Straus, Christian, Marc Zelter, Jean-Philippe Derenne, Bernard Pidoux, Jean-Claude Willer, and Thomas Similowski.Putative projection of phrenic afferents to the limbic cortex in humans studied with cerebral-evoked potentials. J. Appl. Physiol. 82(2): 480–490, 1997.—Respiratory sensations may rely in part on cortical integration of respiratory afferent information. In an attempt to study such projections, we recorded evoked potentials at scalp and cervical sites in 10 normal volunteers undergoing transcutaneous phrenic stimulation (0.1-ms square pulses, intensity liminal for diaphragmatic activation, series of 600 shocks at 2 Hz). A negative cerebral component of peak latency (12.79 ± 0.54 ms; N13) was constant, and a negative spinal component (7.09 ± 1.04 ms; N7) could also be recorded, all results being reproducible over time. Monitoring of cardiac frequency, skin anesthesia, and stimulation adjacent to the phrenic nerve made the phrenic origin of N7 and N13 the foremost hypothesis. Increasing stimulation frequency and comparison with median nerve stimulation provided arguments for the neural nature of the signals and their cerebral origin. Recordings from intracerebral electrodes in a patient showed a polarity reversal of the evoked potentials at the level of the cingulate gyrus. In conclusion, phrenic stimulation could allow one to study projections of phrenic afferents to the central nervous system in humans. Their exact site and physiological meaning remain to be clarified.
- phrenic nerve
- cerebral cortex
- respiratory afferents
- control of breathing
the act of breathing is mainly automatic, the permanent and vital cyclic contraction of the respiratory muscles being governed by activities of brain stem neuron groups that react to different stimulations to maintain homeostasis. This automatic regulation can be disrupted for voluntary tasks that involve respiratory muscles, including voluntary apnea and various respiratory maneuvers, but disruptions are most often for nonrespiratory or indirect respiratory actions. In humans, speech gives a particularly frequent and complex example of a voluntary perturbation of automatic respiration that intimately depends on a fine tuning of respiratory muscle action and coordination (38).
The volitional control of respiratory muscles implies their motor cortical representation. This motor representation has long been described in humans (13) and can be explored by techniques such as electrical stimulation (16), magnetic stimulation (24, 30), or functional imaging (5). Several structures are involved, principally the primary motor area but also other regions within the premotor cortex and the supplementary motor area (5, 13). Respiration has also important behavioral and emotional aspects, in view of which it is interesting to note that motor responses of the diaphragm can be evoked from stimulation of deeper cortical structures such as the limbic region (27). The behavior of such a complex motor system necessarily relies on the integration of afferent information, and indeed the central nervous system permanently receives and processes messages arising from numerous respiratory-related structures, including the upper and lower airways, the lung, the chest wall, and the respiratory muscles. It has long been appreciated that this “afferent” side of control of breathing is of particular pathophysiological and clinical relevance because it is the basis for load perception and probably for some aspects of dyspnea.
The technique of evoked potentials provides a convenient tool to study afferent pathways and the corresponding cortical structures. From the respiratory point of view, cortical potentials have been related to events such as inspiratory occlusion (7) or pressure pulse stimuli applied at the mouth (33). These types of stimuli do not allow one to identify the precise source of the causative afferents. In particular, the specific muscular messages, which may well play a very peculiar role as sources of dyspnea (20), cannot be disentangled from those of other sources. However, this can be achieved by specifically stimulating either the muscles themselves (15) or the phrenic nerve, which conveys diaphragm-originating information (14). Projections of phrenic nerve-mediated afferents have been demonstrated to the cerebral cortex in the cat (2, 9), and modifications of the corresponding evoked potentials by diaphragm fatigue have been described (2). In humans, Gandevia and Macefield (15) have evoked potentials from needle electrical stimulation of intercostal muscles. They were able to record on the scalp a negative component that had a latency that ranged from 21.3 to 38 ms and had a vertex maximum. In the same study, the authors attempted to elicit potentials from phrenic nerve stimulation, but they were unable to do so in a reproducible manner. Only in one of six experiments conducted in three subjects did they observe a negative scalp component in response to phrenic nerve stimulation (Fig. 5 in Ref. 15), with an onset latency of 11 ms and a peak latency of ∼13–14 ms. Recently, Zifko et al. (37) have described, in a short report, somatosensory evoked potentials, the characteristics of which suggested an afferent projection of the phrenic nerve close to the cortical motor representation of the diaphragm.
