Vestibular-mediated synaptic inputs and pathways to sympathetic preganglionic neurons in the neonatal mouse

Abstract

To assess when vestibulosympathetic projections become functional postnatally, and to establish a preparation in which vestibulosympathetic circuitry can be characterized more precisely, we used an optical approach to record VIIIth nerve-evoked synaptic inputs to thoracic sympathetic preganglionic neurons (SPNs) in newborn mice. Stimulation of the VIIIth nerve was performed in an isolated brainstem–spinal cord preparation after retrogradely labelling with the fluorescent calcium indicator Calcium Green 1-conjugated dextran amine, the SPNs and the somatic motoneurons (MNs) in the thoracic (T) segments T2, 4, 6, 8, 10 and 12. Synaptically mediated calcium responses could be visualized and recorded in individual SPNs and MNs, and analysed with respect to latency, temporal pattern, magnitude and synaptic pharmacology. VIIIth nerve stimulation evoked responses in all SPNs and MNs investigated. The SPN responses had onset latencies from 90 to 200 ms, compared with much shorter latencies in MNs, and were completely abolished by mephenesin, a drug that preferentially reduces polysynaptic over monosynaptic transmission. Bicuculline and picrotoxin, but not strychnine, increased the magnitudes of the SPN responses without changing the onset latencies, suggesting a convergence of concomitant excitatory and inhibitory synaptic inputs. Lesions strategically placed to test the involvement of direct vestibulospinal pathways versus indirect pathways within the brainstem showed that vestibulosympathetic inputs in the neonate are mediated predominantly, if not exclusively, by the latter. Thus, already at birth, synaptic connections in the vestibulosympathetic reflex are functional and require the involvement of the ventrolateral medulla as in adult mammals.

Key points

• When the body is tilted from horizontal to vertical, blood tends to accumulate in the legs, and blood pressure may fall in the upper body and head, a phenomenon known as orthostatic hypotension.

• The vestibulosympathetic reflexes triggered by stimulation of vestibular afferents help to counteract orthostatic hypotension, especially during its initial onset.

• We used an ex vivo preparation of the brainstem and spinal cord of the neonatal mouse, and high-throughput optical recording to assess when vestibulosympathetic projections become functional postnatally.

• Vestibulosympathetic synaptic connections are present in the mouse, and are already functional at birth. The organization of the projections includes many of the key features seen in adult mammals.

• The demonstration of similarity between vestibulosympathetic pathways in mice and other mammals is exciting because mouse models provide the possibility to use powerful molecular genetic techniques to make discoveries that may be very relevant for human vestibulo-autonomic dysfunction.

Introduction

When the body is tilted from horizontal to vertical, blood tends to accumulate in the legs, and blood pressure may fall in the upper body and head, a phenomenon known as orthostatic hypotension. Normally, compensatory mechanisms including the baroreceptor reflex, contraction of leg muscles and the negative thoracic pressure created by inspiration work in synergy to counteract orthostatic hypotension. However, the earliest phase of compensation occurs too quickly to be explained by these mechanisms, indicating an anticipatory mechanism that reads body tilt more rapidly. It is now well established that the vestibular system has influence over the sympathetic nervous system and cardiovascular control (reviewed in Yates, 1996; Yates & Bronstein, 2005; Wilson et al. 2006; Holstein et al. 2011). Thus, vestibulosympathetic reflexes triggered by stimulation of vestibular afferents are now considered to constitute the rapid compensatory component that counteracts orthostatic hypotension.

In the first study that proposed a connection between the vestibular system and the sympathetic nervous system, it was shown that bilateral transection of the vestibular nerves in cats greatly compromised their ability to compensate for orthostatic hypotension (Doba & Reis, 1974). More recently this effect has been shown to have a regional character (reviewed in Kerman et al. 2000a), with compensation in hindlimb vasculature much more dependent on intact vestibular afferents than compensation in forelimb vasculature (Kerman et al. 2000b; Wilson et al. 2006). Studies in rodents and humans have indicated that attenuation of vestibular afferent signals in the elderly and in post-flight astronauts may be responsible for the orthostatic intolerance that these subjects experience (Buckey et al. 1996; Ray & Monahan, 2002; Etard et al. 2004). Several studies in cats have shown that electrical or natural stimulation of vestibular afferents influences the firing rate of sympathetic nerves (sympathetic chain, muscle sympathetic efferents, cervical, splanchnic and renal nerves) through combined excitatory and inhibitory synaptic inputs (reviewed in Yates, 1992; Kerman et al. 2000a). Binaural low-frequency sinusoidal galvanic vestibular stimulation (sGVS) in rats, which preferentially activates otolith-responsive vestibular pathways, leads to a transient drop followed by oscillations in blood pressure and heart rate (Cohen et al. 2011). Similar results (using electrical, caloric or natural stimulation) have been obtained in human studies (reviewed in Ray & Carter, 2003; Carter & Ray, 2008; Holstein et al. 2012).

While there is ample evidence that vestibular inputs influence the sympathetic nervous system, the underlying neural pathways that mediate this are not completely understood. Lesion studies in the cat have identified a region within the vestibular nuclear complex situated partly in the medial nucleus and partly in the inferior nucleus that is necessary for vestibulosympathetic control (Yates et al. 1993; Yates & Miller, 1994). However, there is currently no evidence suggesting a direct monosynaptic connection between vestibular neurons and sympathetic preganglionic neurons (SPNs). A model has been proposed based on anatomical and electrophysiological evidence in which vestibulosympathetic influence is mediated by vestibular neurons that project onto a network of neurons in the rostral and caudal ventrolateral medulla (RVLM and CVLM) as well as the nucleus of the tractus solitarius (NTS; reviewed in Yates, 1996; Balaban & Porter, 1998; Yates & Bronstein, 2005; Holstein et al. 2011). Neurons in the RVLM appear to be the principal source of the descending axons that mediate the vestibulosympathetic responses (Lee et al. 2007; Abbott et al. 2009), and have been shown to make monosynaptic and polysynaptic excitatory (glutamatergic) and inhibitory (GABAergic) connections onto SPNs (Deuchars et al. 1995, 1997). Another potential pathway involved is a vestibulo-raphespinal projection that has been demonstrated anatomically and electrophysiologically (Bacon et al. 1990; Yates et al. 1993; Porter & Balaban, 1997; see however Iigaya et al. 2007).

