• Regional blood flow was measured with Doppler ultrasonic probes in anaesthetized rabbits. We used focal microinjections of pharmacological agents to investigate medullary pathways mediating ear pinna vasoconstriction elicited by electrical stimulation of the spinal tract of the trigeminal nerve or by pinching the lip, and pathways mediating mesenteric vasoconstriction elicited by electrical stimulation of the afferent abdominal vagus nerve.
• Bilateral injection of kynurenate into the rostral ventrolateral medulla reduced arterial pressure and prevented the mesenteric vasoconstriction and the rise in arterial pressure elicited by abdominal vagal stimulation. However, kynurenate did not prevent ear pinna vasoconstriction or the fall in pressure elicited by trigeminal tract stimulation. Similar injections of muscimol also failed to prevent the trigeminally elicited cardiovascular changes.
• Injections of kynurenate into the raphe–parapyramidal area did not diminish trigeminally elicited ear vasoconstriction or the depressor response. However, injections of muscimol substantially reduced or abolished the trigeminally elicited ear vasoconstriction, without affecting the depressor response. raphe–parapyramidal muscimol injections also entirely abolished ear vasoconstriction elicited by pinching the rabbit's lip.
• The trigeminal depressor response does not depend on either the rostral ventrolateral medulla or the raphe–parapyramidal region.
• Mesenteric vasoconstriction elicited by stimulation of the afferent abdominal vagus nerve is mediated via the rostral ventrolateral medulla, but ear vasoconstriction elicited by lip pinch or by stimulation of the trigeminal tract is mediated by the raphe–parapyramidal region. Our study is the first to suggest a brainstem pathway mediating cutaneous vasoconstriction elicited by nociceptive stimulation.

These regionally differentiated changes in vascular tone presumably occur via net excitation or inhibition of appropriate sympathetic preganglionic neurones in the spinal cord, reflecting patterned vascular control via brainstem neurones with axons descending to the spinal cord. The detailed brainstem location of vasomotor neurones controlling the different peripheral vascular beds is still being investigated. Neurones in the rostral ventrolateral medulla (RVLM) regulate mesenteric, renal and skeletal muscle vascular flow (see reviews by Dampney, 1994; Blessing, 1997), so that inhibition of these cells could mediate trigeminally elicited increases in flow to these beds. Neurones in the RVLM have also been implicated in the control of ear pinna blood flow in rabbits (Ootsuka & Terui, 1997) and in the control of cutaneous sympathetic nerve activity in cats (Dampney & McAllen, 1988; McAllen & May, 1994a,b). Excitation of appropriate RVLM cells could therefore mediate trigeminally elicited cutaneous vasoconstriction. However, we have recently discovered in rabbits that major changes in cutaneous blood flow can be elicited from the rostral medullary raphe–parapyramidal region (Blessing et al. 1999; Nalivaiko & Blessing, 1999). Although skin blood flow has not been measured in previous studies, there is evidence that the parapyramidal region may influence both nociception and cardiovascular function (Lovick, 1987; Siddall & Dampney, 1989).

Thus trigeminally elicited cutaneous vasoconstriction in rabbits could be mediated via descending spinal pathways originating in the raphe–parapyramidal region rather than in the RVLM. We have now examined this possibility in the anaesthetized rabbit. We elicited cutaneous vasoconstriction by electrical stimulation of the trigeminal tract and by a more natural nociceptive stimulus, pinching the rabbit's lip. To interrupt cardiovascular reflex pathways in the RVLM and in the raphe–parapyramidal region we used intramedullary injections of kynurenate to induce blockade of excitatory amino acid (EAA) receptors, and intramedullary injections of muscimol to induce a general neuronal inhibition. We tested the effectiveness of RVLM kynurenate-induced EAA blockade by examining a second cardiovascular reflex, a rise in AP associated with mesenteric vasoconstriction elicited by electrical stimulation of the central end of the vagus nerve just above the diaphragm. This second reflex is likely to be mediated via the RVLM because afferent abdominal vagal stimulation vigorously excites the great majority of bulbospinal barosensitive vasomotor neurones located there (Gieroba & Blessing, 1994; Gieroba et al. 1995).

