In previous studies we have shown that is it possible to investigate some of the cortical pathways involved in control of the pharynx and upper oesophagus using transcranial magnetic brain stimulation in conscious human subjects (Aziz et al. 1994, 1995; Hamdy et al. 1996, 1997). In the present paper we extend these findings by showing, first, how the pathways from each hemisphere interact to control these mid-line structures, and second, how the excitability of the pathways is influenced by afferent input from the face, pharynx and oesophagus.
Interaction between pathways from each hemisphere
The present data confirm that single stimuli applied over either hemisphere can elicit EMG responses both in the pharynx and oesophagus. In addition, the experiments with pairs of stimuli suggest that spatial facilitation can occur between the responses from each hemisphere. Thus, single stimuli applied to either hemisphere, which on their own were too small to evoke any EMG response, could evoke clear activity when given together. At suprathreshold intensities, the size of responses to pairs of stimuli was larger than the sum of the responses to each stimulus given alone, and the latency was shorter. These results are compatible with a shared population of brainstem interneurones or motoneurones receiving combined input from both hemispheres. Thus, as described in animal experiments, fibres from the motor and pre-motor cortex of both cerebral hemispheres probably converge, via interneurones of the central pattern generator (CPG), onto motor nuclei of the V, IX, X and XII cranial nerves (Car, 1970; Hockman et al. 1979; Miller, 1986). This overlap of inputs from each hemisphere may well explain why swallowing can be maintained in a large percentage of patients after unilateral cerebral stroke (Barer, 1989). It also suggests that both cerebral hemispheres are involved in initiating volitional swallows, possibly following pre-initiation processing (intention to initiate motor activity) in other cortical swallowing regions such as the insula (Mesulam & Mufson, 1982), or in supplementary motor areas (Penfield & Rasmussen, 1950), or via ascending afferent feedback (Miller, 1972).
There were subtle differences in hemispheric summation for the pharynx and oesophagus. Spatial facilitation for the pharyngeal responses was clear even with subthreshold stimuli, whereas this was not the case for the oesophagus. In addition, although facilitation was maximal at an interstimulus interval of 1 ms in both muscles, some degree of facilitation was seen at all three intervals in the pharynx, but not in the oesophagus. It is unclear why these differences occurred, although they may be due to the fact that cortical input to the pharynx is probably larger than to the oesophagus (Hamdy et al. 1996) and that input to the pharynx can induce distal inhibition of motor activity within the oesophagus (Jean, 1990), which could have influenced the oesophageal response to the second cortical stimulus. If facilitation were caused by the arrival of descending volleys at shared interneurones or motoneurones, then a 1 ms ISI would produce larger amplitude facilitation than a 5 or 10 ms ISI, because the excitatory post synaptic potentials (EPSPs) would sum on their rising phases. Facilitation at 5 and 10 ms would be correspondingly smaller because summation would occur on the falling phase of the initial EPSP. It is interesting to note that it was only when an interval of 5 ms was used that the latency of the response to combined stimulation was shorter than the response to single stimuli given alone. The reason for this is probably that at 5 ms, the first EPSP raises the excitability of shared brainstem neurones nearer to threshold so that they can respond more quickly to the arrival of the second input. This is compatible with the fact that the latency decrement is similar to that seen when responses are elicited in pre-active rather than relaxed muscle. No latency decrease was evident with an ISI of 1 ms, but this is probably because the maximum possible reduction (i.e. 1 ms) was within the noise of our measurements of onset latency in these muscles.
It is of interest that when the two hemispheres were stimulated sequentially (at ISIs of 5-10 ms), no convincing inhibition was seen. This is in contrast to the effects seen in hand muscles, where sequential magnetic stimuli of each hemisphere, at ISIs of 5-30 ms, inhibited the motor response, possibly via transcallosal interactions that may exist, at least for limb muscles, to ensure strictly unilateral movement (Ferbert, Priori, Rothwell, Day, Colebatch & Marsden, 1992). Our data indicate that the inter-hemispheric interactions for mid-line structures such as the pharynx and oesophagus, which are represented on both hemispheres and have bilateral projections to their motor nuclei, differ from those which have predominantly unilateral representation, although in the absence of direct recordings from the cortico-bulbar tracts, some transcallosal inhibition cannot be completely excluded, given that increased excitability of downstream neurons (at ISIs of 1-5 ms) could have masked any reduction in motor cortex output.
