Distinct parafacial regions in control of breathing in adult rats

Recently, based on functional differences, we subdivided neurons juxtaposed to the facial nucleus into two distinct populations, the parafacial ventral and lateral regions, i.e., pFV and pFL. Little is known about the composition of these regions, i.e., are they homogenous or heterogeneous populations? Here, we manipulated their excitability in spontaneously breathing vagotomized urethane anesthetized adult rats to further characterize their role in breathing. In the pFL, disinhibition or excitation decreased breathing frequency (f) with a concomitant increase of tidal volume (VT), and induced active expiration; in contrast, reducing excitation had no effect. This result is congruent with pFL neurons constituting a conditional expiratory oscillator comprised of a functionally homogeneous set of excitatory neurons that are tonically suppressed at rest. In the pFV, disinhibition increased f with a presumptive reflexive decrease in VT; excitation increased f, VT and sigh rate; reducing excitation decreased VT with a presumptive reflexive increase in f. Therefore, the pFV, has multiple functional roles that require further parcellation. Interestingly, while hyperpolarization of the pFV reduces ongoing expiratory activity, no perturbation of pFV excitability induced active expiration. Thus, while the pFV can affect ongoing expiratory activity, presumably generated by the pFL, it does not appear capable of directly inducing active expiration. We conclude that the pFL contains neurons that can initiate, modulate, and sustain active expiration, whereas the pFV contains subpopulations of neurons that differentially affect various aspects of breathing pattern, including but not limited to modulation of ongoing expiratory activity.


Introduction
Several brainstem motor nuclei are surrounded by respiratory-related neurons [1,2]. In the case of the facial nucleus, parafacial neurons are essential components of the breathing central pattern generator (bCPG). In particular, parafacial neurons that express the neurokinin-1 receptor (NK1R), the homeobox gene Phox2b, and the glutamate transporter VGlut2, are essential to CO 2 chemoreception [3][4][5][6]; notably, a subpopulation of these neurons have rhythmic respiratory-related activity, both in vitro and in vivo [7][8][9], leading us to postulate that a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 breathing is driven by a dual oscillator system [10]. We identified two neighboring parafacial regions, lateral (pF L ) and ventral (pF V ) that appear to be functionally distinct components of the bCPG [11]. We hypothesized that the pF L is a conditional expiratory oscillator that is inhibited at rest [8,11,12], whereas the pF V provides a generic source of excitatory drive for both inspiration and expiration whose activity depends, at least in part, on CO 2 -related signals [11,[13][14][15]. Furthermore, two parafacial subpopulations, containing Gastrin-Releasing Peptide and Neuromedin B (GRP and NMB, respectively) modulate sighing [16]. Therefore, further subdivision of the parafacial region into functionally distinct nuclei may be warranted, as is the case for other subcortical brain regions, such as the nucleus tractus solitarius, periaqueductal gray, and paraventricular nucleus [17][18][19]. To further investigate the functional contributions of the pF L and pF V , we selectively modulated their excitability and measured the effects on ventilation in spontaneously breathing vagotomized urethane anesthetized adult rats.
We conclude that the: i) pF L contains a functionally homogenous population of excitatory neurons that are tonically inhibited at rest, which following an increase in excitability can initiate and maintain active expiration; ii) pF V contains at least four functionally distinct subpopulations of neurons: three subpopulations that are tonically inhibited at rest, which can separately affect f, modulate active expiration, and modulate basal sigh rate, and one tonically active subpopulation that predominately affects V T . Interestingly no subpopulation of pF V neurons appears capable of directly inducing active expiration; instead the pF V modulates active expiration generated elsewhere, presumably by effects in the pF L and/or (pre)motoneuron pools.

Methods
All protocols were approved by the University of California Los Angeles Chancellor's Animal Research Committee. All experiments were performed in spontaneously breathing vagotomized urethane anesthetized adult Male Sprague-Dawley rats (350-450 g) rats.

