Main

Congenital central hypoventilation syndrome (CCHS) is characterized by multiple respiratory deficiencies, including diminished drive to breathe during sleep (1) and reduced ventilatory responsiveness to hypercapnia (24). Affected individuals do not alter respiratory rate to a forehead cold pressor challenge and show diminished respiratory-related heart rate variation to that stimulation (5); reduced respiratory/heart rate coupling is also found in baseline and sleep conditions (6). The cold pressor findings suggest a nonchemoreceptor respiratory reflex deficit in the syndrome, the trigeminally mediated dive reflex, and indicate failure of integration of respiratory motor output with afferent input. A portion of that failure could stem from autonomic/respiratory interaction deficits, and these deficits may partially derive from maldevelopment of neural visceral areas (e.g. dorsal medullary sites) as a result of mutations of the PHOX2B gene, found in a high proportion of patients with CCHS (7,8). Both chemoreceptor and cold stimuli perception are at least partially intact in CCHS, because affected children arouse from sleep to high CO2 (9) and have perception of forehead cooling. Because the nucleus of the solitary tract (NTS) plays a significant role in both chemoreception and temperature mediation (10), damage to this dorsal medullary structure, as well as its afferent processes, must be incomplete in CCHS, yet ventilation to chemoreceptor challenges is seriously impaired. Voluntary and exercise-related respiratory efforts in CCHS are largely preserved (1113), suggesting that integration of afferent activity with respiratory motor output, rather than a primary chemoreceptor defect, is involved. A ventilatory challenge that did not include chemoreceptor components would be useful to investigate this possibility.

The cold pressor challenge is a means to examine deficient integrative respiratory mechanisms and cardiovascular responses in CCHS, bypassing input from the principal autonomic ganglia that may be affected in the syndrome. The challenge can be administered rapidly, exerts a profound effect on respiratory and cardiovascular patterning, and is noninvasive. We hypothesized that neural responses to a forehead cold pressor challenge, as assessed by functional magnetic resonance imaging (fMRI), would show abnormal patterns in regions that are responsible for the integration of cold afferent stimulation with respiratory and cardiovascular output.

METHODS

Patients.

Thirteen children who had a diagnosis of CCHS according to standard criteria (14) and 14 control subjects (age- and sex-matched with one additional female control) participated in the study. The age of patients with CCHS was 10.9 ± 2.3 years (range 8–15) and of control subjects was 11.0 ± 2.2 years (range 8–15). Six patients with CCHS and seven control subjects were female. Patients who had CCHS and Hirschsprung's disease were excluded to avoid confounding effects from deficient neural pathways of the viscera. All subjects had tracheostomies and were ventilator dependent during sleep but not during waking; the tracheal openings were closed throughout the procedures. This study was approved by the institutional committee on human research; informed consent was obtained from each subject's guardian.

Forehead cold pressor.

Each subject lay supine in the MRI scanner while a baseline recording was collected for 150 s, and for a second acquisition period beginning with a 30-s baseline, after which a cold (3°C) deuterium-filled bag was lowered to the subject's forehead for 120 s. Deuterium, rather than H2O, was used to minimize high signal intensity contributed to images by water. Physiologic signals including blood oxygen saturation (SaO2), the ECG, and end-tidal CO2 (ETCO2) were recorded simultaneously with the fMRI signal (5); the ETCO2 and SaO2 values were obtained in a subset of 13 control subjects and six patients with CCHS, with data from the remaining subjects not collected because of technical issues.

Scanning.

Data were collected using a GE 1.5 Tesla Signa MRI scanner. Masking tape across the forehead and foam pads on either side of the head were used to reduce head motion. For fMRI scans, a gradient echo echo-planar imaging (EPI) protocol (repetition time = 6 s, time to echo = 60 ms, flip angle = 90°, field of view 30 × 30 cm, no interslice gap, and voxel size 2.3 × 2.3 × 5 mm) was used. The EPI protocol uses the blood oxygen level dependent (BOLD) intrinsic contrast to highlight changes in neural activity (15). For each subject, baseline and challenge series were collected. Each series consisted of 24 volumes of 20 oblique image slices collected over 144 s (24 s baseline/120 s challenge). Conventional spin echo T1-weighted images (repetition time = 500 ms, time to echo = 9 ms, field of view = 30 × 30 cm, no interslice gap, voxel size 1.2 × 1.2 × 5 mm) were acquired at the same location and orientation to aid in anatomical identification.