The animal was placed in a Kopf stereotaxic apparatus and the medulla oblongata exposed by incision of the atlanto-occipital membrane. The neck was flexed so that the dorsal surface of the medulla oblongata was horizontal. The animal was then paralysed (vecuronium bromide, 1 mg kg⁻¹i.v.), and mechanically ventilated with oxygen-enriched air, using a Harvard model 681 rodent ventilator. End-tidal CO₂ was kept at 35–40 mmHg. Body temperature was maintained between 38.5 and 39.5°C.

After muscle paralysis, adequate anaesthesia was determined by the absence of any increase in AP in response to possibly painful procedures and by ensuring the absence of a withdrawal reflex to paw squeeze during periods when the return of active respiratory effort indicated that muscle paralysis was no longer present. If anaesthesia was adequate, supplemental vecuronium bromide (0.5 mg kg⁻¹i.v.) was administered to maintain paralysis.

As we have previously noted (Yu & Blessing, 1998), ear pinna blood flow is quite variable in anaesthetized rabbits. Before paralysis, when absence of withdrawal responses can be used to confirm the adequacy of urethane anaesthesia, ear pinna blood flow may suddenly fall to near zero levels, apparently spontaneously or in response to a normally painful stimulus. Addition of halothane to the inspired air sometimes, but not always, restores ear pinna blood flow (see, e.g. the end of the recording in Fig. 1A and the beginning of the recording in Fig. 1B), but it does not prevent further sudden falls in response to administration of a normally painful stimulus. In the present study we adjusted the concentration of halothane so that baseline flow was at an appropriate level for observing increases or decreases in response to intramedullary injections of pharmacological agents, depending on the particular experiment being undertaken.

Figure 1 RVLM kynurenate

For afferent vagal stimulation, the nerve was approached via a left lower thoracotomy, as described previously (Gieroba & Blessing, 1994). One branch (anterior or posterior) of the vagus nerve was dissected from the oesophagus just above the diaphragm. The nerve was cut, and a bipolar silver wire cuff electrode was placed around the central end of the nerve and fixed to the oesophagus with glue. The electrode was surrounded with low melting point wax and the insulated wires were passed through the abdominal wall and positioned subcutaneously. The thoracotomy wound was closed. The nerve was stimulated with 10 s trains of 0.5 ms pulses, 20 Hz, with current (usually about 500 μA) set at twice the threshold intensity required to increase AP by a maximum amount (usually about 20 mmHg).

For injections into the RVLM, the micropipette was inclined 15 deg to the vertical (tip of pipette more rostral) before insertion into the medulla oblongata. Rostrocaudal zero was defined as +0.5 mm rostral to the rostral border of the area postrema in the mid-line. Mediolateral co-ordinates were defined with respect to the mid-line and dorsoventral co-ordinates were defined as the distance between the dorsal surface of the medulla at the point of entry and the position of the tip of the micropipette. Injections were made in the region from −0.5 to +1.5 mm in the rostrocaudal direction, from 2.5 to 3.5 mm lateral to the mid-line, and 4.5–5.5 mm ventral from the point of entry of the pipette into the medulla.

For injections into raphe pallidus and the parapyramidal area, the micropipette was inclined 10 deg to the vertical (tip of pipette more rostral) before insertion into the medulla oblongata. The rostral border of the area postrema in the mid-line was again defined as rostrocaudal zero and as mediolateral zero. Dorsoventral zero was defined as the floor of the fourth ventricle in the mid-line. Injections were made in the mid-line at +0.5 mm in the rostrocaudal direction, 4–4.5 mm ventral to the floor of the fourth ventricle. Injections into the parapyramidal region were made 1 mm lateral to the mid-line and 4–5 mm ventral to the floor of the fourth ventricle, at rostrocaudal +0.5 mm.

Mean AP and mean blood flow were measured off-line during selected 5 s periods using Chart software. Mean vascular conductance for these 5 s periods was obtained by dividing mean flow velocity by mean AP. Analysis of variance with repeated measures and Fisher's protected t test were used to determine the significance of changes in resting parameters induced by intramedullary injections and to determine changes induced by stimulation of the trigeminal tract or the abdominal vagus nerve. We used the same statistical procedure to compare changes in response to stimulation before and after injection of pharmacological agents into the medulla. Values in the tables are means ±s.e.m.