We undertook the present work to look for cerebral and spinal potentials evoked by phrenic nerve stimulation in humans in a systematic manner with a twofold aim. On one hand, we tried to examine confounding factors that could compromise the use of the technique in pathophysiological or clinical assessment of the role of the diaphragm afferents in respiratory sensations and the control of breathing. On the other hand, we tried to understand what could be the nature of the neural structures responsible for the observed potentials, which seemed rather different from previously described ones (15, 37). With regard to this investigation, some observations made in a patient with intracerebral electrodes provided support of the neural origin of the phrenic nerve stimulation-related potentials and pointed to the limbic cortex as their possible source.
The study was carried out on 10 healthy volunteers (6 men, 4 women) aged 24–33 yr. None had a past history of respiratory or neurological disease. Nine were right handed, and one was left handed. During experimental sessions, they were in a supine position, breathing quietly, and were instructed to relax while staying awake.
For different purposes, two patients with neurological diseases were also studied. Patient 1 was studied mainly to address a methodological issue, namely, to rule out the hypothesis that the cerebral potentials recorded were far-field electromyogram (EMG) potentials. Patient 2 was studied to gain topographical information about the source of these potentials.
Patient 1 was a 28-yr-old man in whom an ependymoma of the fourth ventricle had been surgically removed. After surgery, he appeared completely dependent on mechanical ventilation, with a complete absence of respiratory activity after disconnection from the ventilator. Cortical magnetic stimulation showed a normal response of the diaphragm, and the diagnosis of a complete loss of chemosensitivity with a preserved efferent pathway was retained. Phrenic nerve stimulation-related evoked potentials were recorded at this point in time. Spontaneous breathing eventually reappeared, allowing the patient to be weaned from the ventilator. Chronic hypercapnia (arterial 55 Torr) with a diminished CO2ventilatory response (0.9 ml/ Torr, hyperoxic rebreathing technique) persisted as a sequela.
Patient 2 was a 36-yr-old right-handed woman who had been referred to the Salpêtrière Epilepsy Center for severe medically intractable partial epilepsy. The final diagnosis retained was medial temporal lobe epilepsy. A computed tomography scan was normal, magnetic resonance imaging showed a left hippocampus atrophy (1), and positron emission tomography imaging revealed a hypometabolic pattern involving the left amygdala and hippocampus. The implantation of two frontal and two temporal intracerebral depth electrodes was decided because doubt remained concerning the lateralization of the electroencephalographic epileptic zone. Of note, auditory evoked potentials and somatosensory evoked potentials of the median nerve were normal on both sides. At the time of the phrenic nerve stimulation study, the patient received carbamazepine (200 mg/day) and vigabatrin (1,500 mg/day) but still had occasional focal seizures.
The study had been approved according to the French legislation on human biomedical studies by the local Comité Consultatif pour la Protection des Personnes dans la Recherche Biomédicale. All subjects and the two patients were informed in detail of the purpose of the study, the methods used, and potential risks. They all gave written informed consent according to the principles of the Declaration of Helsinski.
Evoked potentials were studied by using a Nicolet Spirit digital system (Nicolet, Madison, WI) featuring automatic artifact rejection. Signals were sampled at 50 kHz, amplified, band-pass filtered (1–1,500 Hz), and averaged. They were stored on hard disk for subsequent analysis (Nicolet Evoked Potentials 1.4” software, Nicolet). Each of the potentials described in resultscorresponds to the averaging of 600 responses of 50-ms epoch after the stimulus onset. Cerebral and cervical response components are labeled from their positive (P) or negative (N) polarity and their peak latency, according to standard conventions.
In healthy subjects and patient 1, cerebral-evoked potentials were recorded by using subcutaneous stainless steel needles inserted into the scalp. Electrodes were positioned in aFz, T3, T4, and Oz according to the 10-20 system modified by Picton et al. (26) (Fig. 1). Spinal potentials were recorded via a pair of surface electrodes (jelly-filled self-adhesive electrodes) placed on the slightly abraded and degreased skin of the neck, each linked as input to a separate channel. The posterior electrode was situated at the level of the second cervical vertebra (C2); the anterior one was at the level of the cricoid cartilage. A single reference surface electrode placed on the left earlobe was used. The earth electrode was on the right earlobe. Electrode impedance was monitored and maintained below 5 kΩ.