Data analysis.

Breathing and heart rates were derived from the physiologic signals and analyzed for changes relative to baseline during the challenge and for differences between patients with CCHS and control subjects. The principal physiologic findings were reported elsewhere (5).

The MRI data were analyzed using SPM (16) and custom software. Preprocessing of the EPI images consisted of slice timing correction, correction for possible motion, spatial normalization to the Montreal Neurological Institute template, spatial smoothing, and removal of any global effects (17). A two-step spatial normalization process was used, as described previously (18,19). The T1-weighted images were also spatially normalized, and an average of all subjects' T1 volumes was used for display purposes.

The preprocessed functional images were analyzed using a priori defined volumes of interest (VOIs) and a whole-brain cluster analysis. VOI analysis allows for examination of responses in structures regardless of the pattern of response, e.g. effects of timing of large, transient responses of interest to understanding breathing interactions may be obscured without trend analysis. The subject-by-subject outlining of relevant areas also ensures that individual differences in structure shape and size are taken into account. In contrast, cluster analysis allows for examination of the whole brain, but the procedure is based on an a priori pattern of expected response. The response pattern for cluster analysis can be modeled, for example, with an “on” and “off” condition (stimulation or no stimulation).

VOI analysis.

Fifteen structures were selected on the basis of classically suggested associations with respiratory or motor control, namely the amygdala; head of caudate; hippocampus; insula; lentiform nuclei; fastigial and dentate nuclei of the cerebellum; vermis; dorsal and ventral pons; dorsal and ventral midbrain; and dorsal, medial, and ventral medulla. Some of these areas, including the dorsal medulla and pons, are particularly affected by PHOX2B expression (20). Each VOI was drawn manually on a subject-by-subject basis, using a baseline-normalized EPI image as reference; examples have been published previously (21). Regions of primary sensory projection of the V3 trigeminal (forehead) area in the ventral posterior medial nucleus of the thalamus and somatosensory parietal cortex were also selected. Bilateral structures were assessed separately by laterality, and the insula was further separated into anterior and posterior areas.

For each VOI, the average time trend of the voxels from preprocessed images was calculated for each subject. These individual time trends were averaged for patients with CCHS and control groups and plotted with between-subject SE bars. The subjects' time trends were analyzed for responses relative to baseline and for differences between groups using repeated measures ANOVA (22).

Cluster analysis.

Cluster analysis consisted of a whole-brain assessment for responses that differed between groups. The EPI images for each subject were modeled on a voxel-by-voxel basis to an earlier determined pattern of continuous blood pressure response to the cold pressure challenge, as measured in six adults and smoothed using a moving-average filter. For baseline, the pattern was a flat line; baseline consisted of the entire baseline series and the first four scans of the challenge series. Arbitrary offsets of overall absolute image intensities between baseline and challenge series were partitioned in the model. A second-level analysis, consisting of a two-sample t test of the parameter estimates in the model at each voxel (23), was performed to detect any regions where the response differed consistently between groups. Because neural responses were not expected to precisely match the model, a relatively low threshold of p < 0.01 (minimum size 6 voxels) was used to allow detection of regions where responses that approximated the model differed between groups. The nature of the response differences within selected clusters was further examined by plotting the average group time courses of all voxels within each cluster.

RESULTS

Physiologic measures.

Respiratory and cardiac rate and variability measures have been described earlier (5); briefly, respiratory rates and respiratory-related heart rate variation did not increase to the cold challenge in patients with CCHS, unlike control subjects, but low-frequency variability increases appeared in both groups. Significantly higher ETCO2 and lower SaO2 levels appeared in patients with CCHS compared with control subjects over the entire recording period, but no changes emerged across the challenge for either group (Table 1).