Ventral approach
Anesthesia was induced with isofluorane and maintained with urethane (1.2-1.7 g/kg; Sigma) in sterile saline via a femoral catheter. Rats were placed supine in a stereotaxic apparatus on a heating pad to maintain body temperature at 37±0.5˚C. The trachea was cannulated. Respiratory flow was monitored via a flow head (GM Instruments), and CO 2 via a capnograph (Type 340: Harvard Apparatus) connected to the tracheal tube. Paired electromyographic (EMG) wires (Cooner Wire Co.) were inserted into genioglossal (GG), diaphragmatic (Dia), and oblique abdominal muscles (Abd). Anterior neck muscles were removed, a basiooccipital craniotomy exposed the ventral medullary surface, and the dura was resected. After bilateral Distinct parafacial regions in respiratory control vagotomy, exposed tissue around the neck and mylohyoid muscle were covered with dental putty (Reprosil; Dentsply Caulk) to prevent drying. Rats were left for 30 minutes for breathing to stabilize. At rest, ventilation consisted of alternating active inspiration and passive expiration. Once stabilized, solutions of drugs in micropipettes were pressure injected (100-200 nL) bilaterally using a Picospritzer II (General Valve Corp.) controlled by a Master 8 pulse generator (AMPI) into the pF L or pF V (Fig 1). To reduce disruption of the tissue, solutions were injected at~50 nL/min. To ensure parity of injections of different drugs, i.e., AMPA, B+S, A +N, and consistency between both sites, i.e., pF L and pF V , the bilateral injections of a drug were performed~2 mins apart. The timing between; the 2 injections of AMPA (119 ± 16 sec), the 2 injections of B+S (121 ± 10 secs), and the 2 injections of A+N (121 ± 15 secs) were not statistically different (F [2,47] = 0.01; p = 0.98; 2-way ANOVA), and no differences were found between the timings of the 2 injections in the pF V (121 ± 8 secs) and the 2 injections in the pF L (120 ± 13 secs; F [1,47] = 0.0004; p = 0.98; 2-way ANOVA). The timing between the 2 injections of Glu before (122 ± 13 secs) and after (120 ± 13 secs) vagotomy, were also not statistically different (p = 0.8; paired T-test). After each injection rats were allowed 30-45 minutes for drugs to take effect and washout, and for baseline recordings to stabilize before the next injection. The pF L is defined as the area ventral to the lateral edge of the facial nucleus, juxtaposed to the spinal trigeminal tract [11]. The pF V is defined as the area ventral to the caudal half of the facial nucleus, at a central location between the pyramidal tract and the spinal trigeminal tract [11]. Coordinates: lateral from the basilar artery, rostral from the rostral hypoglossal nerve rootlet, and dorsal from the ventral surface (in mm); pF V : 1.8, 0.6, 0.1, and pF L : 2.5, 0.9, 0.2.
In one set of experiments, a ventral approach to the medulla was performed in vagus-intact rats. After a resting period to allow breathing to stabilize, rats received 100-200 nL bilateral injections of glutamate (10 mM; Sigma) administered at~50 nL/min into the pF V (Glu pFV ), following which breathing was allowed to recover. After breathing returned to baseline levels, rats were bilaterally vagotomized at the mid-cervical level. Breathing was allowed to stabilize (~30-60 mins), following which rats received a second bilateral injection of Glu pFV .
Care was taken to reduce any transient effects of mechanical stimulation when placing the pipette into the tissue. As experimental controls to determine whether insertion of the pipette and injection of solution per se had effects, we tested the effects of saline injections.