Effect on resting AP, and mesenteric and ear conductance Bilateral injection of kynurenate (10 nmol in 200 nl) into the RVLM caused a gradual fall in AP so that by 5 min after the injection mean AP had decreased to approximately 50 mmHg (Table 1A, Fig. 1). This decrease was associated with an increase in superior mesenteric vascular conductance, but with no significant change in ear pinna conductance (Table 1A).

Table 1 Kynurenate and muscimol RVLM injections: effects on resting cardiovascular parameters and on responses to stimulation of the trigeminal tract and the afferent abdominal vagus nerve

Effects on AP, and mesenteric and ear conductance evoked by trigeminal tract stimulation Before injection of kynurenate into the RVLM, electrical stimulation of the trigeminal tract caused a fall in AP, associated with a substantial and brisk decrease in ear pinna conductance and with an increase in mesenteric vascular conductance (Table 1A, Fig. 1A). Bilateral injection of kynurenate into the RVLM did not substantially reduce any of these trigeminally elicited cardiovascular effects (Table 1A, Fig. 1C).

Effects on changes in AP, mesenteric conductance and ear pinna conductance evoked by abdominal vagal stimulation Before bilateral injection of kynurenate into the RVLM, electrical stimulation of the central end of the afferent abdominal vagus nerve caused a rise in AP associated with a slight rise in ear pinna conductance and a fall in mesenteric conductance (Table 1A, Fig. 1B). Bilateral injection of kynurenate completely abolished the rise in AP and the fall in mesenteric conductance normally elicited by abdominal vagal stimulation (Table 1A, Fig. 1D), but ear pinna conductance still increased slightly (Table 1A, Fig. 1D).

Table 2 l-Glutamate and GABA injections into the RVLM: effects on resting cardiovascular parameters

Figure 2 RVLM L-glutamate and GABA

When baseline ear blood flow was zero or at near zero levels, unilateral injection of GABA (100 nmol in 100 nl) into the RVLM (2.5–3.5 mm lateral to the mid-line) increased ear vascular conductance (data not shown), as previously shown by Ootsuka & Terui (1997), and illustrated for bilateral injection of muscimol, a GABA_A receptor agonist, into the RVLM in Fig. 3B. When baseline ear blood flow was at moderate or high levels, injection of GABA actually decreased ear pinna vascular conductance in a uniphasic manner, indicating that ear pinna vasoconstriction rather than vasodilatation accompanied the hypotensive response (Table 2, Fig. 2B). Mesenteric vascular conductance increased in response to unilateral injection of GABA into the RVLM (Table 2, Fig. 2B).

Figure 3 RVLM muscimol

Effect on resting AP and ear conductance Bilateral injection of muscimol (1 nmol in 100 nl) into the RVLM caused AP to fall to approximately 35 mmHg (Table 1B, Fig. 3). After injection of muscimol into the RVLM, ear pinna conductance increased if the pre-injection baseline flow was low, as illustrated in Fig. 3B. Muscimol increased conductance in three out of five animals but the variability of the response and the high baseline flow in the other two animals meant that overall no significant change in ear conductance was recorded. It is also important that, as noted above, the AP fell to around 35 mmHg, a level which may compromise cerebral perfusion.

Effects on AP, mesenteric and ear conductance evoked by trigeminal tract stimulation Even after the substantial fall in AP produced by the muscimol injections, trigeminal tract stimulation still caused a further fall in AP (Table 1B, Fig. 3C) and trigeminal tract stimulation continued to elicit a prompt fall in ear pinna blood flow (Fig. 3C). Overall, the trigeminally elicited fall in ear pinna conductance was not reduced in magnitude by bilateral injection of muscimol into the RVLM (Table 1B). We did not measure mesenteric flow in the muscimol experiments.