Table 1 ETCO2 and Sao2 measured during baseline and during the first and second minutes of the challenge, with SE in a subset of subjects (13 control and six CCHS)

VOI analysis.

Summary data for 23 VOIs are shown in Table 2; time trends from a subset of sites are shown in Fig. 1. Several sites showed lateralized responses in both groups (Table 2); bilateral comparisons were made for regions that showed lateralized responses, whereas regions with similar responses on the left and right side or medial structures were presented as one structure. The most common finding was an increased signal in the control group with diminished extent of responses in the CCHS group. Thus, CCHS responses tended to be less pronounced, more transient, absent, or a modest change in the opposite direction compared with control subjects, as indicated by the data from the head of caudate, lentiform nuclei, cerebellar dentate and fastigial nuclei and vermis, and dorsal and ventral midbrain in Fig. 1. The region of forehead representation in the primary sensory cortex showed an initial decline in both groups but a return to baseline in CCHS patients, with a sustained fall in control subjects. The ventral posterior medial nucleus of the thalamus, the primary somatosensory nucleus for the head, showed a signal increase in both groups, but the increase was larger in control subjects over patients with CCHS. In contrast to the control pattern, the medial and ventral medulla showed a larger transient increase in patients with CCHS, lasting 60 s. The dorsal and ventral pons showed no group differences. Of the VOIs not in Fig. 1, the left amygdala showed comparable increases in both groups, but the right side showed no change in either group. The hippocampus and posterior insula showed lateralized responses. The left anterior insula remained at baseline levels in both groups, and the right anterior portion showed a comparable decline in patients with CCHS and control subjects. No change occurred in either group in the dorsal medulla.

Table 2 Summary of VOI analysis
Figure 1
figure 1

Average time trends of 14 VOIs for CCHS and control groups, with SE bars. Time points of significant response (increase or decrease) relative to baseline for each group are indicated, as well as time points of significant group difference (repeated measures ANOVA p < 0.05). Baseline is defined as the combined baseline series and initial 30 s of the challenge series. VPM Thal, ventral posterior medial thalamus; Par. Som. Cortex, parietal somatosensory cortex (forehead).

The lentiform nuclei showed a transient response, with the control group having higher signals during the first minute. The dentate and fastigial nuclei patterns were notable in that the group differences developed principally in the last minute of the challenge.

Cluster analysis.

Table 3 shows all clusters of significant group difference, and Fig. 2 illustrates selected clusters overlaid on an average of all subjects' anatomical T1-weighted images. Figure 3 shows time trend patterns of signal changes from selected clusters.

Table 3 Clusters of voxels with a significant (p <0.01) difference in response between groups (14 control and 13 CCHS subjects), where the response is modeled as a typical pattern of blood pressure increase
Figure 2
figure 2

Clusters of group difference in response, as determined by modeling image series time course to a typical pattern of blood pressure increase, color-coded according to statistical significance (t value) and overlaid onto the average of all 27 subjects' anatomical T1 images. Regions of control > CCHS are in warm (yellow-red) colors and regions of CCHS > control are in cool (blue-green) colors, coded according to significance level (key bottom right). Locations of axial (A), sagittal (B), and coronal (C) views are indicated in sketch at bottom left. Note that the “warm” vs “cool” color coding refers to the group difference direction, not difference in extent of response, i.e. the blue “CCHS > control” colors in the cortical regions indicate a response decline in control cases but not necessarily greater extent of response in patients with CCHS compared with control subjects (as illustrated in Fig. 3).

Figure 3
figure 3

Time trends from selected clusters of group difference (Fig. 2 and Table 2); mean (± SE) for control subjects and patients with CCHS shown for baseline and challenge series. Cluster overlays follow the same conventions as Fig. 2.