Data analysis and statistics
EMG signals and airflow measurements were collected using preamplifiers (P5; Grass Instruments) connected to a Powerlab AD board (ADInstruments) in a computer running LabChart software (ADInstruments), and were sampled at 400 Hz/channel. High pass filtered (>0.1 Hz) flow head measurements were used to calculate: tidal volume (V T , peak amplitude of the integrated airflow signal during inspiration; pressure sensors were calibrated with a 3 mL syringe); V T is expressed as mL. Inspiratory duration (T I , beginning of inspiration until peak V T ), expiratory duration (T E , peak V T to the beginning of the next inspiration), and f (1/[T I +T E ]); T I and T E , are expressed in secs (s), and f is expressed as breaths per minute (BPM). Minute ventilation (V e ) was calculated as f x V T , and is expressed as mL/min. EMG data were integrated (τ = 0.05 s; R Dia EMG , R GG EMG , and R Abd EMG in arbitrary units (a.u.)) and the peak amplitude of each signal computed for each cycle.
To obtain control values, all parameters were averaged for 20 respiratory cycles preceding each injection. To measure drug effect, 20 cycles were averaged during a period where the injection had its greatest effect on the airflow channel. Measurements were only made of the initial response to the drug, usually within the first 5 minutes following the 2 nd injection, at a similar time as the expanded traces in the figures (marked in each figure by a black arrow with a black dotted line). Care was taken to avoid measurements where reflexive changes had taken place, for example, where the drug caused an initial decrease in breathing followed by a compensatory increase in breathing as the compound wore off. In these cases, measurements were taken at the peak effect during initial decrease and not during the reflexive increase that followed. Data was analyzed offline and exported to Excel™ (Microsoft) for further analysis. All statistical tests were performed using Igor Pro™ (WaveMetrics), except 2-way ANOVAs which were performed in OriginPro™ (OriginLab).
As described above, for each rat we calculated the average of 20 cycles preceding the stimulus ( control ), and the average of 20 cycles during the stimulus ( stimulus ). Both groups, the control injections for B+S pFV . Examples of sighs are marked with arrowheads labelled with #. Post-inspiratory burst Abd EMG s are marked with arrowheads labelled with †. Bi) Rest. Bii) Following B+S pFV . Grey vertical boxes demark phases of each breath: inspiration (I; light gray), post-inspiration (Post: medium grey), and pre-inspiration (Pre: Dark gray). Sigh marked by #. Post-Inspiratory Abd EMG marked by †. C) Comparison between ventilation at rest (Rst) and after B+S pFV injection. Lines connect data from individual experiments, box and whisker plots show combined data. Data are normalized to highest value for each parameter, i.e., f, T I, T E , V T , GG EMG , Dia EMG , or Abd EMG regardless of whether it belonged to control or B+S pFL group. Ã : p < 0.05. Abbreviations defined in Fig 2. https://doi.org/10.1371/journal.pone.0201485.g003 Distinct parafacial regions in respiratory control values and their associated stimulus value for every rat, were combined into a single data set. To facilitate graphical comparisons data was normalized to the highest value in the data set regardless of whether it belonged to control or stimulus group. Therefore, the highest value in the data set, whether it be control or stimulus , was 1.0.
We define active expiration as the epoch of appearance of burst activity in expiratory muscles, i.e., abdominals, that leads to forced air outflow, typically during late expiration, and consequently, increased V T in the following inspiration. We define sighs by their characteristic augmented V T caused by a second inspiratory effort that occurs before the initial eupneic inspiration is complete. These augmented breaths result from largely from high amplitude inspiratory Dia EMG .
Data were not normally distributed. Data were therefore analyzed using the non-parametric 2-sided Wilcoxon signed-rank test with a significance level of P 0.05 and reported as median and interquartile range (IQR). Data are displayed as box and whisker plots for comparison of groups, and as line graphs for individual experiments. There were 8 biological repeats and no technical repeats in all data sets, with 2 exceptions: Statistical outliers were excluded from the data if they failed both Pierce's criterion and Grubb's test; this led to the removal of 1 outlier from the Abd EMG data from the AMPA pFV and AMPA pFL data sets.