Table 3 Kynurenate and muscimol raphe–parapyramidal injections: effects on resting cardiovascular parameters and on responses to electrical stimulation of the trigeminal tract and to pinching the lip

Figure 4 raphe–parapyramidal kynurenate

Muscimol injection (1 nmol in 100 nl, 1 mid-line injection in 4 rabbits plus an additional injection at 1 mm lateral on each side in 4 rabbits) into the raphe–parapyramidal region caused a small fall in AP in some animals but in the group data muscimol did not significantly alter AP (Table 3B). Ear conductance increased markedly if flow was low at the time of injection (Fig. 5C) but showed little change if baseline flow was already high. Overall, in eight animals, we recorded a significant increase in ear pinna vascular conductance after injection of muscimol into the raphe (Table 3B). Muscimol injection did not affect the fall in AP elicited by trigeminal stimulation (Table 3B, Fig. 5D). However, muscimol abolished or substantially reduced the trigeminally elicited ear vasoconstriction. Flow still fell in each animal, but the fall in flow was principally secondary to the fall in AP, so that the trigeminally elicited fall in ear pinna vascular conductance was substantially reduced by the muscimol injections (Table 3B, Fig. 5D). This change from the pre-muscimol control response was also apparent when we measured the effect of trigeminal tract stimulation on the pulse amplitude of the ear blood flow signal before and after muscimol injection into the raphe–parapyramidal area. Before injection of muscimol, trigeminal stimulation reduced the pulse amplitude of the ear blood flow signal (see insets to Fig. 5A) to 31 ± 10 % of the pre-stimulation value (mean ±s.e.m., n = 8 rabbits, P < 0.01). After injection of muscimol, trigeminal stimulation no longer reduced the pulse amplitude of the ear blood flow signal: stimulation now increased the pulse amplitude to 140 ± 13 % of the pre-stimulation value (n = 8 rabbits, P < 0.01), as illustrated in Fig. 5D (insets).

Figure 5 raphe–parapyramidal muscimol

In the control situation before injection of muscimol, a firm pinch of the rabbit's lip caused a marked reduction in ear pinna conductance accompanied by a slight fall in AP (Table 3B, Fig. 5B). After injection of muscimol into the raphe–parapyramidal area, the marked fall in ear pinna conductance induced by lip pinch was completely abolished (Table 3B, Fig. 5D). The small fall in AP was reduced or converted to a small pressor response, so that, taken overall, lip pinch produced no significant change in AP after injection of muscimol into the raphe–parapyramidal area (Table 3B).

Figure 6 Injection and stimulation sites

Bilateral RVLM kynurenate injections dilated the mesenteric vascular bed, but not the ear pinna vascular bed, even though each injection was 200 nl and the amount of kynurenate injected (10 nmol on each side) was in excess of that shown to prevent activation of EAA receptors in rats (Guyenet et al. 1987). As shown in Fig. 6B, the injection spread throughout the RVLM region. The contrasting effects of kynurenate on mesenteric and ear pinna beds suggest different control mechanisms for mesenteric and ear pinna presympathetic motoneurones in the RVLM. In the anaesthetized rabbit, RVLM vasomotor neurones controlling the ear pinna bed do not appear to receive a tonically active EAA input. It may even be that ear pinna presympathetic motoneurones present in the RVLM lack EAA receptors, since we found that injecting L-glutamate into the RVLM failed to decrease ear pinna conductance after any of the 25 injections performed.

Quite variable and inconsistent effects on ear pinna conductance with injection of L-glutamate into the RVLM were also noted by Ootsuka & Terui (1997) who explained their results by postulating mixed excitatory and depolarization blockade effects of L-glutamate. However, this does not seem likely in our experiments since L-glutamate injections reliably increased AP and decreased mesenteric conductance. It is interesting to note that Key & Wigfield (1994) found that electrical stimulation of the RVLM in rats increased AP within seconds, but that tail temperature was reduced only after some minutes, at a time when circulating adrenal medulla catecholamines (see below) could have contributed to cardiovascular changes. When injections of dl-homocysteic acid were used to stimulate neurones in the RVLM, tail temperature did not fall until 18–30 min after the injection, much longer than would be expected for the occurrence of flow changes initiated by the chemical excitation of neurones in the RVLM. In the present experiments, L-glutamate induced vasoconstriction in the ear pinna bed after a delay of about 20 s following the RVLM injection. Adrenal medullary catecholamines are secreted following pharmacological excitation of RVLM neurones (Ross et al. 1984; McAllen, 1986; Natarajan & Morrison, 1999), and the time course of the L-glutamate-induced ear vasoconstriction in our experiments suggests that this delayed effect may reflect the actions of these humoral agents.

