Role of ventral medullary catecholaminergic neurons for respiratory modulation of sympathetic outflow in rats

Sympathetic activity displays rhythmic oscillations generated by brainstem inspiratory and expiratory neurons. Amplification of these rhythmic respiratory-related oscillations is observed in rats under enhanced central respiratory drive or during development of neurogenic hypertension. Herein, we evaluated the involvement of ventral medullary sympatho-excitatory catecholaminergic C1 neurons, using inhibitory Drosophila allatostatin receptors, for the enhanced expiratory-related oscillations in sympathetic activity in rats submitted to chronic intermittent hypoxia (CIH) and following activation of both peripheral (hypoxia) and central chemoreceptors (hypercapnia). Pharmacogenetic inhibition of C1 neurons bilaterally resulted in reductions of their firing frequency and amplitude of inspiratory-related sympathetic activity in rats in normocapnia, hypercapnia or after CIH. In contrast, hypercapnia or hypoxia-induced enhanced expiratory-related sympathetic oscillations were unaffected by C1 neuronal inhibition. Inhibition of C1 neurons also resulted in a significant fall in arterial pressure and heart rate that was similar in magnitude between normotensive and CIH hypertensive rats, but basal arterial pressure in CIH rats remained higher compared to controls. C1 neurons play a key role in regulating inspiratory modulation of sympathetic activity and arterial pressure in both normotensive and CIH hypertensive rats, but they are not involved in the enhanced late-expiratory-related sympathetic activity triggered by activation of peripheral or central chemoreceptors.

It has long been recognized that vasoconstrictor class of sympathetic fibers show respiratory-related oscillations [1][2][3] . This respiratory-sympathetic interaction underpins the production of rhythmic fluctuations in arterial pressure (Traube-Hering waves), which are derived from phasic constriction of the arterial tree [4][5][6] . An enhancement of these rhythmic oscillations is important for mediating sympatho-excitatory responses to cardio-respiratory reflex activation under increased central respiratory drive 2,7-9 and crucial for the sympathetic over activity observed in animal models of hypertension 6,10,11 . Despite the clear importance of this rhythmic component to control the level of sympathetic activity and cardiovascular function, its central origin remains elusive.
Vasoconstrictor sympathetic preganglionic neurons located in the intermediolateral cell column (IML) of the spinal cord receive synaptic excitation from supraspinal pre-sympathetic neurons located in the rostral ventrolateral medulla (termed here RVLM neurons) [12][13][14] . The RVLM neurons comprise glutamatergic C1 (catecholaminergic) and non-C1 (non-catecholaminergic) phenotypes 15,16 that are believed to contribute to the generation of sympathetic outflow, since bilateral inactivation of the RVLM reduces sympathetic activity and arterial pressure to levels similar to those observed after transection of the spinal cord, at least in the anaesthetized state 17,18 . RVLM C1 and non-C1 neurons are both modulated by respiration 19 and both show increased respiratory-related activity during hypoxia or hypercapnia 20,21 . In this regard, our previous studies have demonstrated that rats submitted to chronic intermittent hypoxia (CIH) exhibit active expiration 10 and an increase in the firing frequency of expiratory-modulated pre-sympathetic RVLM C1 neurons 19,22 , which was also reflected in sympathetic post-ganglionic activity 10 . Indeed, we proposed that this expiratory-related sympathetic activity 1 may contribute to the sustained hypertension induced by CIH in rats 23 . Although depletion or acute silencing of the majority of C1 neurons reduced the sympatho-excitation induced by activation of peripheral 24 or central 25 chemoreceptors in rats, the contribution of C1 and non-C1 RVLM neurons for generating expiratory-related increases in sympathetic activity evoked by CIH or activation of peripheral or central chemoreceptors has not yet been characterized.
Herein, we acutely silenced C1 neurons by application of the insect peptide allatostatin (Alst) following cell-specific targeting with a lentiviral vector to express the inhibitory Drosophila Alst receptor (AlstR) 26 in the RVLM of rats. Using both in vivo (conscious) and in situ (arterially perfused brainstem preparations) rats, we determined the contribution of the C1 neuronal population for the generation of expiratory-modulated sympathetic activity of rats under enhanced central respiratory drive induced by either CIH or activation of peripheral or central chemoreceptors.

