Local application of the NK1R agonist, substance P (SP), into the preBötC in vitro increases the frequency of endogenous respiratory-related output; similar application of the μOR agonist [D-Ala²,N-Me-Phe⁴,Gly⁵-ol]enkephalin acetate (DAMGO) or the γ-aminobutyric acid type B receptor (GABA_BR) agonist baclofen decreases respiratory frequency (5–7). Similar responses are seen in vivo (8). Because these drugs affect frequency, they must act either directly on preBötC rhythmogenic neurons or indirectly on other preBötC neurons that provide modulatory signals to the rhythm generator. In vitro respiratory-related output does not require chlorine-mediated postsynaptic inhibition (2, 3, 5, 9). If peptide neuromodulators affect respiratory frequency by direct actions on only inhibitory neurons, as has been suggested for opioid actions in the midbrain (10), then peptide receptor-expressing neurons could not be responsible for rhythm generation. Thus, we examined whether the effects of SP and DAMGO on respiratory rhythm in vitro were affected by blocking GABA type A receptor (GABA_AR) and glycine receptor activation (11). Pressure injection of SP (2 pmol) into the preBötC of rhythmically active brainstem slices from rats increased respiratory frequency to 192% ± 5% of control, whereas injection of DAMGO (1.5 pmol) decreased frequency to 64% ± 1% of control (Fig. 1). GABA_BR activation has similar effects (5, 7). Bath application of the GABA_AR antagonist bicuculline and the glycine receptor antagonist strychnine to these slices did not change the baseline respiratory frequency or the effects on respiratory frequency of SP or DAMGO (Fig. 1B). Thus, the preBötC NK1R-, GABA_BR-, and μOR-expressing neurons that mediate agonist frequency-modulating effects are likely excitatory and are either directly involved in respiratory rhythm generation or are nonrespiratory neurons that modulate rhythm-generating circuits.

We examined the immunohistochemical expression pattern of NK1R, μOR, and GABA_BR in the ventrolateral medulla of adult rats (12). Within the ventrolateral respiratory column, GABA_BR expression was widespread. However, both NK1R and μOR expression were concentrated within the preBötC and absent from the adjacent Bötzinger complex (BötC) (1) and rostral ventral respiratory group (1) (Fig. 2); NK1R expression in adult mouse brainstem was similar (13). The preBötC NK1R⁺ neurons were segregated from both motoneurons and catecholaminergic sympathetic neurons as revealed by double immunohistochemistry for choline acetyl-transferase (ChAT) or tyrosine hydroxylase (TH) (13, 14).

To determine whether GABAergic and peptidergic effects on breathing are due to convergent inputs onto individual neurons, we looked for μOR or GABA_BR colocalization in NK1R⁺ neurons. We found NK1R⁺/μOR⁺ neurons and processes (Fig. 3A) and NK1R⁺/μOR⁻ neurons and processes (of 88 preBötC NK1R⁺ neurons examined in detail, 53 coexpressed μOR) (Fig. 3B). We found NK1R⁺/GABA_BR⁺ (Fig. 3C) as well as NK1R⁺/GABA_BR⁻ neurons (of 160 preBötC NK1R⁺ neurons examined in detail, 118 coexpressed GABA_BR). We also found NK1R⁻/GABA_BR⁺ neurons (Fig. 3C) and a few μOR⁺/NK1R⁻ processes (Fig. 3B) and somas (13). Thus, GABAergic and peptidergic synaptic inputs converge on subsets of preBötC neurons. Moreover, in the ventrolateral respiratory column, neurons that are targets of these convergent modulatory inputs that affect respiratory frequency are present only in the preBötC. We propose that the expression of NK1R⁺ neuronal somata in the ventrolateral respiratory column defines the anatomic extent of the preBötC.

These results do not differentiate whether these preBötC neurons generate or simply modulate respiratory rhythm. We tested the responsiveness of behaviorally identified respiratory neurons to neurokinin and opioid agonists (15). In the preBötC of rhythmically active slices from mice, inspiratory neurons fall into two categories (types 1 and 2) based on their electroresponsive properties and membrane potential trajectories (16, 17) (Fig. 4A). After blockade of action potential driven synaptic transmission by 2 μM tetrodotoxin (TTX), bath application of SP depolarized type 1 neurons 7.7 ± 1.5 mV (n = 4) (Fig. 4B), whereas type 2 neurons depolarized much less [2.7 ± 0.8 mV (n = 4) (Fig. 4B)]. Input resistance was not significantly changed by SP in type 1 or type 2 neurons (Wilcoxon signed rank test). Bath application of DAMGO (in TTX) hyperpolarized all type 1 neurons tested (7.9 ± 1.8 mV; n = 5) and reduced their input resistance to 67% of control value but had no effect on either the membrane potential or input resistance of type 2 neurons (n = 6) (Fig. 4C). The current-voltage (I-V) curves of type 1 neurons before and after DAMGO (under voltage clamp in the presence of TTX) intersected at −102 mV with normal [K⁺]_o concentration (n = 5), whereas when [K⁺]_o was increased by 10 mM, the intersection point shifted to − 66 mV (n = 3; P < 0.05, Mann-Whitney rank sum test) (Fig. 4D). Taken together, these results suggest that the effect of DAMGO was mediated by activation of a K⁺ conductance.

Most models for respiratory rhythm generation predict that depolarization of rhythm-generating neurons will speed up rhythm and that their hyperpolarization will slow it down (2, 3, 18). Type 1 neurons, which are rhythmogenic (2, 16, 17), are the only preBötC inspiratory neurons depolarized by both SP and thyrotropin-releasing hormone (16), peptides that increase respiratory frequency, and hyperpolarized by DAMGO, a peptide that decreases respiratory frequency; this is consistent with these models and the hypothesis (2) that type 1 neurons are responsible for both rhythmogenesis and frequency control.