The choice of this montage was the result of a preliminary step involving ∼50 experiments performed in ∼20 subjects. During this process, various recording sites (including some close to the vertex) and various reference electrode position were tried, with a great number of combinations. The montage that gave the clearest and most reproducible results with a reasonable “no artifact” degree of confidence was finally retained, despite its seemingly surprising nature (see discussion).
In patient 2, evoked potentials were recorded from four platinum eight-contacts Spencer-type electrodes (electrode diameter 1 mm, contact length 2.5 mm, intercontact distance 9 mm, total electrode length 30 cm; Ad-Tech Medical Instruments, Racine, WI) that had been bilaterally implanted in the brain of the patient (Fig. 2) by stereotaxic neurosurgery (32). Their exact position had been checked by using volumic magnetic resonance imaging. Two electrodes were at the level of the internal part of the frontal lobes, the two others being in the temporal lobes at the level of the hippocampus and amygdala. Location of the eight contacts of the right and left temporal electrodes was as follows, from a toh: amygdala (a), hippocampus head (b), head-body junction of the hippocampus (c), hippocampus body (d), hippocampus tail (e), white matter (f,g), and occipitotemporal cortex (h) (Fig.2 A). Contacts of the left and right frontal electrodes were distributed as follows, from1 to8: orbitofrontal cortex (1,2), anterior part of the cingulate gyrus (3,4,5), mesial first frontal gyrus (6,7), and frontal cortex at its convexity (8) (Fig.2 B). Signal processing and evoked potentials analysis were performed following the same principles for the depth electrodes as for the scalp electrodes.
EMG signals were recorded and analyzed by using the aforementioned Nicolet Spirit apparatus and associated softwares. Right hemidiaphragm EMG (EMGdi) was obtained from two skin-taped disposable silver cup electrodes positioned on the chest in the sixth or seventh intercostal spaces at the level of the anterior axillary line. Right sternomastoid EMG (EMGsm) was recorded from two similar electrodes placed ∼3–5 cm below the mastoid as a means to evaluate a possible role of a sternomastoid action potential in the evoked potentials.
We used one ECG lead to record cardiac frequency during phrenic nerve stimulations as a means to evaluate a possible simultaneous vagal activation that could contribute to the cerebral potentials.
Phrenic nerve stimulation was delivered at a 2-Hz frequency, transcutaneously, via a constant-current stimulator included in the Nicolet Spirit machine and connected to a bipolar electrode with a 2-cm interelectrode distance. Square electrical pulses (0.1 ms) were used. Their intensity was chosen as to be liminal for diaphragmatic activation. It ranged from 7 to 20 mA, and it was not further increased, first, for obvious reasons of comfort and, second, because this was not needed for the purpose of the study (seediscussion). Cathode position was kept proximal throughout the experiments. Right phrenic stimulation was performed beyond the sternocleidomastoid and in front of the scaleni muscles. The effectiveness of the stimulation was attested by clear “clinical” diaphragmatic contraction and a motor response on EMGdi (M wave). Maximal care was taken not to activate the brachial plexus during phrenic stimulation. This was checked by constantly verifying the absence of visible or palpable contraction of muscles innervated by the brachial plexus and by asking the subjects to signal even light paresthesia. When such an event occurred, the stimulation point was adjusted so as to maintain the diaphragm response and suppress the parasitic contraction or paresthesia. No attempt was made to control for the phase or the respiratory cycle during which stimulations occurred. Indeed, with several hundred stimulations delivered at random over several minutes, it appeared likely that a putative effect, spontaneous breathing, of phrenic nerve stimulation-elicited afferent messages would cancel itself. Of note, during the preliminary experiments, derivations particularly sensitive to a possible electrically induced eye-blink artifact were studied and did not show a pattern resembling such artifact.
Other afferents originating from cervical anatomic structures could be involved in the cerebral potentials evoked by transcutaneous phrenic nerve stimulation such as cervical skin, sternomastoid, and scaleni muscles or the brachial plexus. That is why, to individualize artifactual components, several other sites of stimulation were used. Sternomastoid and scaleni muscles proximal to the phrenic nerve were stimulated by slightly moving the electrode from the previously defined phrenic spot. In the rest of the paper, this procedure is termed paraphrenic stimulation. It was used in all subjects. The brachial plexus was stimulated at the Erb’s point in four subjects (1, 3, 6, and7). Finally, we also attempted to evoke cerebral potentials in one subject (7) by phrenic stimulation during skin anesthesia, first by Freon cooling to block conduction in the large myelinated fibers and second by subcutaneous infiltration of 2% lidocaine to block conduction in the small myelinated and unmyelinated fibers.