Although these studies have provided a description of neural substrates sufficient and in some cases necessary to mediate the vestibulosympathetic reflex in the cat, less is known about the organization of this reflex in other mammalian species. A few studies have begun to investigate the vestibulosympathetic reflex in the rat, demonstrating its contribution to the maintenance of blood pressure during gravitational stress (Gotoh et al. 2004; Tanaka et al. 2006; Abe et al. 2008) and its response to sGVS (Cohen et al. 2011). Other studies have demonstrated the expression of activity-dependent immediate-early gene expression in subpopulations of medial, descending and superior vestibular nuclei neurons in response to sGVS in the rat (Holstein et al. 2012), and shown that some neurons in the descending vestibular nucleus and caudal half of the medial vestibular nucleus contain imidazoleacetic acid-ribotide, a putative neurotransmitter for blood pressure regulation (Friedrich et al. 2007; Martinelli et al. 2007). Recent efforts have also been made to characterize the sympathetic effects of neurochemically defined subpopulations of RVLM neurons (Burke et al. 2011). An important advance would be to characterize vestibulosympathetic circuitry in the mouse because of the opportunities for transgenic manipulation, including optogenetic recording and stimulation approaches (reviewed in Zhang et al. 2007).

An additional issue that has not yet been addressed is the ontogeny of the vestibulosympathetic reflex and its underlying circuitry. This can also be characterized in the mouse, which lends itself well to both anatomical and physiological studies of bulbospinal projections in embryos and neonates (Auclair et al. 1999; Pasqualetti et al. 2007; Szokol et al. 2008; Szokol & Perreault, 2009, and see below).

Here, we begin to investigate the functional organization of the murine vestibulosympathetic connections using an ex vivo preparation of the brainstem and spinal cord of the neonatal mouse that permits high-throughput optical recording of synaptically evoked calcium responses in spinal neurons (Szokol et al. 2008, 2011; Szokol & Perreault, 2009; Kasumacic et al. 2010). We have found that electrical stimulation of the VIIIth nerve evokes calcium responses in SPNs throughout the thoracic cord. A variety of lesions within the brainstem and upper cervical cord indicate the involvement of an indirect descending pathway with a relay within the ventrolateral medulla, as has been proposed in adult mammals. The inability to induce the same responses by direct stimulation of the lateral vestibulospinal tract (LVST) supports this conclusion. Long response latencies similar to those demonstrated in the adult cat (Yates et al. 1995), together with pharmacological experiments using selective application of mephenesin to the spinal cord, indicate that the spinal component of the indirect pathway may also be polysynaptic. An increase in response magnitude during pharmacological blockade of GABA receptors within the spinal cord indicates the presence of both inhibitory and excitatory components. Preliminary accounts of this work have been reported elsewhere (Kasumacic et al. 2009, 2010).

Methods

Animals and brainstem–spinal cord preparation

Experiments were performed on newborn mice of the ICR strain (n= 70), at postnatal days (P) 0–5. Mice were deeply anaesthetized with isoflurane (Abbott, Scandinavia AB, Solna, Sweden) and decerebrated at the level of the superior colliculus. Decerebration was followed by rapid dissection of the brainstem and spinal cord using the same procedure as described in Kasumacic et al. (2010). All efforts were made to minimize the number of animals used in accordance with the European Communities Council directive 86/609/EEC, and the National Institutes of Health guidelines for the care and use of animals. All procedures were approved by the National Animal Research Authority in Norway (Forsøksdyrutvalget).

Retrograde labelling of SPNs and somatic motoneurons (MNs) with fluorescent calcium indicator

After dissection the preparation was placed in oxygenated artificial cerebrospinal fluid (aCSF) containing in mm: NaCl, 128; KCl, 3; d-glucose 11; CaCl₂, 2.5; MgSO₄, 1; NaH₂PO₄, 1.2; Hepes, 5; NaHCO₃, 25. SPNs and somatic MNs in the thoracic cord (T2, T4, T6, T8, T10 and/or T12 segments) were labelled retrogradely by applying crystals of Calcium Green 1-conjugated dextran amine (CGDA; 3000 MW; Life Technologies Ltd, Paisley, United Kingdom) to the cut ventral roots (Glover 1995). Retrograde labelling continued in the dark at room temperature (23–25°C) for 3 h.

Optical recording of calcium responses

Brainstem–spinal cord preparations were transferred to a Sylgard-coated chamber and pinned ventral side up with stainless steel (Minuten) pins inserted through the peeled dura mater. Prior to transferring to the recording chamber, an oblique transection (about 45 deg from the horizontal) was performed at the level of the segment containing the CGDA-labelled neurons (see Szokol & Perreault, 2009; Szokol et al. 2011 for more details). The transected end was slightly bent and positioned on a trapezoidal block of Sylgard glued to the bottom of the chamber so that the obliquely cut surface lay horizontally under the objective (Fig. 1A). In some preparations, SPNs and somatic MNs were labelled in three different thoracic segments (n= 9). In these preparations, oblique cuts were made sequentially starting first with the most caudal segment. Subsequent oblique cuts were performed only after completing optical recording from the caudalmost segments. During recording, labelled neurons were imaged using a 40× water immersion objective (LUMPlanFl, 0.8 NA, Olympus, Norway) on an upright epi-fluorescence microscope (Axioskop FS 2, Carl Zeiss, Oberkochen, Germany) equipped with a 100 W halogen lamp and excitation (band-pass 450–490 nm) and emission (long-pass 515 nm) filters, and the preparation was perfused continuously with oxygenated aCSF at a rate of 7.5 ml min⁻¹ (total volume exchange every 2 min). Fluorescence intensity changes elicited by stimulation (Ca²⁺ responses) were registered using a cooled CCD camera (Cascade 650, Photometrics, Texas Instruments, USA) mounted on a video zoom adaptor, using the image-processing software Metamorph 7.7 (Universal Imaging Corporation, Molecular Devices, USA). Image series were routinely acquired at 4 frames s⁻¹ and a binning of 3 × 3. However, in some experiments (n= 10) we used a higher frame rate (100 frames s⁻¹) and higher binning (6 × 6) to record latencies of the responses at higher temporal resolution. High frame rate recordings could not be performed for long periods because they require higher intensity illumination, which rapidly photobleaches the CGDA.

Figure 1. Brainstem–spinal cord preparation and effective stimulation parameters.

A, schematic representation of the brainstem–spinal cord preparation indicating the segments and neuron groups studied. Inset shows an image of the glass suction electrode placed on the proximal end of the cut VIIIth nerve for electrical stimulation. Shown to the right are schematic illustrations of the obliquely cut spinal cord with the transverse surface facing up towards the objective. Inset shows an image of CGDA-labelled SPNs, pseudocoloured as described in Methods, in a transverse section of the T10 segment. Scale bars: 100 μm; D, dorsal; M, medial. B and C, effective stimulation parameters for evoking Ca²⁺ responses in sympathetic preganglionic neurons (SPNs) in T10. The graph in B displays the magnitudes of the responses produced by stimulating the VIIIth nerve at threshold current (T) with a 5 s train as a function of frequency. Each response is an average of 4 experiments (4–8 SPNs per experiment). The graph in C displays the magnitude–frequency curves when stimulating at 2T with a 5 s train, and the response magnitudes as a function of the train duration when stimulating at 2T and 5 Hz (individual open circles). Each response is an average (total of 4 experiments, 4–9 SPNs per experiment) and is expressed as a percentage of the response at 5 Hz. All values are means ± standard deviations.