Results
Location and efficacy of viral gene transfer. Catecholaminergic brainstem neurons express transcriptional factor Phox2 and can be targeted using lentiviral vectors (LVV) to express a gene of interest under the control of an artificial promoter -PRSx8 25,27,28 . We examined the efficacy of viral-transduction of ventrolateral medullary C1 neurons by immunofluorescent co-localization of AlstR-GFP and tyrosine hydroxylase (TH) and absence of Phox2b expression (a marker of the ventral medullary CO 2 -sensitive neurons) after bilateral microinjections of PRSx8-AlstR-GFP-LVV (Fig. 1A). GFP immunostaining revealed bilateral clusters of AlstR-GFPexpressing RVLM neurons between 11.6 and 12.8 mm caudal to Bregma [200 µm rostral to 1000 µm caudal to the caudal pole of the facial nucleus (0); Fig. 1B]. Between 0-500 µm caudal to the facial nucleus, where most C1 pre-sympathetic neurons reside 29 , AlstR-GFP expression was found bilaterally in 69 ± 2% (n = 67 rats) of all identified C1 (TH-positive/Phox2b-negative) RVLM neurons, confirming effective targeting of the C1 neuronal population. We also found that C1 neurons projected to sympathetic preganglionic neurons in the thoracic (T8-T10) IML, as reflected by the distribution of varicosed axons intermingled with choline acetyltransferase (ChAT)-immunoreactive neurons (n = 4; Fig. 1C). Additional Phox2b-positive/GFP-expressing non-C1 neurons were scattered within ventral medulla beneath and caudal to the caudal border of the facial nucleus (5 ± 1% of GFP-expressing neurons) corresponding to the Retrotrapezoid Nucleus (RTN; Fig. 1B), as described before 30 . Only 1.9 ± 0.7% of GFP-expressing neurons in the ventral medulla lacked either TH or Phox2b expression (non-C1 neurons).

Discussion
In the present study, we used lentiviral vectors to express the inhibitory Drosophila AlstR in a population of sympatho-excitatory pre-sympathetic neurons containing TH (C1 neurons) in the RVLM in order to determine their contribution for the generation of expiratory-related sympathetic activity in CIH rats and also that induced by activation of peripheral or central chemoreceptors in rats. In in vivo and in situ preparations, acute inhibition of C1 neurons resulted in substantial reductions in arterial pressure, Traube-Hering waves, HR, in baseline and reflex-evoked inspiratory-related sympathetic activity recorded from C1 neurons and from postganglionic nerves. These findings suggest that C1 neurons contribute to the inspiratory-sympathetic coupling and for the maintenance of baseline sympathetic outflow and arterial pressure. However, our data also indicate that pre-sympathetic RVLM C1 neurons are not essential for generating the expiratory-related increases in sympathetic activity evoked by CIH, and activation of either peripheral or central chemoreceptors.