The effects of increased phrenic nerve stimulation frequency were assessed in four subjects (1, 3, 6,and 7), and in four others (2, 4, 5, and8) the phrenic stimulation-related potentials were compared with the known somatosensory potentials evoked by median nerve stimulation at the wrist (10-12, 18).
Sham stimulation was performed in all subjects to assess the possibility of artifacts related to spontaneous breathing.
In five subjects (1, 3, 4, 6, and7), reproducibility of the cerebral- and spinal phrenic-related evoked potentials was assessed by a second experimental session, at a 3- to 6-wk interval.
Influence of reference electrode position.
Because the montage chosen at the end of the preliminary experiments was atypical in some aspects, the effects of changing the reference electrode position was assessed in three subjects (reference on the inion, third thoracic vertebra, and right hand).
Latencies are expressed as means ± SD.
Phrenic Nerve Stimulation
The phrenic nerve stimulation resulted in a diaphragmatic muscle action potential of biphasic shape and normal latency (7.51 ± 1.00 ms). It was well tolerated despite the large number of stimuli delivered. The sternomastoid muscle was inconstantly activated by phrenic nerve stimulation, and it was never activated in subject 7, as demonstrated by a permanently silent EMGsm. It proved easy to avoid brachial plexus costimulation. Both good tolerance and ease in dissociating phrenic nerve and plexus brachial activation were probably consequences of the use of low-intensity stimulation. In no case was phrenic stimulation associated with a decrease in cardiac frequency as assessed by R-R interval measurements on the concomitant ECG monitoring.
In all of the healthy volunteers, phrenic nerve stimulation evoked a negative scalp component with a peak latency of 12.79 ± 0.54 ms (N13; Figs. 3 and4). In four subjects (3, 4, 5,and 6), N13 was present at all recorded sites. Insubjects 1 and 2, an EMG artifact due to failure of relaxation by the subject made its identification impossible at T4. N13 was not found at Oz in subjects 1 and 7 or at T3 in subject 2 and at aFz in subject 8, despite a satisfactory recording quality. The amplitude of N13 generally varied from one recording site to another one in a given subject (Fig. 3,top).
In the one subject in whom it was applied, complete cutaneous anesthesia did not modify the potentials evoked by phrenic nerve stimulation.
In three of the four subjects in whom it was tested (1, 6, and7), a stimulation frequency of 15 Hz resulted in a marked reduction of N13 amplitude (Fig. 3,bottom).
In patient 1, no potentials could be recorded at any site, despite a normal diaphragm EMG response.
In patient 2, several of the contacts of the depth electrodes recorded negative or positive components,the time characteristics of which were close to what had been observed with scalp electrodes in the volunteers. At the left frontal electrode, a positive component with a peak latency of 12 ms was recorded at two contacts located in the anterior part of the left cingulate gyrus (3 and4), whereas the potential was globally negative in the upper contacts (5, 6, 7, and 8) at the corresponding latency. Therefore, a polarity reversal of the phrenic nerve stimulation-related evoked potential was observed at the anterior part of the cingulate gyrus. No potential, either positive or negative, could be seen in the orbitofrontal cortex (contacts 1 and 2) at this latency (Fig. 5). A similar pattern was recorded with the right frontal electrode. However, the positive component with a peak latency of 12 ms extended from the orbitofrontal cortex up to the right cingulate gyrus (contacts 1, 2, 3, and 4). From the superior part of the cingulate gyrus up to the convexity (contacts 5, 6, 7, and8), the corresponding potential was negative. At the right temporal electrode, a negative potential with a peak latency of 11.6 ms was present at the four posterior contacts (e,f, g, and h) from the tail of the hippocampus to the occipitotemporal cortex. The four anterior contacts (a,b, c, and d), from the amygdala to the body of the hippocampus, remained silent. No potential was recorded at the left temporal electrode, possibly because of the hippocampal abnormality evidenced by magnetic resonance imaging and positron emission tomography.
A negative component with a peak latency of 7.09 ± 1.04 ms (N7) was evoked at the C2 level in seven of the healthy volunteers. N7 was associated with a concomitant positive component at the cricoid cartilage recording site in three cases (subjects 2, 3, and7).