In agreement with Ootsuka & Terui (1997), we found that injections of GABA into the RVLM can increase ear vascular conductance. In our study this increase occurred only when the resting pre-injection level of ear blood flow was at low levels. When resting flow was at moderate or high levels, injection of GABA into the RVLM actually decreased ear pinna vascular conductance. Our findings with RVLM injections of muscimol (a GABA_A agonist) are consistent with the effects of GABA itself. When resting ear blood flow was at a low level, injection of muscimol into the RVLM caused a marked increase in ear pinna flow. As illustrated in Fig. 3, the increase in ear pinna blood flow tended to occur in an ‘all-or-none’ fashion, with no clear inverse relationship between the ear flow signal and the AP signal.

Thus pharmacological agents known to cause net excitation or inhibition of RVLM vasomotor neurones produce the predicted results for the mesenteric bed, but findings for the ear pinna vascular bed are much more complex, suggesting different control mechanisms for RVLM neurones regulating the ear pinna bed. It appears from our findings that RVLM neurones regulating ear pinna vascular tone do not have the requisite EAA receptors to be excited by L-glutamate, and the rapid changes in ear pinna conductance induced by stimulation of GABA receptors suggest that regulation of RVLM ear pinna vasomotor neurones is different from that of neurones regulating the mesenteric bed.

The failure of kynurenate to block trigeminally elicited ear vasoconstriction prompted us to carry out further experiments in which we produced a more general inactivation of RVLM neuronal function using focal injections of muscimol. Bilateral injections of this agent reduced AP to approximately 35 mmHg, confirming its effective inhibition of RVLM bulbospinal vasomotor neurones. However, even at such low levels of AP, it was clear that electrical stimulation of the trigeminal tract continued to elicit a prompt fall in ear blood flow to near zero levels. Our combined evidence thus suggests that the brainstem pathway for trigeminally elicited ear vasoconstriction does not involve important synaptic relays in the RVLM.

The original descriptions of the trigeminal depressor response emphasized its similarity to baroreceptor-mediated falls in AP (Kumada et al. 1978). Initially, less emphasis was placed on possible nociception-related functions of the trigeminal response, but a later study (Terui et al. 1981) demonstrated initiation of the trigeminal depressor response and renal sympathoinhibition by nociceptive facial stimuli, including pinching of the lip.

Our present results with injections of muscimol into the raphe–parapyramidal region support this conclusion. Muscimol increased ear pinna vascular conductance, without causing major changes in AP, confirming that the resting discharge of neurones in the raphe–parapyramidal region tonically constricts the ear pinna vascular bed in our experimental situation. This conclusion is supported by our observation that muscimol-induced blockade of neuronal function in the raphe–parapyramidal region substantially reduces or prevents ear pinna vasoconstriction elicited by electrical stimulation of the trigeminal tract or nociceptive pinching of the lip. It appears that raphe–parapyramidal neurones are essential for the occurrence of the ear vasoconstriction in response to nociceptive stimuli. Ear pinna vasoconstriction in response to nociceptive stimulation occurs together with vasoconstriction in the tail bed, in a highly correlated manner (Nalivaiko & Blessing, 1999). Taken together, therefore, our results suggest that the vigorous cutaneous vasoconstriction which is elicited by nociceptive stimulation depends on neural circuitry in the raphe–parapyramidal region.

It is also clear that there is no clear demarcation between the mid-line raphe neurones and surrounding cells in more lateral regions, dorsal to the pyramids and the rostral portion of the inferior olive, and reaching down to the ventromedial border of the medulla, between the pyramids and the facial nucleus. This conclusion also applies to the detailed study of the raphe nuclei in the rabbit brainstem by Felten & Cummings (1979). A wealth of other neurotransmitter and connectivity-based evidence, including experiments in rabbits, strengthens the view that, at least for the present, these more lateral neurones can be considered together with the raphe neurones (Howe et al. 1983; Bowker et al. 1988; Haselton et al. 1988; Blessing, 1990). We therefore adopt the concept, introduced by Helke and colleagues (Helke et al. 1989), of ‘the parapyramidal region’ and we use ‘raphe–parapyramidal’ region as an umbrella term to refer to the region of the medulla involved in our present injection experiments. There is no clear characteristic which defines a neurone as belonging to raphe pallidus or raphe magnus. By convention, raphe magnus is more rostral and dorsal, and raphe pallidus is more caudal and ventral. Thus, according to this terminology, our injections involved both raphe pallidus and the more caudal portion of raphe magnus, as well as the parapyramidal area.