Our histological analysis revealed that ~70% of TH-immunoreactive/Phox2b-negative C1 RVLM neurons located 0-500 µm caudal to the facial nucleus were transduced bilaterally, suggesting that lentiviral targeting results in substantial AlstR-GFP expression among catecholaminergic neurons, which is consistent with that described by others 11,25,31,32 . The expression of AlstR within the RVLM region was also confirmed by the immediate silencing of the firing frequency of whole cell recorded transduced C1 GFP-positive medullary RVLM neurons following Alst applications whereas firing was unperturbed in non-C1 RVLM, GFP-negative cells (Figs 2 and 3). Although the PRSx8 promoter is selective for C1 neurons it is also active in neighboring central chemoreceptors CO 2 /[H + ]-sensitive Phox2b-positive RTN neurons 30,33 . Indeed, we found a scattering of GFP-expressing neurons showing Phox2b staining in the RTN, caudal to the caudal pole of the facial nucleus. However, all axonal GPF-positive varicosities in the thoracic IML (T8-T10) were TH-immunoreactive (data not shown), suggesting that the transduced Phox2b neurons were not spinally projecting and unlikely to influence sympathetic activity directly. Also, AlstR activation did not change either baseline inspiratory (PN) frequency or amplitude or active expiration recorded in the AbN nor did it affect reflex evoked active expiration following stimulation of either peripheral or central chemoreceptors; these data suggest that RTN Phox2b-positive neuronal population was not infected sufficiently to cause a functional response following Alst application. In addition, RTN Phox2b-positive neurons do not alter sympathetic outflow or arterial pressure when optogenetically stimulated 34 . Therefore, based upon the previous information, we are confident that the observed changes in sympathetic outflow, HR and arterial pressure in response to AlstR activation in both in vivo and in situ are mediated by inhibition of RVLM C1 neurons. of fast synaptic transmission. Perfusion of Alst produced again a rapid and reversible silencing of this neuron. (D) The same inspiratory-modulated pre-sympathetic neuron labeled in situ with biocytin, exhibiting GFP and TH immunofluorescence (a C1 neuron). Scale bar: 20 µm.
The role of sympatho-excitatory RVLM C1 neurons in regulating baseline respiratory-related sympathetic activity, Traube-Hering waves and cardiovascular function was addressed previously in normotensive and spontaneously hypertensive (SH) rats 11,24,25 . The experimental approach used in the present study caused a significant reduction in baseline sympathetic outflow, by suppressing mainly inspiratory-modulated activity, but also    its small peak during post-inspiration, HR, Traube-Hering waves and arterial pressure in conscious and in situ preparations. Recent studies involving selective optical stimulation of genetically targeted RVLM C1 neurons confirmed that their activation increased respiratory-related sympathetic activity 31 and MAP 32 . However, our observation of cardiovascular effects differs from recent work reporting very little change in MAP after acute C1 neuronal optical inhibition in conscious rats 35 . Both approaches have certain limitations and this discrepancy may reflect differences in strains (Wistar vs Sprague Dawley) and age (juvenile vs adult) of rats, topographical distribution of transfected C1 neurons (faster vs slowly conducting transfected C1 neurons 29 ), methods for neuronal manipulations (long-lasting pharmacogenetic inhibition vs acute optogenetic inhibition), conscious state, and the amount of stress or recent manipulations/anesthesia (e.g. the use of anesthesia was used to connect the optical fibers one hour before experiments 35 ) of rats. RVLM neurons exhibit different patterns of activity entrained by the respiratory pattern generator, with peaks of activity modulated by either I or PI 19,22,[36][37][38] . We showed that Alst inhibited TH-positive inspiratory-modulated, but not TH-negative post-inspiratory-modulated RVLM bulbo-spinal barosensitive pre-sympathetic neurons. We propose that the peak of tSN activity during post-inspiration is also generated by RVLM inspiratory-modulated pre-sympathetic C1 neurons. The conduction velocity delays both in the pathway from the RVLM site to IML and delays in the efferent nerve traffic to the site of tSN recording (T8-T10) may shift the peak of tSN activity from inspiration to post-inspiration. These properties may explain why the inhibition of inspiratory-modulated RVLM C1 neurons using Alst also reduced the peak of tSN activity during post-inspiration. The expiratory-related sympathetic discharge (tSN late-expiratory burst) evoked by CIH, activation of peripheral or central chemoreceptors, was associated with a selective enhancement of the frequency of spontaneous excitatory post-synaptic events and the firing frequency of post-inspiratory-modulated non-C1 RVLM neurons, in the same phase of the respiratory cycle (late-expiratory activity) 19 . Given that MAP remained higher after C1 neuronal inhibition in CIH rats versus controls, this suggests that an important component of the CIH-evoked expiratory-related sympathetic activity and hypertension is caused by the bulbo-spinal non-catecholaminergic post-inspiratory-modulated RVLM neuronal activity; a finding not dis-similar to that we described in the SH rats 36 . In this regard, Menuet et al. 11 demonstrated that deleting a large proportion of C1 neurons attenuated, but did not prevent, the development of hypertension in SH rats, suggesting that non-C1 neurons may also contribute to the generation of sympathetic over activity and high blood pressure in this animal model. We acknowledge that we cannot rule out other brainstem sources of expiratory-related neuronal firing from, for example, the A5 cell group of the pons 39 , spinal pre-or even post-ganglionic neurons 40 .
