (Manzke et al. 2003) and qualitatively similar to the conclusion by Manzke et al. that there is a large presence of noninspiratory NK1R-ir neurons in the preBötC.

We tested the null hypothesis that these independent measurements reflect the same underlying fraction of NKR+/NK1R+ neurons in the preBötC. A virtual preBötC in silico containing 40% NKR+/NK1R+ neurons (the pooled fraction of NKR+/NK1R+ neurons detected using all methods excluding the control ISP experiment) and 60% NKR/NK1R neurons was used to randomly sample 14, 28, 31, 32, and 13 neurons (with replacement) corresponding to the experiments above. Each sample was repeated 10,000 times. We tallied the results in a frequency histogram and found that, for a population containing 40% NKR+/NK1R+ neurons, drawing empirical samples of 42.9, 41.9, 50, 34.4, and 30.7% were statistically indistinguishable (P ≫ 0.05). Finally, we considered the possibility that the early inspiratory neurons we found in the neonatal mouse preBötC are phenotypically the same as the preI neurons recorded in adult rats in vivo (Guyenet and Wang 2001); again, the fraction of NKR+/NK1R+ neurons was statistically indistinguishable (8/22, 36.4% vs. 11/32 34.4%, P ≫ 0.5).

In contrast, we evoked ISP in 13/15 (86.7%) inspiratory neurons with gap junctions intact (Fig. 4). In resampling simulations, this outcome (i.e., drawing a sample fraction of 86.7% NKR+ neurons) occurred by chance <1% of the time, so we rejected the null hypothesis at P < 0.01. Figure 7 plots the sample mean fraction of NKR+ neurons with 95% credible intervals to show the consistency between the fraction of NKR+ neurons detected with imaging experiments, immunohistochemistry, RT-PCR, and voltage-clamp experiments in the presence of CBX compared with the much larger number of NKR+ neurons with measurable ISP with gap junctions unblocked.


Our data suggest that the preBötC comprises ∼40% NKR+ inspiratory neurons in rodents. Nonetheless, SP may exert widespread excitatory effects caused by gap junctions that activate ISP in both NKR+ and NKR neurons. The functional roles of NKR+ and NKR neurons may overlap because both subsets showed early and late inspiratory phenotypes, and both respond to SP-mediated modulation (Gray et al. 1999) in the absence of gap junction blockers. Therefore the NKR expression per se may not be a reliable means to classify preBötC neurons functionally. Destruction of NKR+ neurons disrupts normal respiratory physiology (Gray et al. 2001; McKay et al. 2005). Because NKR+ neurons exhibit both early inspiratory and late inspiratory phenotypic properties, as well as expiratory and nonrespiratory phenotypes, the loss of all of these neuron types must be considered when interpreting the functional consequences of lesion or natural NKR+ cell death.

Biophysics of ISP in inspiratory neurons

ISP is measurable in early and late inspiratory neurons using doses of SP that have clear respiratory effects in previous studies (Gray et al. 1999; Pagliardini et al. 2005; Pena and Ramirez 2004). ISP does not depend on extracellular Ca2+ and is TTX-insensitive, and Na+ is the dominant inward charge carrier (Pena and Ramirez 2004). Its reversal potential (ESP) is –19 mV, so we conclude that K+ is also a charge carrier. ESP was the same with K-gluconate patch solution and Cs+-based patch solution that substantially raised the Cl reversal potential, so Cl is not a charge carrier for ISP. We observed ISP in the presence of combined Na+, Ca2+, and K+ blockers, which suggests that ISP arises from a single type of mixed cation channel.

Tail current analysis would enable detection of any component of ISP that slowly inactivates during the steady-state I-V protocol. Because ISP tail currents and the steady-state ISP were identical (Fig. 4C, inset), we concluded that there was no significant voltage-dependent component of ISP that inactivates on the time scale of 100–500 ms. This contradicts the hypothesis that ISP is a TTX-insensitive Na+ current (Pena and Ramirez 2004) that can give rise to negative slope resistance and bursting properties (Delmas et al. 1997).

SP increases excitability through the closure of K+ channels in hypoglossal motoneurons (Yasuda et al. 2001) and C1 neurons that are situated at the ventral border of the preBötC (Blessing 1997; Li and Guyenet 1997). Our data set did not contain C1 neurons because ISP never reversed at EK and was unaffected by intracellular and extracellular K+ channel blockers.