Phrenic nerve stimulation-related cerebral and cervical potentials were studied on two occasions in five subjects (1, 3, 4, 6, and 7) over a 3- to 6-wk interval.
N13 average latency was 12.84 ± 0.69 ms on the 1st study day vs. 12.56 ± 1.55 ms on the 2nd day (no significant difference, cross correlation).
The N7 spinal component was found in three of five subjects on the 2nd study day and was of similar latency to that of the 1st day (6.27 ± 1.35 vs. 5.57 ± 2.16 ms; no significant difference, cross correlation).
Influence of reference electrode position.
N13 persisted with the different references tested but with changing amplitudes (Fig. 4). Displacement of the reference from the earlobe resulted in a dramatic increase in the rate of signal rejection from the automatic averaging system, which, depending on the subject, could fall from <10 to >50%.
No cerebral component was identifiable in the absence of actual stimulation.
Paraphrenic stimulation, which involved skin, sternomastoid, and scalene muscles, resulted in signals that were radically different in shape and latency from those evoked by phrenic stimulation (Fig.6). This was also the case for brachial plexus stimulation at the Erb’s point (Fig. 6). Paraphrenic and brachial plexus stimulations were not performed inpatient 2 because, with the patient’s tolerance and cooperation being good but limited, we opted to multiply recordings from the intracerebral electrodes. We were confident enough that we obtained selective and good quality phrenic nerve stimulation.
Somatosensory evoked potentials of the median nerve (subjects 2, 4, 5, and8 and patient 1) obtained by stimulation of the right median nerve at the wrist evoked a cortical negative component N20 (peak latency 20 ms) at the recording scalp sites C3 (subject 5) or C3 and T3 (subjects 2, 4, and 8). Such results are typical when a cephalic reference electrode is used (18). No other median nerve stimulation-related component could be identified and particularly no N18 component (11; seediscussion).
In this study, spinal and cerebral potentials have been reproducibly evoked by phrenic nerve stimulation. The characteristics of the cerebral potentials are dissimilar from what has recently been described by Zifko et al. (37), with sites of recording and latencies being different between the two studies. Some features of the evoked potentials described here and the polarity reversal observed between two contacts of a depth cerebral electrode placed in the vicinity of the cingulate gyrus suggest that phrenic afferents could project to the limbic cortex in humans. Before this possibility and its potential applications to pathophysiological models or to patients with acute or chronic diseases are discussed, at least two points warrant discussion. One is the reality of the neural nature the potentials described, and the other is their phrenic origin.
Nature and Origin of the Phrenic Nerve Stimulation-Related Potentials
To begin with, it should be noted that the completely artifactual nature of our results is made unlikely by the following.1) There was the absence of detectable component when, all experimental conditions being otherwise identical, no stimulation was performed. This attests that quiet breathing was not responsible for N13.2) Our study of somatosensory potentials evoked by electrical stimulation of the median nerve at the wrist consistently provided results matching classic data (cortical N20; see Significance of the N13 Cerebral Potential) (18). This substantiates the technical quality of the recordings.
Two points remain to be examined. The first one is the reality of the neural origin of the observed deflections. As is often the case in such studies, there is and there can be no proof that the potentials recorded are not far-field EMG potentials. The fact that the amplitude of N13 was not homogeneous (Fig. 3) is against this hypothesis, but it has to be noted that the tracings were relatively similar at the different recording sites. Under our experimental conditions, the diaphragm, the sternomastoid muscle, and muscles innervated by the brachial plexus could be suspected to create such an artifact. However, the time course of the diaphragmatic M wave induced by phrenic nerve stimulation was very different from that of N13, with an onset latency shorter by 1.5–2 ms and a peak latency shorter by >3 ms. N13 was present in subjects in whom the sternomastoid was not activated by phrenic nerve stimulation (e.g., subject 7) and despite the consistent absence of brachial plexus activation. When there was a sternomastoid contraction, the time course of the corresponding M wave was radically different from that of N13. Finally, in patient 1, who was specifically studied to address methodological issues, phrenic nerve stimulation did not evoke cerebral potentials despite a normal motor diaphragmatic response. Thus the hypothesis of a far-field EMG potential does not seem very probable. In addition, the shape and latency of the recorded potentials, as well as their presence at T3, T4, and Oz, were against a possible eye-blink artifact, be it electrically induced or of the startle type.