Electrical stimulation

The entire eighth (VIIIth) cranial nerve was stimulated electrically with a suction electrode made from borosilicate glass (Harvard Apparatus, Kent, UK; cat. no. 30–0056). Nerve stimulation consisted of 5 s trains (200 μs pulse duration) at 5 Hz, with a current strength between 60 and 250 μA. In an additional series of experiments (n= 18), we stimulated the LVST group, which is the source of the LVST, lying principally within the lateral vestibular nucleus (LVN; Díaz et al. 2003; Pasqualetti et al. 2007; and references therein) using a monopolar tungsten microelectrode (Parylene-coated, shaft diameter 0.254 mm, tip diameter 1–2 μm, impedance 0.1 MΩ at 1 kHz; WPI, USA). To gain access to the LVN, we first carefully removed the cerebellum and the dura mater on the dorsal surface of the brainstem. For LVST group stimulation, we used the same parameters as for VIIIth nerve stimulation, except that current strengths were lower (50–200 μA, mean = 65 μA). The locations of stimulation sites in the LVN were confirmed histologically after making electrolytic lesions (see Szokol & Perreault, 2009 for details). In some of these preparations (n= 4), we compared directly the location of the electrolytic lesion to the location of the LVST neurons by labelling these retrogradely on the side opposite to the stimulation by applying crystals of rhodamine dextran amine (RDA; 3 kDa; Life Technologies Ltd, Paisley, United Kingdom) to their transected axons in the ventral and ventrolateral funiculi at C1 immediately after making the electrolytic lesion. In 3 out of 18 preparations we also blocked glutamatergic synapses selectively in the brainstem using kynurenic acid (see below).

Pharmacological experiments

After isolating the brainstem and the spinal cord with a Vaseline barrier made at the C1 level, bicuculline (20 μm; Tocris, Bristol, UK), picrotoxin (40 μm; Sigma, St. Louis, USA), strychnine (0.2 μm; Sigma) or mephenesin (1 mm; Sigma) were superfused into the compartment containing the spinal cord; or kynurenic acid (5 mm, Sigma) was superfused into the compartment containing the brainstem. The tightness of the barrier was verified by adding the dyes Phenol Red or Fast Green to one of the compartments at the end of the experiment. Recordings were started 10 min after drug application.

Spinal cord, medial longitudinal fasciculus (MLF) and medullary lesions

Spinal cord hemisections were made just rostral to the C1 ventral root, and longitudinal lesions of the midline spanned rostrocadually across two segments (T9 and T10). Lesions of the MLF were made from the dorsal surface of the brainstem by transecting the fasciculus bilaterally at about 200 μm rostral to the obex. The MLF lesion extended about 150 μm from each side of the midline, encompassing the entire mediolateral extent of the MLF in neonatal mice (Paxinos et al. 2007; and our own anatomical observations). The MLF lesions effectively eliminate medial vestibulospinal tract (MVST) projections as well as many reticulospinal projections originating rostral to the lesion. Large medullary lesions (Fig. 2) were designed to ablate the regions containing the ipsilateral and contralateral CVLMs and RVLMs, all contralateral vestibulospinal tracts, while sparing the ipsilateral LVST. This was done by first removing nearly the entire contralateral medulla down to C1, thus eliminating the CVLM, RVLM and all vestibulospinal tracts on that side. Then, a large portion of the lateral ipsilateral medulla was removed spanning from below the LVST group (whose location we are very familiar with through anatomical mapping in several previous studies in addition to this one) to the C2/C3 boundary, extirpating the region containing the CVLM and RVLM according to their locations in the rat and mouse (Willette et al. 1983; Paxinos & Watson, 1986; Lipski et al. 1996; Tolentino-Silva et al. 2000), sparing medially the LVST and the MLF. Optical recordings of calcium responses in SPNs and somatic MNs (see below) were always performed both before and after each of these two lesions. The various lesions were made using either fine iridectomy scissors (spinal cord hemisections, midline lesions, medullary lesions) or a custom-made wedged blade (midline lesions, MLF lesions), and their locations and spatial extents were confirmed histologically.

Figure 2. Histological demonstration of large medullary lesions designed to eliminate the CVLM and RVLM.

This lesion was designed to eliminate the region containing the CVLM and RVLM on both sides (and any descending axons originating from the RVLM) but spare axons descending in the LVST on one side. A, lesioned brainstem and cervical cord as seen in whole-mount, ventral view. B, left, transverse sections (50 μm) taken at the levels indicated in A, stained with methylene blue to show the extent of the remaining tissue through which the LVST projects. Levels 4 and 5 are at the rostral end of C1 and caudal end of C2, respectively. right, corresponding sections from an intact brainstem.

Histology

Preparations were fixed overnight in 4% paraformaldehyde, cryoprotected in 20% sucrose, embedded in OCT (Tissue-Tek Sakura, Alphen aan den Rijn, The Netherlands) and frozen. Serial cryostat (Leica, Wetzlar, Germany CM3050S) sections (50 μm) made in the transverse, sagittal or frontal plane were stained with Methylene Blue (0.05 % w/v Methylene Blue, Sigma) and mounted on Super Frost glass slides with a gelatin–glycerol medium.

Data analysis

The timing markers for the VIIIth nerve stimulation and the gating pulse from the CCD camera were recorded at 5 kHz (Digidata1320A; Molecular Devices, Sunnyvale, USA). To synchronize the electrical and the optical recordings, a digital pulse from the Digidata unit was sent to the CCD camera. Using the MetaMorph software, circular digital apertures of identical size and shape (regions of interest or ROIs) were placed manually over all the individual neuron somata that were clearly in focus. To compensate for variability in the CGDA labelling intensity between the different neurons, the change in fluorescence (ΔF) in each ROI was normalized to the baseline fluorescence F₀ before the stimulation (ΔF/F₀= (F−F₀)/F₀). The ROI files were converted to text files using a custom-made program (FileConvert) and imported into Clampfit 9.2 (Molecular Devices) where the data were expressed as waveforms. The magnitudes of the responses were measured by calculating the areas under the waveforms that exceeded the mean value of the pre-stimulus baseline by 2 SD. Alignment of optical waveforms and stimulus markers in the illustrations was done manually using Corel Draw X5 (Corel Corp, Canada). Unless indicated otherwise, all data are presented as grand means across preparations ± SEM.