The cluster analysis principally showed larger and more sustained signal responses in control subjects than in patients with CCHS, as with the VOI analysis. Cortical regions where control subjects demonstrated greater responses than patients with CCHS included the superior surface of the right temporal lobe (Fig. 2 C5–C6), and the mid- and posterior cingulate (B3–B5, C4–C6). Cortical regions where control responses declined while CCHS signals remained unchanged emerged in the frontal cortex extending to the anterior cingulate (B3–B5, C7), the superior frontal cortex (C7–C8), and the left prefrontal cortex (A6–A7, C7). The time courses of the clusters of group difference in cortical regions showed differences throughout the challenge period (Fig. 3). The right posterior insula showed increased signals in control subjects over patients with CCHS (A4–A5), together with mid portions of the hippocampal formation (B7, C5), dorsal thalamus (B2, C5), and dorsal hypothalamus (B2, C5). Signals in the dorsal hypothalamus and insula increased early in the response and later declined in control subjects but not patients with CCHS (Fig. 3). In the cerebellum, the control group showed several regions of greater signal over the CCHS group, including the vermis (Fig. 2, A1, B3–B4), dentate nucleus (A2, B2), and quadrangular lobule of the cerebellar cortex (B3, C1–C3). The time course of the cerebellar cortex cluster differed between groups only in the second minute of the challenge, with similar signal increases in the first minute occurring in both groups (Fig. 3).

DISCUSSION

Sensory, motor, and integration areas showed deficits in patients with CCHS to the cold pressor challenge. Diminished cardiac variability and breathing responses emerged in the patient group, and, whereas control subjects showed significant responses within discrete sites in the cerebellum, basal ganglia, and cortical regions, the most common deficit found in patients with CCHS was reduced change in signal within these areas. However, larger signal increases were found in medial medullary inhibitory areas and the ventral medulla of patients with CCHS. A region expected to be affected by mutations in PHOX2B transcription, the dorsal medulla, showed little response in patients with CCHS or control subjects. We speculate that the diminished and altered physiologic and neural responsiveness in CCHS results from enhanced action of medullary inhibitory areas, from aberrant cerebellar influences on projection sites, particularly in the frontal cortex and indirect to basal ganglia, and from a generalized unresponsiveness in neural sites as a result of deficient cardiovascular reactivity. PHOX2B mutations, because of their effect on autonomic processes, may play a significant role in the reduced reactivity.

Timing.

The cold application normally elicits an increase in breathing rate and enhances rapid, respiratory-related heart rate variation. Both characteristics were missing in patients with CCHS, although slower (<0.1 Hz) heart rate variation changes were similar in the two groups. Respiratory influences on heart rate were already diminished during baseline conditions in the patient group (5); a failure to increase rapid heart rate variability changes to hypercapnia in patients with CCHS has been reported (17). Cerebellar structures have been implicated in other rapid-physiologic-adjustment deficits, including somatic movements (24) and cognitive switching tasks, such as those in autistic disorders (25), a syndrome associated with reduced Purkinje cell size (26) and impaired cerebellar-cerebral projections (27). Cognitive disorders of an autistic nature do not always accompany CCHS, although patients with CCHS do show below-average motor hand coordination performance (28). The aberrant cerebellar responses in CCHS appeared in dorsal anterior and deep nuclei structures, in addition to vermal areas usually affected in autism (29). The diminished rapid heart rate changes to breathing suggest that the loss of rapid coordination of physiologic functions in CCHS extends to autonomic aspects.

PHOX2B targets.

An unexpected finding was the absence, in either group, of responses from the dorsal medulla, including the NTS or the dorsal pons, containing the locus coeruleus. The absence of a response in dorsal medullary areas probably stems from the nature of the cold pressor stimulation, which was not a chemoreceptor challenge, and apparently did not recruit NTS neural action in a major way in either control subjects or patients with CCHS. Both the NTS and the locus coeruleus would be expected to be affected by mutations of PHOX2B (20,30). Mutations would likely also affect autonomic ganglia associated with the 7th, 9th, and 10th cranial nerves (20); the actions of these peripheral structures, however, could not be assessed in this study. Localized responses within the nucleus ambiguus, the primary effector nucleus for rapid cardiac rate changes (31), and a branchial arch motor nucleus likely to be affected by PHOX2B developmental issues, also could not be examined with the resolution of the current MRI procedures. It is likely that defects in sympathetic ganglia resulting from PHOX2B mutations contribute to aberrant sympathetic patterns found in CCHS. The loss of cerebrovascular reactivity, found earlier by assessing global magnetic resonance signal changes over the brain (32), could also stem from such mutations.