Activation of central or peripheral chemoreceptors evoked significant increases in tSN during inspiration, but mainly during expiration 7,8,41 . Although, there is evidence that hypercapnia increases discharges of C1 neurons in vivo 42 , recent data suggest that RVLM C1 neurons do not appear to be responsible for the increases in sympathetic outflow evoked by central actions of CO 2 25 . In this regard, following C1 neuronal inhibition in the present study, the baseline level of tSN activity remained higher during hypercapnia than during normocapnia, even after reductions in inspiratory-and post-inspiratory-related sympathetic activity. Expiratory increases in sympathetic activity in response to activation of central chemoreceptors were not affected after C1 neuronal inhibition. The parafacial Respiratory Group (pFRG) has been proposed to be one of the sources of this expiratory modulation 8 . The pFRG contains neurons that are silent during breathing at normoxia and normocapnia 43 , but evoked during activation of central 44,45 or peripheral chemoreceptors 46 , exhibiting a pattern of activity that is strongly correlated with AbN active expiration and the late-expiratory related sympathetic discharges. These late-expiratory neurons of the pFRG could provide excitatory drive not only to expiratory abdominal motoneurons in the spinal cord to produce active expiration 45 , but also to bulbo-spinal barosensitive post-inspiratory-modulated non-C1 pre-sympathetic RVLM neurons, culminating in increases of sympathetic activity correlated with AbN active expiration observed during hypercapnia and hypoxia. In this regard, Dempsey et al. 47 identified monosynaptically connected input neurons from pFRG to RVLM bulbo-spinal pre-sympathetic neurons, which may contribute to the generation of sympatho-excitation during late-expiration under conditions of enhanced central respiratory drive. Future electrophysiological experiments are needed to evaluate the contribution of pFRG late-expiratory neurons for the generation of expiratory-related sympathetic activity observed in rats under hypercapnia, hypoxia and after CIH to ascertain whether this neural coupling mechanism is common to other forms of neurogenic hypertension.
As far as the C1 neurons appear to be critically involved in the generation of the baseline respiratory modulation of sympathetic activity 11,25 , we tested the possibility that this neuronal population is involved in the processing of the inspiratory-and expiratory-modulated sympathetic reflex responses. We verified that inhibition of C1 neurons reduced the inspiratory, but not the expiratory, sympatho-excitatory responses to peripheral chemoreflex activation. Therefore, as observed in chronic intermittent activation of peripheral chemoreceptors (e.g. CIH), these findings indicate that the C1 neurons play no major role in modulation of the expiratory sympathetic reflex response to acute activation of the carotid bodies using KCN. Previous studies from our laboratory demonstrated that depression of post-inspiratory activity (elicited by inhibition of either the nucleus tractus solitarii or Bötzinger Complex) reduced the magnitude of the expiratory-related sympatho-excitatory response to peripheral chemoreflex activation 7,48 . We propose that inhibitory RVLM-projecting neurons in the caudal ventrolateral medulla, with expiratory-modulated activity under resting conditions and inhibited during activation of peripheral chemoreceptors 49 , may also disinhibit RVLM neurons at expiration during hypoxia. These proposed connections are not yet proven and still require further experimental verification.
In summary, our data confirm that the C1 neuronal population significantly contributes to the maintenance of both sympathetic tone, through its inspiratory modulation, and arterial pressure. Although C1 neurons do not appear to mediate the expiratory-related sympathetic activity and hypertension evoked by CIH, this study provides important guidance for further studies seeking to understand brainstem mechanisms underlying increases in sympathetic outflow and cardiovascular dysfunctions which accompany disruptions in gas exchange in patients with obstructive sleep apnea(OSA) and hypertension, for example. Our study supports an important role of expiratory-related oscillations in the generation of active expiration and their significant contribution for the development of hypertension and sympatho-excitatory responses to cardio-respiratory reflex activation under conditions of increased central respiratory drive. The neurons responsible for these effects may provide a novel and highly selective target for treatment of hypertension in conditions of CIH and perhaps in patients with OSA. Viral vectors. The LVV system used here was HIV-1-derived and pseudo-typed with the VSV-G envelope 50 .