SP has widespread excitatory effects on inspiratory neurons in vitro (Gray et al. 1999; Pena and Ramirez 2004; Yamamoto et al. 1992), and we evoked ISP in 86.7% of inspiratory neurons in the absence of CBX. However, NKR expression seems to be much less prevalent: ∼42% of both early and late inspiratory neurons in neonatal mice were TMR-SP+, 34% of preI neurons identified in adult rats in vivo were NK1R-ir (Guyenet and Wang 2001), and 31% of preBötC neurons were NK1R+ as identified with single-cell RT-PCR (Manzke et al. 2003). The latter measurements are consistent with our ability to evoke ISP in only 50% of inspiratory neurons after blocking gap junctions, suggesting that gap junctions are involved in evoking ISP in NKR neurons. It is conceivable that the slightly higher (but not statistically significant) difference between our TMR-SP+ fraction and the meta-analyzed NK1R-expression data can be attributed to other tachykinin receptors that can bind TMR-SP but do not show NK1R immunoreactivity. However, this is unlikely because respiratory-related neurons in NK1R−/− mice do not respond to SP (Ptak et al. 2000).

A cationic current that reverses at –11 mV (in ACSF with 9 mM external [K+]) is coupled to muscarinic receptor activation in preBötC neurons (Shao and Feldman 2000). This current is very similar to ISP: it is TTX-insensitive, its activation is voltage- and Ca2+-independent, and Na+ and K+ are the principal charge carriers. This suggests that muscarinic and neurokinin receptors may open the same underlying class of cation channels (Pena and Ramirez 2004; Shao and Feldman 2000), but this remains to be tested.

Putative roles of NKR+ and NKR inspiratory neurons in respiratory rhythmogenesis

The majority of our neurons showed early inspiratory activity patterns and small CM. The early latency, small size, and incremental discharge trajectory are characteristic of propriomedullary glutamatergic interneurons that putatively serve in a rhythmogenic capacity (Guyenet and Wang 2001; Stornetta et al. 2003a; Wallen-Mackenzie et al. 2006). Thirty-six percent of these early inspiratory neurons were NKR+. Given their discharge pattern and NKR expression, these neurons are probably glutamatergic and are unlikely to be GABAergic or glycinergic (Stornetta et al. 2003a,b; Wang et al. 2001), although we cannot rule out some of these NKR+ neurons belonging to a class of GABAergic neurons involved in sympathetic control of blood pressure (Wang et al. 2002). NKR+ early inspiratory neurons are unlikely to contain cardiovagal preganglionic motoneurons in the external division of the nucleus ambiguous because choline acetyl-transferase was never co-detected with NK1R expression in adult rat preBötC neurons (Wang et al. 2001). The fraction of NK1R-ir early inspiratory-like neurons (called preI by the authors) in adult rats in vivo is also near 36% (Guyenet and Wang 2001), so we conclude that the fraction of NKR+ rhythmogenic neurons in the preBötC is consistent in neonates and adults.

Large CM and late inspiratory discharge pattern are characteristics consistent with glutamatergic bulbospinal neurons that putatively serve in a premotor capacity (Guyenet et al. 2002; Rekling et al. 1996; Stornetta et al. 2003b). Because 67% of late inspiratory neurons were NKR+, SP-SAP lesions may ablate a larger percentage of late inspiratory (putative premotoneurons) compared with early inspiratory neurons. However, because early inspiratory neurons are more numerous and show membrane properties consistent with a role in rhythmogenesis, NKR-targeted lesions will probably cause a greater total reduction in NKR+ rhythmogenic-like neurons. However, the destruction of a large fraction of NKR+ respiratory premotoneurons must be considered as a factor in explaining apneas resulting from SP-SAP lesions (Gray et al. 2001; McKay et al. 2005).

NKR neurons with early and late inspiratory discharge properties probably incorporate some cardiovagal preganglionic and pharyngeal motoneurons (Bieger and Hopkins 1987; Rekling and Feldman 1997; Rekling et al. 1996), as well as respiratory premotoneurons (Guyenet et al. 2002). Inspiratory neurons within the preBötC may also be GABAergic (Kuwana et al. 2006) or glycinergic (Shao and Feldman 1997). Therefore some NKR neurons are either inhibitory or motor-related neurons that presumably do not directly contribute to rhythmogenesis.