The second point to examine is the nature of the afferents contributing to the potentials evoked by phenic nerve stimulation. Indeed, we are well aware that, the stimulus through surface electrodes being all but selective, there can be no absolute proof that sensory fibers of structures other than the phrenic nerve are activated by the applied electrical stimulation. This is the reason why we have tried to collect as many arguments as possible against possible artifacts.
That the recorded potentials were at least in part due to phrenic nerve-conveyed diaphragmatic afferents is substantiated by the fact that phrenic nerve stimulations always induced a diaphragm contraction. The intensity of stimulation, comparable to the one reported by Zifko et al. (37) in their previously mentioned study, was thus above the threshold of the motor fibers. For motor and sensory nerves, this corresponds to two to three times the threshold of sensory fibers and is adequate to obtain somatosensory-evoked potentials (18). It should, however, be noted that Zifko et al. indicated that there was no significant difference in phrenic nerve stimulation-related potentials when stimulation was performed by using needle electrodes placed in the vicinity of the nerve and when it was performed transcutaneously.
Cutaneous receptors and sternomastoid and scaleni muscles.
Paraphrenic stimulation was associated with scalp deflections, the shape and latency of which had little to see with those elicited by phrenic nerve stimulation (Fig. 6). Furthermore, no marked difference in N13 was observed before and after complete subcutaneous anesthesia in the subject (7) in whom it was performed. We are aware of the fact that performing skin anesthesia in more subjects would have given more strength to this point. However, we felt that one observation was suggestive enough in view of the possible consequences of accidental phrenic nerve damage. In a similar manner, the results of paraphrenic stimulation provide arguments against a significant contribution of the muscles surrounding the phrenic nerve in the neck.
As previously mentioned, we took care to limit coactivation of the brachial plexus when stimulating the phrenic nerve. We did not encounter particular difficulties in doing so. This is probably because of the morphological characteristics of our subjects, all of whom had a relatively thin neck, and of the low stimulation intensity used, which contributed to limit the spread of the stimulus. The criteria that were used to monitor brachial plexus stimulation (seemethods) were not very sensitive. However, it is our experience that liminar and selective phrenic nerve stimulations fail to elicit EMG responses, as studied with needle electrodes in muscles innervated by the brachial plexus, and we are hence confident that the brachial plexus probably did not contribute much to the recorded potentials. This is reinforced by the marked difference in shape and latency between the “phrenic” N13 and the tracings obtained when the brachial plexus was stimulated at the Erb’s point (Fig. 6).
Because of their anatomic proximity, the vagus nerve afferents are other potential candidates to interfere with the potentials recorded after phrenic nerve stimulation. There are several arguments against this possibility. First, to our knowledge, there are no reports of vagal effects in the numerous phrenic nerve stimulation studies published with stimuli similar to the one we used (see, e.g., Refs. 3,31). In this regard, no decrease in heart rate was observed in our subjects. Second, vagus nerve stimulation-related evoked potentials are much different from those we obtained with phrenic stimulation (35). In the study by Tougas et al. (35), although recording conditions were different, these potentials had a much slower first negative component, at 71.7 ± 12.7 ms.
Finally, the observations made in patient 1 (absence of phrenic nerve stimulation-related potentials despite a normal diaphragm motor response in a patient with documented alterations in respiratory afferences; seeresults) also support the actual phrenic origin of the described potentials.
Significance of the N7 Spinal Potential
Together with a diaphragm action potential and a scalp potential in aFz, a spinal potential was evoked in seven subjects by phrenic nerve stimulation at the level of C2. Its peak latency was shorter than that of the concomitant sternomastoid motor action potential, ruling out the possibility that N7 reflected contraction of this muscle. Of note, N7 was present in subject 7 in whom phrenic stimulation was specific enough to leave EMGsm silent.
The nature of the N7 generator will certainly need further evaluation, but it can be discussed in two ways. On one hand, N7 could be interpreted as a volley ascending the dorsal column (18). On the other hand, the fact that the spinal potential probably has a horizontal dipole, as indicated by the recording of an anterior positivity (cricoid cartilage electrode) simultaneously to the posterior negativity (C2 electrode), suggests that N7 could correspond to postsynaptic activity generated in the posterior horn and depends on a medullary relay of the afferent information carried by the phrenic nerve. The dual nature of the N7 spinal generator can also be postulated (12).