Disinhibition of pF L or pF V affect breathing pattern (Figs 2-5, Table 2)
Disinhibition of pF L neurons by the GABAergic antagonist bicuculline and the glycine antagonist strychnine (B+S pFL ) can induce active expiration [8,11], which we confirm here. Bilateral injection of B+S pFL (n = 8) decreased f and T I , increased T E , V T , R Dia EMG , and inspiratoryrelated R GG EMG activity, and induced rhythmic expiratory bursting in R GG EMG and R Abd EMG (Fig 2), the latter a signature of active expiration, q.v., [8,11]. Bilateral B+S pFL had no effect on minute ventilation (V e ) due to a compensatory increase in V T in response to the changes in f elicited by the antagonism of the inhibitory receptors.
Disinhibition of pF V neurons by unilateral injection of bicuculline increases V T with a reciprocal decrease in f in awake rats [21]. Furthermore, pF V appears to facilitate active expiration through projections to abdominal and genioglossus motoneurons, but does not itself induce active expiration [11]. We therefore expected that pF V disinhibition with a cocktail of bicuculline and strychnine (B+S pFV ) would increase V E , as well as alter abdominal and gengioglossus activity, but would not induce active expiration. Bilateral injections of B+S pFV (n = 8) increased f, decreased T I , did not alter T E , and decreased V T and R Dia EMG . Bilateral B+S pFV in anesthetized rats did not alter V E due to a compensatory decrease in V T in response to an increase in f elicited by the antagonism of inhibitory receptors, which is the opposite response to unilateral injection of bicuculine in the same region in awake rats [21]. pF V disinhibition had multiple effects on genioglossus activity, increasing inspiratory-related R GG EMG and  inducing both pre-inspiratory and post-inspiratory R GG EMG activity (Fig 3). In 6 out of 8 experiments, B+S pFV also induced high amplitude post-inspiratory Abd EMG activity (Fig 3A †  and 3Bii †), which while rhythmic was slow, occurring every 10 ± 1 breaths. This pattern of high amplitude post-inspiratory Abd EMG activity was distinct from active expiration that Lines connect data from individual experiments, box and whisker plots show combined data. Data are normalized to highest value for that parameter, i.e., f, T I, T E , V T , GG EMG , Dia EMG , or Abd EMG regardless of whether it belonged to control or Saline pFV group. Abbreviations defined in Fig 2. https://doi.org/10.1371/journal.pone.0201485.g005  Distinct parafacial regions in respiratory control occurs between every inspiration at a lower amplitude (see Fig 2 and [8,11]. Interestingly, coincident with high amplitude post-inspiratory Abd EMG bursts, there was inhibition of GG EMG activity, showing co-ordination between GG EMG and ABD EMG during expiration ( Fig  3B). In all experiments, B+S pFV also induced sighs i.e., augmented breaths with high amplitude inspiratory Dia EMG followed by prolonged T E (Fig 3A# and 3Bii#); sighs were rhythmic but slow, occurring every 12 ± 1 breaths. The high amplitude post-inspiratory Abd EMG activity was not coordinated with sighing.
To test for any nonspecific effects of pF V or pF L injections on breathing, we injected saline into both regions. In anesthetized rats, saline injections in the pF L (n = 8) did not alter f, T I , T E , V T , R Dia EMG , GG EMG , R Abd EMG , or V E (Fig 4). In anesthetized rats, saline injections in the pF V (n = 8) did not alter f, T I , T E , V T , R Dia EMG , R GG EMG , R Abd EMG , or V E (Fig 5).

Excitation of either pF L or pF V affects breathing pattern (Figs 6-8, Table 2)
Photostimulation of pF L neurons elicits active expiration [8]. We predicted that excitation of the pF L with the glutamatergic agonist AMPA (AMPA pFL ) would also elicit active expiration. Bilateral injections of AMPA pFL (n = 8) decreased f and T I , and increased T E , V T , R Dia EMG , inspiratory-related R GG EMG activity and R Abd EMG (Fig 6), the latter a signature of active expiration, q.v., [8,11]. Like B+S pFL , bilateral injections of AMPA pFL did not affect V e , presumably due to a compensatory increase in V T in response to the decrease in f.