In rabbits there are direct projections from the raphe- parapyramidal area included in our injections to the intermediolateral column of the spinal cord (Haselton et al. 1988). raphe–parapyramidal neurones are amongst the first to be labelled via retrograde transneuronal transport from the sympathetic nerves supplying vessels in the rat tail (Smith et al. 1998), a cutaneous bed similar to the ear pinna in the rabbit. In rabbits there is a dense serotonin (5-HT) innervation of sympathetic preganglionic neurones projecting to the superior cervical ganglion (Jensen et al. 1995) and these terminals probably derive from the 5-HT cell bodies in the raphe–parapyramidal region of the rabbit (Howe et al. 1983). It is our hypothesis that the vigorous cutaneous vasoconstriction which can be elicited from the raphe–parapyramidal region in the rabbit (Blessing et al. 1999) is mediated by direct spinal projections of presympathetic motoneurones located there. We consider that these same bulbospinal neurones are likely to mediate the ear pinna vasoconstriction elicited by nociceptive pinching of the lip and by trigeminal tract stimulation in the present experiments. There is some indirect evidence that the responsible neurones might include 5-HT cells in the rabbit (White et al. 1986). These raphe-spinal neurones may also be relevant to cutaneous blood flow changes and brown adipose tissue metabolic effects occurring as part of temperature regulation in rats (Morrison et al. 1999; Morrison, 1999; Rathner & McAllen, 1999).

Raphe nuclei in the ventral portions of the rostral medulla and caudal pons, as well as the immediately ventrolateral regions, are thought to participate in CNS suppression of reflex withdrawal responses (e.g. the tail flick response) normally initiated by nociceptive stimuli (Fields & Heinricher, 1989; Mason, 1999). This effect is thought to occur via the well-established projection of the more rostral raphe–parapyramidal neurones (raphe magnus cells) to the dorsal horn of the spinal cord (Basbaum et al. 1978). The same neuroanatomical study (in the cat) reported a bilateral projection from raphe magnus to the intermediolateral column of the thoracic spinal cord. Whether individual spinally projecting raphe–parapyramidal neurones innervate both the dorsal horn and the sympathetic preganglionic neurones remains to be established.

Figure 7 Proposed neural pathways

The raphe nuclei and the RVLM may function as two complementary presympathetic control centres, as suggested by Morrison and colleagues (Morrison, 1999; Morrison et al. 1999). The RVLM appears to regulate baroreceptor-modulated ‘cardiovascular’ sympathetic outflow, with particular involvement of vascular beds supplying the abdominal viscera and skeletal muscle, and consequent major effects on AP. The raphe and parapyramidal region may regulate ‘non-cardiovascular’ functions involving changes in cutaneous sympathetic nerve activity and blood flow, and associated temperature-related changes in brown adipose tissue metabolism, without major engagement of baroreceptor-vasomotor reflexes and without major changes in AP (Habler et al. 1994; Johnson & Gilbey, 1994; Rathner & McAllen, 1999). Vigorous cutaneous vasoconstriction with little associated change in AP also occurs when the conscious individual detects a salient environmental stimulus, a response which depends on the functional integrity of the amygdala (Yu & Blessing, 1999), and it may be that raphe neurones are part of an amygdala-brainstem-spinal link mediating this cutaneous blood flow response to alerting stimuli.

Codes of animal ethics emphasize that adequate anaesthesia eliminates sudden increases in AP which might otherwise occur in response to normally painful stimuli. Cutaneous blood flow is much more reactive and variable, even in anaesthetized animals receiving additional halothane. Similar cutaneous flow changes occur in anaesthetized humans, and it may be that resting levels of skin blood flow and changes with nociceptive stimuli may provide an indication of patient well-being during surgical anaesthesia (Mashimo et al. 1997; Shimoda et al. 1998). Our study demonstrates that the CNS pathway regulating skin blood flow in response to trigeminal tract and nociceptive cutaneous stimulation includes neurones in the raphe–parapyramidal region.

Our research was supported by the National Health and Medical Research Council, by the National Heart Foundation of Australia and by the Neurosurgical Research Foundation of South Australia. Dr Y.-H. Yu, Ms Robyn Flook and Ms Sarah Kennedy provided technical assistance.