Significance of the N13 Cerebral Potential
Again keeping in mind that in such studies the hypothesis a far-field EMG potentials can never be completely ruled out, there seem to be arguments of various strength for the cortical nature of the N13 generator. However, N13 probably does not correspond to a somatensory projection to the parietal cortex but, as will be seen, could rather be a reflection of a limbic representation of the diaphragm.
First, increasing stimulation frequency to 15 Hz considerably reduced the amplitude of N13 in three subjects, in whom it, in fact, disappeared. Such a phenomenon has been shown to be related to the cortical nature of a component (28), although other explanations (e.g., fiber desynchronization) are possible (28). This argument has been retained by Gandevia and Macefield (15) to support the cortical nature of the potentials evoked by intercostal muscle stimulation. The persistence of N13 with increasing phrenic nerve stimulation frequency in subject 3was, therefore, a matter of great surprise. The only difference between this subject and the rest of the group was a considerable level of training and skillfullness with respect to voluntary control of diaphragm contraction. Although this is extremely speculative, the persistence of N13 despite increasing frequency in this subject could be the reflection of possible effects of training on some characteristics of somatosensory evoked potentials (25, 29) and on the effectiveness of certain categories of synapses (36).
Second, comparison with the reasoning and discussion found in the literature about the N18 and N20 components evoked by median nerve stimulation (10) seems to us an element in favor of the cortical or immediately subcortical origin of N13. When a cephalic reference electrode at the earlobe is used, a negative component (peak 20 ms; N20) is the first to be detected after median nerve stimulation in the parietal derivations that are contralateral to the stimulation (18). It is thought to correspond to a near-field potential, namely, to have a cortical origin. Conversely, when evoked potentials from median nerve stimulation are recorded by using a montage including a noncephalic reference, N20 is preceded by an earlier negative component (N18), which is detected all over the scalp (11). It is generally admitted that N18 is a far-field component arising from activation of neuronal structures in the brain stem or in the thalamus (18). In the four subjects in whom we stimulated the median nerve at the wrist and recorded cerebral potentials with the same montage used to detect N13 (earlobe reference), we could not find N18. It is, therefore, likely that our recording conditions were not appropriate to record far-field potentials, and thus that the source of the phrenic N13 is “nearer” than the source of N18. Of note, N13 corresponds to a rather short latency projection. It is 7–8 ms shorter than the median nerve latency of 20 ms, although, considering the differences in path length and assuming similar types of paths (polysynaptic), the difference in latencies should not be >3–5 ms (37). This could be a sign that a brain structure deeper than the somatosensory cortex is recorded. It must be emphasized that this reasoning assumes similarities between N13 and N18 and N20 in terms of origin or electrical field behavior, which may or may not exist.
Third, the observations in patient 2point to the limbic cortex and more precisely the cingulate gyrus as a projection of phrenic-mediated afferents. In this patient, all the potentials recorded at the different electrode sites, be they positive or negative, had peak latencies similar to these observed on the scalp in healthy volunteers. Unfortunately, direct comparison could not be made because the scalp was not accessible in this patient at the time of the study because of bandages that the clinician and nurses in charge of the patient did not want to be removed for reasons of asepsis. After the electrodes had been removed, the patient did not wish to participate in a research study again. It should be noted that auditive and somatosensory evoked potentials had been studied before the implantation of depth electrodes and had been found normal on both sides. It is, therefore, likely that the patient’s disease had no important effect on evoked potentials. The lower contacts of the frontal electrodes recorded a positive potential with a peak latency of ∼13 ms (P13), and a synchronous polarity reversal could be observed at contact 5 (N13) (Fig. 2). Therefore, the source of the P13-N13 dipole must have laid in a plane passing betweencontacts 4 and5. The fact that the four posterior contacts of the right temporal electrode recorded a negative potential with a peak latency of 11.6 ms, whereas the four anterior ones were silent, could indicate that the latter were located exactly in the plane of the electrical generator and that the former were above it. Combined with the frontal electrodes pattern, this would define the direction of the plane in such a way that it encompasses, among others, four anatomic structures that could possibly be relevant as the P13-N13 electrical generator. These structures are the thalamus, the insular cortex, the hippocampus, and the cingulate gyrus. The possibility of recording evoked potentials originating in the thalamus is generally considered not very likely (18). The absence of any positive potential ∼12–13 ms at the temporal electrodes makes the insular cortex as the source of N13 a weak hypothesis. A similar reasoning applies to the hippocampus, for which a polarity reversal would have been expected between two of the contacts of the temporal electrodes. The anterior part of the cingulate gyrus, therefore, remains the main hypothesis for the origin of N13. This is substantiated by the anatomic position of the frontal electrode contacts 4 and5.