Cerebellum.

Cerebellar Purkinje neurons normally inhibit deep nuclei; regions within the cerebellar cortex, dentate, and fastigial deep nuclei showed higher signals in control subjects, presumably representing enhanced Purkinje influences on the deep nuclei that project to more rostral sites. Excitation of the fastigial nucleus augments breathing, and lesions reduce the ventilatory response to hypercapnia and hypoxia (3335). Cardiovascular functions are also modified by fastigial action (36,37). The impaired recruitment of the fastigial nucleus to the cold pressor challenge may contribute to the failure to respond appropriately with augmented breathing and respiratory-related heart rate variation.

The dentate nucleus also showed signal deficits in patients with CCHS late in the challenge. The dentate nucleus, once considered a motor-only–related nucleus, contains a nonmotor portion that projects to thalamic and prefrontal and dorsal parietal cortex regions and apparently serves cognitive, decision-making, or visual-spatial functions (38). We could not differentiate motor from nonmotor portions of the dentate nucleus with the resolution offered by the scanner. The normal interaction of cerebellar nuclei with basal ganglia, frontal motor, and somatosensory parietal cortex likely results in the enhanced caudate and lentiform signals in control cases; some of the timing issues in CCHS may be related to cerebellar/basal ganglia deficits.

The deficiencies in cerebellar responses may partially result from repetitive exposure to hypoxia, a consequence of ineffective ventilation during sleep or relaxation periods in these children. Climbing fibers from the inferior olive are especially sensitive to hypoxia, with resultant damage to Purkinje cells (39), which normally inhibit deep nuclei (40). The release of inhibition to deep cerebellar nuclei, the major output from the cerebellum, can exert considerable influences on more rostral projection sites.

Cingulate and insula.

The cingulate cortex in patients with CCHS showed diminished responses and was of particular interest. Unlike the response to expiratory loading, in which nearly the entire structure showed group differences (21), only a defined region within the mid-to-caudal cingulate showed a lower response in CCHS patients. The affected cingulate region borders an area previously identified with sympathetic nervous system activation (41). The patients with CCHS did not differ markedly in response to the cold pressor challenge on slow variation in heart rate, although more rapid, presumably parasympathetically related variation was substantially diminished. The right insula deficits may act in concert with the cingulate cortex to mediate aberrant autonomic responses; insular structures show significant lateralized effects on autonomic outflow, with the left insula principally involved in parasympathetic control and the right insula in sympathetic action (42,43). Other cardiovascular syndromes with substantial loss of respiratory-related variation, such as adult heart failure (44), also show considerable right insular damage (45), and adult patients with obstructive sleep apnea show aberrant right insular responses to cold pressor challenges (19). Lateralization of brain responses to autonomic challenges in adult control subjects has been noted earlier (46,47) and occurs in multiple limbic structures. The collective findings emphasize that autonomic control patterns may be regulated by one side of the brain over the other, an issue of theoretical importance and of practical significance in conditions of early neural damage.

Enhanced responses in CCHS.