Results

Optical recordings of calcium responses from individual SPNs in the intermediate motor column of the thoracic segments were obtained in a total of 74 animals (T2, T4, T6, T8 and T12, n= 3 for each; and T10, n= 65). Recordings were made from ipsilateral SPNs (n= 44 animals) and/or contralateral SPNs (n= 42 animals). In 9/74 animals, recordings were obtained from SPNs in more than one segment (T2, T6 and T12, n= 3; T4, T8 and T10, n= 3; T2 and T4, n= 3). In 6 out of the 65 animals where T10 SPNs were recorded, we also recorded from the somatic MNs in the medial motor column of the same segment. A total of 470 neurons (422 SPNs and 48 somatic MNs) were analysed in the present study.

Determination of effective stimulation parameters for VIIIth nerve-evoked responses in SPNs

Stimulation of the VIIIth nerve with 5 s trains at 5 Hz readily evokes reproducible Ca²⁺ responses in somatic MNs of the thoracic cord in the neonatal mouse (Kasumacic et al. 2010). To determine whether these parameters are also effective in evoking responses in SPNs, we systematically varied stimulation parameters (Fig. 1B and C). With 5 s train stimulation, the magnitudes of the evoked responses in the SPNs increased with increasing stimulus frequency (Fig. 1B). While 5 s trains at 1 Hz were often sufficient to evoke Ca²⁺ responses in SPNs, the responses were usually small (sometimes falling below the detection level) and required high-threshold currents (200–320 μA, mean = 273 ± 32 μA; Fig. 1B). In contrast, 5 s trains at 5 Hz always evoked easily detectable responses and required substantially lower threshold currents (100–150 μA, mean = 110 ± 13 μA). Further increase in stimulus frequency to 10 or 20 Hz did not substantially decrease the threshold current. Stimulation at twice the current threshold (2T) for evoking a measurable increase in CGDA fluorescence (as defined in Methods) showed that 5 s trains were more effective than trains of shorter duration (shown as individual white circles in Fig. 1C). Because increasing the train duration to 10 s only increased response magnitudes by about 30%, we settled on 5 s as an appropriate train length.

Altogether these experiments indicate that 2T stimulation of the VIIIth nerve with 5 s trains at 5 Hz reliably produces responses in SPNs (as in somatic MNs), and therefore all experiments described below were performed using these specific stimulus parameters.

Potential contribution of an auditory component to the responses in SPNs was considered negligible on two grounds: (1) lesion of the MLF, which carries the pontine reticulospinal axons that mediate the spinal component of the acoustic startle reflex (Yeomans & Frankland, 1996), did not affect responses (this and other lesion experiments are described below); and (2) although noise can elicit increases in sympathetic tone through stress reactions in rodents (Alario et al. 1987), we have not found any reports that demonstrate unequivocally an auditory input to the CVLM or RVLM (see, however, Blair, 1991).

Pattern and latency of responses in SPNs

As shown in Fig. 3, VIIIth nerve stimulation evoked Ca²⁺ responses in ipsilateral and contralateral SPNs in all the thoracic segments investigated (total of 422 SPNs). Response magnitudes tended to increase with more caudal segmental location. To examine this further, in six preparations where SPNs were labelled in multiple thoracic segments (see above), we compared the response magnitudes in the SPNs of the most rostral segment (T2 or T4, pooled) with those in the most caudal segment (T10 or T12, pooled). The response magnitudes in the SPNs of T10 and T12 were significantly larger than those in the SPNs of T2 and T4 (Mann–Whitney, U= 15, P= 0.010).

Figure 3. Responses evoked in sympathetic preganglionic neurons (SPNs) in different spinal segments.

Responses recorded at 4 frames s⁻¹ of ipsilateral (right) and contralateral (left) SPNs in segments T2, T4, T6, T8, T10 and T12 during stimulation of the VIIIth nerve (grey vertical bars, 5 s trains at 5 Hz and 2T). Each waveform shows the average of recordings from 4 SPNs.

To estimate the latencies of the vestibular-evoked responses in SPNs, we recorded at a higher temporal resolution (100 frames s⁻¹ rather than 4 frames s⁻¹, n= 10 animals) and measured the time from the onset of the stimulation train to the onset of the responses. As shown in Fig. 4, the vestibular-evoked responses had a mean latency of 131 ± 13 ms in ipsilateral SPNs (n= 4 animals for a total of 11 SPNs; black circles in Fig. 4A1 and top trace in Fig. 4A2) and 132 ± 21 ms in contralateral SPNs (n= 4 animals for a total of 13 SPNs; black circles in Fig. 4B1 and top trace in Fig. 4B2). The difference in response latencies between the ipsilateral and contralateral SPNs was not statistically significant (Mann–Whitney, U= 72, P= 0.976). In addition to their long latencies, the responses in SPNs rose slowly, reaching peak amplitude several seconds after the onset of stimulation. Because CGDA should be able to follow transients that rise within 1–2 ms or slower (Naraghi, 1997), these features suggested that vestibular-evoked responses in thoracic SPNs were mediated predominantly by polysynaptic pathways.

Figure 4. Latencies and waveforms of vestibular-mediated responses in thoracic sympathetic preganglionic neurons (SPNs) and somatic motoneurons (MNs).

Cumulative distribution of latencies measured in ipsilateral (A1) and contralateral (B1) SPNs (black circles) and somatic MNs (grey circles) recorded at 100 frames s⁻¹. Both ipsilateral and contralateral somatic MNs responded with significantly shorter latencies than their sympathetic counterparts. There was no significant difference in the responses of ipsilateral vs. contralateral somatic MNs or ipsilateral vs. contralateral SPNs (see main text for statistics). A2, waveforms of Ca²⁺ responses recorded in ipsilateral SPNs (top, black waveform) and somatic MNs (bottom, grey waveform) in T10 during nerve VIII stimulation. Each waveform is the average of responses from 6 neurons. B2, waveforms of Ca²⁺ responses recorded in contralateral SPNs (top, black waveform) and somatic MNs (bottom, grey waveform) in T10 during nerve VIII stimulation.

We then compared the latencies of the responses in SPNs and the responses in the somatic MNs of the same segment. Mean response latencies were 67 ± 4.1 ms in ipsilateral somatic MNs (n= 4 for a total of 12 MNs; grey circles in Fig. 4A1 and bottom trace in Fig. 4A2) and 72 ± 3.8 ms in contralateral somatic MNs (n= 5 for a total of 16 MNs; grey circles in Fig. 4B1 and bottom trace in Fig. 4B2). These response latencies in ipsilateral and contralateral somatic MNs were not statistically different (Mann–Whitney, U= 112, P= 0.47), but both were shorter than the response latencies in the SPNs of the same side (Mann–Whitney, U= 2, P < 0.0001 for ipsilateral SPNs vs. MNs; and U= 13, P < 0.0001 for contralateral SPNs vs. MNs). The difference in response latency between SPNs and somatic MNs suggested that separate pathways from vestibular nuclei to spinal cord mediated the responses in these two classes of thoracic neurons.