Relevance of a Phrenic Projection to the Cingulate Gyrus
A deep and median location for the source of N13 would explain some of the differences in our observations and those by Zifko et al. (37). These authors, using a C3-P3-CP1-CP3 montage with a Fz reference, found a N17 phrenic nerve evoked potential with a maximal amplitude at CP3. They hypothesized that this potential corresponded to the projection of phrenic afferents to a sensory cortical representation of the diaphragm located close to its primary motor area (5, 13, 16, 22). The main source of difference between the two studies is probably the type of montage used. The montage chosen by Zifko et al., with relatively close electrodes placed over the vertex and a median reference, is probably more appropriate to record near-field potentials than is the montage used in our study, which would be more efficient to record potentials arising from a relatively more distant structure. The “cingulate gyrus hypothesis” would also explain the shorter latency observed in our study [13 vs. 17 ms in (37)] and why N13 could be observed with a relatively similar pattern at the different scalp electrodes.
A projection of phrenic afferents to the cingulate gyrus is consistent with the current concept that respiratory muscle afferents involve two separate pathways (8). In the first one, which would correspond to the potentials recorded by Zifko et al. (37), neural information arising from mechanoreceptors enters the spinal cord, ascends in the dorsal column, relays in the brain stem, projects to the thalamus, and is transmitted through a thalamocortical pathway to the sensorimotor cortex. The function of this pathway could relate to the proprioceptive control of respiratory muscles and the integration of movements originating in the motor cortex. The second pathway involves ascending afferents conveyed by the vagus nerve and its branches and possibly phrenic afferents. They relay from the brain stem to the amygdala before projecting to the mesocortex. In line with the involvement of the amygdala in visceral nociceptive processes (4), this circuit may deal with respiratory nociception and some behavioral aspects of respiration. Several facts link the diaphragm to the mesocortex. First, stimulation of the cingulate gyrus can produce various respiratory effects in animal (19), and the existence of a respiratory excitatory zone in this region has been evidenced in humans (27). There are case reports that describe the association of respiratory events with neurological lesions in the cingulate gyrus determining emotional disorders (23, 34). Second, two recent studies using different techniques have provided further arguments for the involvement of the human limbic system in the ventilatory response to CO2. Functional magnetic resonance imaging allowed Gozal et al. (17) to show that CO2 response was associated with changes in brain stem pattern of activation and, in some subjects, with changes in “limbic regions of the rostral brain.” Positron emission tomography also demonstrated significant increases in regional cerebral blood flow in the cingulate gyrus during CO2-stimulated breathing in conscious humans (6). Because of the very principle of the CO2 rebreathing method, which undissociatively tests both the metabolic afferent messages and the corresponding efferent responses, the “sensitive” or “motor” nature of the areas delineated by functional imaging techniques cannot be ascertained. With our observation taken into account, it is suggested that some of the changes depicted in the limbic system by these techniques corresponded to its afferent activation, after transduction of the hypercapnia associated changes in respiratory mechanics into a neural code by diaphragm receptors.
In conclusion, we believe that the potentials described in this study could well be the reflection of a projection of phrenic afferents to the limbic cortex in humans. Caution is obviously needed in interpreting these findings, which warrant further investigation and the use of different models. However, they could contribute to a better understanding of the role of respiratory muscles afferent in nociceptive perceptions such as dyspnea.
The authors are extremely grateful to Drs. Claude Adam and Didier Dormont for the invaluable help that they provided by allowing them to study patient 2 and for their contribution to the description of the case and of the magnetic resonance imaging data. We thank Profs. Alain Lockhart and Alain Harf for their support and Dr. François Bellemare for stimulating discussions about phrenic nerve stimulation evoked potentials a few years ago. Finally, we are truly indebted to Dr. Duncan Newton for help with English.
Address for reprint requests: T. Similowski, Service de Pneumologie et de Réanimation, Groupe Hospitalier Pitié-Salpêtrière, 47-83 Blvd. de l’Hôpital, 75651 Paris cédex 13, France (E-mail:).
This work was supported in part by the Association pour le Développement et l’Organisation de la Recherche en Pneumologie (Paris, France), and C. Straus was funded by the Fondation pour la Recherche Médicale (Paris, France).
- Copyright © 1997 the American Physiological Society