Excitation of pF V neurons by injection of glutamate increases phrenic nerve discharge amplitude and induces sighing in urethane anesthetized, paralyzed, artificially ventilated, vagotomized cats [22]; photostimulation of pF V neurons leads to increased sighing and respiratory frequency in conscious rats [23]. We predicted that excitation of the pF V with AMPA (AMPA pFV ) would increase ventilation and sighing. Bilateral injection of AMPA pFV (n = 8) increased f, decreased T I , did not alter T E , and increased V T , R Dia EMG , and inspiratory-related R GG EMG , but neither induced expiratory-modulated GG EMG nor Abd EMG (Fig 7). Unlike B +S pFV , bilateral injections of AMPA pFV increased V E due to increases in both V T and f. In 5 out of 8 rats, before AMPA pFV caused V T to reach maximal amplitude it induced 1-2 sigh like events, but with no associated GG EMG or Abd EMG activity (data not shown).
The lack of induction of sighing could have been due to either the increased V T in vagotomized rats, or due to the lack of activation of other glutamatergic receptors, e.g., NMDA, mGluR, etc, in addition to AMPA receptors. To explore these possibilities, in separate experiments, we injected glutamate into the pF V (Glu pFV ) of anesthetized rats before and after vagotomy. Before vagotomy (n = 8), bilateral Glu pFV decreased f, increased T I , T E , V T , R Dia EMG , inspiratory-related R GG EMG (Fig 8), and sigh rate, but neither induced expiratory-modulated GG EMG nor Abd EMG (Fig 8). Bilateral injections of Glu pFV did not affect V E due to a compensatory decrease in f in response to an increase in V T , elicited by the activation of glutamate receptors. Following vagotomy, bilateral Glu pFV increased f, decreased T I , did not alter T E , and increased V T , R Dia EMG , and inspiratory-related R GG EMG , but neither induced expiratorymodulated GG EMG nor Abd EMG (Fig 8), similar to AMPA pFV (Fig 7). Like AMPA pFV , bilateral injections of Glu pFV increased V E due to increases in both V T and f. In 3 out of 6 vagotomized  (I; light gray), post-inspiration (Post: medium grey), and pre-inspiration (Pre: Dark gray). C) Comparison between ventilation at rest (Rst) and after AMPA pFL injection. Lines connect data from individual experiments, box and whisker plots show combined data. Data are normalized to highest value for each parameter, i.e., f, T I, T E , V T , GG EMG , Dia EMG , or Abd EMG regardless of whether it belonged to control or AMPA pFL group. Ã : p < 0.05. Abbreviations defined in Fig 2. https://doi.org/10.1371/journal.pone.0201485.g006 Distinct parafacial regions in respiratory control rats, before Glu pFV caused V T to reach maximal amplitude it induced 3-6 sigh-like events but with no associated GG EMG or Abd EMG (data not shown).

Reduced excitation of pF V and pF L have different effects on breathing (Figs 9 and 10, Table 2)
Many, if not most or all, pF L neurons are silent at rest [8,24]; not surprisingly, hyperpolarizing pF L neurons at rest does not affect ventilation [11]. We predicted that reduction of pF L excitability with local injection of a cocktail of the glutamatergic antagonists AP-5 and NBQX (A+N pFL ) would not affect breathing. Bilateral injections of A+N pFL (n = 8) had no effect on f, T I , T E , V T , R Dia EMG , or R GG EMG ; Abd EMG silent at rest, remained so after A+N pFL (Fig 9). Bilateral injections of A+N pFL did not affect V E as it neither affected V T nor f.