Mephenesin applied selectively to the spinal cord abolishes vestibulosympathetic responses

To investigate further the polysynapticity of the descending connections mediating the vestibulosympathetic responses, we examined the effects of mephenesin, a drug that diminishes the efficacy of synaptic transmission in both mono- and polysynaptic pathways, but preferentially in the latter (Floeter & Lev-Tov, 1993; Vinay et al. 1995; Juvin & Morin, 2005). The aim of the experiments was to determine the degree to which the spinal portion of the pathway is sensitive to mephenesin and might therefore be polysynaptic. In four animals, the brainstem was isolated from the spinal cord by a Vaseline barrier at C1, and vestibular-evoked responses in T10 SPNs were recorded before and after selectively applying mephenesin (1 mm) to the spinal cord. As shown in Fig. 5, mephenesin completely eliminated the responses in both ipsilateral and contralateral SPNs (total of 17 SPNs). Vestibular-evoked responses in thoracic somatic MNs are mediated mostly by axons descending in the LVST with little involvement, if any, from circuits in the brainstem, and they are known to have a monosynaptic component (Kasumacic et al. 2010). Thus, as a positive control for a differential action of mephenesin on monosynaptic and polysynaptic connections in the spinal cord, in three of the preparations we also examined the effect of mephenesin on the vestibular-evoked responses in T10 somatic MNs. Responses in these somatic MNs were substantially diminished but not completely eliminated by mephenesin (not shown, total of 32 MNs). These results substantiate the idea that vestibulosympathetic responses in thoracic SPNs are mediated predominantly by polysynaptic vestibulospinal pathways, and that the spinal portion of the pathway is itself polysynaptic.

Figure 5. Mephenesin applied to the spinal cord abolishes vestibulosympathetic responses.

Responses evoked in T10 sympathetic preganglionic neurons (SPNs) by VIIIth nerve stimulation (2T, 5 s, 5 Hz) before (control) and after application of 1 mm mephenesin to the thoracolumbar spinal cord. Responses were recorded at 4 frames s⁻¹, and each waveform is the average of responses from 5 SPNs.

Effects of GABA_A and glycine receptor antagonists on vestibulosympathetic responses

It has been suggested that the descending pathways that mediate the vestibulosympathetic responses in SPNs involve combined inhibitory and excitatory inputs (Kerman et al. 2000b). However, this has been deduced from extracellular recordings of sympathetic nerve fibre activity without substantiation at the synaptic level. With calcium imaging, excitatory synaptic connections are revealed by postsynaptic depolarization and subsequent activation of voltage-sensitive Ca²⁺ channels. Convergent inhibitory synaptic connections that would counteract excitatory synaptic potentials and diminish Ca²⁺ responses can be revealed indirectly by the increase of Ca²⁺ response that occurs in the presence of inhibitory neurotransmitter receptor blockers. Thus, to test for vestibular-mediated inhibitory inputs onto SPNs, we examined the effects of the GABA_A receptor antagonists bicuculline and picrotoxin, and the glycinergic receptor antagonist strychnine on the responses evoked in SPNs of the T10 segment (n= 18 animals). As shown in Fig. 6A and B, bicuculline (20 μm) and picrotoxin (40 μm) each increased the response magnitudes in ipsilateral SPNs by more than 100% (Mann–Whitney, bicuculline: U= 29, P= 0.03; picrotoxin: U= 16, P= 0.01), whereas strychnine (0.2 μm) had no significant effect (Mann–Whitney, strychnine: U= 11, P= 0.65). Figure 6A and C shows similar effects on the responses evoked in contralateral SPNs (Mann–Whitney, bicuculline: U= 29, P= 0.04; picrotoxin: U= 16, P= 0.01; strychnine: U= 8, P= 0.55). These data demonstrate a clear contribution from GABAergic inhibitory synaptic inputs. Recordings at 100 frames s⁻¹ (not shown) showed no significant difference in the latencies of the responses before and after blockade of GABAergic inputs (ipsilateral SPNs: Mann–Whitney, bicuculline: U= 18, P= 0.04; picrotoxin: U= 20, P= 0.18; strychnine: U= 20, P= 0.61; contralateral SPNs: bicuculline: U= 18, P= 0.16; picrotoxin: U= 20, P= 0.2; strychnine: U= 20, P= 0.73).

Figure 6. Effect of bicuculline, picrotoxin and strychnine applied to the spinal cord on vestibulosympathetic responses.

A, waveforms showing responses recorded at 4 frames s⁻¹ in ipsilateral (top) and contralateral (bottom) sympathetic preganglionic neurons (SPNs) in T10 before and during application of 20 μm bicuculline, 40 μm picrotoxin or 0.2 μm strychnine to the thoracolumbar spinal cord. Each waveform is an average of responses in 5 SPNs. B and C, response magnitudes in ipsilateral (B) and contralateral (C) T10 SPNs during application of bicuculline, picrotoxin or strychnine normalized to control responses. Bicuculline and picrotoxin lead to an increase in response magnitudes on both sides, whereas strychnine has no effect (see main text for statistics).

Pathways mediating the vestibulosympathetic responses

It cannot be assumed that network architecture in newborn mammals is the same as in adults. To investigate the axonal pathways that mediate the vestibulosympathetic responses in the neonatal mouse, we first considered whether, in contrast to adult mammals, there might be direct vestibulospinal projections to SPNs by assessing the extent to which direct stimulation of the LVST generated responses in ipsilateral SPNs (Fig. 7). We then investigated the effects of various types of lesions in the brainstem and cervical cord on responses in ipsilateral and contralateral SPNs (Fig. 8).

Figure 7. Direct stimulation of the lateral vestibulospinal tract (LVST) group within the LVN evokes small responses in sympathetic preganglionic neurons (SPNs) that are blocked by kynurenic acid applied selectively to the brainstem.

A1 and A2, frontal sections showing the distribution of vestibulospinal neurons and axons labelled retrogradely with RDA from C1, and the location of an electrolytic lesion (white arrow) used to determine the location of the stimulation site in one of the preparations in which the LVST was stimulated directly. A3, schematic showing the distribution of all stimulation sites (black squares) relative to retrogradely labelled LVST group neurons. Scale bars: 50 μm. B, responses evoked in ipsilateral SPNs (black) by stimulation (5 s, 5 Hz) within the LVST group. The responses were smaller in magnitude and shorter in duration relative to responses elicited by stimulation of the VIIIth nerve. These responses were blocked by 5 mm kynurenic acid. Responses in the ipsilateral somatic motoneurons (MNs; grey) were similar relative to the responses elicited by stimulation of the VIIIth nerve. These responses were not blocked by 5 mm kynurenic acid. C, latencies of responses in SPNs (black circles) and somatic MNs (grey circles) elicited by stimulation within the LVST group. cMVST, contralateral medial vestibulospinal tract; iMVST, ipsilateral medial vestibulospinal tract.