The patients with CCHS did not show uniform patterns of reduced signal response over control subjects. Both the medial and the ventral medulla showed enhanced responses in patients with CCHS over virtually no change in control subjects. The motor suppression areas of the medial medulla (4850) may be exerting inhibitory or dysfacilitatory control over a number of neural sites, contributing to the muted or absence of responses in multiple sites and leading to reduced respiratory and cardiovascular action. The medial medullary areas serve important roles in maintaining motor tone relative to blood pressure levels and interact closely with vestibular and cerebellar projections in that role (51). Influences from dorsolateral pontine regions on the medial medulla have been suggested as regulating state-related medial medullary atonia (49); however, little change was found in dorsal pontine areas between the two groups here. We speculate that the medial medullary responses likely arose from cerebellar-medial medullary interactions (5255) in these waking studies. The influences from medial and ventral medullary areas may assist in mediating diminished CCHS responses through motor control areas at multiple levels of the neuraxis, including the basal ganglia (left and right lentiform nuclei) and midbrain sites.

Muted responses in CCHS.

The greatest proportion of regional differences between patients with CCHS and control subjects developed from muted, shorter term, or absent signal changes in patients with CCHS in neural regions where signals in control subjects rose. Several processes may contribute to the muted responses, among which is the possibility of reduced overall cerebral reactivity in patients with CCHS; such a possibility stems from the demonstration of reduced global BOLD signal to ventilatory challenges in the syndrome (32). The enhanced signals in unique sites within patients with CCHS, however, suggest that the reduced responsiveness does not result entirely from reduced cerebral reactivity. The frontal cortex time-trend plots and the demonstrated projections to those sites suggest that components of the responses derived from influences of more active cerebellar deep nuclei in control subjects, resulting in lowered signals, while values in patients with CCHS remained the same or rose slightly. The primary somatosensory cortex showed an immediate response in the same direction and extent as the control group but no sustained response. Thus, an initial sensory cortical response seems to be comparable in the two groups, but influences from cerebellar and other sources that mediate the integrative sensorimotor aspects to cold may affect the late cortical pattern. The absence or diminished late response in certain regions of patients with CCHS may reflect enhanced responses of other sites, particularly medullary inhibitory areas described earlier.

Limitations.

It was not possible to separate neural abnormalities that resulted from potential hypoxic damage from those that were caused by preexisting developmental deficiencies. Recent progress in MRI technologies may offer the possibility of such determination.

The 6-s period required to acquire each functional image volume limited the ability to measure very rapidly changing neural patterns. Thus, patterns of neural responses related to the diminution of very fast heart rate responses to breathing or other sudden physiologic changes were restricted. Such patterns could be examined by reducing the field of view and by implementing recently developed faster scanning procedures; the objective of this study was to provide a survey of potential changes over as large a brain area as possible.

The cold application has the potential to modify the global cerebral circulation with resulting global effects in the BOLD signal. We addressed this issue by removing all global components from the data (17), leaving the regional responses to the stimulation.

The findings represent data from a restricted sample, namely patients who were ventilator dependent only during sleep. Neural responses may differ from patients who are continuously ventilator dependent and thus might exhibit even more extensive dysfunction. However, examination of such cases was precluded by logistic issues of scanner safety.

SUMMARY

Localized areas of the brain, principally involving cerebellar, medullary, basal ganglia, frontal, insular, temporal, and cingulate cortex regions, showed different response patterns to facial cold stimulation, a nonchemoreceptor challenge, in patients with CCHS compared with control subjects. The aberrant responses included muted or no change in signals compared with significant responses in control subjects or an increase in signals where none were normally present and may underlie the failure to elevate respiratory rate and enhance respiratory-related heart rate variation to the manipulation. Although PHOX2B mutations likely contribute to multiple autonomic deficiencies in CCHS, a structure expected to be affected by such mutations showed little change in response to the cold pressor application. Neural responses to the cold pressor challenge have the potential to reveal deficiencies in other processes that may not be targeted directly by the mutation. Among the affected sites, primary sensory-mediating structures in the thalamus and somatosensory cortex showed reduced signal changes in patients with CCHS. Some of the reduced signal responsiveness may have resulted from diminished cerebral reactivity in patients with CCHS; other contributions are likely to have developed from medial medullary inhibitory or dysfacilitatory influences or from reduced control by cerebellar projections. The increased responses in medullary sites argue against a generalized loss of cerebral reactivity and suggest a selective active suppression of aspects of breathing and rapidly changing integrative mechanisms.