Interaction between afferent input and responses to cortical stimulation
In a previous paper we reported the effect of suprathreshold vagal or trigeminal stimuli on the response to cortical stimulation (Hamdy et al. 1997). The present paper extends these findings by using a range of different intensities of conditioning shock. Single suprathreshold stimuli applied to the vagus or trigeminal nerves elicit both an intermediate and a late reflex response in the pharynx and oesophagus by exciting afferent pathways from the face and neck, which then converge on interneurones of the nucleus of the tractus solitarius in the brainstem (Aziz et al. 1995; Hamdy et al. 1997). The former reflex response is small and has a latency of 20-30 ms; the latter is larger and longer lasting, with a latency to onset of about 50-70 ms. When the cortical stimulation is timed so that the evoked responses occur at the time of the late reflex (ISI, 50-100 ms), then facilitation is clear, and the onset latency of the cortical component is reduced. Presumably, neuronal pools in the brainstem are facilitated by the reflex input, and become more responsive to descending input from cortex. The situation is more difficult to understand when the interstimulus interval is short (20-30 ms), and cortical responses occur during the initial part of the reflex. At these times, the latency is reduced but there is no facilitation of the size of the cortically evoked responses. In order to account for this discrepancy between the effect on latency and amplitude, we proposed that reflex input could have had two effects: excitation at the brainstem and suppression at the cortex. The result would be that a given stimulus evoked a smaller descending volley which impinged on more excitable downstream structures.
The present data show that, as might be expected, threshold stimuli have effects similar to, but smaller than, those seen with suprathreshold intensities. In general, the time course of effects is similar, except that facilitation begins earlier at higher intensities. There were, however, apparent differences between the cranial nerves in effecting facilitation: for example, an increase in the response amplitude was observed following threshold conditioning of the trigeminal nerve but not the vagus nerve. This may indicate that whilst afferent inputs from both sites have comparable effects in exciting brainstem (swallowing) circuitry, ascending vagal input to the cortex is largely inhibitory (Rutecki, 1990) unless more vigorous stimuli are applied. Furthermore, the data also show that facilitation is not evident at subthreshold intensities. The implication is that afferent input evoked by such stimuli does not reach the population of neurones excited by descending cortical input. This may be because there are several synapses in the afferent pathway before convergence with cortical inputs.
Single stimuli applied, in pilot studies, to either the pharynx or oesophagus had no effect on the responses evoked by cortical stimulation. Because of this we used a train of repetitive stimuli, and gave the cortical test shock 100 ms after the end of the train, a time at which there had been maximum effects in the experiments using cranial nerve conditioning shocks. Under such conditions, stimulation of the pharynx or oesophagus demonstrated both frequency- and site-specific properties. At the highest frequencies, pharyngeal stimulation facilitated cortically evoked responses in both pharynx and oesophagus, whereas oesophageal stimulation had only limited effects on responses from the oesophagus alone. This indicates that pharyngeal sensation has a more powerful facilitatory effect on the cortical swallowing pathways than oesophageal sensation. These differences probably reflect the differing properties of the fibres innervating these regions. Sensation from the pharynx is carried largely by the glossopharyngeal nerve and the superior laryngeal nerve (SLN), stimulation of which is the most potent trigger of swallowing (Jean, 1990), whereas that from the upper oesophagus is conveyed via the recurrent laryngeal nerve, which does not trigger reflex swallowing (Miller, 1982; Jean, 1990). In addition, it is recognized that convergence of afferent fibres from the pharynx, in the nucleus of the tractus solitarius of the brainstem swallowing centre, is much more extensive than that from the upper oesophagus (Sessle, 1973).
Our observation that the highest stimulation frequencies of the pharynx also produced the greatest facilitation suggests that there may exist, at least for the pharynx, a frequency-dependent pattern for afferent feedback. In support of this, animal studies have indicated that the facilitation of reflex swallowing by SLN and glossopharyngeal nerve stimulation is also frequency dependent, with an optimal frequency of 30-50 Hz (Sinclair, 1971; Miller, 1972; Jean & Car, 1979; Weerasuriya, Bieger & Hockman, 1980).
Although there was a clear shortening of response latency, repetitive stimulation of the pharynx or oesophagus had no consistent effects on the amplitude of cortically evoked responses. This is similar to the results seen with cranial nerve conditioning at short intervals. As in the latter case, it might be that conditioning with pharyngeal or oesophageal stimuli excited brainstem motoneurones, whilst the motor cortex was inhibited. Direct excitation of brainstem vagal neurones seems likely in view of the strong projections from the pharynx and oesophagus to the CPG of the brainstem (Jean & Car, 1979). The mechanism of a possible cortical inhibition is less obvious. It is plausible that the same stimuli could produce direct inhibition of motor cortex. Alternatively, since the duration of the sensory stimulation was at least 2.5 s, it could be that subjects volitionally attempted to suppress (at a cortical level) the reflex swallowing which such stimuli can induce (Miller, 1982). In support of the first explanation, animal data have demonstrated that repetitive stimulation of pontine swallowing regions, receiving afferent input from the pharynx, inhibits cortical swallowing neurones, a finding that was not observed with similarly applied single stimuli (Sumi, 1972a). In addition, it is also known that repetitive stimulation of the human vagus nerve can suppress the frequency of epileptiform seizures, inferring a reduction in cortical excitability (Rutecki, 1990). While future animal studies may elucidate further the exact relationship between the short latency responses explored in our study and those activated during swallowing, it is possible that cortical inhibition may ensure that once brainstem CPG is activated, cortical discharge is suppressed, so that reflex swallowing can occur without interruption by other volitional commands to swallowing musculature.