Figure 8. Effect of spinal cord, medial longitudinal fasciculus (MLF) and medullary lesions on vestibulosympathetic responses.

A, effects of ipsilateral and contralateral hemisections, midline spinal lesions, bilateral MLF lesion and medullary lesions on responses in ipsilateral and contralateral SPNs in T10 segment. Responses were recorded at 4 frames s⁻¹. Each waveform is an average of responses in 5 SPNs. B and C, response magnitudes in ipsilateral (B) and contralateral (C) T10 SPNs following spinal cord, MLF or medullary lesions, normalized to control responses. The spinal cord lesions had differential effects on the ipsilateral and contralateral SPNs, the MLF lesion had no effect on responses on either side, the contralateral medullary lesion alone had an effect similar to that of the contralateral hemisection, and the combined contralateral plus ipsilateral medullary lesions abolished responses (for details see main text).

To directly stimulate the LVST, we inserted a stimulation electrode into the LVN where the LVST group, source of the LVST, resides. Electrode placement was made based on external landmarks that we had correlated with the location of the LVST group in earlier anatomical and physiological studies of embryonic and neonatal mice (Pasqualetti et al. 2007; Kasumacic et al. 2010). Stimulation sites were marked by electrolytic lesions, and in a few preparations we confirmed correct electrode placement by retrogradely labelling the LVST group at the end of the experiment. Because the LVST group lies in close proximity to and partly overlaps the VIIIth nerve entry zone, stimulation of the group also risks stimulating vestibular afferents, an obvious confounding factor. Thus, we first asked whether direct stimulation of the LVST group generated the same response pattern as VIIIth nerve stimulation in thoracic segments. We found that, compared with stimulation of the VIIIth nerve, stimulation within the LVST group generated responses in somatic MNs of similar magnitude, but smaller and shorter responses in SPNs. This suggested straightaway that the pathways mediating VIIIth nerve-mediated responses in somatic MNs and SPNs must differ, as has been shown in adult mammals, as LVST stimulation mimicked VIIIth nerve stimulation for somatic MN responses but not for SPN responses. Because the small SPN responses could have been due either to activation of the LVST or of a small number of passing afferents, we next sought to stimulate the LVST group while blocking all excitatory synaptic transmission within the brainstem with kynurenic acid. This would be expected to eliminate any afferent contribution to the responses. We found that kynurenic acid (5 mm) applied selectively to the brainstem completely eliminated the small SPN responses to LVST stimulation, and diminished but did not eliminate the somatic MN responses. Washout of the kynurenic acid from the brainstem compartment was partial in 3 out of 4 experiments, but in the remaining experiment we were able to reverse these effects completely. These experiments suggested that if a direct LVST-to-SPN connection exists, it provides only a minor contribution.

Because direct LVST stimulation provided little support for a direct vestibulospinal component, we performed the following lesions to better characterize the nature of what obviously was predominantly an indirect projection: (1) ipsi- or contralateral cervical hemisection at C1, which interrupted transmission along all ipsi- or contralaterally descending pathways (n= 4); (2) midline lesion at T9–T10, which interrupted transmission along segmental, midline-crossing axons or dendrites (n= 4); (3) bilateral MLF lesion at medullary levels, which interrupted transmission via the MVSTs (ipsilateral (i)MVST and contralateral (c)MVST) and all other MLF-projecting descending pathways originating from above the level of the lesion (n= 20 animals); and (4) medullary lesions designed to eliminate brainstem circuitry encompassing the CVLM and RVLM as well as descending projections from the RVLM on both sides, but to preserve descending projections from the ipsilateral LVST (as well as from the iMVST and raphe nuclei, n= 4). Figure 8A shows examples of responses in ipsilateral and contralateral SPNs before and after these various lesions.

Ipsilateral and contralateral hemisections at C1 each diminished vestibular-mediated responses in ipsilateral SPNs (to 44 ± 7% and 72 ± 13%, respectively, n= 4) and contralateral SPNs (to 92 ± 13% and 49 ± 11%, respectively, n= 4; Fig. 8B and C). Responses were not affected by bilateral MLF lesions. Thus, responses in SPNs on either side of the spinal cord are mediated by a combination of ipsilaterally and contralaterally descending pathways that course lateral to the MLF (within the brainstem), with the major contribution coming from the pathway descending on the same side as the SPN.

Thoracic midline lesions diminished responses only in contralateral SPNs (to 78 ± 2%, n= 4).

The medullary lesions designed to assess whether the responsible descending projections derive from the RVLM or from a direct LVST projection were made in two steps. The first, which removed nearly the entire contralateral medulla, produced an SPN response decrement similar to that seen after C1 contralateral hemisection (n= 4; Fig. 8B and C). The second, which removed the lateral part of the ipsilateral medulla (and hence the entire ipsilateral CVLM and RVLM, as well as projections into and out of these regions), led to a complete loss of all SPN responses but did not eliminate responses in somatic MNs (n= 4). This showed that the ipsilaterally descending projection must include elements within the region that contains the CVLM and RVLM, and that neither the ipsilateral raphespinal nor the LVST projection is sufficient to elicit SPN responses.

To summarize, the vestibular-evoked responses in ipsilateral and contralateral SPNs were mediated respectively by predominantly ipsilateral and predominantly contralateral indirect descending pathways that include the ipsilateral region containing the CVLM and RVLM (Fig. 9). We did not test whether responses in contralateral SPNs depend on the contralateral region containing the CVLM and RVLM. In contrast to responses in ipsilateral SPNs, responses in contralateral SPNs also depend on excitatory commissural projections at thoracic levels. These may arise from commissural collaterals from descending axons, excitatory commissural interneurons (CINs) or both.

Figure 9. Summary diagram of brainstem vestibulosympathetic circuitry supported by the present experiments.

Summary of the brainstem circuit believed to underly the vestibulosympathetic response (A), its relationship to the lateral vestibulospinal tract (LVST) group and tract (B), and the impact of the medullary lesions described under Fig. 2 on the various projections involved (C). Flow of impulse traffic is shown by arrows, with dotted lines indicating connections that are not well characterized in the literature (caudal ventrolateral medulla (CVLM) to rostral ventrolateral medulla (RVLM)) or only postulated from our results (RVLM to RVLM). The precise descending trajectory of spinally projecting RVLM neurons has not been described; it is shown here in the MLF, but may lie further lateral. Vestibulosympathetic impulses derive from neurons in the medial vestibular nucleus (MVN) and inferior vestibular nucleus (IVN) MVN and IVN, but only the MVN is shown here for simplicity. B, the LVST group and the proximal portion of its derivative tract lie dorsal to the medullary reticular formation. The vestibular afferents have been removed from B and C for clarity. C, although not shown here, the ipsilateral lesion actually extends down to the border between cervical segments C2 and C3.