By contrast, pF V neurons are active at rest, providing excitatory drive for quiet breathing [25][26][27][28][29]; hyperpolarizing pF V neurons reduces ventilation [5,11,13]. We predicted that reduction of pF V excitability with local injection of AP-5 and NBQX (A+N pFV ), would reduce ventilation. Bilateral A+N pFV (n = 8) increased f, decreased T I and T E , V T , R Dia EMG , and R GG EMG ; Abd EMG , silent at rest, remained so after A+N pFV (Fig 10). Bilateral injections of A+N pFV did not affect V E due to a compensatory increase in f in response to a decrease in V T , elicited by the activation of glutamate receptors. That no injection into the pF V induced active expiration is indicative that the injectate did not spread to the adjacent pF L , likewise since A+N pFL did not induce any changes in breathing, this indicates the injectate did not spread to the adjacent pF V .

Discussion
Since the putative identification of a conditional expiratory oscillator in the rostral medulla [10,12,30], attention has focused on regions surrounding the facial nucleus as its location [8,11,15,24,31]. We identified two functionally separate parafacial regions: the pF V and pF L [11]. We propose that the pF V provides a critical generic drive to breathe, driving inspiration at rest and facilitating both inspiration and expiration when chemosensory drive increases [11,15], and that the pF L is silent at rest, but once activated, drives active expiration [8,11]. Additionally, there appears to be a third parafacial region, more dorsocaudal, containing neurons expressing gastrin releasing peptide that modulates baseline sigh rate [16]. Thus, there appear to be several distinct parafacial regions contributing to the bCPG. To further investigate the role of parafacial neurons, and the neuronal composition of parafacial regions at the ventral medullary surface, we pharmacologically altered the excitability of pF V and pF L neurons and measured the effects on breathing.

Further support of the hypothesis of the pF L as the source for active expiration
Antagonizing ionotropic glutamate receptors with A+N pFL did not alter any respiratory parameter, i.e., no change in f, T I , T E , V T , Dia EMG , GG EMG , or Abd EMG , supporting our previous observation that these neurons are silent at rest, q.v., [8,11]. In the pF L , excitation (with AMPA) or disinhibition (by antagonizing ionotropic GABA and glycine receptors with B+S) Fig 7. AMPA pFV increases f and V T , but does not induce post-inspiratory activity in either abdominal muscles or in pre-and post-inspiratory activity gengioglossus muscles. A) Integrated traces from a single experiment. Black arrows at bottom indicate epochs in expanded traces (Bi and Bii), gray arrows at top indicate unilateral injections for AMPA pFV . Bi) Rest. Bii) Following AMPA pFV . Grey vertical boxes demark phases of each breath: inspiration (I; light gray), postinspiration (Post: medium grey), and pre-inspiration (Pre: Dark gray). C) Comparison between ventilation at rest (Rst) and after AMPA pFV injection. Lines connect data from individual experiments, box and whisker plots show combined data. Data are normalized to highest value for each parameter, i.e., f, T I, T E , V T , GG EMG , Dia EMG , or Abd EMG regardless of whether it belonged to control or AMPA pFV group. Ã : p < 0.05. Abbreviations defined in Fig 2. https://doi.org/10.1371/journal.pone.0201485.g007 Distinct parafacial regions in respiratory control decreased f with a compensatory increase in V T and inspiratory Dia EMG and GG EMG , and onset of expiratory bursting on GG EMG and Abd EMG , i.e., active expiration [8,11]. Thus, these excitatory neurons have presumptive projections to neurons in the preBötzinger Complex (preBötC) or Bötzinger Complex (BötC) [32,33] that inhibit inspiration during expiration, i.e., reciprocal inhibition, and to excitatory premotoneurons in the caudal ventral respiratory group (cVRG) that project to abdominal muscle motoneurons [34][35][36]. Given the delayed increase in V T following the induced decrease in f, a direct excitatory projection from the pF L to the preBötC appears unlikely, but rather suggests an indirect pathway related to controlling pCO 2 , perhaps via the pF V . These observations are consistent with our hypothesis that the pF L is a conditional expiratory oscillator with neurons that are tonically inhibited at rest that can be turned on either by disinhibition and/or excitation.