In vivo gene transfer. Post-weaned male rats (50-55 g) were anaesthetized with a ketamine (75 mg kg −1 i.p.)/ xylazine (5 mg kg −1 i.p.) mixture. The depth of anesthesia was checked at regular intervals (20-30 min) by assessing the withdrawal reflex response to noxious pinching of the tail or hind paw. Animals were placed in a stereotaxic frame (tooth bar −3.5 mm below the inter-aural line; David Kopf, Tujunga, USA) and two microinjections (different rostrocaudal levels separated by 300 μm) per side of either PRSx8-AlstR-GFP-LVV or PRSx8-GFP-LVV (50 nl each, over 5 min) were delivered into RVLM (Picospritzer II; Parker Instruments, Cleveland, USA). The injection pipette was angled at 25° and injections were made −3.7 mm ventral from calamus scriptorius and ± 1.7 mm lateral from the midline. Post-surgery, rats were treated with one prophylactic dose of analgesic and antipyretic flunixin meglumine (1 mg kg −1 ; Schering-Plough, Rio de Janeiro, Brazil) and 0.1 ml of veterinary antibiotic (1.2 million i.u.; Fort Dodge, Campinas, Brazil) via intramuscular injections.
Chronic Intermittent Hypoxia. Five days after the LVV injections, rats were divided into 2 groups: CIH treatment (6% of O 2 for 30-40 s, every 9 minutes and 8 hours a day) and normoxia (20.8% of O 2 ) for 10 days as this CIH protocol generates forced expiration and increases in expiratory-related burst of sympathetic activity 19 .

Cardiovascular measurements in conscious animals.
At the end of the hypoxic or normoxic protocols, animals were re-anesthetized. A stainless-steel guide cannula (13 mm long, 0.6 mm o.d., 0.4 mm i.d.) was implanted into the lateral cerebral ventricle (−0.6 mm to Bregma, 1.5 mm lateral to the midline and −3.6 mm ventral to dura mater) and a catheter (PE-10 connected to PE-50; Clay Adams, Parsippany, USA) inserted into the abdominal aorta, through the femoral artery, for pulsatile arterial pressure measurements. MAP and HR were processed from PAP by a data acquisition system. Post-surgery, rats were treated again with analgesic and antipyretic and veterinary antibiotic. Two days later, the arterial catheter was connected to a pressure transducer (MLT0380; ADInstruments, Sidney, Australia), and in turn, to an amplifier (ML221; ADInstruments

In situ perfused preparations of rats and neurophysiological recordings. Control and CIH rats
were prepared for in situ perfused preparation 51 at the end of the CIH and control protocols. The ventral medullary surface was exposed using a ventral approach 19 for neuronal recordings in the RVLM. Peripheral chemoreceptors were stimulated by injections of KCN (0.05%, 50 μl; Sigma-Aldrich) into the descending aorta via a side branch of the perfusion cannula. Central chemoreceptors were stimulated by hypercapnia raising CO 2 to 10% with 90% O 2 in control rats after acute bilateral denervation of carotid bodies. Alst (1 µM) was added to the perfusate and fresh perfusate was used to washout the peptide 33 .