Discussion

Using calcium imaging, we have shown that vestibulosympathetic synaptic connections are present in the mouse, and are already functional at birth. We have further shown that the projections are organized as in the adult mammal, in the sense that they involve circuitry within the brainstem critically dependent on neurons present in the region containing the CVLM and RVLM, with no obvious contribution from direct vestibulospinal pathways (although see below). A novel finding not previously documented in adult mammals is an indication that the spinal component of the pathway may be polysynaptic, which we have obtained through selective pharmacological treatment of the spinal cord with mephenesin. The differential effects of this treatment on responses in SPNs and thoracic somatic MNs accords with the longer response latencies of SPNs and the known monosynaptic component of vestibulospinal inputs to somatic MNs. Lastly, we show that the vestibulosympathetic responses occur within a context of convergent excitation and inhibition, as they are augmented by pharmacological blockade with GABA_A blockers. One potential explanation for this is that vestibular stimulation activates mixed excitatory/inhibitory synaptic inputs, as previously deduced (but not directly demonstrated) in the adult mammal (but see below).

Technical considerations related to latency measurements

Several technical issues regarding the use of Ca²⁺ imaging to characterize synaptic connectivity have been addressed in our previous studies of brainstem–spinal cord projections (Szokol et al. 2008, 2011; Szokol & Perreault, 2009; Kasumacic et al. 2010) and will not be discussed here. However, there is one point that warrants additional discussion. By using relatively high frame rate imaging (100 frames s⁻¹) we have been able to estimate the latencies of the evoked responses with a temporal precision of ±10 ms. Although this does not permit us to definitively gauge the number of synapses involved in polysynaptic responses, as might be possible with electrophysiological approaches, it does allow us to distinguish clearly the response latencies of SPNs (about 130 ms) from those of somatic MNs (about 70 ms). We can therefore conclude that the former must involve more synaptic relays or longer total conduction times or both. As discussed previously by us and others, conduction times cannot be assessed accurately in the neonatal rodent spinal cord because descending volleys cannot be readily monitored (Floeter & Lev-Tov, 1993; Szokol et al. 2008, 2011). However, because we saw a residual response in somatic MNs but not SPNs after selective treatment of the spinal cord with mephenesin, at least part of the difference may lie in a greater degree of polysynapticity in the spinal portion of the vestibulosympathetic projection. Mephenesin is thought to act generally by increasing activation thresholds, an effect that would be compounded by the addition of neuronal relays. Although the effect is not specific for types of neurons or receptors, it appears to be preferential for polysynaptic pathways. Of course, much of the longer response latency in SPNs is explained by the more complex brainstem circuitry involved in the vestibulosympathetic responses compared with the vestibulosomatic responses.

Although we saw no change in response latencies following selective GABA_A receptor blockade in the spinal cord, the temporal uncertainty of the response latency measurements does not allow us to preclude the possibility that inhibitory events (which can be invisible to Ca²⁺ recording if occurring in isolation) might precede excitatory events, as seen in a few instances in the adult cat (Kerman et al. 2000b). Thus, our latency measurements may be overestimates if the actual responses involve inhibitory inputs that fall within about 10 ms before excitatory inputs.

Comparing vestibulosympathetic responses in the newborn mouse and adult mammals

Response latencies

We have recorded responses in ipsilateral and contralateral sympathetic MNs during electrical stimulation of the VIIIth nerve, with response latencies averaging about 130 and 140 ms, respectively. These latencies are similar to those that have been recorded from the splanchnic nerve of the adult cat during VIIIth nerve stimulation (Yates et al. 1993), but are substantially longer than those recorded from the splanchnic nerve of the adult rat during VIIIth nerve stimulation (about 45 ms; Pan et al. 1991). In general, longer latencies are to be expected in neonates due to lack of myelination and immaturity of synapses. The large discrepancy between latencies recorded in adult cats and adult rats could in principle be explained by different conduction distances (about fourfold longer in cats than rats) if the descending axons involved are slowly conducting. Indeed, Deuchars et al. (1995) have estimated that conduction velocities in the RVLM–SPN pathway in the rat are within the range for C fibres (about 0.5 m s⁻¹).

Pathways mediating vestibulosympathetic responses

Based on work in the cat, Yates (1996) proposed a model for vestibulosympathetic circuitry in which head pitch-sensitive receptors activate neurons in the medial and inferior vestibular nuclei, which then via a relay in the CVLM activate neurons in the autonomic control centre in the RVLM, which have been shown to project to the spinal cord and synapse onto SPNs. More recently, the model has been augmented by the demonstration of a direct connection from vestibular nuclei to the RVLM (reviewed in Holstein et al. 2011). Much of the foundation for this model is based on the attempted elimination of various parts of the brainstem through electrolytic or neurochemical lesions (Pan et al. 1991; Yates & Miller, 1994; Yates et al. 1995). Although very indicative, the use of lesions has several caveats. First, because it is hard to control the extent of lesions in vivo, it can be difficult to conclude that the observed effects are due to damage to the target region or nucleus and not other nearby structures. Second, lesioned regions that do not result in changes in the vestibulosympathetic responses cannot, without further investigation, be excluded as potential components of the reflex if the reflex involves parallel pathways. The LVN, for example, has been excluded as a potential component of the vestibulosympathetic reflex based on lesion studies, a conclusion supported by the argument that the medial and rostral descending vestibular nuclei are more suitable candidates because they have a somewhat larger proportion of pitch-sensitive cells than the LVN (Yates, 1996) and respond preferentially to sGVS (Holstein et al. 2012). However, Lee et al. (2007), using transynaptic labelling with pseudorabies virus to trace the source of supraspinal inputs to SPNs, found labelling in the LVN that looked no less direct (based on labelling latency) than the labelling they found in the RVLM (see fig. 2 in their paper).

Because brainstem–spinal cord circuitry may not be mature in newborn mammals, we carried out a series of lesion experiments to assess the pathways involved in the vestibulosympathetic responses in the neonatal mouse. Because we made them in ex vivo preparations, we had much better control over lesion placement than is the case for in vivo lesions. In particular, we wanted to test whether direct vestibulospinal connections to SPNs exist. Our findings provide no support for the involvement of a direct vestibulospinal projection. Direct stimulation of the LVN/LVST produces only very short and small responses in SPNs, and lesions that spare the LVST but eliminate the CVLM and RVLM abolish the responses. MVST axons do not reach the thoracic cord and do not recruit thoracic somatic MNs in the neonatal mouse (Kasumacic et al. 2010), but could in principle engage propriospinal interneurons to activate SPNs. However, lesions of the iMVST and cMVST (made by transecting the MLF) do not affect the responses. We have not stimulated directly the raphe nuclei but, according to Iigaya et al. (2007), only few raphespinal neurons, if any, have functional synaptic connections with SPNs in the neonatal rat.