Multifunctional role of the pF V
A+N pFV injected into the pF V to lower its excitability, decreased V T and inspiratory-related muscle activity, likely via projections to the preBötC and/or the rostral ventral respiratory group (rVRG) [37]. The associated delayed increase in f could again be explained as intrinsic to the slower time course of chemosensory feedback to maintain pCO 2 . As no change in phase durations or f were seen, it appears unlikely that this excitatory drive to inspiration was mediated by rhythmic preBӧtC neurons [38]. Rather, this observation is consistent with our hypothesis of a subpopulation of tonically active pF V neurons that provides facilitative drive to phrenic and/or other inspiratory pump motoneurons to affect V T , but do not contribute directly to regulating f or inspiratory drive to genioglossal motoneurons [11]. Instead it is more likely that the pF V affects V T through its projections to the rVRG [39], the premotor bulbospinal relay to the phrenic nucleus for inspiratory drive [40], as this will alter V T without directly altering other inspiratory parameters, i.e., f and GG EMG . B+S pFV to increase pF V excitability, increased f, most likely through projections to the preBӧtC [38,41], presumably to the same neurons that lead to an increase in f following optogenetic photostimulation of the pF V [42,43]. B+S pFV also increased inspiratory-related GG EMG , likely through pF V projections to the parahypoglossal region (pXII) [39], which appears to be the premotor relay for inspiratory drive to the XII nucleus [44]. Though B+S pFV attenuated Dia EMG and V T, this appeared secondary to the reduction in f and thus was most likely due to chemosensory feedback to control pCO 2 . This further supports our hypothesis of a subpopulation of tonically suppressed pF V neurons that provide facilitative drive to modulate f, but does not contribute directly to V T .
Unlike B+S pFV , AMPA pFV potentiated V T and Dia EMG activity, most likely through excitation of the neurons that were attenuated by A+N pFV and project to the rVRG. AMPA pFV also increased f and inspiratory-related GG EMG most likely through excitation of neurons that project to the preBӧtC and parahypoglossal region that were activated following B+S pFV . As B +S pFV and AMPA pFV each led to different patterns of breathing with neither similar to the effects of activating the pF L , we suggest that there are at least two relevant pF V subpopulations, one expressing inhibitory receptors and one that does not, and that both of these subgroups are distinct from the pF L . https://doi.org/10.1371/journal.pone.0201485.g008

Distinct parafacial regions in respiratory control
Similar to stimulation of pF V neurons in awake behaving vagus intact rats [23] and in vagotomized urethane anesthetized cats [22], disinhibition with B+S pFV elicited sighs in vagotomized rats (Fig 3Bii#), as did excitation with Glu pFV in vagus-intact rats (Fig 8A). In vagotomized rats the amplitude of normal breaths is considerably larger than vagus-intact rats, with the consequence that sighs are masked. Accordingly, when V T was low, i.e., in vagusintact rats or following a reduction in amplitude caused by B+S pFV in vagotomized rats, sustained increases in sigh activity could be seen. This confirms our recent study showing a cluster of neurons in the pF V that release bombesin-like neuropeptides that affect sighing through the activation of cognate receptors in the preBötC [16].
Hyperpolarizing pF V neurons during hypercapnia and hypoxia affects the amplitude of Abd EMG and GG EMG , but not V T or f [11], likely through direct projections to the cVRG [15] and parahypoglossal region [39]. Interestingly, B+S pFV induced high amplitude post-inspiratory activity on both GG EMG and Abd EMG , likely through the same projections, supporting our previous finding that the pF V provides excitatory drive to expiratory premotor nuclei independent of its projections to the preBӧtC [11]. Interestingly, no perturbation of pF V excitability induced active expiration, while hyperpolarization of the pF V reduces active expiration during chemosensory stimulation [11,13]. We conclude that the pF V provides can modulate expiratory activity generated elsewhere, but cannot itself induce active expiration.