PN, tSN and AbN nerves and the ECG were recorded simultaneously using bipolar electrodes mounted on separate 3D micromanipulators (YOU-1; Narishige, Tokyo, Japan). Left PN discharges were recorded from the central end. Right thoracic/lumbar AbN (T13-L1) were isolated from abdominal muscles and cut distally, and their central activity was recorded. The efferent activity of the tSN was recorded from the sympathetic chain (T8-T10). All signals were amplified, band-pass filtered (0.1-5 kHz; 1700 amplifier, A-M Systems, Sequim, USA) and acquired with an A/D converter (5 kHz; CED 1401, Cambridge Electronic Design, Cambridge, UK) controlled by a computer running Spike 2 software (Cambridge Electronic Design). All nerves were recorded in absolute units (μV) and analyses were performed off-line from rectified and integrated signals (time constant: 50 ms). Baseline PN burst frequency was assessed. Active expiration was determined when the AbN activity at the end of expiration was bigger (µV) than the AbN activity earlier in expiration. Based upon absolute values of tSN (µV), we determined a percentage scale, in the range from 0 to 100, in order to compare the levels of baseline tSN across animals during different experimental conditions in control and CIH rats. This percentage scale was determined for each preparation and was based on the maximal activity observed during ischemia (pump-off) as 100% and the electrical noise level as 0% obtained at the end of each experiment after the death of the preparation. PN-triggered averages of tSN were generated from 1-2 min epochs and divided into three parts: I (coincident with inspiratory PN discharge), PI (first half of expiratory phase) and E2 (second half of expiration). The tSN activities during each respiratory phase were assessed by the measurement of individual area under the curve and normalized by the total area (100%), i.e. sum of areas. PN frequency, HR and PP responses to chemoreflex activation were assessed by the difference between baseline and the peak of response observed after the stimulus (ΔPN, ΔHR and ΔPP). The tSN (during inspiration and expiration) and AbN expiratory responses to peripheral chemoreflex activation were assessed by the measurement of the area under the curve, in a time window of ≤ 10 s after the stimulus, and expressed as percentage values (ΔtSN and ΔAbN in percentage) in relation to the activity before the stimulus.
Blind whole cell patch-clamp recordings were obtained from RVLM pre-sympathetic neurons with electrodes filled with a solution containing in mM: 130 K-gluconate; 4.5 MgCl 2 ; 14 trisphosphocreatine, 10 HEPES; 5 EGTA; 4 Na-ATP; 0.3 Na-GTP; 0.2% biocytin (Molecular Probes, Grand Island, USA) 36,52 . RVLM pre-sympathetic neurons were characterized by their inhibitory response to baroreflex activation (by electrically stimulating ADN; 0.2 ms, 50 Hz, 2 s) and the presence of antidromic responses evoked by stimulation (0.2 ms; 1 Hz) of spinal segments T8-T10 (DS2A, Digitimer, Welwyn Garden City, UK) 53 . Current-clamp experiments were performed using an Axopatch-200B integrating amplifier (Molecular Devices, Sunnyvale, USA) and pClamp acquisition software (version 10.0, Molecular Devices). All data were low-pass filtered at 5 KHz and digitized at the rate of 20 KHz (Digidata 1550, Molecular Devices). The intrinsic pacemaker property of RVLM pre-sympathetic neurons was evaluated in the presence of blockers of fast synaptic transmission in the perfusate (2.5-6.0 mM kynurenic acid, 20 µM bicuculline and 1 µM strychnine; Sigma-Aldrich). A liquid junction potential of 15 mV was corrected off-line. All-points histograms were constructed using 1 min of membrane potential values recorded in each experimental condition and the resting membrane potential value was considered to be the value at which the cells spent most of the time as defined previously 19 . Histology. Immunohistochemistry was performed on sections from the brainstem and spinal cord of control and CIH rats to evaluate the location of labelled RVLM neurons and detection of: TH, Phox2b, ChAT and GFP expression following viral gene transfer. Animals used for neuronal tracing studies were sacrificed 6 weeks after virus injections to optimize fluorophore labeling of axonal projections to the IML. Standard perfusion and post-fixation protocols were performed as previously published 54 . Immunohistochemistry protocols were performed as previously published for fluorescence immunohistochemistry 54  Statistical Analyses. Results are expressed as means ± SED and two-way ANOVA followed by Bonferroni's post hoc test were used to compare tSN activity among phases of the respiratory cycle in in situ preparations, MAP and HR in conscious rats exposed to CIH and control conditions before and after Alst. Student's paired t test was used for direct comparisons of neuronal activity, PP, Traube-Hering waves, HR, tSN, PN and AbN before and after Alst in the same experimental group (GraphPad Prism 4, La Jolla, USA).