Thus, as in adult mammals, the vestibulosympathetic control of SPNs in the neonatal mouse involves predominantly, if not exclusively, indirect projections that are channeled to the spinal cord through the region that contains the CVLM and RVLM (see summary diagram in Fig. 9).

Vestibulosympathetic responses are clearly bilateral in adult cat, adult rat and neonatal mouse. This is likely due in part to commissural collaterals from RVLM axons to the intermediolateral region of the thoracic spinal cord, as shown in the adult rat (Moon et al. 2002; and see below). Our lesion experiments indicate in addition a bilateral influence within the brainstem, as ipsilateral hemisection at C1 did not abolish responses in contralateral SPNs. Clearly, there must be a commissural connection at some level of the brainstem vestibulosympathetic circuit that channels activity from vestibular afferents to the contralateral RVLM. This could occur either through vestibulo-vestibular connections, connections between the two sides of the lateral reticular formation, connections between the two RVLMs, or commissural connections transcending these levels. A more comprehensive set of lesion experiments could be envisioned to test these possibilities once sufficient information about the trajectories of the relevant axons is available.

We show further that mephenesin, when applied to the spinal cord, completely abolishes the vestibular-evoked responses in SPNs, but not in somatic MNs. This indicates that the pathways mediating vestibulosympathetic responses and vestibulosomatic responses must differ, with the latter relying strongly on direct vestibulospinal projections. This is consistent with the findings of Miyazawa & Ishikawa (1985), who showed that a lesion of laterally descending fibres at the T3 level in the adult cat abolished vestibulosomatic responses without affecting vestibulosympathetic responses.

It has been reported from experiments in the adult cat that vestibulosympathetic responses are composed of inhibitory as well as excitatory components (Kerman et al. 2000b). We find that application of bicuculline and picrotoxin, but not strychnine, selectively to the spinal cord leads to an increase in the magnitude of vestibulosympathetic responses. This indicates the presence of an inhibitory component in the response mediated by GABA (through GABA_A receptors) but not glycine. Whether this component is part of the vestibulosympathetic pathway or a tonic input from other sources remains to be determined. As discussed previously (Kasumacic et al. 2010), an increase in vestibular-mediated Ca²⁺ responses after blockade of GABA_A receptors could be attributed to a direct increase in the excitability of the target neurons and/or to disinhibition. To evaluate these possibilities it will be important in future experiments to determine whether the inhibitory component is mediated by a direct descending GABAergic projection, by intercalated GABAergic spinal interneurons, or by tonically active GABAergic inputs, and to characterize in more detail the actions of GABA (including those mediated by GABA_B receptors; Wang et al. 2010) on SPNs and on excitatory and inhibitory interneurons associated with the vestibulosympathetic pathway.

Segmental distribution

Several studies have shown that vestibulosympathetic effects are larger in hindlimb than forelimb vasculature, indicating a regional pattern within the vestibulosympathetic reflex (Kerman et al. 2000a; Wilson et al. 2006; Sugiyama et al. 2011). The way this is achieved in the spinal cord is not well understood, but could involve the well-characterized segmental mapping of sympathetic outflow in which more caudal thoracic segments contain greater proportions of SPNs projecting caudally within the sympathetic trunk (reviewed in Forehand et al. 1994). This topographic efferent arrangement can be easily accessed by descending axons through the distinctive intrasegmental segregation within each segmental cohort of SPNs projecting rostrally versus caudally in the sympathetic trunk (Forehand et al. 1994, 1998). Here, we have shown that vestibulosympathetic responses as assessed with Ca²⁺ imaging are larger in caudal than in rostral thoracic segments; indeed, there appears to be a trend of increasing magnitude that suggests a rostrocaudal gradient of inputs to SPNs. Such a gradient could clearly contribute to a larger response in the hindlimb. It will be important in this respect to investigate how these different response magnitudes are generated, including assessing the density of RVLM-derived axon collaterals and synaptic terminals and characterizing electrophysiologically the synaptic drive in the different thoracic segments.

Spinal circuitry involved in mediating vestibulosympathetic responses

Our mephenesin experiments suggest that transmission of vestibular information to SPNs involves multiple synapses within the spinal cord. The potential role of spinal interneurons in vestibulosympathetic circuitry has recently been addressed by Miller et al. (2009), who found interneurons in the thoracic region that respond to vestibular nerve stimulation and to head tilts. Whether these interneurons are necessary or sufficient components of the vestibulosympathetic reflex circuit has not been demonstrated. It is not clear from our experiments, for example, whether the mixed excitatory/inhibitory nature of the vestibulosympathetic responses arises through convergence of excitatory and inhibitory descending projections or through the engagement of excitatory and inhibitory spinal interneurons by purely excitatory descending projections. Further investigation of RVLM inputs to spinal interneurons is therefore warranted.

Our finding that midline lesion of the thoracic spinal cord effectively reduces responses in contralateral SPNs implies in addition that a proportion of the ipsilaterally descending RVLM axons either extend collaterals across the midline at the thoracic level or innervate spinal CINs, which relay the effects to the opposite side. Anatomical evidence exists for commissural axon collaterals (Moon et al. 2002), but the connectivity of RVLM axons to spinal interneurons including CINs has yet to be investigated.

Future perspectives

Although it is clear that there are functional synaptic connections between the vestibular system and SPNs in the mouse already at birth, it is not clear whether the vestibulosympathetic reflex has achieved a mature organization at birth. Future studies should be aimed at investigating the physiological function of the reflex in the newborn mouse in vivo, as has been done in the adult cat (Doba & Reis, 1974). Moreover, the specific connectivity of the brainstem and spinal cord circuitry involved is not yet fully worked out, a challenge that can be addressed advantageously in the neonatal mouse through combined anatomical, transgenic and optical approaches that target specific neuron populations.

Glossary

2T

twice the current threshold

aCSF

artificial cerebrospinal fluid

CGDA

Calcium Green 1-conjugated dextran amine

CIN

commissural interneuron

cMVST

contralateral medial vestibulospinal tract

CVLM

caudal ventrolateral medulla

iMVST

ipsilateral medial vestibulospinal tract

LVN

lateral vestibular nucleus

LVST

lateral vestibulospinal tract

MLF

medial longitudinal fasciculus

MN

motoneuron

MVST

medial vestibulospinal tract

RDA

rhodamine dextran amine

ROI

region of interest

RVLM

rostral ventrolateral medulla

SPN

sympathetic preganglionic neuron

sGVS

sinusoidal galvanic vestibular stimulation

T

thoracic

Author contributions

N.K., J.C.G. and M-C.P. designed and planned experiments. N.K. and J.C.G. carried out experiments. N.K., J.C.G. and M-C.P. analysed data, wrote the manuscript and approved the final version of the manuscript.

Author's present address

M-C. Perreault: Department of Physiology, Emory University School of Medicine, Whitehead Biomedical Research Building, 615 Michael Street, Atlanta, GA 0322, USA.