Interestingly, most manipulations which changed either f or V T led to compensatory changes, presumably to regulate V E to control pCO 2 to within the normal range. For example, reducing excitation in the pF V reduced activity of neurons that influence diaphragmatic (pre) motoneurons, which are constitutively active at rest. Thus, this manipulation reduced V T , but had no effect on f as the pF V neurons that influence f were supressed at rest and therefore their activity could not be affected by A+N; this allows for other brain regions to affect preBötC rhythmogenic neurons to increase f to compensate for the reduction in V T . Only one manipulation, glutamatergic activation of the pF V (with either AMPA or Glu) changed V E . We believe that this is because glutamatergic activation of the pF V RTN leads to activation of the tonically supressed neurons that activate preBötC rhythmogenic neurons; furthermore this manipulation also excites the neurons that are active at rest that influence diaphragmatic (pre)motoneurons, consequently altering both f and V T simultaneously.

Summary
We propose that there are at least 6 subpopulations of parafacial neurons (Fig 11). The pF L is a conditional expiratory oscillator, with a functionally homogeneous population of neurons that drive active expiration (Fig 11). By contrast, the pF V provides a critical generic facilitatory drive to breathe, and consists of at least 4 functionally distinct subpopulations of neurons: i) a tonically active subpopulation that drives V T via the diaphragm; ii) one subpopulation of tonically suppressed neurons that modulate f; and; iii) a second subpopulation of tonically suppressed neurons that provide drive to abdominal and genioglossus expiratory motor pools, iv) a subpopulation of bombesin-peptide, i.e., NMB, neurons of the hypothesized peptidergic sigh circuit [16]. In addition, there is a 6 th subpopulation bombesin-peptide, i.e., GRP, neurons in the dorsocaudal parafacial (pF DC ) that also can modulate basal sigh rate [16].  (I; light gray), post-inspiration (Post: medium grey), and pre-inspiration (Pre: Dark gray). C) Comparison between ventilation at rest (Rst) and after A+N pFL injection. Lines connect data from individual experiments, box and whisker plots show combined data. Data are normalized to highest value for each parameter, i.e., f, T I, T E , V T , GG EMG , Dia EMG , or Abd EMG regardless of whether it belonged to control or A+N pFL group. Abbreviations defined in Fig 2. https://doi.org/10.1371/journal.pone.0201485.g009 Distinct parafacial regions in respiratory control  (I; light gray), post-inspiration (Post: medium grey), and pre-inspiration (Pre: Dark gray). C) Comparison between ventilation at rest (Rst) and after A+N pFV injection. Lines connect data from individual experiments, box and whisker plots show combined data. Data are normalized to highest value for each parameter, i.e., f, T I, T E , V T , GG EMG , Dia EMG , or Abd EMG regardless of whether it belonged to control or A+N pFV group. Ã : p < 0.05. Abbreviations defined in Fig 2. https://doi.org/10.1371/journal.pone.0201485.g010 Fig 11. Schematic of minimal bCPG, which consists of 4 essential components. 1) preBötzinger Complex (preBötC) drives inspiration by exciting inspiratory premotor neuronal populations projecting to inspiratory muscles, e.g., diaphragm and tongue, and inhibits pF L ; 2) parafacial Dorsocaudal (pF DC ) contains GRP positive neurons contributing to basal sigh rhythm. 3) pF L drives active expiration by exciting expiratory premotor neuronal populations projecting to expiratory muscles, e.g., abdominals and tongue, and excites neurons that inhibit preBötC, either in preBötC or in BötC (not shown); 4) pF V contains neurons and glia that contribute to CO 2 / pH regulation and integrates sensory afferents affecting breathing, including basal sigh rate, via excitatory connections to preBötC and breathing premotor and motor neurons. pF V contains 4 subpopulations: i) tonically active neurons that modulate V T and diaphragm bursting at rest; ii) tonically suppressed neurons that modulate f; iii) NMB positive neurons that affect basal sigh rate, and; iv) tonically suppressed neurons that provide rhythmic drive to abdominal and genioglossus expiratory motor pools producing active expiration. https://doi.org/10.1371/journal.pone.0201485.g011