Nevertheless, because of their discharge phenotype and sensitivity to SP (with gap junctions intact), we propose that many early inspiratory NKR neurons are also rhythmogenic interneurons analogous to NKR+ glutamatergic early inspiratory interneurons (Guyenet et al. 2002; Stornetta et al. 2003a). However, we cannot be certain of the transmitter type in NKR early inspiratory neurons and thus cannot exclude the possibility that some of these neurons have nonrhythmogenic functions.

Furthermore, it is difficult to ascertain how our early and late inspiratory phenotypes map to respiratory phenotypes in larger brain stem preparations or in vivo, which is problematic from the standpoint of nomenclature, because the pattern of activity may change with further levels of embedded neural circuitry. However, we provide simple names for distinct phenotypes in slices, and our dichotomy may be useful to distinguish putatively rhythmogenic and premotor neurons in this context.

Estimating the size and composition of the neonatal preBötC

The preBötC in rats is remarkably constant in size during early neonatal development and extends for ∼200 μm in the rostral-caudal axis of rats (Ruangkittisakul et al. 2006; Smith et al. 1991). Given somatic diameter of ∼10 μm for preBötC neurons (Stornetta et al. 2003a; Wang et al. 2001), we can offer a rough estimate of the population size of inspiratory neurons in the preBötC. If we assume one neuron-layer per 10 μm of tissue in the sagittal plane, account for a bilaterally distributed preBötC, and use our measured average of six inspiratory neurons per plane, then the neonatal rodent preBötC contains ∼240 inspiratory neurons. This assumes that the rostro-caudal extent of the neonatal mouse preBötC matches that of the rat.

We counted 22/28 (78.6%) neurons with early inspiratory discharge pattern and small CM, in which 8/22 (36.3%) were TMR-SP+. We found 6/28 (21.4%) neurons with late inspiratory pattern and large CM in which 4/6 (66.7%) were TMR-SP+. We thus estimate that the preBötC contains ∼189 early inspiratory neurons, of which 69 are NKR+ and 120 are NKR, and 51 late inspiratory neurons, of which 34 are NKR+ and 17 are NKR.

Physiological significance: a prediction for recovering respiratory function after NKR+ neuron loss

In neonatal mice (our results) and adult rats (Guyenet and Wang 2001), ∼36% of rhythmogenic-like neurons showed evidence of NKR expression. We postulate that ∼64% of putative rhythmogenic inspiratory neurons and ∼33% of premotor-like neurons may survive SP-SAP lesions or diseases that ablate NKR+ neurons and impair breathing (Gray et al. 2001; McKay et al. 2005). Our estimates for population sizes will facilitate graded cell-destruction simulations in mathematical models of the preBötC that reflect the approximate numbers of NKR+ and NKR neurons with respective rhythmogenic-like and premotor-like phenotypes. Models of this type may elucidate the mechanism by which graded neuron destruction perturbs rhythmogenesis and may help clarify the different effects of destroying rhythmogenic versus premotor neurons.

Stable breathing behavior is impaired by NKR+ neuron loss in the preBötC and may be a result of a breakdown in fundamental rhythmogenic mechanisms. However, strengthening the excitatory synaptic transmission between NKR preBötC neurons may restore respiratory function, assuming that NKR preBötC neurons are glutamatergic and interconnected (Guyenet et al. 2002; Rekling et al. 2000; Stornetta et al. 2003a,b). Augmenting excitatory synaptic strength could be accomplished using cyclothiazide (Funk et al. 1995) or ampakines (Ren et al. 2006) that enhance ionotropic glutamate receptors or by enhancing the role of metabotropic glutamate receptors by targeting specific intracellular signaling cascades coupled to their activation. This prediction arises from the hypothesis that a limited number of synaptically interconnected constituent neurons in the preBötC can maintain rhythmic function by periodically evoking burst-generating intrinsic membrane properties that are only available in the context of behavior through ionotropic and metabotropic glutamate receptors (Feldman and Del Negro 2006; Rekling and Feldman 1998; Rekling et al. 1996; Wallen-Mackenzie et al. 2006).


This study was supported by National Science Foundation Grant IOB-0616099, the Suzann Wilson Matthews Faculty Research Award, and the Jeffress Memorial Trust.


  • 1 The online version of this article contains supplemental data.

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


View Abstract